Blinding diseases like retinitis pigmentosa and macular degeneration arise from loss of photoreceptors over time, concurrent with aging in some disorders. The effect of prior degeneration on the functional states of retinal immune populations could thus influence the susceptibility of the retina to subsequent insults and cell loss over one’s lifespan. Here, we find that photoreceptor degeneration is associated with an increase in the total number of myeloid cells in the retina, and that these cells remain morphologically, transcriptionally, and functionally heterogeneous.
Dramatic increase in CD45+ cells in the retina following degeneration
In previous work, the large increase in myeloid cells in the retina during photoreceptor degeneration in the
Arr1−/− mouse appeared to be the result of both infiltration of monocytes from the periphery and proliferation of myeloid cells within the retina [
9]. Here we show that the number of myeloid cells eventually falls as degeneration wanes, but never quite returns to normal pre-degeneration levels. Because loss of photoreceptors during degeneration results in a dramatic decrease in the volume and total number of cells in the retina, even this relatively small increase in total immune cells results in a considerable increase in the density of immune surveillance within the retina. Our laser damage experiments show that this higher density of surveillance does not seem to affect the time course of the microglial response to subsequent injury compared to previous work in otherwise healthy retinas [
19]. Whether or not the change in cell number or composition accounts for the apparently poor responsiveness of the monocyte-derived macrophages remains an open question. Interestingly, our EdU experiments show that roughly half of the myeloid cells in the post-degeneration mosaic were born when degeneration was most active (Fig.
1D), but we do not know whether these cells all locally proliferated or to what extent they entered the retina as EdU + newly born monocytes. Our lineage tracing experiments show that roughly half of the immune cells are monocyte-derived (Fig.
4C), but whether or not an equal fraction of microglia and monocyte-derived cells proliferated to re-populate the retina remains a question for future studies.
Monocyte-derived cells remain in the retina following degeneration
Our flow cytometry data here, in combination with previous work, suggest that these monocyte-derived macrophages arise from monocytes that invaded the retina during the first few days of degeneration, rather than resulting from continued invasion of monocytes throughout the entire course of degeneration. Newly born monocytes express high levels of Ly6C, which then decreases over the cell’s lifetime, especially if the monocyte differentiates into a monocyte-derived macrophage [
29]. In previous work, flow cytometry results showed that the number of Ly6C
high cells in the
Arr1−/− retina peaked between 24 and 48 h during the height of photoreceptor degeneration, which corresponded well to the time course of the increase in Müller-cell derived CCL2 expression (monocyte chemoattractant protein) within the retina [
10]. In this study, we examined the number of Ly6C
high cells in the retina at longer times, after degeneration had slowed and found that the number of Ly6C
high returned to non-detectable baseline levels by 10 days (Fig.
2E). Immunohistochemistry and Sholl analysis revealed that these monocyte-derived cells had processes, though were somewhat less ramified than the resident microglia (Fig.
5A–C). Thus although the integrity of the blood retinal barrier has not been explicitly examined in this model, together these results suggest that the monocyte invasion is limited to the peak period of degeneration, and that these monocytes rapidly differentiate into microglia-like macrophages that persist in the retina for at least several more weeks.
The monocyte-derived macrophages that take up residence in the retina after photoreceptor degeneration adopt a ramified morphology that is spatially intermingled with the endogenous microglia (Fig.
4A), similar to the monocyte-derived cells that infiltrate the retina following NaIO
3 administration to damage the RPE [
15]. Unlike the macrophages in other ocular compartments with high turnover rates [
37], these monocyte-derived macrophages in the retina are long-lived, and our morphological analysis shows that they are smaller and less complex that resident microglia on average (Fig.
5A–C). The extent of morphological difference between resident and monocyte-derived macrophages may vary by disease. For example, in the NaIO
3 model of RPE damage, resident and monocyte-derived macrophages were morphologically indistinguishable [
15].
Although monocyte-derived macrophages adopt a largely microglial phenotype significant differences in gene expression remain. One such difference readily detectable at the protein level is MHCII. Immunohistochemically, the MHCII
+ cells observed in the degenerated retina (e.g., Figure
1B) mostly correspond to the monocytic cells (ratio of 6:1 monocyte to microglia-derived cells; Fig.
5D). MHCII
+ resident macrophages have been observed occasionally in other models of retinal/RPE injury [
15,
37]. Notably, here we find that the relatively few resident cells that express high levels of MHCII proteins are morphologically indistinguishable from monocyte-derived macrophages (Fig.
5F–H). Thus, while MHCII staining may be a useful tool for qualitative assessment of peripheral infiltration that is much easier and cheaper than fate-mapping experiments, we emphasize that, like morphology, MHCII expression alone is not sufficient to determine the lineage of a given cell.
Monocyte-derived, or other non-microglial macrophages, have been observed to similarly adopt microglia-like phenotypes and remain in the retina in the absence of continuing or active neuronal loss in previous work. However, those instances have been situations in which there is physical damage [
22], disruption of blood–tissue barriers [
15,
37], or a pharmacological intervention that ablates the resident microglia [
6]. Monocytic cells have been detected in other photoreceptor degenerative disorders, including widespread light damage [
20] and age-related macular degeneration [
27], although it has not yet been investigated if monocytes permanently engraft into the retina in these instances. To our knowledge, this is the first demonstration of monocyte-derived macrophages taking up long-lived residence in the retina following cell autonomous degeneration. The “new resident’’ monocytic cells adopt a microglia-like morphology and gene enrichment program aligned with a monocyte-derived macrophage phenotype; however, they fail to functionally respond to a subsequent insult as well as the resident microglia. The molecular basis of the apparent functional laxity of the monocyte-derived cells remains to be determined.
In the
Arr1−/− experiments presented here, monocyte-derived macrophages adopted a microglia-like phenotype over the course of days, consistent with evidence from other studies that the retinal environment promotes a microglia-like phenotype for myeloid cells. For example, after pharmaceutical ablation of retinal microglia, non-microglial immune cells from the periphery invade and adopt morphologies and gene expression reminiscent of microglia [
6,
17,
22]. Non-microglial repopulation following microglia ablation has not been observed in the brain thus far [
7]. However, following whole body irradiation and bone marrow or hematopoietic stem cell transplantation, non-microglial macrophages can invade the brain and develop similar microglia-like morphologies and some shared gene expression [
28]. This study identified a non-microglia-specific gene expression program expressed following LPS challenge [
28], and some of these genes were found to be similarly expressed in the small cluster of inflammatory macrophages identified here (Cluster #4), but are notably absent in the more microglia-like cells (Cluster #3, Fig.
3C). All together the CNS seems to promote a microglia-like phenotype for mononuclear phagocytes, although it remains unclear to what extent this phenomenon may differ between the brain and retina.
A specific disease-associated microglia (DAM) phenotype first observed at the single-cell level in a mouse model of Alzheimer’s disease [
11] has been observed in a wide range of neurodegenerative disorders [
4] and similar changes in microglial gene expression have been observed in several models of retinal degeneration [
21,
24,
37]. Many DAM-associated genes were indeed expressed by activated microglia in this study, and interestingly these genes were also expressed in the microglia-like macrophages cluster (Fig.
3C). This phenotype of activation was not observed in the small cluster of pro-inflammatory monocyte-derived macrophages, indicating that this is not a pan-activation phenotype in this model.
Unexpectedly, many of the most highly expressed genes in the monocyte-derived macrophage cluster in our single-cell dataset have recently been identified as key genes expressed by various macrophage subsets in peripheral CNS tissues, such as the meninges and choroid plexus [
35]. However, to what extent these conserved gene expression programs are indicative of shared functions, similar environment, or similar cell lineages remains unknown.
Heterogeneous resolution of neuroinflammation across models of acute degeneration
In the field of retinal degeneration, activation and resolution of the immune response has been investigated in a model of RPE injury that causes photoreceptor loss [
15] and following a corneal alkali chemical burn that leads to retinal ganglion cell death [
22]. In both models, expression of genes associated with immune activation are detected well after cell loss ceases, suggesting that at least some of the retinal immune cells do not return to rest [
15,
22]. Because neither study utilized single-cell transcriptomics, the degree of heterogeneity in those post-degeneration immune populations remains unclear. However, the resolution of the immune response to neuronal loss has been investigated at a single-cell level in a model of facial nerve axotomy [
32,
33]. Although this model does not result in widespread neuronal loss like the retinal
Arr1−/− model, the loss of neurons does drive the activation and proliferation of nearby microglia, and the immune response resolves approximately 30 days after injury. In this case, single-cell sequencing revealed that although the majority of microglia return to a resting phenotype, a small number of microglia continue to express many activation-related genes like the retinal degeneration models described above, and similar to the mildly activated microglia phenotype we observe in the single-cell dataset described here (Fig.
3).
Curiously, this persistent, mildly activated phenotype observed in all the above models has similarities to changes observed in microglia during aging. In the aging retina, microglia accumulate in the subretinal space, increase in number, and increase their expression of activation- and inflammation-related genes [
3,
14]. Typically, aged microglia are not as responsive to injury as young microglia [
14], yet here we find that microglia are still capable of robustly responding to focal injury even after all photoreceptors are lost (Fig.
6B,
C). The similarities of microglia in neurodegenerative disorders and those in aging are noticeable and could have implications for further immune challenges or therapeutic interventions with age, especially after potential repeated challenges throughout life.
In summary, we find that photoreceptor degeneration causes long-lasting changes to the retinal immune environment, lasting well past when neuronal loss has completed. Although we find that these retinal macrophages can still respond to further retinal insult, the retinal immune environment does not entirely return to rest, and the microglial and monocytic cells respond differently to subsequent injuries and insults. Determining the precise molecular signals and pathways involved in these distinct subpopulations and their functional long-term differences, if any, will be the focus of future studies. Of particular interest will be determining if this altered immune state contributes to late-stage retinal remodeling, and if these immune cells affect therapeutic interventions such as gene therapies and cell transplants.