Background
Innate immune cells mount an early response to stress, injury, and infection in central nervous system (CNS) tissue, including the retina [
1], and are important contributors to CNS development and homeostasis [
2‐
4]. A substantial literature attributes a wide range of innate immune functions in the CNS to microglia, the tissue-resident myeloid cells of the CNS [
5‐
8]. Tissue-resident macrophages are found in many tissues, and their origins continue to be studied. In the case of microglia, the embryonic yolk sac was found to be their origin [
9‐
12]. Langerhans cells appear to be derived in part from the fetal liver [
11,
13]. Most other tissue macrophages originate in bone marrow [
14]. Regardless of their origin, these macrophages were recruited to and live in tissue niches in which multiple factors, including chemokines, cytokines, and corresponding receptors, play an important role in maintaining their presence and regulating their activity [
15]. The vascular endothelial cells that mediate passage of cells from the circulation into tissues use combinations of chemokines and chemokine receptors [
16‐
18], especially CCL2/CCR2, to facilitate entry of innate immune cells into tissues including the kidney [
19], lung [
20], gut [
21,
22], brain [
23,
24], and retina [
25]. A growing literature describes populations of self-renewing tissue resident macrophages, and efforts to understand their origins and distinct functions [
26]. For tissue-resident macrophages, such as microglia, that originate from embryonic sources no longer present in adult animals, identifying how these cells are maintained and/or replaced over the lifetime of the host has been a significant challenge.
Our interest in antigen presenting cells of the retina that could contribute to T cell-mediated autoimmune retinitis led us to examine the retina for cells possessing the antigen-presenting capabilities of dendritic cells. Using conventional dendritic cell markers, others found evidence for substantial numbers of dendritic cells in the inflamed retina associated with experimental autoimmune uveoretinitis, a model for the retinal autoimmune disease [
27]. In the normal retina, a small number of candidate dendritic cells was found by staining for Dec-205 [
28] and 33D1 [
29]. We found that CD45
+ myeloid cells isolated from normal retina functioned poorly as antigen presenting cells and also inhibited the antigen presenting activity of splenocytes in cocultures, yielding T cells with signs of anergy [
30]. This led our studies to the CD11c
DTR/GFP mice which permits tracking and depletion of a candidate antigen presenting cell population via green fluorescent protein (GFP) and diphtheria toxin receptor (DTR) [
31]. Using these mice, we reported that quiescent murine retina contains a small, local population of CD45
medCD11b
hiLy6G
loGFP
hi cells that resembled dendritic cells in their ability to upregulate expression of MHC class II [
32], and to process and present cognate antigen to antigen-specific naive T cells [
32,
33]. Depletion of retinal GFP
hi cells in CD11c
DTR/GFP mice by intraocular administration of diphtheria toxin eliminated the local antigen-specific T cell response in the retina [
34]. The number of GFP
lo microglia (CD45
medCD11b
+Ly6G
loGFP
lo cells) remaining after diphtheria toxin ablation was unchanged and did not support an Ag-specific T cell response. The numbers of retinal GFP
hi cells expanded in response to different types of retinal injury and stress, and the cells migrated to specific sites of injury [
32,
35,
36].
More recently we reported in two injury models using CD11c
DTR/GFP mice that the expanded retinal GFP
hi cells were derived from microglia; appearing in response to the stress of cone photoreceptor degeneration in CD11
DTR/GFP × RPE65
−/− double transgenic mice [
35] and in response to optic nerve injuries [
37]. Results from ablation, parabiosis, fate-mapping, and optic nerve transection experiments showed that the GFP
hi cells in the retina were not derived from the circulation, but were rather derived from resident retinal microglia and/or microglia recruited from the optic nerve. In contrast to retinal microglia, many of the microglia from the optic nerve were also GFP
+ and Ki67
+ [
37]. These findings suggested that there were two niches of retinal innate immune cells; a relatively stable microglia niche and a specialized niche of microglia, as represented by the GFP
hi cells in CD11c
DTR/GFP mice, that can function as dendritic cells and whose numbers dynamically expand and contract in response to stimuli. However, in contrast, we have observed that following certain severe injury protocols, such as those involving radiation bone marrow chimeras, retinal microglia were substantially replaced by circulating bone marrow-derived cells [
32]. Given the recent recognition of non-parenchymal CNS macrophages in meninges, perivascular spaces, and choroid plexus that are distinct from microglia [
38], the contaminating presence of non-parenchymal myeloid cells may confuse the analysis of either the putative parenchymal microglia or the GFP
hi myeloid cells we have found in the retina, optic nerve, and brain of CD11c
DTR/GFP mice. Importantly, the murine retina can be isolated without contamination by meninges and choroid plexus, simplifying the analysis of parenchymal microglia.
In this study, we explore the basis for the diverse observations regarding the origins of replacement myeloid cells and their responses in CNS tissue. Given the highly specialized and regulated environment of the retina, we asked whether bona fide microglia and their monocytic replacements acted similarly within the environment of the retina. To do this, we used two distinctly different means to ablate retinal microglia and asked how the myeloid cells that repopulated retina responded to stress. Using CD11cDTR/GFP mice in conjunction with other transgenic mice, we found that the retinal microenvironment exerted a potent influence on the morphology and function of GFPhi and GFPlo cells derived from either CNS microglia or adult bone marrow as they occupied the retinal myeloid cell niches. Replacement retinal myeloid cells from the circulation were largely GFPlo until stimulated by a retinal injury which induced GFP expression and migration to the site of injury. Since the vast majority of studies of CNS myeloid cells have been done in the brain, some of our experiments were done in parallel in the brain and optic nerve and revealed significant differences.
Discussion
The retinal microenvironment has a potent influence on the biology of its resident myeloid cells. New myeloid cells recruited to the retina whether from retinal progenitors, stores in adjacent CNS tissue, or from circulating monocytic cells, are constrained to substantially mimic the morphology and behavior of its resident, parenchymal myeloid cells, the microglia. Using the CD11c
DTR/GFP mouse which labels a subset of the microglia with GFP, the data in this study along with our previous reports has led us to propose that the retina maintains two niches of microglia. In the quiescent retina, the majority of microglia are CD45
medCD11b
+Ly6G
−Ly6C
−F4/80
+Iba-1
+ and do not express GFP (GFP
lo). This niche of microglia is stable in number and carries out the functions associated with tissue macrophages that are necessary for the development and health of the retina. The second smaller niche of microglia is also CD45
medCD11b
+Ly6G
−Ly6C
−F4/80
+Iba-1
+ but is characterized by expression of GFP (GFP
hi). This population of microglia dynamically expands and contracts in response to stimuli and can carry out the T cell activation functions of dendritic cells. It should be noted that all murine retinas have these two microglia niches but only with the CD11c
DTR/GFP mouse can they be distinguished. While there is controversy about the existence of GFP
hi cells in the retina of CD11c
DTR/GFP mice [
64,
65], we find their presence to be obvious in the data we [
32‐
37] and others [
66,
67] have presented. Others have observed that myeloid cells recruited from the circulation often assume the appearance of microglia once inside retina [
68,
69]. By exploring different strategies for repopulating the CD11c
DTR/GFP mouse retina, we found that recruited myeloid cells could also express GFP when activated by injury. While circulating myeloid cells recruited to the retina cannot be considered true or bona fide microglia, nonetheless, they look and act like microglia.
The retina has a number of unique properties that make it particularly well suited to study myeloid cell maintenance, replacement, and function in CNS tissue. First, the entire retina is easily removed, cleaned of adjacent tissue, and analyzed in its entirety as a unit. In this way, the total number (and changes to that number with treatment) of any given cell type in a retina can be analyzed directly. Second, the retina lacks meninges, and thus is a truer representation of neural parenchyma. While microglia occupy neural parenchyma of the brain, there is a substantial contamination of circulation-derived macrophages found in the perivascular/Virchow-Robbins space, subdural meninges, and choroid plexus [
38]. Third, retina has a thin, flat structure that is amenable to full-thickness microscopy as flat mounts. Fourth, fluorescent markers can be visualized by non-invasive imaging through the pupil in vivo. Fifth, with a sufficient stimulus, the parenchyma can be repopulated with recruited myeloid cells able to take up long-term residence in the retina. We used strategies to ablate resident retinal myeloid cells and promote their replacement either from circulating myeloid cells or from CNS precursors and then compared the replacement myeloid cells to resident retinal microglia. While we observed differences in the dynamics and function of the replacement myeloid cells to resident microglia depending on the method of ablation and/or stimulation, we note that myeloid cells recruited from the circulation are very similar in function and phenotype to both the static and dynamic (as represented by GFP
lo and GFP
hi cells in the CD11c
DTR/GFP retina) resident microglia. We also note differences and similarities between the retina and other CNS tissue concerning myeloid cell repopulation after ablation.
One of the most striking differences between the retina and other CNS tissue was the kinetics of myeloid cell repopulation following tamoxifen-induced DTA ablation. In CD11c
DTR/GFP:CX3CR1
YFP-creER:ROSA
DTA mice, the short-term replacement of both GFP
lo and GFP
hi microglia in the retina significantly lagged compared to the brain and optic nerve. The rapid repopulation of brain microglia after ablation has been attributed in part to the activity of CD34
+ and CD117
+ progenitors [
49]. However, we found minimal evidence for local progenitors in the retina. Even after injury we observed only a small, transient number of cells that were either CD34
+, CD117
+, or SCA-1
+, and none of the individual cells stained for more than one marker, nor did any layer of the retina have cells positive for more than one marker. Thus, differences in putative progenitor cell numbers between brain and retina could account for the initially slow replacement in the retina. Another factor in CNS microglia repopulation could be the survival of residual microglia. A recent study proposed that after ablation, all microglia reappearing in the brain were derived from the few remaining survivors and not the circulation or de novo from resident progenitors [
70]. Similar results have been reported in the retina [
69], with the possibility that myeloid cells infiltrating from the adjacent ciliary body also contributing to retinal microglia repopulation [
71]. While identifying sources of replacement microglia, these studies do not account for the different rates of microglia repopulation in various CNS tissues. If expansion of progenitors or residual microglia were replenishing the niches after ablation there should be evidence of cellular proliferation within the retina. However, using Ki67 as a marker for proliferation, the only instance in which we observe a significant number of retinal Ki67
+ cells was with bone marrow transfer plus injury, and even then, the proliferating cells were of donor origin. We have also observed that an ONC can induce the appearance of numerous Ki67
+ CD11b
+ cells within optic nerve but few within the retina [
37]. These results suggest that the retina differs from optic nerve and brain in that retina lacks the environment that supports myeloid cell proliferation in the absence of frank inflammation.
Repopulation of microglia in the retina and brain after tamoxifen-induced DTA ablation was characterized by a transient spike in GFP
hi cells. This was a rapid and relatively greater change in the brain (about 8-fold over background at day 8 post-tamoxifen) compared to the retina (about 3-fold over background at day 47 post-tamoxifen). While not accounting for the difference in timing in the GFP
hi cell response between retina and brain, we speculate that a lack of interplay between CD115 and its ligands CSF-1 and IL-34 contributes to its transient nature. CNS microglia numbers are reduced in the absence of CSF-1 and IL-34 [
49,
52‐
58] suggesting the importance of these ligands in maintaining normal numbers of microglia. We observed that CD11b
+GFP
hi cells had much lower levels of CD115 expression and that there was no change in either CSF-1 or IL-34 expression in tamoxifen-induced DTA depleted versus control retinas. Our results imply that CD115 and CSF-1/IL-34 levels were enough to replace and/or maintain GFP
lo microglia levels but could not support a sustained increase in GFP
hi microglia following tamoxifen-induced DTA ablation. The low levels of CD115 on GFP
hi cells would also account for their transient nature after ONC without microglia ablation despite a slight increase in retinal CSF-1 levels after injury.
Although concerns about the relevance and effects of irradiation have limited the use of radiation bone marrow chimerism strategies, we believe the approach has some merit. If microglia are self-renewing [
46] or proliferating [
72], or if there is a local microglia progenitor [
49], they all may be radiosensitive, providing a strategy to induce their turnover and replacement by the grafted bone marrow. Repopulation of microglia in bone marrow chimeras differed from tamoxifen-induced DTA ablation in that brain and retina had a similar slow rate of replacement by donor myeloid cells. In tamoxifen-induced DTA-ablated brain and retina, the total number of CD11b
+ cells declined sharply from normal levels before recovery whereas, in bone marrow chimeras, the total number of retinal CD11b
+ cells remained close to normal with a gradual replacement of host microglia by donor myeloid cells. We found this to be the case for both retinal microglia niches as donor bone marrow cells slowly replaced both the resident GFP
lo and GFP
hi microglia. This is consistent with other reports that found the replacement of retinal resident microglia by donor-derived myeloid cells was an extended process in the absence of further manipulations [
42].
The other major difference between tamoxifen-induced DTA and radiation-ablated retinal microglia was the ability of the replacement microglia, particularly the GFPhi cells, to appear in response to injury. Attempts to stimulate recruitment or elevation of GFPhi cell numbers by ONC were unsuccessful after tamoxifen-induced DTA ablation, even at 47 days post-tamoxifen when recovering retinas contained a supranormal level of GFPhi cells and a subnormal level of GFPlo microglia. In contrast, the outcome of injury was quite different if the ONC was performed in radiation bone marrow chimeric mice. Even when there was only a partial recovery of retinal GFPhi cells after bone marrow transfer, an ONC simulated a significant response including a 90% reduction in recipient retinal microglia by 7 days post-ONC, and a large increase in donor-derived myeloid cells. A significant portion of the donor-derived cells, approximately 35%, were GFPhi. The overall magnitude of the ONC response was similar to that seen in a normal, non-chimeric CD11cDTR/GFP retina, but was greater than 95% derived from circulating donor cells compared to less than 1% in normal mice. Further, approximately half of the retinal myeloid cells in chimeric mice after ONC were CD45hi, suggesting their recent influx into the retina. This vigorous, donor-derived GFPhi cell response to ONC in chimeric mice could be observed to at least 128 days post-bone marrow transfer. We believe the ability to generate a GFPhi cell response to ONC in radiation versus tamoxifen-induced DTA-ablated mice is 2-fold. First, in bone marrow chimeric mice, circulating cells recruited to the retina are CD115hi and not expressing GFP, thus amenable to proliferation and/or expression of GFP upon injury stimulation. In contrast, after tamoxifen-induced DTA ablation, new retinal microglia are likely being recruited from the optic nerve and skewed compared to normal towards cells that are already CD115loGFPhi through at least 47 days post tamoxifen. Under these conditions, an ONC cannot enhance the rate of migration from the optic nerve.
These results are consistent with other reports demonstrating circumstances by which circulating monocytes [
73], or other progenitor cells distinct from microglia could be recruited into the retina [
68,
74,
75]. However, our results also suggest that regardless of the identity of the circulating precursor found to enter the retina in radiation bone marrow chimeric mice, their phenotype and function came to resemble the endogenous microglia. At 5 weeks, post-transfer retinal myeloid cells in B6 recipients given CD11c
DTR/GFP bone marrow were largely CD45
medCD11b
+Ly6G
−Ly6C
−F4/80
+Iba-1
+GFP
lo with a subpopulation being GFP
hi. They were highly ramified and otherwise indistinguishable from microglia. Following ONC, there was a sharp increase in donor origin GFP
hi cells, and they formed a close association with the injured retinal ganglion cells and their axons similar to that seen in non-chimeric mice.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.