Introduction

In the mammalian retina, a rapid and coordinated immune response to infection or injury is important for maintaining tissue homeostasis. This is especially critical in the eye since mature retinal neurons do not typically regenerate, resulting in long-term functional losses for the host.^1,2 The retina is considered immune-privileged and is equipped with a resident population of innate immune cells including microglia.^3,4 Under homeostatic conditions, microglia are distributed primarily in the inner retina, residing within nerve fiber layer (NFL), inner plexiform layer (IPL) and outer plexiform layers (OPL).^5,6, generally avoiding the nuclear layers of the retina. Microglia tile the retina and like their counterparts in the brain, exhibit long, thin cellular processes that continually probe the neuro-glial microenvironment.^4,7,8

In addition to phagocytosing debris^9,10, regulating synaptic maintenance^11,12 and removing dead tissue^13,14, microglia can secrete chemokines to recruit other leukocytes to help fight infection and repair damaged tissue.^4,15–17 For a number of injuries, one of the first systemic responders recruited and activated by microglia are neutrophils.^18,19 Neutrophils comprise a large fraction of leukocytes in murine blood (7.4 – 33.9%, “young” C57BL/6J mice²⁰). They assist in maintaining tissue homeostasis by neutralizing foreign agents, regulating the immune response, and processing dead tissue via phagocytosis^21,22. Beyond these well-known functions, they are also implicated in becoming activated in conditions of shock and trauma²³. Under inflammatory conditions, the spatiotemporal interplay between microglia and neutrophils is poorly understood. A missed window of interaction is highly problematic in histological study where a single time point only reveals a snapshot of the temporally complex immune response, which changes dynamically over time. Here, we use in vivo imaging to overcome these constraints.

Documenting interactions between immune cells in the retina over time has been challenged by insufficient resolution and contrast to visualize single immune cells in the living eye. Recently, advances in AOSLO imaging have provided micron-level resolution and enhanced contrast for imaging individual immune cells in the retina^7,24, and do so without requiring extrinsic dyes. This is achieved by using near infrared (NIR) light, to which the eye is less sensitive, and combining this with phase-contrast approaches^25–27. In addition, AOSLO provides multi-modal information from confocal reflectance, phase-contrast and fluorescence modalities, which can reveal a variety of cell types simultaneously in the living eye. Here, we used confocal AOSLO (<6 µm axial resolution^28,29) to track changes in reflectance at cellular scale. Phase-contrast AOSLO provides detail on vascular wall, single blood cells^30–32, PR somata³³, highly translucent retinal cells, and is particularly suited to image resident and systemic immune cells.^7,24 Simultaneous fluorescence imaging provides the ability to study fluorescently-labeled cells^26,29,34 and exogenous dyes^30,35 throughout the living retina. These modalities used in combination have recently provided detailed images of the retinal response to a model of human uveitis.^24,36 Together, these innovations now provide a platform to visualize, for the first time, the dynamic interplay between many immune cell types, each with a unique role in tissue inflammation. We combine these innovative modalities with conventional histology as well as commercial OCT to reveal the progressive nature of the cellular response to acute retinal injury.

As this work is in its early inception, we focus on one of many immune cell interactions, the interplay of two critical first responders to injury: neutrophils and microglia. To begin unraveling the complexities in this response, we examine the innate immune response to a deep retinal laser ablation model. Using AOSLO, we track and characterize the spatial, axial and temporal changes in microglia, neutrophils and retinal structure within hours, days and months after acute laser exposure.

Methods

Mice

All experiments herein were approved by the University Committee on Animal Resources (protocol #: UCAR-2010-052E) and according to the Association for Research in Vision and Ophthalmology statement for the Use of Animals in Ophthalmic and Vision Research as well as institutional approvals by the University of Rochester. C57BL/6J (#000664, Jackson Labs, Bar Harbor, ME) mice were used to track the retinal phenotype after laser exposure.

Heterozygous CX3CR1-GFP (#005582, Jackson Labs) mice were used to track GFP-expressing microglia. Mice with heterozygous transgenic expression of tdTomato in neutrophils (“Catchup” mice) were provided by the foundry lab of M. Gunzer.³⁷ 18 mice (10 male, 8 female) in total were used for in vivo imaging. 10 additional mice (5 male, 5 female) were used for ex vivo histology. Age for all mice used for this work were between post-natal weeks 6 - 24.

Preparation for in vivo imaging

Mice were anesthetized with intra-perotineal injection of ketamine (100 mg/kg) and xylazine (10 mg/kg). The pupil was dilated with 1% tropicamide (Sandoz, Basel, Switzerland) and 2.5% phenylephrine (Akorn, Lake Forest, IL). A custom contact lens (1.5 mm base curve, 3.2 mm diameter, +10 diopter power, Advanced Vision Technologies, Lakewood, CO) was fitted to the eye. For a subset of experiments, 50 mg/kg fluorescein (AK-FLUOR, Akorn, Decatur, IL) was administered by intra-perotineal injection to confirm vascular integrity and perfusion status after injury. During AOSLO imaging, anesthesia was supplemented with 1% (v/v) isoflurane in oxygen and mice were maintained at 37°C via electric heat pad. Eye hydration was maintained throughout imaging with regular application of saline eye drops (Refresh tears, Allergan, Sydney, Australia) and lubricating eye gel (Genteal, Alcon Laboratories Inc. Fort Worth, TX) in combination with rigid contact lens described above. Mice were placed in a positioning frame with 6-degrees of freedom to allow for stable animal positioning and to aid in retinal navigation relative to the beam of the AOSLO and OCT.

In vivo AOSLO imaging

Four light sources were used for AOSLO imaging. A 904 nm diode source (12 µW, Qphotonics, Ann Harbor, MI) was used for wavefront sensing. A second 796 nm superluminescent diode (196 µW, Superlum, Cork, Ireland) was used for reflectance imaging, including confocal and phase contrast modes.^29,30 A third 488 nm light source (56 µW, Toptica Photonics, Farmington, NY) was used to visualize GFP-positive microglia in CX3CR1-GFP mice. A fourth 561 nm light source (95 µW, Toptica Photonics) was used to visualize tdTomato-positive neutrophils in Catchup mice. All light sources were fiber-coupled and were axially combined through the AOSLO system in a free-space design²⁹. Fast (15.4 kHz) and slow (25 Hz) scanners create a raster scan pattern, which is relayed through a series of afocal telescopes to and from the eye. A Shack-Hartmann wavefront sensor (consisting of a lenslet array and a Rolera XR camera, QImaging, Surrey, Canada) measures the aberrations of the eye and a 97 actuator deformable mirror (ALPAO, Montbonnot-Saint-Martin, France) provided the wavefront correction. Reflected 796 nm light was collected with a photomultiplier tube (H7422-50, Hamamatsu Photonics, Hamamatsu, Japan). All confocal reflectance images were captured with a 30 µm pinhole (1.3 Airy Disc Diameters, ADD). Phase-contrast was achieved by displacing the pinhole relative to the principal axis of the detection plane as previously described.²⁶ Fluorescence was captured with a photomultiplier tube (H7422-40, Hamamatsu) either coupled with a 520Δ35 band-pass filter (FF01-520/35-25, Semrock, Rochester, NY) for GFP emission, or a 630Δ92 band-pass filter (FF01-630/92-25, Semrock) for tdTomato emission. All fluorescent images were captured with a confocal 50 µm pinhole (2.1 ADD). Image field sizes were either 4.98° x 3.95° or 2.39° x 1.94°. NIR and visible imaging channels were made coplanar, compensating for longitudinal chromatic aberration by independently focusing each light source so it projected onto the same axial structure in the retina. Through-focus acquisitions were acquired by sequentially changing the focus from nerve fiber layer through to the PR outer segments by using the defocus term on the custom adaptive optics control software. This strategy allowed depth imaging while keeping all visible and NIR sources conjugate to the same axial plane.

As described previously, red blood cell (RBC) imaging was achieved by combining phase-contrast imaging with a strategy to arrest the slow galvanometer scanner and let the resonant scanner project a single “line” on the retina. This enabled RBC flux imaging by positioning this line orthogonal to the direction of flow for single capillaries. As blood cells moved through capillaries, they were “self-scanned” producing images of RBCs in space/time images³⁰ (Figure 4 – figure supplement 1b + f). Scan angle for this modality was 0.71°.

PR laser damage model and post-injury time points for imaging

488 nm light (Continuous wave Laser diode, ±4 nm bandwidth, Toptica Photonics) was used to create an acute laser injury. Laser light was projected through the AOSLO and focused onto the PR outer segments. To create damage, 785 µW of 488 nm light was focused onto the PR outer segments for 3 minutes in a single line on the retina subtending 24 x 1 µm to concentrate the power to a small focal region. Laser injuries were placed between 5-15° from the optic disc. To avoid absorption confounds, we avoided placing lesions beneath large retinal vessels. For experiments examining neutrophil involvement, lesions were placed less than ∼100 microns away from retinal veins and within microns of capillaries to increase the chance of extravasation through these preferred pathways in the retina^38–40. As many as four such lesions were placed per retina (Figure 1). We chose to use the above protocol as an established way to create modest, consistent and reproducible outer retinal damage.

Figure 1.

Throughout this work, we assessed the effects of the laser injury for the following time points: baseline/control, 1 day (defined as 18 – 28 hours post-damage), 3 days, 7 days and 2 months.

In vivo SLO, and OCT imaging

To confirm global ocular health and changes imparted by the laser damage, a commercial Heidelberg Spectralis system (Heidelberg, Germany) was used to acquire scanning laser ophthalmoscopy (SLO) and optical coherence tomography (OCT) images. 30° and 55° fields were used for SLO acquisitions. The 30° field was used for OCT acquisitions. The wider imaging fields allowed for global views not only of damage locations, but also adjacent, non-damaged retina. For some experiments, fluorescein angiography was captured by imaging the retina within 10 minutes of fluorescein administration (details above). As an additional advantage, the fluorescence mode of the Spectralis enabled wide-field images of GFP positive microglia (Figure 5a, right).

OCT was acquired with the following settings: ART 16 (automatic real-time tracking) was used to average B-scans to improve signal. To capture several damage locations in a single field, we used a coarse scan area (61 B-scans over a 1.02 x 0.85 mm window, Figure 1c). A dense 3D data cube was also captured (513 x 171 µm, 49 progressive B-scans, Figure 2a) to provide the most detailed information regarding the axial nature of the laser damage. “Follow-up” mode, which allows HRA software to return to the same retinal location, was used whenever possible. To reveal the cross-sectional profile for each lesion, several adjacent B-scans were spatially averaged (∼30 µm).

Figure 2.

Preparation for ex vivo imaging

Mice were euthanized by CO₂ asphyxiation followed by rapid cervical dislocation. Eyes were enucleated and placed in 4% paraformaldehyde (PFA, diluted from: #15714-S, Electron Microscopy Sciences, Hatfield, Pennsylvania) in 1x phosphate buffered saline (PBS, #806552, Sigma-Aldrich, St. Louis, MO) within 5 minutes of asphyxiation. Intact eye globes were left in PFA fix for one hour at room temperature. Eyes were dissected to remove the cornea, lens and vitreous. Each eye cup was placed in one well of a 24-well plate containing 0.5 ml of 0.8% PFA and left overnight at 4°C. The retina was separated from the retinal pigmented epithelium/choroid with attention to preserve inner retinal layers “up” orientation. If not applying antibody, the tissue was directly flat mounted (see below). For antibody staining, the retina was placed in 1x BD perm/wash buffer (#554723, BD Biosciences, Franklin Lakes, NJ) with 5% donkey serum (#D9663, Sigma-Aldrich) diluted in PBS for overnight incubation at room temperature with gentle shaking. The following was performed in the dark. Ly-6G-647 antibody (1:200, #127610, RRID:AB_1134159, BioLegend, San Diego, CA) and DAPI (1:500 of 10 mg/ml stock, #4083, Cell Signaling Technology, Danvers, MA) were diluted in 1x perm/wash buffer and retinas were incubated for 3 days at room temperature with gentle shaking. Retinas were washed with PBS three times over three hours. The retina was cut into 4 radially symmetrical petals, flat-mounted on a glass slide in ∼40 µl of Vectashield mounting buffer (H-1000-10, Vector labs, Newark, CA) with a #1.5 cover slip (#260406, Ted Pella Inc., Redding, CA), sealed with nail polish and stored at 4°C until imaged.

Endotoxin-Induced Uveitis protocol

To serve as a positive control and show evidence of known neutrophil invasion, we adopted the endotoxin-induced uveitis model to confirm that fluorescent neutrophils could be observed in vivo with AOSLO. This was performed in two mice. The endotoxin-induced uveitis model has been described previously.^36,41 Briefly, we performed intravitreal injections of lipopolysaccharide (LPS, #L4391, Sigma-Aldrich). Mice were anesthetized with ketamine and xylazine. A 34 gauge Hamilton needle was used to deliver 1 µL (1 ng) of LPS diluted in PBS into the vitreous space posterior to the limbus. 1 day post-LPS injection, mice were either imaged with AOSLO (Catchup mice) or collected for ex vivo histology (Ly-6G-647 stained C57BL/6J mice) to confirm fluorescent neutrophil presence in the neural parenchyma.

Ex vivo confocal imaging

Whole-mount retinas were imaged with a Nikon A1 confocal microscope (Melville, NY). DAPI (405 nm ex, 441Δ66 nm em), CX3CR1-GFP (488 nm ex, 525Δ50 nm em), and Catchup/anti-Ly-6G-647 (635 nm ex, 665Δ50 nm em) were simultaneously imaged. Z-stacks at control or laser-damaged locations were acquired with a 60x oil objective, producing images that were 295 x 295 µm. Z-stack axial sectioning was performed with either 0.1 or 0.5 µm step size. Larger (796 x 796 µm) 60x z-stacks were acquired by blending several 60x z-stack acquisitions (3×3, 15% overlap) using the native Nikon NIS Elements software.

AOSLO image processing

To correct for residual motion from heart rate and respiration, AOSLO images were registered with a custom cross-correlation-based frame registration software.^42,43 Motion correction was also applied to simultaneously collected fluorescence images. After registration, confocal, phase-contrast and fluorescence AOSLO images were temporally averaged (250 frames, 10 seconds). Blood perfusion maps were computed by calculating the standard deviation of pixel intensity over 30 seconds³⁴ (Figure 4).

Ex vivo confocal image processing

Imagej FIJI (free NIH software)⁴⁴ was used to view and manage data files. Confocal z-stacks were re-sliced (X-Z dimension) to visualize fluorescence depth profiles.

INL + ONL nuclei quantification

Raw DAPI z-stacks were used for manual counting of nuclei (n = 10 mice, 3 unique regions per time point). Analysis regions were circular, with a 50 µm diameter (roughly corresponding to the size and shape of the damage region seen with confocal AOSLO, Figure 3e) and analyzed in depth producing volumetric cylinders through the INL or outer nuclear layer (ONL, Figure 3d). Nuclei were counted manually using the “Cell Counter” plugin in Imagej (Author: Kurt De Vos, Imagej version 1.53q).

Figure 3.

PR + Microglia nuclei volume quantification

A single DAPI-stained z-stack (OPL + ONL) from a CX3CR1-GFP mouse was used to quantify nuclear volume of PR’s (n = 20) and microglia (n = 14). Data was centered at a lesion location 3 days post-laser exposure. Imagej was used to manually measure the en-face diameter of nuclei in their short and long axis. We averaged the two measurements and assumed spherical shape for analysis. These measurements aided in the differentiation of PR’s, invading microglia and PR phagosomes.

Quantification of RBC flux in capillaries

Using the line-scan approach across single capillaries^30,45,46, we quantified RBC flux within an epoch of 1 second. This spanned several cardiac cycles.^32,47 To improve SNR, space-time images were convolved with a Gaussian spatio-temporal filter (σ = 7.5 pixels, 0.33 µm). This strategy did not interfere with spatial resolution as pixels oversample the optical point-spread of the AOSLO by >20x.³⁰ RBC flux was determined by manually marking blood cells using the “Cell counter” plugin in Imagej.

Neutrophil density quantification

Ly-6G-647-stained retinal tissue was used to manually count neutrophils within montaged 796 x 796 µm z-stacks. Cells were counted at control (n = 4 locations, 2 mice) and lesioned (n = 4 locations, 2 mice) locations using the “Cell Counter” Imagej plugin. All values are reported as mean ± SD. There were no data points omitted from any of the analysis reported in this work.

Results

Characterization of deep focal laser damage

Four complementary imaging modalities were used to evaluate the nature and localization of the focal laser damage induced by 488 nm light: wide field SLO, OCT, AOSLO and post-mortem histology. Each is reported in turn below.

Laser damage induces focal hyperreflective lesions in outer retina imaged with wide-field SLO

To observe global and focal retinal health, we used commercial SLO. 1 day post-488 nm light exposure, both NIR and blue reflectance modalities showed hyperreflective lesions, most apparent at a deeper retinal focus position. When SLO focus was brought to the inner retina, lesions were not visible and retina appeared otherwise healthy (Figure 1b). Despite the deep retinal damage phenotype, fluorescein angiography did not reveal dye leakage (Figure 1b, bottom), indicating the blood retinal barrier remained intact.

OCT B-scans reveal outer retinal hyperreflection without inner retinal damage

To assist in determination of which retinal layers are damaged by the 488 nm laser, we used OCT. Within 30 minutes post-laser exposure, OCT revealed a zone of hyperreflection within the ONL (Figure 2 – figure supplement 1). 1 day post-lesion, the focal hyperreflection remained localized to the ONL (∼50 microns wide, Figure 1c, 2a). With OCT, the retina laterally adjacent to damage foci appeared normal. The hyperreflective phenotype persisted through 7 days post-damage and was cleared by 2 months (Figure 2a). There did not appear to be any cellular excavation, “cratering” or evidence of edema for any post-exposure time point assessed. Despite the axially localized hyperreflective phenotype, Bruch’s membrane appeared intact for all time points assessed. This was evidenced by lack of fluorescein leakage at the site of lesion. Additionally, retinal vessels appeared normal in OCT B-scans aligning with findings from fluorescein angiography, vascular perfusion and histology (discussed below).

AOSLO reveals outer retinal damage with confocal and phase-contrast modalities

Confocal AOSLO provided micron-level detail of the lesioned area. Confirming OCT findings, we observed hyperreflective changes localized within the ONL that were, by 1 day, the brightest at the OPL/ONL interface, shown in Figure 2b. The AOSLO has an axial resolution near 6 µm²⁹ and thus evidence of retinal capillaries at the deepest vascular stratification acted as an axial landmark for the OPL. At this plane, the lesions manifest as an elliptical region with the long axis in the direction of the linescan used to create the lesion. Tracking this region over time, showed most prominent hyperreflectivity at 1 day post lesion, with diminishing size and brightness by day 3 and 7. The hyperreflective phenotype was largely diminished by 2 months post-exposure (Figure 2b).

Using phase-contrast AOSLO also allowed visualization of translucent cells within the retina which enabled us to image PR somas of the ONL^26,33. Normally, the ONL is comprised of 8-12 PR somata that are axially stacked^33,48. We found that the dense packing of individual somata was disrupted 1 day post-exposure (∼50 µm ovoid, Figure 2c), suggesting degradation or ablation of the cell membrane.

Confirmation of outer retinal cell loss using post-mortem histology

To assess the extent of cell loss caused by focal laser exposure, we performed DAPI staining on whole-mount retinal tissue using the same time points assessed previously for in vivo imaging. 1 day after laser exposure, we observed mild thickening of the ONL compared to unexposed locations only microns away. At 3 and 7 days, local ONL thinning was observed (Figure 3a). En-face planes at multiple layers within the ONL revealed a loss of PR nuclei in the outer aspect of the ONL for 3 and 7 day time points (Figure 3c). The inner ONL exhibited little evidence of cell loss (Figure 3b), illustrating the precise axial confinement of the laser damage being induced by this method. By 2 months, the lesion’s overall appearance and ONL thickness returned to baseline (Figure 3a-c). This histological finding corroborates the OCT findings observed in vivo.

ONL nuclear counts were reduced at locations of laser exposure. In comparison to control locations (501317 nuclei/mm² ± 46198 nuclei/mm², mean ±SD, similar to previous work⁴⁹), at 3 and 7 days post-exposure, we found a corresponding reduction in the density of ONL nuclei of 18% (408965 nuclei/mm² ± 25621 nuclei/mm²) and 27% (367542 nuclei/mm² ± 9038 nuclei/mm²) respectively. Calculated in comparison to control locations, the 3 and 7 day data resulted in p-values of 0.17 and 0.07, respectively (Student’s, paired two-tailed, t-test). 2 months after damage, PR nuclear densities were indistinguishable from that of control (Figure 3f, gray bars). Despite losses of PR nuclei, total cell nuclei within the INL remained unchanged for all time points (Figure 3f, black bars). These data indicate that the laser exposure focally ablated PR’s whilst leaving inner retinal cells intact.

Retinal vasculature unaffected by deep retinal lesion

A concern with laser lesions of this type is that it may coagulate large vessels or capillaries. Motion contrast^27,30 images revealed that the vasculature remained perfused from hours to months after laser damage suggesting that acute or long term changes are not imparted by the laser injury (Figure 4). None of the primary vascular stratifications within the NFL, IPL and OPL of the mouse retina³⁴ showed stopped flow as a result of the laser lesion, reinforcing the findings above regarding the axial confinement of the damage to the outer retina.

Figure 4.

In addition to perfusion status, phase-contrast AOSLO also permitted analysis of single blood cell flux within capillaries^30,50. We tracked capillary flux at different retinal depths within and above damage sites (IPL, OPL, 1 capillary each, Figure 4 – figure supplement 1a). Blood cell flux for capillaries within lesion locations were within the range of normal flux for the C57BL/6J mouse⁵⁰ (Figure 4 – figure supplement 2b, c). Flux tracked from hours to days in these capillaries changed synchronously, displaying positive linear correlation (R² = 0.59, Figure 4 – figure supplement 1c, d). This suggested any such changes in flux were a property of systemic perfusion, rather than locally imparted changes in flow due to the lesion. As an additional control, we evaluated blood flux in two distant capillaries; one located 10-15° temporal of the optic disc (lesion location) and another of equivalent eccentricity on the nasal side (control, Figure 4 – figure supplement 1e). Both capillaries displayed similar flux values from minutes to 2 months post-injury (Figure 4 – figure supplement 1f, g) resulting in positive linear correlation (R² = 0.78, Figure 4 – figure supplement 1h). Taken together, these findings suggest lesions do not rupture or stall blood flow in capillaries, nor do they appreciably impact total RBC delivery in the capillary network.

PR laser injury promotes a robust response in nearby microglia

To observe the microglial response to PR laser injury, we imaged fluorescent microglia in CX3CR1-GFP mice with both SLO and AOSLO. 1 day after injury, SLO revealed bright, focal congregations of microglia at injury locations. This was in contrast to undamaged locations which maintained a distribution resembling lateral tiling (Figure 5a, b). The global visualization of microglia was augmented by high resolution fluorescence AOSLO, providing enhanced detail of the microglial response to laser injury, especially cellular process detail^7,51. Whereas AOSLO imaging of microglia in the healthy retina displayed a distributed array of microglia with ramified processes (Figure 5c, left), laser damaged locations showed a congregation of cells 1 day post-injury with less lateral ramification (Figure 5c, right). Within hours of the laser exposure, we did not observe a photo-bleaching or death of regional microglia suggesting that while the laser exposure was sufficient to damage PR’s, it left inner retinal microglia intact.

Figure 5.

A standing question in OCT/confocal AOSLO interpretation is whether microglia contribute or directly produce the hyperreflective phenotype seen in axial B-scans and en-face fundus images^52,53. With confocal AOSLO, we find that the hyperreflective phenotype is visible as early as 30 minutes and becomes slightly larger and brighter by 90 minutes. During these time points, simultaneously imaged microglia remained ramified and maintained a tiled arrangement, indicating that microglial are not the initial source of the lesion-induced hyperreflective appearance (Figure 6 – figure supplement 1).

We returned to the same laser-damaged locations to capture microglial appearance at baseline, 1, 3, 7 day and 2 month time points with AOSLO to track the natural history of the microglial response to PR damage. At 1 day post-injury, damage locations displayed aggregations of microglia. The surrounding microglia displayed process polarization with extensions projecting toward the injury (Figure 6). By days 3 and 7, microglia exhibited fewer lateral projections and somas have migrated into the ONL where they do not normally reside (Figure 6). By 2 months post-injury, the hyperreflective phenotype was absent and microglia once again occupied only the inner retina. The axial and lateral microglial distribution at 2 months after lesion was similar to baseline (Figure 6). Additionally, cell morphology once again showed radially symmetrical branching projections, similar to those prior to laser injury.

Figure 6.

With phase-contrast imaging targeting the ONL, we documented a rare event of putative pseudopod extension at a lesion site (Figure 5 – video 1). Given the axial complexity of microglia in this layer (Figure 9 – figure supplement 2c) it is now possible that microglial process dynamics may be revealed with this label-free approach.

Despite robust microglial involvement, neutrophils do not extravasate

While we observed a robust microglial response to PR damage, there was no evidence of neutrophil involvement at the times we examined (1, 3, 7 day, and 2 month follow-up).

In vivo fluorescence imaging allowed us to track neutrophils with AOSLO in “Catchup” mice³⁷. In healthy mice, we observed a sparse population of circulating neutrophils flowing quickly within the largest retinal vessels (Figure 7 – video 1). We also observed neutrophils moving through single capillary branches (Figure 7 – video 2). This provided evidence that neutrophils could be directly observed in vivo, where they are moving at exceptionally fast speed³². Within capillaries, we found that neutrophils show deformation within the small confines of the capillary lumen, often resulting in tube or pill-shaped neutrophil morphology seen both in vivo and ex vivo (Figure 7b, top).

Figure 7.

After inducing laser lesion, we found no evidence of neutrophil aggregation or extravasation for any time point assessed (Figure 8). There was also a notable lack of rolling/crawling neutrophils (or any putative leukocyte) in large arterioles or venules surrounding the injury (Figure 8 – video 1). The deepest retinal neutrophils, closest to the ONL lesion, were occasionally detected within the deepest retinal capillaries. However, these neutrophils stayed within the retinal vasculature, as evidenced by their pill-shaped morphology and passage routes that follow the known vascular paths seen in confocal/phase contrast imaging (Figure 8 – video 2). Just like healthy mice, leukocytes, including neutrophils often impede flow due to their large size (13.7 µm⁵⁴) as they must compress through capillaries <7 µm in diameter⁵⁰. However, we found no evidence of permanently stopped flow, as the rare stalls would re-perfuse similar to those previously characterized in healthy mice (Figure 7a).⁵⁰

Figure 8.

As a positive control, we used the endotoxin-induced uveitis model in Catchup mice to show evidence that we could image extravasated neutrophils. This model is known to induce a strong neutrophil response. With fluorescence AOSLO, we found aggregation of Catchup-positive neutrophils within the retinal parenchyma, providing confirmation that extravasated neutrophils can be imaged with this modality. Moreover, we found that neutrophils that have extravasated into the retinal parenchyma tended to have a more spherical morphology rather than the compressed, pill-shaped morphology of neutrophils within capillaries (Video 2).

Ex vivo analysis confirms in vivo findings

To confirm our in vivo findings and make certain that no neutrophils evaded detection with in vivo imaging, we examined fluorescent microglia and neutrophils in laser-damaged retinal whole-mounts imaged with confocal microscopy. Both the spatial and progressive nature of the microglial response to PR damage corroborated in vivo findings (Figure 6 vs Figure 9 – figure supplement 1). Similarly, we found a general lack of neutrophil response.

Ex vivo: Microglia display dynamic changes in morphology in lesion areas

Without laser damage, microglia exhibited a tiled distribution and stellate morphology with highly-ramified branching patterns (Figure 9a). 1 day after laser exposure, microglial somas aggregated to the lesion location. They began to migrate into the ONL and the most notable phenotype was a change from the ramified, and pancake-flat morphology seen in the healthy retina, to a dagger-like axial morphology. (Figure 9a + Video 1). At 3 and 7 days after lesion, microglia migrated deeper into the outer retina and remained aggregated with an axially elongated phenotype. 2 months after laser lesion, microglial somas were no longer found in the ONL. Instead, microglia redistributed similar to that of the healthy retina (Figure 9a), once again co-stratifying predominantly with the NFL and plexiform layers of the retina.

Figure 9.

Whereas we found a robust response of deep microglia at the OPL to PR injury, we observed little migratory response of the microglia in the NFL or IPL (Figure 9 + Figure 9 – figure supplement 1) suggesting axial and lateral constraints on the extent of microglial recruitment.

Microglia form PR-containing phagosomes

A detailed examination of the microglial involvement with PR somata in response to PR injury revealed unique microglia-PR interactions within the ONL. In cross-section, PR cells comprise a large portion of the retina reaching from the outer-segment tip, inner segments, somata and spherule/pedicle synaptic contacts at the OPL. Likewise, we saw microglial involvement with all of these layers. Within 1 day, we found microglial processes interspersed within the dense aggregation of PR somata within the ONL. Amoeboid cells enveloped the somata of PR’s and phagocytosis was detected 1, 3 and 7 days post-laser exposure (Figure 9 – figure supplement 2). Confocal microscopy revealed GFP-positive processes surrounding PR nuclei, engulfing multiple somata (Figure 9 – figure supplement 2a, c). By 3 days, microglial processes and somas were found to migrate deeper, now with processes extending into the distal portions of the PR cells including the inner/outer segments. By two months microglial processes and somas had retreated out of the ONL.

The DAPI nuclear stain combined with confocal microscopy not only helped us discern retinal neurons, but also allowed us to differentiate between microglia and phagocytosed PR’s. Two features were different from PR and microglial nuclei. First, whereas PR’s displayed a uniform, homogeneous nuclear fluorescence, microglial nuclei showed a mottled, and heterogenous DAPI appearance (Figure 9 - figure supplement 2a). Second, microglia nuclei were nearly 3x larger in volume compared to healthy and damaged PR nuclei. Microglia had nuclear volume of 110 ± 42 µm³ and PR nuclei were 35 ± 5 µm³ (p < 0.001, Figure 9 – figure supplement 2b). PR nuclei within microglial phagosomes displayed similar nuclear volume compared to adjacent PR’s in undamaged locations (Figure 9 – figure supplement 2a).

Ex vivo: At lesion sites, neutrophils remain within the retinal vasculature

Corroborating in vivo AOSLO findings we did not find neutrophil aggregation or extravasation in response to laser damage at any of the time points examined. Ex vivo microscopy revealed that neutrophils were found within the vascular network and did not extravasate into the neural retina. Very few neutrophils were found in lesioned retinas, comparable to the paucity of neutrophils found in healthy retinas. The few detected neutrophils were remnants of those found within the capillary network at point of death. Detailed Z-stacks at 60x magnification revealed “pill-shaped” morphology similar to that seen in vivo, where neutrophils are compressed within capillaries. In one exceptional example (1 day post-damage, Figure 9c,d), we observed two neutrophils in a single z-stack, one in a capillary (1) and one at a capillary branch point (2), but none were observed to have left the confines of the vasculature suggesting they were not recruited by activated microglia.

Retinal neutrophil concentrations were quantified from larger z-stacks (796 x 796 µm, Figure 10a). Control locations (n = 2 mice, 4 z-stacks) had 15 ± 8 neutrophils per mm² of retina whereas lesioned locations (n = 2 mice, 4 z-stacks) had 23 ± 5 neutrophils per mm² of retina (Figure 10b). The difference between control and lesioned groups were not statistically significant (p = 0.19).

Figure 10.

Finally, when we rendered confocal Z-stacks in cross-sectional view, we discovered that the few fluorescent neutrophils present were found colocalized within the tri-layered vascular stratifications of the mouse retina ^34,55,56. None were found in non-vascular layers (Figure 9). Taken together, all ex vivo data indicates that neutrophils remain localized within vessels, suggesting that neutrophils do not extravasate or aggregate in response to this laser lesion model, regardless of the robust phagocytic microglial response. These data corroborate the in vivo findings.

Discussion

Summary

In many retinal diseases, resident microglial populations are found to exhibit cross-talk with systemic immune cells^15,18. The complex temporal dynamics between resident and systemic immune cells unfolds from seconds-to-months and has been poorly characterized due to lack of resolution, sufficient contrast and a non-destructive imaging approach that can track changes over time. Here, our engineering advance overcomes these limitations by using phase-contrast and fluorescent imaging with adaptive optics to visualize the interaction of multiple cell types. With this advanced imaging technology, we reveal the absence of neutrophil involvement in response to an acute injury, despite progressive changes in microglial activity and morphology. Here we discuss the implications of such findings in the context of retinal damage and more broadly, to retinal disease.

Focal retinal lesions for tracking the immune response

Light damage offers a number of advantages for creating acute damage in the retina. It offers placement precision that is better controlled than other acute and invasive methods, such as incision/poke injuries⁵⁷, retinal detachment⁵⁸ and chemical injury^59,60. In the mouse, focal light dosage comes with greater axial confinement within the retina by nature of the high numerical aperture of the mouse eye²⁹ which imposes minimal collateral damage to layers above or below the plane of focus (Figure 1a). The laser exposure protocol employed here was highly reproducible as ∼95% of light exposures resulted in retinal damage of similar size and hyperreflective phenotype.

Seminal studies in the brain have used focal light exposure to induce acute microglial response to targeted areas.^61,62 Since then, it has been a popular damage model. Focal lesions in the retina have been conducted in a number of popular animal models including mouse^63,64, macaque^65–67 and cats.^68,69 Laser damage is also clinically relevant as it is both experienced with accidental laser exposure^70,71 as well as purposeful ablation used in photocoagulation and as a therapy for retinal ischemic disease, retinopathy of prematurity⁷² and diabetic retinopathy⁷³. Thus our results shed light toward the immune response that may be imparted due to phototoxic damage. More widely however, this approach may be considered when investigating how native microglia may signal to the systemic immune system in response to any acute injury or chronic photoreceptor loss in aging/diseased retinas.

Titration of the 488 nm laser leads to mild retinal damage

While it is beyond the scope of this report to catalog the vast parameter space in which light may impose damage in the retina, it is noteworthy to discuss that the exposure intensity, subtended angle and duration of the CW light source we used delivered an intentionally mild damage to the retina. Using dosages too high would run the risk of collateral damage to the blood retinal barrier or Bruch’s membrane⁷⁴ which could cause a choroidal neovascularization phenotype⁷⁵. Either of these would be undesirable for studying the native/systemic immune response as it would potentially confound the traditional extravasation pathways of the inflammatory system^24,76. Instead, we chose our laser exposure condition to impart a weak, but reproducible loss in photoreceptors while minimizing collateral damage to the cells above and below the plane of focus.

With OCT, we see the primary damage phenotype localized to the OPL/ONL bands (Figures 1, 2) with visibly undetectable change in the anatomy of the inner retina. Histology corroborated this finding since the only layer showing quantifiable cell loss was the photoreceptor layer (Figure 3).

Progressive PR loss and morphology of the lesion tracked over time

The amount of light we delivered did not immediately ablate photoreceptors as the peak cell loss wasn’t observed until 3-7 days after laser exposure suggesting a progressive damage phenotype of apoptosis or necroptotic death rather than direct ablation. Histology showed a ∼27% reduction in PR nuclei 7 days post lesion. The total number of PR nuclei within the lesion location returned to values similar to baseline by 2 months, which was surprising given that PRs do not regenerate. Our interpretation is that adjacent PRs spill into the region of loss, similar to what several have described previously^65,77. Corroborating the histological finding of mild cell loss, is the in vivo OCT data which shows no evidence of local edema, cavitation or excavation of the ONL after PR loss (Figure 2a). The return of the appearance, thickness and redistribution of cells within the ablated location indicates that retinal remodeling occurs within a 2 month window. The concurrence of microglia in these outer retinal bands (where they are normally absent in health⁷⁸) support the hypothesis that microglia localized to this region to facilitate a phagocytic injury response and perhaps contribute toward synaptic remodeling¹¹.

The hyperreflective phenotype does not arise from microglia or neutrophils

Light damage is known to create a hyper-reflective bands in OCT imaging ^66,79,80. A common speculation is that the increased backscatter may arise from the local inflammatory cells that activate or move into the damage location. In our data, confocal AOSLO and OCT revealed a hyperreflective band at the OPL and ONL after 488 nm light exposure (Figure 2a, b). However, we found that the hyperreflective bands appeared within 30 minutes after the laser injury. This phenotype preceded any detectable microglial migration toward the damage location (Figure 2 – figure supplement 1 and Figure 6 – figure supplement 1). We thus conclude that the in vivo hyperreflective phenotype is not caused by microglial cell activity or aggregation.

Direct activation of microglia from 488 nm light exposure was minimal

It is conceivable that 488 nm light used for either imaging (56 µW) or imparting damage (785 µW) might activate the GFP-containing microglia in the Cx3CR1 mouse model used here. However, several lines of evidence speak against this possibility. 1) We did not observe photobleaching of CX3CR1 positive cells in response to the damage suggesting the fluorescence imaging light was insufficient to damage microglia. 2) The focal plane was adjusted using the optics of the AOSLO, focusing the light such that the dose was most axially concentrated onto the outer retina when performing retinal injury. Thus, the light dosage received by the microglia above was defocused and less than the targeted PR layer (Figure 1a). 3) Histology showed no evident necrotic/apoptotic microglial morphologies and were similar to cells at undamaged locations.

488 nm laser lesion does not photocoagulate or alter retinal circulation

Despite imparting damage to the photoreceptors, the damage regime used here did not alter the perfusion of the retinal circulation. We show three independent measures that blood flow is uninterrupted, despite quantifiable PR loss and activation of microglia. 1) Fluorescein angiography (Figure 1b, bottom) displays intact vasculature passing through and above lesion locations. The absence of fluorescein leakage indicates that the lesion did not compromise the blood-retinal barrier. Additionally, there was no indication of a perforated Bruch’s membrane. 2) AOSLO motion contrast vascular maps ^25,30,34 displayed persistent blood perfusion inside vessels near laser lesion sites (Figure 4). This indicated that single cell perfusion through the capillaries above and adjacent to the regional damage was maintained 3) Capillary line scans indicate that RBC flux was not modified at lesion locations and fell within the normal range^30,50 (Figure 4 – figure supplement 1). Altogether, these three lines of evidence indicate that the 488 nm light exposure does not impart perfusion changes within the retinal vasculature.

Microglial/neutrophil discrimination using label-free phase contrast

In this work, we use fluorescence to validate the presence and dynamics of microglia and neutrophils in response to PR damage. We have found that microglia and neutrophils show a number of differences that may be useful for identifying their type without the need for fluorescent labels. For example: under healthy conditions, these cells exhibit morphological differences: retinal microglia exhibit processes that emanate from the soma, exhibiting an arbor that actively surveys a large patch of retina, while neutrophils are spheroid, sometimes presenting as cylinders when squeezing through retinal capillaries. Another difference lies in their motility: neutrophils remain within the vasculature, often times traveling at high speeds while microglial somas generally remain stationary as they monitor the local microenvironment. With a strong base-of-knowledge about the morphological and behavioral features to expect for each cell type, label-free imaging could discriminate not just immune cells, but other retinal cells as well in the future.²⁴ This is particularly promising when considering translating these observations into the human population where dyes and fluorescent agents are used sparingly.

Resident microglia do not need systemic neutrophils for resolution of mild laser-damage

The CNS and retina, unlike other peripheral tissues, cannot suffer from excess inflammation as there may be dire functional consequences. Therefore it is possible that microglia protect against exorbitant inflammation by modulating the recruitment of systemic inflammatory cells.^4,15–17 Of these, neutrophils are often one of the first systemic responders. Despite their helpful roles in other tissues, neutrophils can secrete neurotoxic compounds that could present a danger to the CNS.⁸¹ Given their conflicted role in the body, we ask the question: to what extent do neutrophils respond to acute neural loss in the retina? Retinal cells are lost with age⁸² and disease⁸³ and yet, for the organism, visual perception must persist. In this laser lesion model, we ablate 27% of the photoreceptors in a 50 µm region. This acute damage model can begin to provide clues about how the immune system responds when photoreceptors are lost in disease.

We find that microglia undergo a rapid and progressive response to this injury. We show evidence of PR phagocytosis (Figure 9 - figure supplement 2), interaction with neighboring microglia (Figure 9 – figure supplement 1) and they are also axially positioned to facilitate retinal remodeling (Figure 9). Furthermore, throughout the temporal evolution of the microglial response, we find no evidence of neutrophil recruitment despite the damage being within 10s of microns from retinal vessels that carry systemic neutrophils. At the onset of the neutrophil extravasation cascade, endothelial cells in the vicinity of inflamed tissue typically elevate the expression of adhesion molecules, facilitating the adherence and extravasation of circulating neutrophils.⁸⁴ Furthermore, neutrophils are dependent on priming events as prerequisite to further activation and engagement of their effector functions.^85,86 Based on our data, we suggest that although microglia show a strong and lasting activation, at no time point from seconds-to-months, are the damage-associated molecular patterns or chemotactic gradients strong enough to recruit neutrophils in response to this damage. This is evidenced in our data from two key observations: 1) in response to laser injury, we saw no examples of systemic leukocytes rolling in vessels adjacent to injury locations (Figure 8 – video 1). 2) Using fluorescence, we did not observe adherent or extravasated neutrophils adjacent to imparted photoreceptor loss (Figure 8). This may suggest that the region of insult is too small, or that activated microglia are not sufficient for recruiting neutrophils with damage of this magnitude. This could mean that a minimum threshold of neural damage must be met before neutrophils will respond to damage of this type. Such a strategy would benefit the CNS. It is possible that resident microglia facilitate the necessary phagocytic and retinal remodeling response despite release of cytokines from damaged retinal cells that would normally recruit systemic immune cells in peripheral tissues.

Future work will explore whether there is a threshold magnitude of neuronal cell loss required for recruitment of systemic cells that is unique to the retina. Building on this work, next studies will examine more severe or widespread injury regimes that may provide stronger activating molecular signals and interact with a larger population of systemic cells.

Microglia may inhibit neutrophil activation

Microglia may be involved in a system that protects the CNS from propagating a much larger systemic response, potentially exacerbating disease pathologies that would compromise overall CNS function. From this work, both in vivo and ex vivo data corroborate the finding that microglia become reactive, whereas no evidence of neutrophil recruitment is observed. In another damage model⁸⁷, they report that tissue-resident macrophages may exhibit the capacity to cloak tissue micro-damage. This offers the possibility that resident immune cells, such as retinal microglia, can handle small insults without inducing a chemokine cascade that may invoke a larger systemic response that could further damage the precious retinal tissue.⁸⁸ Regardless of the mechanism, we find that despite a robust microglial activation that lasts for weeks, at no time point do they recruit neutrophils. The nuance of this interaction likely represents the fine balance that facilitates a helpful local response within the CNS that does not impart a more widespread cytokine storm that may otherwise exacerbate retinal damage. Further work will explore whether microglia exhibit a cloaking response in the retina, inhibiting neutrophil or other immune cell extravasation/chemotaxis toward lesion sites. We expect such work to be pivotal in understanding the balance that is broken or left unchecked in conditions of autoimmune disease and the umbrella of diseases that comprise the uveitic response which is a direct threat to lifelong vision.⁸⁹

Conclusion

Here, we have applied innovative in vivo imaging at the microscopic scale to reveal the cellular immune response to a retina in jeopardy. The dynamic environment of the retina includes a native population of resident microglia and systemic immune cells delivered through the vasculature. These two lines of defense work in concert in the mammalian body and are critical for maintaining retinal homeostasis. In this work we directly study the interaction between microglia and neutrophils, two major classes of immune cells that are implicated in inflammatory initiation, escalation, propagation and debris removal in response to an acute geographical injury in the retina. Using cutting-edge retinal imaging modalities, we find that resident microglia become locally activated and regionally responsive to focal laser lesion. Cells migrate away from their stratified locations near plexiform layers of the retina and migrate toward the axial site of damage within hours to weeks after injury. However, systemic neutrophils, which are typically regarded as first-line responders to tissue damage, are not recruited to this damage despite neutrophils flowing within 10s of microns away from the location of damage (Figure 8 – video 2). Beyond the context of this specific finding, we share this work with the excitement that AOSLO cellular level imaging may reveal the interaction of multiple immune cell types in the living retina, a unique tissue in its own regard. By using fluorophores associated with specific immune cell populations, the complex immune cell dynamics that orchestrate the immune response may be examined in this specialized tissue and may reveal further insights to the interactions of single immune cells in the living body in a non-invasive way.

Acknowledgements

Thanks to Colin Chu, Justin Elstrott and Tiffany Heaster for useful intellectual conversations. Thanks to Minsoo Kim for generously transferring the Catchup mouse strain.

Impact statement

Systemic neutrophils do not respond to laser-induced photoreceptor damage despite the robust response from resident microglia

Financial support

Research was supported by the National Eye Institute of the National Institutes of Health under Award No. R01 EY028293, and P30 EY001319. Work was also supported by a collaborative grant from Genentech Inc, a Career Development Award and Career Advancement Award (Schallek) and Unrestricted Grant to the University of Rochester Department of Ophthalmology from Research to Prevent Blindness, New York, New York and the Dana Foundation David Mahoney Neuroimaging Award (Schallek)

Conflict of Interest statement

Portions of this work were funded by a collaborative grant from Genentech, Inc. (Elstrott) to examine the extent to which immune cells could be studied in the deep retina using adaptive optics. Schallek also has six patents held through the University of Rochester on adaptive optics technology.

Data availability statement

All data collected (∼3 TB) is archived at the University of Rochester and is available upon request.

Figure supplements

Figure 2 - figure supplement 1

Figure 4 – figure supplement 1

Figure 6 – figure supplement 1

Figure 9 – figure supplement 1

Figure 9 – figure supplement 2