Human donor eye tissue was obtained from the Manchester Eye Tissue Repository, an ethically approved Research Tissue Bank (UK NHS Health Research Authority reference no.15/NW/0932). Eye tissue was acquired after the corneas had been removed for transplantation and explicit consent had been obtained to use the remaining tissue for research. Guidelines established in the Human Tissue Act of 2004 (United Kingdom) and the tenets of the Declaration of Helsinki were adhered to. All ocular tissue was processed within 49 hours postmortem and AMD status graded in accordance to the Age-related Eye Disease Study (AREDS) classification. The Manchester Eye Tissue Repository was subsequently transferred to the University of Tübingen and renamed the Helmut Ecker Eye Tissue Resource (HEETR, Tübingen, Germany, local ethical approval 370/2021BO1). We analyzed the ocular globes of patients diagnosed with early (2 patients) or late (3 patients) AMD. In the experiments we used age-and sex-matched controls (3 patients) without any known ocular vascular or inflammatory pathology, nor RPE or retinal lesions. All information with regard of the donors is compiled in Supplementary Table S1 .

Mice were bred in-house and maintained on a 12-hour light and 12-hour dark cycles, standard environmental conditions and unrestricted access to food and water.

Cx3cr1 ^GFP/GFP mice, hereinafter referred as Cx3cr1, were examined at 8 and 12 months of age. C57BL/6J mice, hereinafter referred as WT, served as sex- and age-matched controls. Male and female littermates were used in a 1:1 ratio. Cx3cr1 were bread on a C57Bl6/J background. Mice were routinely monitored for mutations in Crb1 (RD8) and tested negative (23).

At the end of the experiment, mice were anesthetized with isoflurane and euthanized through cervical dislocation. Eyes and spleens were rapidly excised and kept in PBS supplemented with 0.5% BSA (Sigma Aldrich) and 2mM EDTA (Sigma Aldrich) at 4°C. The neurosensory-retina and RPE-choroid complexes were later dissected and analyzed separately depending on the experiment.

We used a principal component analysis (PCA) for dimensional reduction using the R package factoextra (version 1.0.7). Top 12 features that contributed to the first (Dim. 1) and second (Dim. 2) PCA dimensions were extracted (stats, version 3.6.2). For visualization, the plugin ggplot was used.

All experiments were repeated at least 3 times. N (biological replicates) is indicated in each respective figure legend throughout the manuscript. Values are expressed as the mean ± SD unless indicated otherwise. Statistical analysis was performed using GraphPad Prism 9 and R Studio. In case of single comparisons of two groups, normality was assessed by Kolmogorov-Smirnov test and outliers excluded with Grubb´s test. In groups with n<5 normality could not be tested, and a non-parametric distribution was assumed. Accordingly, groups were tested with two-tailed student´s t-test or two-tailed Mann-Whitney U test depending on the assumption of normal distribution of the data points.

To determine the functional state of the immune system and the potential presence of T cells in the retina during AMD, we performed histological analyses of the RPE and choroid of patients with early and late AMD compared to age- and sex-matched controls ( Supplementary Table S1 ). Overall, we observe a significant increase of memory T cells (CD3⁺CD45RO⁺) in patients with early AMD. These frequencies returned to basal levels at more advanced late AMD stages ( Figures 1A, B ). The findings convey that a recruitment of memory T cells occurs at early stages of AMD and prior to manifesting the degenerative hallmarks of AMD.

We sought to investigate if the recruitment of memory T cells occurs prior to degenerative features of AMD in a murine mouse model. Cx3cr1-deficient mice manifest naturally occurring RPE degeneration and drusen formation in the course of aging. 8-month-old animals exhibit a pro-inflammatory state without apparent morphological changes in RPE. 12-month-old animals show a pro-inflammatory state with apparent RPE degeneration. We analyzed ocular samples of Cx3cr1 (referred to as Cx3cr1 in the following) at 8 and 12 months of age. While analyzing RPE-choroid complexes, we detected a significant increase in the number of T cells (CD3⁺) compared to C57Bl/6J control ( Figures 1C-E ). More specifically, T cells were largely detected at the vicinity of patches of RPE seemingly in transition to degeneration, not inside but in the close periphery of the degenerative area. ( Figure 1D ). T cells in the Cx3cr1 retina displayed a wide morphological variance and consolidating activated phenotypes ( Figure 1F ). Furthermore, orthogonal sectioning and three-dimensional reconstruction of the confocal images confirmed that T cells were intimately associated to the basolateral membrane of RPE, suggesting a migration from the choroid and adherence to the RPE ( Figures 1G, H ).

Collectively, our findings support that T cells accumulate during early AMD prior to degeneration in our mouse model, and that these cellular populations are associated to the RPE.

Next, we sought to investigate the diversity of T cell phenotypes in AMD by flow cytometry. We evaluated the neuroretina-RPE-choroid complex to capture organ-specific immune signature at 8 (basal time-point) and 12 months of age (prior to degeneration) in Cx3cr1 mice and compared to C57Bl/6J control. For comparison with the systemic immune signature, we included the spleen as a systemic lymphoid organ. At 8 months of age, we did not observe differences in the ocular immune phenotypes of Cx3cr1 mice at the T cell level compared to control ( Figures 1I, J ), nor systemically ( Figures 1K, L ). Interestingly, with age (12 months), the ocular profiles of C57Bl/6J mice showed a significant reduction of frequencies of T cells (CD3⁺) and T helper cells (CD3⁺ CD4⁺), whereas the frequencies of CD3⁺ increased significantly in Cx3cr1 ( Figure 1J ). Importantly, these dynamics were only observable in the ocular samples, and the CD3⁺ and CD4⁺ T cell systemic populations were significantly decreased in both C57Bl/6J and Cx3cr1 ( Figure 1L ). Furthermore, we found a significant increase of cytotoxic T cells (CD3⁺ CD8⁺) specifically in the eyes of Cx3cr1 during aging ( Figure 1J ). Conversely, cytotoxic T cells remained stable in the spleen ( Figure 1L ). Together, these changes led to a significantly lower ratio of CD4⁺/CD8⁺ T cells in C57Bl/6J and Cx3cr1 mice, suggesting a lymphocyte immunosenescence and active recruitment of cytotoxic T cells prior to the destructive features of retinal degeneration.

Complementarily, we evaluated potential sources of innate immune inflammation, yet no differences were found regarding myeloid cell frequencies (CD45⁺ CD11b⁺) at these time-points ( Supplementary Figures S3A–B ). We also evaluated populations of T cells potentially interacting with the innate response, but we could not observe changes in γδT cells (CD3⁺TCRγδ⁺) in the eye, albeit a decrease of these populations was observed in the spleen of control animals ( Supplementary Figures S3A–B ).

Taken together, our data show an influx of memory T cells, specifically cytotoxic T cells to the eye prior to the manifestation of retinal degeneration in human and mice.

We used dimensionality reduction approaches to determine whether the T cell profiles in the context of experimental early AMD are organ specific. Using a principal component analysis (PCA) of hierarchically gated T cell subpopulations, we show a specific immunophenotype in the ocular samples (neuroretina-RPE-choroid) compared to the spleen in both C57Bl/6J and Cx3cr1 mice ( Figure 2A ). PCA analysis of the ocular populations revealed a strong time-point-dependent immunophenotype in the Cx3cr1 mice compared to control ( Figure 2B ). To study the dynamics of T cell subpopulations, we analyzed PCA loadings. Our analysis indicates that changes in memory and activated T cell subsets are important for immune phenotype changes over time ( Figures 2C, D ). Age-related shifts in Cx3cr1 mice were mainly driven by KLRG1⁺ and CD69⁺ subsets of memory T cells (Dim1, Figure 2C ), known markers for effector functions in T cells as well as tissue residency, and participating in premature T cell immunosenescence.

Since the PCA analysis revealed the potential contribution of memory T cell subsets, we subclustered T cells according to their CD44 and CD62L expression. At 8 months of age, Cx3cr1 mice did not present manifestations of retinal degeneration, yet we observed a strong upregulation of central memory T cells (CD44⁺CD62L⁺) within their ocular immunophenotype compared to C57Bl/6J controls ( Figure 2E ). Importantly, these changes were exclusively found in the eye and not systemically ( Figure 2F ). In C57Bl/6J control mice, central memory T cells were hardly detectable. Similarly to the central memory T cells, we found an early upregulation of naïve T cells (CD44^-CD62L⁺) at 8 months in Cx3cr1 mice. Over time, central memory and naïve T cell frequencies decreased. On the other hand, the number of tissue-resident memory T cells (CD44⁺CD62L⁻CD103⁺CD69⁺) increased significantly, while this population tended to decrease in controls. These dynamics of naïve memory T cell regress and tissue-resident memory T cell upregulation was even detected systemically ( Figure 2F ).

Our T cell analysis suggests a role for an early hit of the memory T cell system in AMD pathogenesis before the occurrence of degenerative changes in the retina. Thus, we consider the adaptive immune system as one of the causative elements in the etiology of retinal degeneration in this model.

We sought to investigate regulatory T cells (Treg) and Th17 cells as prototypic anti- and pro-inflammatory T cells in central nervous system (CNS) inflammation, respectively. Recently studies reported alterations of these T cell subsets in AMD patients and animal models (16). In accordance with the literature, we observed a retina-specific increase of Th17-like cells (CD3⁺ CD4⁺ CD196⁺) in Cx3cr1 mice, whereas no frequency alteration was detected in C57Bl/6J ( Figure 2E ). No trend was detected in systemic CD196⁺ Th17-like populations ( Figure 2F ).

Focusing on Treg (CD3⁺ CD4⁺ CD25^high), we observed an upregulation at 8 months, paralleled by the upregulation of central memory and naïve memory T cells ( Figure 2E ) in Cx3cr1 mice. Over time, ocular Treg decrease in Cx3cr1, whereas conversely a systemic Treg increase became apparent ( Figure 2F ). No dynamic in ocular nor systemic Treg population was detected in C57Bl/6J.

Overall, this pattern indicates an early anti-inflammatory response that precedes the disease phenotype in the retina. Importantly, this anti-inflammatory pattern was lost in the diseased retinal states.

Considering the changes in T cell surface markers KLRG1 and CD69 in the aging retina, we sought to characterize the activation status of cytotoxic and helper T cells. Specifically, within cytotoxic T cells, CD69 and KLRG1 known markers for cell activation and tissue residency were regulated in retinal immune populations.

In retinal tissue of Cx3cr1, we observed an increase KLRG1⁺ as well as CD69⁺ KLRG1⁺ double positive cells over time. These changes were not found in C57Bl/6J; in contrast, we rather found a decrease of KLRG1 from the timepoint 8 months to 12 months ( Figures 3A, B ). In comparison, splenic CD8⁺ cells did not exhibit an alteration of neither KLRG1 nor CD69 ( Supplementary Figure S4B ), indicating a local retinal activation of the specific immune system in Cx3cr1 mice. For T helper cells, we found a similar pattern of activation and antigen experience ( Figures 3E, F ). Systemically, we observed an upregulation of activated CD69⁺ T helper cells in Cx3cr1 but no alteration in C57Bl/6J ( Supplementary Figure S4F ).

Since KLRG1 and CD69 are predominantly expressed in effector memory T cells, we focused our analysis on these subtypes. In C57Bl/6J and Cx3cr1, the number of cytotoxic effector memory T cells increased during aging ( Figures 3C, D ). This dynamic was specific for retinal populations, since the systemic analysis exhibited stable frequencies of these memory populations ( Supplementary Figure S4D ). Although both groups expressed the activation marker CD69 at 12 months to a greater extent, only Cx3cr1 mice displayed an elevation of KLRG1 within the cytotoxic effector memory T cell population. In contrast, C57Bl/6J mice showed a decrease KLRG1 expression over time. Additionally, we found an increase in CD69⁺ KLRG1⁺ cells over time in Cx3cr1 mice. These activation status effects were reflected in the systemic analysis of CD8⁺ effector memory T cells in Cx3cr1 mice, whereas a downregulation of CD69⁺ KLRG1⁺ population in C57Bl/6J controls became apparent ( Supplementary Figure S4D ). Regarding T helper cells, no changes in the frequency or activation markers of effector memory cells were detected in C57Bl/6J ( Figures 3G, H ). However, the frequency of retinal CD4⁺ effector memory population in Cx3cr1 increased over time. Retinal KLRG1⁺cells as well as CD69 KLRG1 co-expressing populations within the CD4⁺ effector memory T cells were dynamically upregulated from timepoint 8 to 12 months. Systemically, we observed an increase in the number of CD4⁺ effector memory T cells in Cx3cr1 mice. Additionally, a systemic upregulation of KLRG1 and CD69 can be observed in the CD4⁺ effector memory compartment of the Cx3cr1 mice. ( Supplementary Figure S4H ) This again contrasts with the unaltered systemic effector memory T cells of C57Bl/6J, corroborating a systemic impact of retinal activated T cells in the Cx3cr1 animals.

In summary, our flow cytometric analysis reveals upregulation of markers for tissue residency and effector function of T cells specifically in the retina/RPE/choroid complex of Cx3cr1 mice.

Immunophenotyping of T cell populations suggests the recurrence of interactions among different cells, including memory T cells and other pro- and anti-inflammatory subpopulations. To provide further evidence for this active immunomodulatory scenario specific to the retina and choroid, we profiled gene expression of relevant genes involved in inflammation in neuroretina and in RPE-choroid complex samples.

Globally, the inflammatory signature found in the neuroretina was modest, showing only an upregulation of transcription factors Il23 and Foxp3 in 12-month-old Cx3cr1 mice ( Figure 4A ). In the RPE-choroid complexes, where most observed degenerative features take place, we observed a transcriptomic activity ∽10-fold higher to baseline ( Figure 4B ). Furthermore, we observed an age-dependent dynamic shift (8 months to 12 months of age) only in Cx3cr1 mice, and not in C57Bl/6J control. The most expressed interleukins found in aged Cx3cr1 RPE-choroid samples were Il1b, Il2, Il6, Il7, Il8 (CXCL1), Il15 and Il23. Tcrb was strongly upregulated along with genes coding for the adhesion and activation molecules Icam1 and Itgb2 (CD18). Moreover, aging also influenced the expression of genes known to contribute to AMD pathogenesis, including C3, Cfh, Pgf, Tnf, Tgfb and CCL2 (1, 5). Other genes such as Ctla2a were also increased over time in the Cx3cr1 RPE-choroid, whereas the expression of Rbp3 was significantly decreased.