Background
Diabetes mellitus, a metabolic disorder resulting in high blood glucose levels, causes systemic complications, including chronic kidney disease, fatty liver, cardiovascular disease, and vision loss due to diabetic retinopathy (DR) [
1]. Although the exact mechanisms of DR pathogenesis are not fully elucidated, long-term hyperglycemia is associated with local retinal inflammation, weakened blood vessels leading to vascular damage, exudate production, hemorrhages, ischemia, and neuronal damage [
2]. In addition to diet changes and insulin administration, intravitreal delivery of pharmacologic agents to target angiogenesis and surgical scarred tissue removal are common treatment modalities [
2]. Unfortunately, current clinical approaches only slow vascular damage, but do not restore vision loss [
2].
Maintenance and function of the blood retinal barrier (BRB) is supported by the dynamic communication between the diverse cell types comprising the retinal neurovascular unit (NVU) which include glial cells (müller glia, astrocytes, and microglia), vascular cells (endothelial cells and pericytes), and various neural cell types (amacrine cells, horizontal cells and ganglion cells). Prolonged hyperglycemia leads to BRB breakdown resulting in the three classical hallmarks of DR, (1) vascular damage, (2) neurodegeneration and (3) inflammation. As vascular damage develops in the diabetic retina, VEGF-mediated angiogenesis occurs producing poorly perfused blood vessels that escalate damage to the diabetic retina creating an ischemic environment and BRB breakdown [
3,
4]. Understanding neurodegeneration in the diabetic retina may provide pathways of intervention to protect neurons as they constitute the major cell type responsible for processing and executing visual information.
Resident retinal macrophages (microglia), as phagocytes of the central nervous system (CNS), coordinate synapse development in the early postnatal retina and in adulthood support the survival and maintenance of neurons [
5]. Microglia also guide blood vessel formation and maintain vascular integrity in the retina [
5,
6]. In the diabetic retina, microglia rapidly respond to hyperglycemia, leukostasis and vascular leakage [
7,
8]. Hyperglycemia induces vascular damage that leads to leakage of serum proteins and danger associated molecular patterns (DAMPs) into the retina, creating a loop of inflammation perpetuated by microglia proinflammatory cytokine production. However, the exact mechanisms of microglia-mediated neuroprotection or neurotoxicity in the retina remain unclear.
To clarify the contribution of retinal microglia to disease progression we sought to transiently deplete and repopulate microglia using a genetic approach, CX3CR1
CreER:R26
iDTR transgenic mice, and a pharmacological approach via CSF-1R antagonist (PLX-5622) in a streptozotocin-induced diabetic model [
9,
10]. Neither model resulted in complete elimination of the Iba1
+ retinal microglia cells, but lead to significant reduction in the overall density of Iba1
+ cells. Three-day treatment with diphtheria toxin in CX3CR1
CreER:R26
iDTR mice resulted in depletion of ~ 65% of retinal microglia. Treatment with PLX-5622 for 2 weeks resulted in ablation of ~ 74% of retinal microglia, whereas 2-week treatment with diphtheria toxin (DTx) in CX3CR1
CreER:R26
iDTR mice induced robust retinal EdU
+ microglial proliferation. Detailed morphological analyses of microglia following the depletion regimens revealed morphologically ameboid cells present in the diabetic retina. However, in contrast to these phenomics, a recovery period in diabetic tissues reverted microglia back to a ramified morphology most commonly present in the non-diseased retina. We further determined that 2-week recovery period in diabetic CX3CR1
CreER:R26
iDTR mice was associated with neuroprotective cues to prevent neuronal loss and decrease fibrinogen extravasation into the diabetic retina. Similar observations were noted in PLX-5622-treated CX3CR1-WT mice. Additionally, mRNAseq analysis of diabetic, PLX-5622-treated retinas revealed a retinal transcriptome with reduced expression of genes associated with DR pathogenesis, microglial activation and complement activation and synaptic pruning in microglia depleted and repopulated groups. Together this data supports the idea that microglial reprogramming can be used to enhance a homeostatic microglia cell population to regulate RGC loss and support vascular repair in the diabetic retina.
Discussion
To clarify the role of microglia in DR initiation and progression, we utilized two models to deplete microglia, a spatially and temporally controlled conditional model using CX3CR1
CreER:R26
iDTR mice, and a pharmacological model, PLX-5622, in
CX3CR1-WT mice [
9,
10]. Analysis of brain, spinal cord and retinal CX3CR1
CreER:R26
iDTR tissues revealed CNS regional differences in the extent of microglial depletion, with greater susceptibility in the retina following acute DTx treatment. Prolonged DTx treatment induced significant Iba1
+Edu
+ retinal microglia proliferation, whereas PLX-5622 microglial depletion yielded robust and proportional microglial depletion in brain and retinal tissues. Among both models, we found that DTx treatment in CX3CR1
CreER:R26
iDTR mice and PLX-5622 treatment in CX3CR1-WT mice correlated with reduced pathogenic angiogenesis, vascular damage, and neuronal cell loss in diabetic retinas. Furthermore, a 2-week recovery after DTx treatment in CX3CR1
CreER:R26
iDTR mice, maintained these neuroprotective and vasculo-protective effects. mRNAseq analysis of PLX-5622 transient microglia depletion and repopulation illuminated potential regulatory pathways related to neuronal and vascular damage and inflammation.
Herein this study, we compared CX3CR1
CreER:R26
iDTR brain and retinal tissues for the degree of microglia depletion following acute DTx treatment to address the possibility that microglia depletion can have various effects depending on the CNS compartment. In comparison to the brain, we found that the retina was more susceptible to acute microglia depletion, with greater than 60% microglia depletion in DTx-treated mice (Fig.
1E–H). We also report that PLX-5622 yields a robust amount of microglia depletion, in the brain by flow cytometric analysis (Additional file
1: Fig. S6B), and the retina by IHC analysis (Fig.
5B, C). In contrast, in the DTR model, DTx treatment did not induce any alterations in the percentage of CD11b
+CD45
LoZombie
– microglia in the brain and spinal cord detected by flow cytometry (Additional file
1: Fig. S6B). The striking difference in the degree of depletion could be due to the pharmacokinetics and half-life of DTx and PLX-5622, and the frequency with which these drugs are delivered. Additionally, following the 2-week DTx treatment, we observed a robust increase in Iba1
+ cells in the diabetic CX3CR1
CreER:R26
iDTR retina (Fig.
3A, B). Previous studies have shown that following depletion, microglia repopulate from a resident pool of depletion-resident microglia and from peripheral infiltrating macrophages that can acquire a microglia-like signature [
17]. Flow cytometric analysis of brain and spinal cord revealed a significant increase in the percentage of Ly6C
+P2RY12
+ mononuclear cells in diabetic mice, with no differences in PBS versus DTx treated mice (Additional file
1: Fig. S3G). However, following a 2-week recovery, DTx-treated mice had a significant reduction in the percentage of Ly6C
+P2RY12
+ mononuclear cells in the brain and spinal cord (Additional file
1: Fig. S3G) and in the number of Iba1
+ cells in the retina (Fig.
3A, B). Furthermore, analysis of DTx treatment in the non-diabetic CX3CR1
CreER:R26
iDTR retina revealed a significant ~ 31% increase in Iba1
+ cells in the non-diabetic DTx-recovery retina, and a significant ~ 89% increase (Student’s
t test,
P = 0.0084) in Ly6C
+P2RY12
+ mononuclear cells in brain and spinal cord by flow cytometry, suggesting that this increase in Iba1
+ cells in the retina could potentially be from infiltrating monocyte-derived macrophages (Additional file
1: Figs. S4 and S5). Moreover, since acute DTx treatment induced microglial depletion and prolonged DTx exposure induced microglial proliferation, these findings raise the question if continued DTx exposure increases proliferation rates in microglia. Additional studies assessing various timepoints following DTx could provide further insights into microglia depletion and proliferation following DTx treatment.
In experimental models of DR and human diabetes, neurodegeneration has been reported to be an early indicator of DR pathogenesis as evident by a robust increase in TUNEL-positive neurons and thinning of the retinal nerve fiber layer (RNFL) [
28‐
30]. Neurodegeneration in DR has been shown to be complement-mediated and complement-based therapies are a strong candidate for treatment of retinal degenerative diseases [
31]. In models of age-related macular degeneration (AMD), C3-expressing microglia and complement deposited proteins colocalized to lesions of photo-oxidative damage perpetrating neurodegeneration and intravitreal injection of C3 small interfering RNA (siRNA) prevented neurodegeneration [
32]. A downregulation in complement-associated gene clusters (
C3ar1,
C1qb,
C1qa, C1qc) in diabetic PLX-5622-treated mice relates to the observations of less NeuN
+RBPMS
+ neuronal cell loss in CX3CR1
CreER:R26
iDTR mice (Figs.
4 and
6). In addition to preserved NeuN
+RBPMS
+ neurons, both models, DTx treatment in CX3CR1
CreER:R26
iDTR mice and PLX-5622 treatment in CX3CR1-WT, were associated with an increase in TUJ1
+ axonal density in the diabetic retina (Figs.
4 and
5). The keratin class of intermediate filaments are the major source of neurofibrils that give axons their fundamental structure therefore we focused on keratin-associated gene clusters since [
33]. PLX-5622 treatment induced an upregulation of keratin-associated gene clusters (
Krt17,
Krt16,
Krt14,
Krt15,
Krt6a,
Krt5 and
Dsp) (Fig.
6) substantiating our findings of increased TUJ1
+ axonal density (Figs.
4 and
5). Although in a model of optic nerve crush, microglia depletion did not alter retinal ganglion cell degeneration and regeneration, data presented here supports other studies highlighting the neuroprotective effects of microglia depletion in retinopathy models, including autoimmune uveitis and excitotoxicity-induced neuronal cell death in which microglia depletion alleviated clinical symptoms and the degree of neuronal cell loss [
34‐
36].
Under inflammatory conditions, reactive astrocyte production of metalloproteases and upregulation of tight junction (TJ) proteins and junctional adhesion molecules (JAMs) strengthens the glial limitans of the blood brain barrier to restrict leukocyte extravasation and adhesion to inflamed venules that exacerbate CNS lesion formation [
37‐
39]. Additionally, astrocyte production of neurotrophic factors also aids in vascular protection and promotes neuroprotection [
40‐
42]. Together the upregulation
GFAP,
Claudin-1,
Claudin-4,
F11r,
Adamts13 and
Csf3 (Additional file
1: Fig. S7) supports the amelioration of vascular damage and fibrinogen deposition visualized by IHC analysis of CD31
+ blood vessels and fibrinogen (Figs.
4 and
5) in both models of microglia depletion used. IHC analysis of GFAP
+ glial responses from astrocytes and Müller glia revealed significant increase in the GFAP
+ percent immunoreactive area in DTx-treated mice (Fig.
4A, B). Further studies characterizing the responses from astrocytes and Müller glia could provide further insight on the effects of microglia depletion to these glial cells that are important in maintaining retinal integrity.
Uncontrolled hyperglycemia diminishes BRB integrity causing neuronal cell apoptosis and vascular damage leading to the activation of the resident professional phagocytes, microglia. Many reports have shown that diabetes results in proinflammatory, phagocytic microglial activation, characterized by an ameboid shape with truncated, retracted cellular processes and microglia-derived IL-1β and NOS2 production [
12,
43,
44]. Intriguingly, DTx treatment resulted in ramified microglia with long cellular processes in diabetic mice, closely resembling non-diabetic controls (Fig.
3E, F). Additionally, diabetic PLX-5622-treated mice displayed a 20% increase in microglial TI, with increased cellular branching, in contrast to ameboid-shaped microglia in diabetic controls (Fig.
5). Consistent with these results, transcriptomic analysis (Fig.
6F, I) revealed a reduction in microglial activation associated gene clusters (
Ly86,
Cx3cr1,
Siglech,
Clec7a,
Trem2,
Fcgr3,
Csf1r,
Fcrg1) indicative that these changes in microglia morphology correlate with transcriptional reprograming. Additionally, it was shown that fibrinogen depletion using the defibrinogenating agent Ancrod, ameliorated reactive microgliosis and release of proinflammatory cytokines in the diabetic retina [
45]. Complementary to these findings, depleting microglia reduced fibrinogen deposition in the diabetic retina (Figs.
4 and
5). Although our data suggest that these depletion regimens reprogram microglia to retain a homeostatic response to the diabetic environment, the length of time it takes for these repopulating microglia to obtain a proinflammatory profile is worth exploration to characterize the duration of these protective effects.
Due to lack of glucose uptake, metabolic dysregulation occurs in peripheral cells leading to the production of DAMPs and proinflammatory serum proteins [
46]. This chronic, low-grade inflammation present in the periphery contributes to the retinal manifestations of disease and contributes to the loop of inflammation present in the retina. In the CD11b-HSVTK and clodronate liposome delivery models of microglia depletion, BBB/BRB damage facilitates peripheral immune cells infiltration to the CNS [
47,
48]. Additionally, the use of CSF-1R inhibitors does not only target microglia, but has also been shown to deplete CCR2
+ monocyte progenitors, bone marrow derived macrophages, hematopoietic progenitor and stem cells in the bone marrow, spleen and blood compartments [
49]. Utilizing the CX3CR1
CreER:R26
iDTR model of microglia depletion, we were able to successfully deplete microglia without affecting peripheral immune cell proportions in the blood (Fig.
2 and Additional file
1: Fig. S3) nor did this model of depletion alter peripheral cytokine responses (Additional file
1: Fig. S3).
Overall, this study highlights differences in the degree of microglia depletion between two commonly used models, defines depletion differences among CNS compartments and also provides a list of potential candidate gene targets to mediate microglial inflammation. This study shows that microglia depletion and replenishment ameliorates microglia-associated retinal inflammation and hallmarks of DR pathogenesis to include neuronal cell loss, pathogenic angiogenesis, vascular damage and inflammation. Microglia reprogramming was validated by morphological and transcriptomic analysis and these results support the idea that transient microglia depletion and repopulation triggers neuro- and vasculo-protective effects in the diabetic retina.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.