1. Immunology and Inflammation
  2. Neuroscience
Download icon

Absence of TGFβ signaling in retinal microglia induces retinal degeneration and exacerbates choroidal neovascularization

  1. Wenxin Ma
  2. Sean M Silverman
  3. Lian Zhao
  4. Rafael Villasmil
  5. Maria M Campos
  6. Juan Amaral
  7. Wai T Wong  Is a corresponding author
  1. National Eye Institute, National Institutes of Health, United States
Research Article
  • Cited 4
  • Views 1,652
  • Annotations
Cite this article as: eLife 2019;8:e42049 doi: 10.7554/eLife.42049

Abstract

Constitutive TGFβ signaling is important in maintaining retinal neurons and blood vessels and is a factor contributing to the risk for age-related macular degeneration (AMD), a retinal disease involving neurodegeneration and microglial activation. How TGFβ signaling to microglia influences pathological retinal neuroinflammation is unclear. We discovered that ablation of the TGFβ receptor, TGFBR2, in retinal microglia of adult mice induced abnormal microglial numbers, distribution, morphology, and activation status, and promoted a pathological microglial gene expression profile. TGFBR2-deficient retinal microglia induced secondary gliotic changes in Müller cells, neuronal apoptosis, and decreased light-evoked retinal function reflecting abnormal synaptic transmission. While retinal vasculature was unaffected, TGFBR2-deficient microglia demonstrated exaggerated responses to laser-induced injury that was associated with increased choroidal neovascularization, a hallmark of advanced exudative AMD. These findings demonstrate that deficiencies in TGFβ-mediated microglial regulation can drive neuroinflammatory contributions to AMD-related neurodegeneration and neovascularization, highlighting TGFβ signaling as a potential therapeutic target.

https://doi.org/10.7554/eLife.42049.001

Introduction

The development of neuroinflammatory changes in the retina is a significant factor in the pathogenesis of multiple retinal disorders including glaucoma (Williams et al., 2017), diabetic retinopathy (Xu and Chen, 2017), and age-related macular degeneration (Guillonneau et al., 2017). Abnormal immune responses arising from physiological changes in microglia, the primary resident innate immune cell in the retina, are thought to drive aspects of disease progression, including neuronal degeneration and pathological neovascularization (Karlstetter et al., 2015; Silverman and Wong, 2018). Under healthy conditions, microglia in the retina integrate a variety of constitutive regulatory signals from other neighboring cells (Fontainhas et al., 2011; Liang et al., 2009), enabling them to perform homeostatic roles in maintaining retinal structure and function (Wang et al., 2016). How retinal microglia transition from a homeostatic physiological state to ones that promote disease progression is however not well understood. Elucidation of molecular mechanisms governing these transitions is likely central to designing strategies for microglial modulation in retinal disease (Arroba and Valverde, 2017; Bell et al., 2018).

TGFβ signaling is a significant influence on the regulation of microglial development and mature function in the brain in vivo (Butovsky et al., 2014; Buttgereit et al., 2016) and promoting microglial survival and specification in vitro (Bohlen et al., 2017). Altered TGFβ signaling in microglia has been linked to pathogenic mechanisms of neurodegenerative disorders in the brain and spinal cord (Lund et al., 2018a; Taylor et al., 2017). In the retina, TGFβ signaling exerts pleotropic effects on multiple retinal cell types that underlie numerous functions ranging from maintaining retinal neuronal differentiation and survival (Braunger et al., 2013; Walshe et al., 2011) to regulating the development and structural integrity of retinal vessels (Braunger et al., 2015; Walshe et al., 2009). However, the specific role of TGFβ signaling to retinal microglia in the regulation of homeostatic vs. pathologic states, and how this may contribute to retinal disease pathogenesis, are not known. Significantly, TGFβ signaling has been implicated in the pathobiology of age-related macular degeneration (AMD), the leading cause of vision loss in older patients in the developed world (Jager et al., 2008) and a condition still lacking comprehension prevention and treatment. Alterations in the levels of TGFβ ligands have been reported in eyes of AMD patients (Tosi et al., 2017; Tosi et al., 2018). Genome-wide association studies have discovered that polymorphisms in TGFBR1, a receptor transducing TGFβ signals in conjunction with TGFBR2, influence the risk for developing AMD (Fan et al., 2017; Fritsche et al., 2013). HTRA1, another significant AMD risk-associated protein, has been thought to confer increased AMD risk by differentially binding to and cleaving intraocular TGFβ−1, altering TGFβ signaling to microglia (Friedrich et al., 2015). These findings have prompted the consideration of TGFβ signaling as a potential target for AMD therapy (Fisichella et al., 2016; Platania et al., 2017). However, how direct TGFβ signaling regulates microglial physiology in the retina to influence inflammatory, neurodegenerative, and neovascular processes in AMD is not elucidated.

Here, we investigate the role of direct and constitutive TGFβ signaling to microglia by inducing microglia-specific ablation of TGFBR2, a receptor required for TGFβ signal transduction, in the adult mouse retina. We found that inhibition of TGFβ signaling in microglia induced abnormalities in microglial homeostasis in the retina, altering overall microglial number, distribution, and morphology. These changes resulted in reduced physical coverage of the retinal plexiform layers by microglial processes, and likely diminished microglial trophic support. TGFβ signaling ablation resulted in a downregulation of microglial ‘sensome’ genes and an upregulation of microglial activation markers. These microglial changes were highly consequential to the maintenance of a healthy retina, inducing widespread Müller cell gliosis and structural and functional degeneration of retinal neurons. Retinal microglia deficient in TGFβ signaling also demonstrated abnormal injury responses that promoted increased choroidal neovascularization in a laser-induced model of injury. Taken together, our findings indicate that constitutive neuron-microglia interactions in the form of TGFβ signaling are necessary in the maintenance of the orderly organization and trophic function of microglia in the retina; in its absence, microglia undergo pathologic transformation in ways that promote retinal changes resembling those observed in AMD pathology. These results provide insight into how abnormal TGFβ signaling in retinal microglia can contribute causally to AMD pathobiology, and raise the possibility that microglia may be modulated via TGFβ signaling as a potential therapeutic strategy.

Results

Constitutive TGFBR2 expression in microglia of the adult mouse retina is specifically ablated in Cx3cr1CreER/+, Tgfbr2flox/flox (TG) mice

We characterized TGFBR2 expression by performing immunohistochemical analysis in flat-mounted retina from two-month old adult Cx3cr1+/GFP mice. CX3CR1-expressing microglia in both the inner and outer plexiform layers (IPL, OPL) demonstrated immunopositivity for TGFBR2 in the cell membranes of somata and ramified processes (Figure 1A), as did CD31+ retinal endothelial cells. These findings were in agreement with RNAseq mRNA expression profiles in specific cells types; in the adult mouse retina (Siegert et al., 2012), Tgfbr2 mRNA expression was high in microglia but low in other retinal neurons (Figure 1B), while in the postnatal mouse brain (Bennett et al., 2016), Tgfbr2 mRNA expression was high in both microglia and endothelial cells, with little or no expression in astrocytes, neurons, and oligodendrocytes (Figure 1C).

TGFBR2 is constitutively expressed in adult mouse retinal microglia and is specifically ablated in retinal microglia of adult Cx3Cr1CreER/+,Tgfbr2flox/flox (TG) mice upon tamoxifen induction.

(A) Immunohistochemical labeling for TGFBR2 (red) in the adult CX3CR1+/GFP mouse retina was localized prominently to CD31-immunopositive vascular endothelial cells (white) and to CX3CR1-expressing, GFP+ microglia cells (green) in both the inner and outer plexiform layers (IPL, OPL). Insets (yellow boxes) show microglia demonstrating colocalization of microglial marker CX3CR1 with TGFBR2. Scale bar = 100 µm. (B) Reference to an atlas of specific cell type transcriptomes from the adult mouse retina highlighted constitutive expression of Tgfbr2 mRNA in retinal microglia, with very low or no expression in different classes of retinal neurons (expression levels >20 correspond to significant expression). (C) Reference to an atlas of specific cell type transcriptomes from the neonatal (P7-17) mouse brain indicated significant levels of constitutive expression in microglial and endothelial cell populations, with considerably lower expression in other brain neuronal and glial cell types. (D) Specific ablation of Tgfbr2 expression from retinal microglia of Cx3Cr1 CreER /+,TGFbR2flox/flox (TG) mice was enabled by tamoxifen (TMX)-induced Cre recombinase activity in CX3CR1-expression microglia, resulting in the genetic excision of exon 4 of the Tgfbr2 gene. CD11b+ microglia (blue points) and CD31+ endothelial cells (green points) were sorted from retinas of untreated TG mice, and from control and TG mice 3 weeks after tamoxifen administration using flow cytometry and analyzed. Cre recombinase-mediated excision of exon 4 of the Tgfbr2 gene from the genomic DNA in microglial, endothelial, and the remaining retinal cell types (purple points) was assessed using qPCR; CD11b+ microglia demonstrated a significant loss of exon 4 relative to the exon 3 in TMX-treated TG animals, but not in untreated TG or TMX-treated control animals. Exon 4 excision was not observed in non-microglial cell types. (E) Quantitative rtPCR analysis demonstrated a corresponding reduction in the transcription of exon 4 of Tgfbr2 mRNA from flow-sorted CD11b+ retinal microglia of TMX-treated TG animals relative to those of TMX-treated control animals and TG animals not treated with TMX. (Graphical data in (D) an (E) are presented as means ± SEM; p values are from one-way analysis of variance (ANOVA) and Sidak’s multiple comparison test, n = 4 animals of mixed sex for each group).

https://doi.org/10.7554/eLife.42049.002

To evaluate the functional significance of constitutive in vivo TGFβ signaling in retinal microglia, we employed a transgenic mouse in which TGFBR2 can be specifically and inducibly ablated in CX3CR1-expressing microglia. We employed the Cx3cr1CreER/+, Tgfbr2flox/flox (termed TG) mouse model in which tamoxifen administration activates CreERT2 recombinase activity, enabling the excision of exon 4 of the Tgfbr2 gene, ablating TGFβ signaling. Following induction, retinal CD11b+ microglia cells and CD31+ endothelial cells were isolated by flow sorting; experimental controls included age-matched TG animals that were not administered tamoxifen, and age-matched transgenic Tgfbr2flox/flox (termed Control) mice that lacked Cre recombinase and which were administered tamoxifen on the same dosing regimen as TG mice. Quantitative PCR analysis of genomic DNA for the targeted exon 4 of the Tgfbr2 gene showed a significant reduction (relative to the preserved exon 3) in CD11b+ microglia isolated from tamoxifen-administered TG animals, but not in CD11b+ microglia from both control groups (Figure 1D). No changes in the relative presence of exon 4 were detected in CD31+ endothelial cells or in the remaining (non-CD11b+, non-CD31+) retinal cell populations for all three experimental groups. Quantitative rtPCR analysis of mRNA isolated from flow-sorted CD11b+ retinal microglia correspondingly demonstrated a marked reduction of exon 4-containing transcripts in tamoxifen-administered TG animals but not in TG animals not administered tamoxifen, nor in tamoxifen-administered control mice (Figure 1E). These results indicate that Tgfbr2 expression can be ablated in an inducible manner in adult TG animals and specifically in microglia among retinal cell types.

Specific in vivo ablation of TGFBR2 in retinal microglia induces rapid morphological transformation and proliferation

We examined microglial morphology and distribution in the retinas of TG mice at different time points following tamoxifen-induced TGFBR2 ablation (1, 2, 5 days; 3 and 10 weeks) in flat-mounted samples. At one day post-tamoxifen administration, retinal microglia showed slight decreases in the length and branching of their processes but still retained a ramified morphology (Figure 2). At 2 to 5 days post-tamoxifen, microglia demonstrated marked reductions in process ramification, and possessed mostly short, stubbly processes. At 3 to 10 weeks post-tamoxifen, microglia transitioned to an elongated cellular morphology that had only a few processes that showed little branching. Interestingly, elongated microglia in TG animals were closely adherent to isolectin B4 (IB4)-labelled retinal blood vessels, with their processes conforming to the branched structure of retinal vessels (Figure 2A,E, Videos 1 and 2). These features were prominently seen at 3 weeks post-tamoxifen in both the OPL (Figure 3A) and the IPL (Figure 3—figure supplement 1A). Microglia in control animals administered tamoxifen were morphologically unchanged, resembling microglia in age-matched wild type animals not administered tamoxifen. Isolated Iba1+ microglia with elongated morphologies were also found in the subretinal space, a zone lacking retinal blood vessels (Figure 3—figure supplement 1B), indicating that the morphological transformations in microglia likely originated from cell-autonomous changes in microglia, rather than indirectly induced by signals from retinal vessels.

Specific TGFBR2 ablation in retinal microglia induces rapid and progressive changes in microglial morphology and distribution.

The time course of morphological changes in retinal microglia following tamoxifen (TMX)-induced ablation of TGFBR2 expression was followed using immunohistochemical analysis in retinal flat-mounts. Panels show changes at the level of the OPL; microglia were labeled using an antibody to IBA1 and retinal vessels labeled with IB4. Gliotic changes in radial Müller glia processes were marked using an antibody to GFAP. At 1 day following TMX administration, a slight reduction in ramification in microglia processes was observed. From 2–5 days post-TMX, a further decrease in microglial ramification and an increase in microglia numbers were detected. From 3–10 weeks post-TMX, retinal microglia transitioned to a branched morphology, demonstrating a close fasciculation with the retinal vasculature. GFAP immunopositivity in Müller glia was prominently upregulated at this time. Scale bar = 100 µm.

https://doi.org/10.7554/eLife.42049.003
Figure 3 with 2 supplements see all
Specific TGFBR2 ablation in retinal microglia induces abnormalities in microglial density, distribution, and morphology.

(A) TG animals administered tamoxifen (TMX) 3 weeks prior, relative to wild type (WT) mice and control mice, demonstrated that TGFBR2 ablation resulted in increased microglial numbers and decreased ramification in the OPL. Scale bar = 100 µm. (B) Analysis of CD11b+ microglia numbers in each animal (two retinas combined) using flow-cytometry showed a significant increase in microglial numbers in TG vs. control animals at 3- and 10 weeks post-TMX. Manual counts of Iba1 +microglia numbers (C) and proliferating Ki67+, Iba1 +microglia (D) in retinal flat-mounts from animals 4 weeks post-TMX demonstrated increases at the levels of the IPL, OPL, and subretinal space (SRS) in TG vs. control retina. (E) TGFBR2-ablated microglia 4 weeks post-TMX showed reduced process ramification and decreased dendritic area (as highlighted in outlines of individual microglial dendritic arbors). TGFBR2-ablated microglia demonstrated branched morphologies (example shown in yellow box, expanded in inset) that showed close adherent contact with IB4-labelled (red) retinal vessels (arrows indicating points of contact). Scale bar = 100 µm. Morphological analysis of individual microglia showed significant decreases in the number of branch points (F) and in the areas of individual arbors (G) of microglia in both IPL and OPL in TGFBR2-ablated microglia. Despite having increased numbers of total microglia, TMX-treated TG retinas have a greater proportion of retinal area not directly occupied by microglial processes (H, areas highlighted in blue), indicating decreased microglial coverage. Graphical data are presented as means ± SEM; p values are from unpaired t-test with Welch’s correction, data points in (C), (D), and (H) represent four individual imaging fields from three animals in each group, those in (F) and (G) represent 16 individual microglia cells from four animals in each group).

https://doi.org/10.7554/eLife.42049.004
Video 1
3D rotation depiction of the morphology and distribution of IBA1-immunolabelled retinal microglia (green) in the OPL with respect to IB4-labeled retinal vessels (white) in TG animals prior to the administration of tamoxifen.
https://doi.org/10.7554/eLife.42049.007
Video 2
3D rotation depiction of the morphology and distribution of IBA1-immunolabelled retinal microglia (green) in the OPL with respect to IB4-labeled retinal vessels (white) in TG animals 2 weeks following the administration of tamoxifen to induce TGFBR2 ablation in retinal microglia.
https://doi.org/10.7554/eLife.42049.008

Morphological transformations in retinal microglia were accompanied by general increases in overall microglial density. Quantification of CD11b+ retinal microglia using flow cytometry showed significant increases in TG mice at 3 weeks and 10 weeks following tamoxifen administration compared with tamoxifen-administered control mice (Figure 3B). Microglia densities, as assessed by cell counting in flat-mounted retinal specimens using immunochemical analyses, also demonstrated increases in all retinal laminae (IPL, OPL, and SRS) (Figure 3C), which corresponded to the emergence of proliferating, Ki67+, microglia. (Figure 3D). Ki67 immunopositivity was absent in subretinal microglia, suggesting that increased subretinal microglia numbers may have resulted from migration of microglia from the inner retinal layers. Monocytic infiltration into the retina was unlikely to have contributed to the increased IBA+ cell numbers as cell-fate mapping of retinal microglia vs. systemic monocytes using TG mice crossed into the Ai14 background (Ma et al., 2017) revealed that IBA1+ cells 4 months following tamoxifen uniformly expressed tdTomato, indicating that systemic monocytes (which at this time had been turned over and replaced by tdTomato-negative cells) had not contributed to the increased numbers of IBA1+ cells induced by TGFBR2 ablation (Figure 3—figure supplement 1C). In addition, we observed that the myeloid cells in the retina demonstrating progressive morphological change in the first week following tamoxifen administration were immunopositive for P2RY12, a marker for endogenous microglia, as well as for Ki67, a marker of proliferating cells (Figure 3—figure supplement 1D). Although P2RY12 immunopositivity was gradually lost after one week following TGFBR2 ablation, these findings indicated that the population of morphologically-transformed myeloid cells in the retina arose from the proliferation and modification of pre-existing endogenous retinal microglia.

As the constitutive presence of microglia in the adult retina is required for ongoing maintenance of retinal synapses (Wang et al., 2016), and may be mediated by repeated microglia-synapse contacts via dynamically motile microglial processes (Lee et al., 2008), we examined how areal coverage in the synapse-rich plexiform layers by microglial processes may be altered following TGFBR2 ablation. We found that following TGFBR2 ablation, individual microglial cells in both the IPL and OPL of TG animals demonstrated marked reductions in the number of branch points per cell (Figure 3E,F) and in the area subtended by the processes of each cell (Figure 3G). Consequently, despite increased microglial density, the proportion of retina lacking direct coverage by microglial processes was significantly greater (Figure 3H), translating to decreased microglia-synapse contact and likely diminished microglial supportive functions.

Specific ablation of TGFBR2 in retinal microglia results in decreased microglial ‘sensome’ function and increased activation

Corresponding to the decreased spatial coverage of the retina by TGFBR2-ablated microglia, we investigated if the ability of retinal microglia to sense environmental signals may be affected by the loss of TGFβ signaling. Previous transcriptomic profiling studies of microglia in the mouse brain have defined a cluster of microglial specific/enriched transcripts encoding proteins that confer the ability to sense environmental signals, collectively referred to as the microglial ‘sensome’ (Hickman et al., 2013). Quantitative RT-PCR analysis of flow-sorted microglia from the TG retinas 2 weeks following tamoxifen administration revealed that mRNA expression levels of ‘sensome’ transcripts such as Cx3cr1, P2yr12, Tmem119, and Siglech, were markedly reduced relative to control mice (Figure 4A). These changes are likely to be a direct consequence of the loss of TGFβ signaling in retinal microglia as in vitro administration of TGFβ ligands (TGFB1, TGFB2) to cultured retinal microglia isolated from wild type mice resulted in upregulation of sensome transcripts, Tmem119 and Siglech (Figure 4B). Downregulation of sensome gene expression with TGFBR2-ablation was also apparent on a protein level; fluorescence associated with EYFP expression as driven by the Cx3cr1 promoter in TG mice, a surrogate marker for the level of Cx3cr1 expression, was significantly reduced by TGFBR2 ablation (Figure 4C,D), which was also associated with decreased TMEM119 immunopositivity in Iba1+ microglia (Figure 4E,F). Microglial responses to endogenous signals include the provision of trophic support to nearby neurons in the form of growth factors, such as BDNF (Parkhurst et al., 2013) and IGF1 (Lalancette-Hébert et al., 2007). We found that mRNA expression of growth factors, Bdnf and Pdgfa, were decreased in retinal microglia following TGFBR2 ablation, while that for Igf1 was unchanged (Figure 4—figure supplement 1). Accordingly, the addition of TGFβ ligands also increased the expression of these growth factors in isolated retinal microglia in culture. Together, these observations indicated that TGFβ-signaling to microglia sustains the microglial homeostatic gene signature and promotes the ability of microglia to sense endogenous signals and exert trophic influences in the retina.

Figure 4 with 1 supplement see all
Constitutive expression of microglial ‘sensome’ genes are downregulated upon TGFBR2 ablation in retinal microglia.

(A) Retinal microglia from control and TG mice were isolated by flow-cytometry 2 weeks following tamoxifen (TMX) administration and mRNA levels of microglial ‘sensome’ genes compared using qPCR. mRNA levels of Cx3cr1, P2yr12, Tmem119, and Siglech were all significantly decreased in microglia from TG vs. control mice. (B) Microglia from the retinas of WT mice were cultured and exposed to media containing TGFB1 (10 or 20 ng/ml), or TGFB2 (10 ng/ml) (media containing 10 ng/ml of BSA served as a control), and mRNA levels of microglial ‘sensome’ genes compared following 24 hr of exposure. mRNA levels of Tmem119 and Siglech were increased by TGFBR2 ligands (TGFB1 or TGFB2), indicating positive regulation of microglial ‘sensome’ genes via TGFBR2-mediated signaling. (C, D) As TG animals contained an IRES-EYFP cassette 3’ to CreERT recombinase in the Cx3cr1 locus, EYFP expression, as regulated by the Cx3cr1 promoter, could be constitutively detected in IBA1-immunopositive retinal microglia in control animals. In TG animals at 3 weeks post-TMX, Cx3cr1-driven EYFP fluorescence was diminished in Iba1+ microglia, indicating downregulation of Cx3cr1 promoter activity. (E, G) Immunohistochemical analysis of TMEM119 showed strong colocalization with Iba1 in microglia of control animals but decreased immunopositivity in TGFBR2-ablated microglia in TG animals. Scale bars = 100 µm. Graphical data in (A), (B), (D) and (F) are presented as means ± SEM; p values in (A), (D), and (F) are from multiple t-tests, while that in (B) are from 2-way ANOVA analysis with Sidak’s multiple comparisons test, * indicate p<0.05 for comparisons relative to control, data points indicate individual biological repeats in (A) and (B), and four imaging fields from three animals in each group in (D) and (F).

https://doi.org/10.7554/eLife.42049.009

As TGFβ signaling has been associated with the induction of a quiescent microglial phenotype in the brain (Abutbul et al., 2012), we evaluated if genetic ablation in microglia within the retina influenced their activation status. We found that TGFBR2 ablation in microglia upregulated mRNA expression of activation markers (MHCII (H2-Aa), Cd68, Cd74), chemotactic cytokines (Ccl2 and Ccl8), and Apoe, a promoter of a proinflammatory, disease-associated microglial phenotype (Kang et al., 2018; Krasemann et al., 2017) (Figure 5A). Cultured WT retinal microglia demonstrated corresponding decreases in ApoE and Ccl2 mRNA levels when TGFβ ligands were added in vitro (Figure 5B). Immunohistochemical analysis of TG mice following tamoxifen administration showed increased immunopositivity for markers of microglial activation, including CD68, MHCII (Figure 5C–F), CD74, F4/80, and CD45 (Figure 5—figure supplement 1A–F) relative to control mice. RT-PCR analysis of mRNA expression in the retina following microglial TGFBR2 ablation also found progressively increasing expression of transcripts found to be enriched in macrophages over that in homeostatic microglia (Saa3, Pf4, Cd5l) (Hickman et al., 2013) (Figure 5—figure supplement 1G). These data indicated that constitutive direct TGFβ signaling is required for the general suppression of microglial activation.

Figure 5 with 1 supplement see all
Expression of genes associated with microglial activation are upregulated on TGFBR2 ablation in retinal microglia.

(A) Retinal microglia from control and TG mice were isolated by flow-cytometry 2 weeks following tamoxifen (TMX) administration and mRNA levels of genes associated with microglial activation and inflammatory chemokines were analyzed and compared using qPCR. mRNA levels for H2-Aa (MHCII), Cd68, Cd74, Apoe, Ccl2, and CCl8 were all significantly increased in microglia from TG vs. control mice. (B) Microglia from the retinas of WT mice were cultured and exposed to media containing TGFB1 (10 or 20 ng/ml), or TGFB2 (10 ng/ml) (media containing 10 ng/ml of BSA served as a control), and mRNA levels of microglial-expressed genes compared following 24 hr of exposure. mRNA levels of Apoe and Ccl2 were decreased by TGFBR2 ligands (TGFB1 or TGFB2), indicating negative regulation of microglial activation genes via TGFBR2-mediated signaling. Immunohistochemical analysis of control vs. TG microglia in retinal flat-mounts showed prominent and significant upregulation of activation markers CD68 (C, D) and MHCII (E, F) in Iba1+ microglia in both the IPL and OPL. Scale bars = 100 µm. (Graphical data in (A), (B), (D) and (F) are presented as means ± SEM; p values in (A), (D), and (F) are from multiple t-tests, while that in (B) are from 2-way ANOVA analysis with Sidak’s multiple comparisons test, * indicate p<0.05 for comparisons to control, data points indicate individual biological repeats in (A) and (B), and four imaging fields from 3 to 4 animals in each group in (D) and (F)).

https://doi.org/10.7554/eLife.42049.011

TGFBR2-deficient microglia induce secondary Müller cell gliosis and neuronal degeneration in the surrounding retina

We examined the consequences of microglia-specific TGFBR2 ablation to the structure and function of the surrounding retina. Following tamoxifen administration in TG mice, we observed an emergence of a radial pattern of GFAP immunopositivity beginning at 5 days which persisted at 10 weeks (Figure 2) that colocalized with glutamine synthetase (GS)-immunopositive Müller cells processes (Figure 6A). GFAP mRNA levels in the retina were also increased at 2 weeks following TGFBR2 ablation and persistent at 8 weeks (Figure 6B). Also, a progressive upregulation of mRNA levels for genes associated with neurotoxic A1 astrocytic gliosis (Liddelow et al., 2017) was induced, while those for genes associated with the neuroprotective form of A2 gliosis were relatively unchanged (Figure 6C). These observations indicate that microglial transformation induced by TGFBR2 ablation led to a rapid and durable induction of gliotic changes in surrounding Müller cells that resemble reactive A1 astrocytic gliosis characterized in the brain under conditions of neurodegeneration and aging (Clarke et al., 2018; Liddelow et al., 2017).

TGFBR2 ablation in retinal microglia induces Müller cell gliosis in the retina.

(A) Immunohistochemical analysis demonstrates upregulation of immunopositivity to GFAP 3 weeks post-TMX in TG animals relative to control animals. GFAP immunopositivity was localized to glutamine synthetase (GS)-labeled Müller cell processes, indicating the induction of Müller cell gliosis. Scale bar = 50 µm. (B) qPCR analysis of retinas isolated from control and TG animals 2 and 8 weeks post-TMX demonstrates a significant upregulation of GFAP mRNA expression following TGFBR2 ablation in retinal microglia. Graphical data are presented as means ± SEM; p values are from one-way analysis of variance (ANOVA) and Sidak’s multiple comparison test, n = 3 animals of mixed sex in each group.(C) RT-PCR analysis of retinal expression of genes associated with A1- and A2-specific astrocytic gliosis following microglial TGFBR2 ablation found progressive upregulation of A1-associated transcripts relative to control, while A2-associated transcripts were relatively unchanged (numbers indicate means, *, **, *** indicate p values < 0.05,<0.01,<0.001 respectively, 2-way ANOVA analysis with Sidak’s multiple comparisons test, data from 3 to 4 animals in each group.).

https://doi.org/10.7554/eLife.42049.013

As chronic proinflammatory microglial activation in the retina has been associated with neuronal degeneration (Langmann, 2007), we investigated if microglial alterations following TGFBR2 ablation resulted in deleterious changes in retinal neurons. Using in vivo optical coherence tomography (OCT) imaging we found total retinal thickness in TG mice decreased progressively with time following tamoxifen administration, falling to 95% of controls at 3 weeks and to 80% at 10 weeks (Figure 7A,B). These changes were contributed to by decreases in retinal thickness in the inner, as well as in the outer retina. Quantitative assessment of the thickness of retinal laminae from histological retinal sections also showed significant decreases in inner and outer nuclear layers, as well as the inner and outer plexiform layers (Figure 7C,D). Analysis in flat-mounted retinal samples demonstrated significant decreases in the density of BRN3A-immunopositive retinal ganglion cells and cone arrestin-positive cone photoreceptors (Figure 7—figure supplement 1). These changes were correlated with the appearance of apoptotic TUNEL +nuclei in all retinal nuclear layers, indicating an induction of neuronal apoptosis (Figure 7E,F). Assessment of retinal function using electroretinography (ERG) revealed that the amplitudes of dark-adapted, rod photoreceptor-dominant responses were significantly reduced in TG vs. control animals, with b-wave amplitudes significantly more reduced than a-wave amplitudes (Figure 7G). For light-adapted, cone photoreceptor-mediated responses, only b-wave amplitudes were significantly reduced, while a-wave amplitudes were unchanged (Figure 7H). Overall, significant decreases in the b-to-a amplitude ratios were observed for both dark- and light-adapted responses, indicating that ablation of TGFBR2 in retinal microglia resulted in some measure of rod photoreceptor dysfunction and also a loss of synaptic transmission in both rod and cone photoreceptors, as previously described for the ablation of microglia in the adult retina (Wang et al., 2016). Taken together, the physiological switch of retinal microglia from a homeostatic mode to a more activated mode upon the loss of microglia TGFβ signaling, is likely causally associated with a loss of microglial support, a dysregulation of inflammatory responses resulting in gliotic changes, and the induction of neuronal and synaptic degeneration.

Figure 7 with 1 supplement see all
TGFBR2 ablation in retinal microglia induces degenerative changes in the retina.

(A, B) In vivo evaluation of retinal structure by optical coherence tomography (OCT) in control animals and in TG animals 3 and 10 weeks following tamoxifen (TMX)-administration showed a preserved lamination in TG animals (insets at higher magnification in yellow boxes) but a progressive and significant reduction in the total retinal thickness relative to controls. Scale bar = 300 µm. Significant reductions in overall thickness were contributed to by reductions in both the inner (measured from vitreal surface to the outer plexiform layer) and the outer retinal layers (measured from the outer plexiform layer to the apical surface of the RPE layer) (p values are from 1-way ANOVA analysis with Tukey’s multiple comparisons test, data points are from 6 eyes of 3 animals). (C, D) Histological analysis of retinal lamina thicknesses in paraffin-embedded sections show significant decreases in the thickness of the inner plexiform layer (IPL), inner nuclear layer (ONL), outer plexiform layer (OPL), and outer nuclear layer (ONL) in TG animals 3 weeks post-TMX relative to controls (p values are from unpaired t-tests with Welch’s correction, data points are from 3 sections from four animals). Scale bar = 50 µm. (E, F) Evaluation for apoptotic retinal cells using TUNEL labeling demonstrated the emergence of apoptotic cells in both the INL and ONL in TG retinas 10 weeks post-TMX. (p values are from unpaired t-tests with Welch’s correction, data points are from 3 sections from four animals). Scale bar = 50 µm. (G, H) Comparison of electroretinographic (ERG) responses between control vs. TG animals 10 weeks post-TMX demonstrated in dark-adapted responses (G) a small but significant decrease in a-wave amplitude and a marked decrease in b-wave amplitudes in TG animals. Light-adapted responses (H) were similar for a-wave amplitude but significantly decreased in b-wave amplitude. The b-to-a amplitude ratios were significantly decreased in TG animals in both dark- and light-adapted responses for a range of flash intensities (p values are from 2-way ANOVA analysis, data points are both eyes of 8 control and 8 TG animals).

https://doi.org/10.7554/eLife.42049.014

Molecular pathways underlying retinal changes induced by microglial TGFBR2 ablation

To further investigate the nature of molecular pathways underlying retinal changes following microglial TGFBR2 ablation, we profiled the mRNA expression levels of 547 immunology-associated genes using targeted multiplex analysis (nCounter, Nanostring). We compared expression in retinas isolated from 4 groups of animals: two groups of control animals with and without tamoxifen administration, and two groups of TG animals administered tamoxifen for either 2 or 8 weeks. Hierarchical clustering revealed similar mRNA profiles between control animals with or without tamoxifen administration, indicating that tamoxifen per se did not exert a major effect on inflammatory retinal gene expression (Figure 8A), while tamoxifen-administered TG animals showed significant differences from control animals. Significantly upregulated genes (>2 fold increase in expression, p<0.05) following either 2 or 8 weeks of tamoxifen administration in TG animals included: (1) markers of microglial activation, such as Cd74, H2-Aa (MHCII), Cd163, Cd40, (2) proinflammatory cytokines, such as Ccl2, Ccl4, Ccl12, and Ccl8, (3) complement components, such as C4a and C3, (4) regulator of inflammatory responses, such as FcγR receptors and Casp1 (Figure 8B). Early downregulation of the microglial sensome gene Cx3cr1 was also detected at 2 weeks. Gene ontology (GO) analysis of differentially expressed genes revealed the differential involvement of pathways in neuroinflammatory signaling and nuclear factor of activated T-cells (NFAT)-mediated signaling, which has been implicated in the regulation of microglial activation (Nagamoto-Combs and Combs, 2010) (Figure 8C). Network analysis indicated that the differentially expressed genes associated with microglial TGFBR2 ablation related to various aspects of cell function including (1) homeostasis of leukocytes, (2) inflammatory response, (3) cytotoxicity, and (4) activation, which may be potentially regulated by interferon-α, interferon-γ, IL1β signaling, and NFΚB-regulated transcription (Figure 8—figure supplement 1).

Figure 8 with 1 supplement see all
Changes in the mRNA expression of immune regulated genes in the retina following microglial TGFBR2 ablation using Nanostring-based profiling.

Four groups of animals (n = 3 animals per group) were analyzed: (1) Control animals not administered tamoxifen, (2) Control animals administered tamoxifen, (3) TG animals 2 weeks after tamoxifen administration, (4) TG animals 8 weeks after tamoxifen administration. (A) Hierarchical clustering of differentially expressed genes showed separate clustering of control and TG animals administered tamoxifen. (B) Volcano plots showing genes that were differentially expressed between control and TG animals administered tamoxifen at 2 and 8 weeks respectively. (C) Gene ontogeny (GO) analysis using IPA demonstrated a number of canonical pathways that were differentially represented between control and TG animals administered tamoxifen at 2 weeks reflecting the activation of neuroinflammatory pathways and pathways involved in immune cell activation and maturation.

https://doi.org/10.7554/eLife.42049.016

TGBR2 ablation in microglia does not affect the normal structure of retinal blood vessels but promotes pathological choroidal neovascularization

As TGFBR2 ablation in the entire retina at an early age (Braunger et al., 2015) or specifically in retinal endothelial cells (Allinson et al., 2012; Schlecht et al., 2017) resulted in abnormal development and structure of retinal blood vessels, we investigated whether TGFBR2 ablation specifically in adult retinal microglia may result in similar effects. We found that 12 weeks following tamoxifen administration in TG mice, despite increases in the expression of inflammatory genes in the retina and the close adherence of deramified microglia to retinal vessels, the blood-retina barrier on fluorescein angiography remained relatively intact, with no obvious vascular leakage or changes in overall retinal perfusion (Figure 9A). In the absence of additional injury, microglial TGFBR2 ablation did not result in histological abnormalities in retinal vascular structure; the arrangement of retinal endothelial cells (marked by CD31 labeling and IB4 staining) and pericytes (marked by NG2 labeling) were similar to that in controls (Figure 9B), and lacked signs of pericyte and endothelial cell loss with spontaneous neovascularization previously described for TGFBR2 ablation in retinal endothelial cells (Allinson et al., 2012; Schlecht et al., 2017). However, when we induced choroidal neovascularization (CNV) using a laser injury model (Campos et al., 2006) and compared the neovascular pathology in tamoxifen-administered control and TG mice, we observed in TG mice an increased number of Iba1+ microglia/macrophages recruited to the site of laser injury, which was associated with a larger RPE layer defect and an increased size of the CNV complex (Figure 9C-F) . This indicated that TGFBR2 loss in microglia, while not inducing vascular change on its own, increased microglia recruitment to retinal injury and promoted CNV growth at the site of microglial aggregation. This data indicates that while microglial TGFBR2 expression was dispensable for the maintenance of normal retinal vasculature, microglia lacking TGFβ signaling can transition to phenotypes that can potentiate pathological neovascularization in the presence of inducing factors.

TGFBR2 ablation in retinal microglia increases pathological choroidal neovascularization (CNV) in an in vivo laser injury model.

(A) In vivo evaluation for abnormalities in retinal vascular permeability using fluorescein angiography was performed in control and TG mice 12 weeks following tamoxifen (TMX) administration beginning at the age of 2 months. No abnormal leakage or vascular structure were detected. (B) Immunohistochemical analysis of endothelial cells (labeled with IB4 and an antibody to CD31) and retinal pericytes (labeled with an antibody to NG2) in retinal vasculature showed normal morphologies and distributions following TGFBR2 ablation in TG mice 12 weeks post-TMX. Scale bar = 100 µm. (C) Control and TG mice 3 weeks post-TMX were subjected to in vivo laser injury in a model of CNV formation. CNV complexes were analyzed in RPE flat-mounts using immunohistochemistry 7 days after laser injury and compared. Scale bar = 200 µm. TGFBR2-ablated TG animals demonstrated a higher recruitment of Iba1+ myeloid cells to the laser injury site (D), which was correlated with a larger laser lesion size (as labeled with phalloidin) (E) and a larger CNV area (F). (p values are from unpaired t-tests with Welch’s correction, data points are from 40 lesions from six animals in each group).

https://doi.org/10.7554/eLife.42049.018

Discussion

TGFβ signaling exerts pleiotropic effects in various tissues that mediate a broad range of regulatory influences on cell survival and inflammation (Fabregat et al., 2014; Travis and Sheppard, 2014). In the retina, TGFβ signaling is constitutively operational under healthy conditions, regulating the maintenance of normal retinal structure and function. TGFβ ligands, TGFβ1, TGFβ2, and TGFβ3, are expressed by multiple retinal cell types, including different classes of retinal neurons, endothelial cells, RPE cells, and retinal microglia (Anderson et al., 1995; Close et al., 2005; Lutty et al., 1993). In particular, Tgfb2 and Tgfb3 mRNA have been detected in amacrine, bipolar, and retinal ganglion cells (Siegert et al., 2012), and TGFB2 protein has been localized to photoreceptors (Lutty et al., 1991). TGFβ receptors are also broadly expressed in different retinal cell types (Obata et al., 1999); specifically, TGFBR2 is expressed in retinal microglia and endothelial cells as shown here. As evidenced by these complex expression patterns, TGFβ-mediated interactions in the retina are diverse and context-dependent; for example, constitutive TGFβ signaling to endothelial cells maintains the structural stability of the choroidal and retinal vascular circulation (Schlecht et al., 2017; Walshe et al., 2009), while that to retinal ganglion neurons promotes their differentiation and survival (Braunger et al., 2013; Walshe et al., 2011). Specific control of TGFβ signaling across these disparate contexts may be enabled by the local nature of interactions, such as through direct cell-cell contact, as well as through the agency of ‘milieu molecules’, such as LRRC33, which can influence localized and selective activation of TGFβ in specific cell types (Qin et al., 2018).

In our study here, we investigated the specific regulatory influence of constitutive TGFβ signaling on microglia in the retina. In the healthy adult retina, microglia are spatially organized in regular, non-overlapping, horizontal arrays concentrated within the synaptic plexiform layers (Santos et al., 2008). This ordered organization of ramified cells, together with the dynamic motility of microglial processes, provides for comprehensive spatial coverage and microglia-synapse contact in the IPL and OPL (Lee et al., 2008), enabling the maintenance of retinal synapses; depletion of retinal microglia results in progressive synaptic degeneration which can be rescued by microglial repopulation (Wang et al., 2016; Zhang et al., 2018). Microglia distribution in the retina also shows laminar specificity, with the outer retina being uniformly devoid of microglia under healthy conditions. This exclusion is functionally significant as the infiltration of microglia into the outer retina, which occurs in aging (Xu et al., 2008) and photoreceptor injury (Ng and Streilein, 2001), is associated with deleterious changes to photoreceptors and RPE cells (Combadière et al., 2007; Ma et al., 2009). We found in the current study that constitutive TGFβ signaling to retinal microglia is necessary for the maintenance of this overall organization, with retinal microglia demonstrating progressive disorganization in number, morphology, and distribution with TGFBR2 ablation. It is likely that TGFβ signaling can serve as a direct signal for microglial homeostasis; also, the induction of microglial ‘sensome’ gene products can enable microglia to orient themselves to environmental guidance cues. With TGFBR2 ablation, the downregulation of key receptors such as Cx3cr1 and P2ry12 can render microglia less responsive to CX3CL1, which regulates microglial activation and distribution (Carter and Dick, 2004; Combadière et al., 2007), and to ATP which promotes microglial morphological ramification and dynamic behavior (Fontainhas et al., 2011; Liang et al., 2009).

We found that this loss of microglial organization in the absence of TGFβ signaling is highly consequential to the function of the retina, such as in its electric response to light stimuli which subserves vision. The induced loss in microglial ramification, which likely resulted in decreased microglia-synapse contact, was associated with significant decrements of synaptic function in the form of abnormal b-to-a wave ratios in ERG responses. The mechanisms underlying microglia-synapse maintenance, while incompletely understood, have been related to local microglia-mediated delivery of neurotrophic factors. For example, genetic ablation of microglia-specific BDNF expression in the brain inhibited learning-related synapse formation and decreased cognitive behavior (Parkhurst et al., 2013). In support of these mechanisms, we observed that TGFBR2-ablation in retinal microglia resulted in a downregulated Bdnf expression, while exogenous TGFβ ligand stimulation conversely increased it in cultured retinal microglia. TGFBR2 ablation in microglia also resulted in abnormal microglial displacement into the subretinal space, akin to that seen with aging, suggesting that aging-related decreases in microglial responses to TGFβ (Rozovsky et al., 1998) may be a mechanism contributing to the misdistribution of microglia in the senescent retina.

We observed that the loss of TGFβ signaling in retinal microglia was also accompanied by the induction of microglial proliferation, increased mRNA expression of microglial activation markers (Cd68, Cd74), antigen presentation molecules (H2-Aa), proinflammatory cytokines (Ccl2, Ccl8), and increased immunopositivity to microglial activation markers (CD68, MHCII, F4/80, CD74, CD45). As also noted in TGFBR2-deficient brain microglia (Buttgereit et al., 2016; Lund et al., 2018a), these changes indicated a transition of resident microglia from a homeostatic to a more proinflammatory phenotype. Our experiments with cell-fate mapping and P2RY12-immunohistochemical analyses indicated that this alteration was predominantly constituted by a transition in resident microglia; we did not detect the infiltration of monocyte-derived macrophages, as was the case in comparable experiments examining the brain (Lund et al., 2018a). The reasons for why monocytic infiltration did not occur in this context are unclear and may be related to the absence of evident breakdown of the blood-retinal barrier, despite increased microglial activation, and also to the absence of an empty myeloid cell niche available to accommodate infiltrating monocytes (Lund et al., 2018b). It is possible that long-lived Cx3cr1-expressing perivascular macrophages resident within the retina may also contribute to the transformed population, but this is likely a smaller contribution, owing to their sparser numbers at baseline (Goldmann et al., 2016; Mendes-Jorge et al., 2009). We observed that retinal microglia following tamoxifen administration gradually acquired immunopositivity for CD206, a marker typically positive for perivascular macrophages. However, prominent proliferation and migration of CD206+ perivascular macrophages present at baseline prior to TGFBR2 ablation were not detected, indicating that true perivascular macrophages are unlikely to contribute substantially to the final population of transformed cells (Figure 3—figure supplement 2).

We found that this altered microglial phenotype induced by TGFBR2 ablation perturbed the regulation of neuroinflammation in the retina, which is mediated in part by microglia-Müller cell interactions (Portillo et al., 2017; Wang and Wong, 2014). We found evidence for widespread secondary gliotic changes in Müller cells that were spatiotemporally-coincident with microglial changes. Interestingly, these featured the selective upregulation of gliosis genes associated with neurotoxic reactive A1 astrocytes (Liddelow et al., 2017), indicating that a neurotoxic influence from Müller cells may be induced by TGFBR2-deficient microglia. These changes in Müller cells may additionally feedback onto nearby microglia via secreted signals to influence their activation (Wang et al., 2011; Wang et al., 2014). In addition, we found that TGFBR2 ablation resulted in a significant upregulation of Apoe in microglia, which has been linked with a reciprocal downregulation of TGFβ signaling in microglia, as well as an induction of a neurodegenerative microglial phenotype (Krasemann et al., 2017), that is also observed in profiling studies of brain microglia in models of aging and Alzheimer’s disease (Kang et al., 2018). Correspondingly, we found that retinal microglia and Müller cells changes were accompanied by progressive retinal thinning and neuronal apoptosis, demonstrating that a deficiency in TGFβ-mediated microglial regulation results in dysregulated neuroinflammation that increases the vulnerability of the retina to neurodegenerative changes.

In a study published following the submission of our manuscript, Zöller et al., (Zöller et al., 2018) using a similar transgenic model, had induced the ablation of exon 2/3 of TGFBR2 in Cx3cr1-expressing cells (Chytil et al., 2002) and described in the brain an upregulation of microglial activation markers, but had failed to detect alterations in microglial density, microglia-specific gene expression or neuronal survival. This contrasts with our findings here and those in previous reports in which TGFBR2 ablation in adult microglia resulted in increased microglial proliferation and numbers, downregulation of microglia-specific genes such as Siglech, and the onset of behavioral phenotypes of degeneration (Buttgereit et al., 2016; Lund et al., 2018a). In Zöller et al., the stability in the expression of microglia-specific genes following TGFBR2 ablation may have arisen from experiments in which mRNA profiling was performed on microglia that had been sorted as CD45low, which may have selected against TGFBR2-ablated microglia (which upregulate CD45 expression) and selected for the fraction of microglia that had not undergone gene recombination. The authors had assayed for neurodegenerative changes using only counts of NeuN+ cortical neurons, a method that is less sensitive than measures of neuronal apoptosis or assays of neuronal function for detecting neurodegenerative changes. Combined with findings in vitro demonstrating a requirement for TGF-β for the expression of a microglia-specific gene signature, and those in vivo in showing that decreased TGF-β signaling to microglia resulted in the downregulation of microglia-specific gene expression and neurodegenerative changes (Butovsky et al., 2014; Qin et al., 2018), it is likely that microglia in the retina require constitutive TGF-β signaling to maintain a microglia-specific gene signature and to prevent a transition to an activated phenotype that helps drive retinal neurodegeneration.

We found that in addition to regulating the homeostatic status of microglia, TGFβ signaling was also important in mediating injury-induced microglial responses in the retina. In a model of laser injury of choroidal neovascularization, we found that an increased number of TGFBR2-deficient microglia was recruited to the area of laser injury, which was associated with a greater area of RPE disruption and an increased size of pathological CNV. Previous studies have found an association between CNV and recruited myeloid cells which can promote neovascularization by the expression of pro-angiogenic factors and inflammatory cytokines (Crespo-Garcia et al., 2015; Li et al., 2017); alterations in the polarization state of these myeloid cells were also influential in the extent of the CNV formed (Kelly et al., 2007; Yang et al., 2016). In this context, our observations posit a potential mechanism by which TGFβ signaling to microglia may regulate the level of chronic inflammation in the aged retina in pathologically significant ways, particularly with respect to AMD. The mechanism underlying the increased risk for AMD development in patients with polymorphisms associated with Tgfbr1 (Fritsche et al., 2013) may relate to decreased levels of TGFβ signaling from the retinal environment to microglia in the AMD retinas, consequently shifting microglia from a homeostatic physiological state to a pathological state (Figure 10). This results in altered interactions between microglia and Müller cell and neurons, increasing the retina’s vulnerability to synaptic and neuronal degeneration and exacerbating the severity of CNV development in response to pathogenic triggers, phenotypes that are both hallmarks of advanced AMD. Pathogenic mechanisms related to polymorphisms in Htra1, another gene associated with AMD risk, have also been related to alterations in the ability of HTRA1 to modulate TGFβ signaling in microglia (Friedrich et al., 2015). Also, in an amyloid-β-induced rodent model for AMD, intraocular delivery of exogenous TGFβ1 resulted in decreased markers of neuronal apoptosis (Fisichella et al., 2016), prompting proposals of modulation of TGFβ signaling as a potential AMD therapeutic strategy (Platania et al., 2017). Analogously, delivery of TGFβ ligands in rodent models of multiple sclerosis (De Feo et al., 2017) and hemorrhagic stroke (Taylor et al., 2017) has been found to facilitate immunomodulation of brain microglia/macrophages to favor structural and functional neuronal recovery. As such, altered TGFβ signaling may constitute an important mechanism underlying the contribution of microglia to AMD pathobiology (Guillonneau et al., 2017).

Schematic showing the role of TGFβ signaling in the regulation of retinal microglial physiology and the consequences of altered TGFβ signaling in the retina.

TGFβ ligands, expressed constitutively by retinal neurons (TGFβ−2 and −3) and retinal microglia (TGFβ−1), signal to TGFBR2-expressing microglia to promote their homeostatic phenotype and to suppress a pathologic phenotype. Conversion between these phenotypes, which are associated with corresponding patterns of gene expression, results in a loss of microglial organization and microglial trophic functions and increased pathological neurodegeneration and neovascularization.

https://doi.org/10.7554/eLife.42049.019

Collectively, these results demonstrate that constitutive TGFβ signaling in retinal microglia is necessary in maintaining the organization of microglia in the healthy retina and is indispensable for their homeostatic function in synapse maintenance. This signaling is also necessary for the regulation of the neuroinflammatory status within the retina as insufficient levels results in aberrant microglial activation that drives chronic inflammatory changes, leading to a transformation of Muller cells to a maladaptive gliotic form, and an increased vulnerability of retinal neurons to neurodegeneration. TGFβ signaling also negatively regulates microglial responses to injury triggers without which exacerbated pathological choroidal neovascularization results. This combination of AMD-related phenotypes that involve chronic inflammation, neuronal degeneration, and pathological choroidal neovascularization, together with the genetic risk for AMD in TGFBR1 polymorphisms, implicate TGFβ regulation of retinal microglia as an influential contributing pathologic mechanism in AMD.

Materials and methods

Experimental animals

Request a detailed protocol

Experiments were conducted according to protocols approved by the National Eye Institute Animal Care and Use Committee and adhered to the Association for Research in Vision and Ophthalmology (ARVO) Statement for the use of animals in ophthalmic and vision research. Animals were housed in a National Institutes of Health animal facility under a 12 hr light/dark cycle with food ad libitum. Transgenic mice in which the Cx3cr1 gene was replaced by a sequence encoding Cre recombinase with a tamoxifen-dependent estrogen ligand-binding domain and a downstream sequence for IRES-EYFP (Cx3cr1 CreER mice, provided by Dr. Wen-Biao Gan, Skirball Institute) (Parkhurst et al., 2013) were crossed with mice possessing loxP sites that flank exon 4 of the transforming growth factor, beta receptor II (Tgfbr2) (Tgfbr2flox/flox, The Jackson Laboratory, #012603) (Levéen et al., 2002) to generate Cx3cr1CreER/+,Tgfbr2flox/flox mice (termed TG mice) which enabled the inducible deletion of exon 4 of Tgfbr2. EYFP expression, as driven by the Cx3cr1 promoter, was used as a marker for Cx3cr1 expression. Transgenic mice in which one copy of the Cx3cr1 gene was replaced by sequences coding for green fluorescent protein (GFP) (designated Cx3cr1+/GFP mice, The Jackson Laboratory, #005582) (Jung et al., 2000) were used to label microglia. Cx3cr1CreER mice were crossed with Ai14 mice harboring a loxP-flanked STOP cassette preventing the transcription of a CAG promoter-driven red fluorescent protein variant (tdTomato) inserted into the Gt(ROSA)26Sor locus (Madisen et al., 2010) (The Jackson Laboratory, #007914) constituted an alternative system to label microglia. These mice were also crossed with Cx3cr1CreER/+,Tgfbr2flox/flox mice to generate Cx3cr1 CreER/+,Tgfbr2flox/flox, Ai14/+ mice to enable cell-fate mapping of microglia vs. monocytes as previously performed (Ma et al., 2017). Cre recombinase activity was induced by tamoxifen administered by oral gavage (10 mg dose twice one day apart). Tgfbr2flox/flox mice (termed ‘Control’ mice) which were administered tamoxifen on the same regimen served as controls to tamoxifen-administered TG mice. All experimental animals were genotyped by gene sequencing to confirm the absence of the rd8 mutation (Mattapallil et al., 2012).

Immunohistochemical and TUNEL analysis of retinal tissue

Request a detailed protocol

Mice were euthanized by CO2 inhalation and their eyes were removed. Enucleated eyes were dissected to form posterior segment eye-cups and fixed in 4% paraformaldehyde in phosphate buffer (PB) for 2–4 hr at 4°C. Eyecups were either cryosectioned (Leica CM3050S) or dissected to form retinal flat-mounts. Flat-mounted retinas were blocked for 1 hr in blocking buffer containing 10% normal donkey serum and 0.5% Triton X-100 in PBS at room temperature. Primary antibodies, which included IBA1 (1:500, Wako, #019–19741), TGFBR2 (1:100, R and D, #AF-241), CD31 (1:100, Bio-Rad, #MCA2388), GFAP (1:200, Invitrogen, #13–0300), TMEM119 (1:100, Abcam, #ab209064), CD68 (1:200, Biorad, #MCA1957), MHCII (I-A/I-E, 1:100, BD Bioscience, #556999), NG2 (1:200, Millipore, #05–710), CD74 (1:100, BD Biosciences, #555318), F4/80 (1:100, Bio-Rad, #MCA497), CD45 (1:100, Bio-Rad, #MCA1388), glutamine synthetase (1:200, Millipore, #MAB302), BRN3A (1:100, Santa Cruz, #SC31984), cone arrestin (1:200, Millipore, #AB15282), choline acetyltransferase (ChAT) (1:100, Millipore, #AB144p), PKCα (1:200, Sigma-Aldrich, #p4334), CCL8 (1:100, Bio-Rad, #AAM62B), Ki67 (1:50, eBioscience, #50-5698-82), P2RY12 (1:100, ThermoFisher, #PA5-77671 and Sigma, #HPA014518), CD206 (1:100, BioRad, #MCA2235GA), were diluted in blocking buffer and applied overnight for sections and flat-mounts at room temperature on a shaker. Experiments in which primary antibodies were omitted served as negative controls. After washing in 1 × PBST (0.2% Tween-20 in PBS), retinal samples were incubated for 2 hr at room temperature with secondary antibodies (AlexaFluor 488-, 568- or 633-conjugated anti-rabbit, mouse, rat or goat IgG) and DAPI (1:500; Sigma-Aldrich) to label cell nuclei. Isolectin B4 (IB4), conjugated to AlexaFluor 568/647 (1:100, Life Technologies), was used to label activated microglia and retinal vessels. Apoptosis of retinal cells was assayed using a terminal deoxynucleotidyl transferase dUTP Nick End Labeling (TUNEL) assay (in situ cell death detection kit, TMR red; Roche) according to the manufacturer’s specifications. Stained retinal samples were imaged with confocal microscopy (Olympus FluoView 1000, or Zeiss LSM 880). For analysis at high magnification, multiplane z-series were collected using 20 × or 40 × objective; each z-series spanned from vitreal surface to the outer plexiform layer (OPL) for retinal flat-mounts, and over a depth of 20 μm for retinal sections, with each section spaced 1–2.5 μm apart. Confocal image stacks were viewed and analyzed with FV100 Viewer Software (Olympus), Zen software (Zeiss), and/or ImageJ (NIH).

Isolation of retinal microglia by flow cytometry

Request a detailed protocol

Enucleated globes were immersed in ice-cold Hank’s balanced salt solution (HBSS) and retinas were isolated by dissection before transfer into 0.2% papain solution including glucose (1 mg/mL), DNAse1 (100 U/mL; Worthington, Lakewood, NJ, USA), superoxide dismutase (SOD) (5 mg/mL; Worthington), gentamycin (1 µL/mL; Sigma), and catalase (5 mg/mL; Sigma, St Louis, MO, USA) in HBSS, and incubated at 8°C for 45 min and then at 28°C for 7 min. The digested tissue was dissociated by trituration and centrifuged at 150G for 5 min at 4°C. The resulting cell pellet was resuspended with neutralization buffer containing glucose (2 mg/mL), DNAse1 (100 U/mL), SOD (5 mg/mL), catalase (5 mg/mL), antipain (50 mg/mL Roche, Indianapolis, IN, USA), d-a-tocopheryl acetate (10 mg/mL; Sigma), albumin (40 mg/mL), and gentamycin (1 mL/mL, Sigma), and again centrifuged at 150 g for 5 min at 4°C. The cellular pellet was resuspended in 100 µL of staining buffer (catalog no. 554656, BD Pharmingen, San Diego, CA, USA) containing an Alexa Fluor 488-conjugated antibody to CD11b (1:50; catalog no. 53-0112-82, eBioscience, San Diego, CA, USA) and incubated for 20 min on ice. The cells were washed twice in 5 mL of staining buffer containing 2 mM ethylenediaminetetraacetic acid (EDTA) and suspended with 0.5 mL of staining buffer. Labeled retinal microglia were isolated by fluorescence-activated cell sorting (FACS)(BD FACSAria II Flow Cytometer; BD, Franklin Lakes, NJ, USA) at the NEI Flow Cytometry Core Facility. Sorted cells were collected into a 1.5 mL Eppendorf tube containing 200 µL of RNAlater solution (Ambion, AM7021) and stored at −80°C for subsequent RNA extraction.

Quantitative PCR analysis

Request a detailed protocol

Quantitative PCR analysis of genomic DNA and mRNA from FACS-sorted cells was performed. Total RNA was extracted from sorted cells (RNeasy Mini kit, Qiagen, Valencia, CA, USA) and used to synthesize cDNA with MessageBooster cDNA synthesis kit (Epicentre) following the manufacturer’s instructions and analyzed with qPCR in CFX96 real time PCR system (BioRad). Levels of mRNA expression were normalized to those in controls as determined using the comparative CT (2ΔΔCT) method. Ribosomal protein S13 (RPS13) was used as an internal control. Oligonucleotide primer pairs used are listed in Supplementary file 1.

mRNA profiling in retinal tissue using Nanostring

Request a detailed protocol

mRNA expression in retinal tissue was profiled and analyzed using the Nanostring platform nCounter Mouse Immunology panel containing 547 immunology-related mouse genes and 14 internal reference controls (Nanostring, Seattle, WA, #XT-CSO-MIM1-12). Briefly, the total RNA from a single retina was extracted using the RNeasy kit (Qiagen, Valencia, CA, USA). A total of 100 ng RNA in a volume of 5 µl was then hybridized to the capture and reporter probe sets at 65°C for 16 hr according to the manufacturer’s instructions. The individual hybridization reactions were washed and eluted per protocol at a NIH Core Facility (CCR Genomics Core, NCI, Bethesda, MD) and the data collected using the nCounter Digital Analyzer (Nanostring). Generated data was evaluated using internal QC process and the resulting data were normalized with the geometric mean of the housekeeping genes using the nSolver 4.0 and nCounter Advanced Analysis 2.0 software (Nanostring). Retina from 4 groups of animals, each comprising three biological repeats, were analyzed: control animals, not administered tamoxifen (Control), control animals, administered tamoxifen for 2 weeks (Control + TMX), TG animals, administered tamoxifen for 2 weeks (TG + TMX, 2 weeks), TG animals, administered tamoxifen for 8 weeks (TG + TMX, 8 weeks). Differentially expressed genes between two comparison groups were defined as those demonstrating a difference in expression level of fold change >2, with a p-value of <0.05 (adjusted p-value, t- test). The unsupervised hierarchical clustering analysis was performed using JMP statistical software (V13, SAS, Cary, NC). Canonical pathway analyses were performed using Ingenuity Pathway Analysis (IPA, Qiagen, Venlo, Netherlands).

Retinal microglia cell culture

Request a detailed protocol

Retinal microglia were isolated from postnatal day (P)20 C57BL/6J wild type mice and heterozygous Cx3cr1+/GFP transgenic mice as previously described (Ma et al., 2013). Briefly, retinal cells were dissociated by digestion in 2% papain, followed by trituration and centrifugation. Resuspended cells were transferred into 75 cm2 flasks containing Dulbecco’s Modified Eagle Medium (DMEM): Nutrient Mixture F-12 media with 10% fetal bovine serum (FBS) (Gibco, Carlsbad, CA, USA) and nonessential amino acids solution (Sigma, St. Louis, MO, USA). Following overnight culture, the medium and any floating cells were discarded and replaced with fresh medium. When the cultures approached confluence and cells begin to show detachment, the culture flasks were shaken gently to detach microglial cells that were then subcultured in 6-well plates. When microglial cultures reached 60–70% confluence, they were exposed to TGFβ ligands for 6 or 24 hr and then harvested for mRNA analysis.

In vivo optical coherence tomographic (OCT) imaging

Request a detailed protocol

Mice were anesthetized with intraperitoneal ketamine (90 mg/kg) and xylazine (8 mg/kg) and their pupils were dilated. Retinal structure was assessed using an OCT imaging system (Bioptigen; InVivoVue Software). Volume scans of 1.4 mm by 1.4 mm centered on the optic nerve (1000 A-scans/horizontal B-scan, 33 horizontal B-scans, average of three frames per B-scan, each spaced 0.0424 mm apart) were captured. Retinal thicknesses in each quadrant of a circular grid of diameter 1.2 mm were measured using the "measure" tool in the manufacturer’s software. Total retinal thickness, measured from the nerve fiber layer to the retinal pigment epithelium (RPE) layer, and outer retinal thickness, measured from the outer plexiform layer to the inner surface of the RPE layer, were obtained. Inner retinal thickness was computed as the difference between total retinal thickness and outer retinal thickness.

Electroretinographic (ERG) analysis

Request a detailed protocol

ERGs were recorded using an Espion E2 system (Diagnosys). Mice were anesthetized as described above after dark adaptation overnight. Pupils were dilated and a drop of proparacaine hydrochloride (0.5%; Alcon) was applied on cornea for topical anesthesia. Flash ERG recordings were obtained simultaneously from both eyes with gold wire loop electrodes, with the reference electrode placed in the mouth and the ground subdermal electrode at the tail. ERG responses were obtained at increasing light intensities over the ranges of 1 × 10−4 to 10 cd/s/m2 under dark-adapted conditions and 0.3 to 100 cd/s/m2 under a rod-saturating background light. The stimulus interval between flashes varied from 5 s at the lowest stimulus strengths to 60 s at the highest ones. Two to 10 responses were averaged depending on flash intensity. ERG signals were recorded with 0.3 Hz low-frequency and 300 Hz high-frequency cutoffs sampled at 1 kHz. The a-wave amplitude was measured from the baseline to the negative peak and the b-wave was measured from the a-wave trough to the maximum positive peak. Statistical comparisons between tamoxifen-treated control and TG mice were analyzed using a two-way ANOVA.

Image analysis

Request a detailed protocol

To analyze the total numbers of microglia in the retina, manual counts of Iba1+ cells were performed over the entire retina in flat-mounted specimens. Microglial counts, as well as the proportion of microglia that were Ki67+, were evaluated in the separate retinal lamina (IPL, OPL, and subretinal space) using high-magnification image stacks captured at consistent retinal regions of interest (ROIs) positioned midway between the optic nerve and the retinal periphery. Morphological analysis of microglia on the parameters of number of branch points/cell and area of dendritic field were performed manually using Image J software (NIH). The intensity of immunohistochemical labeling was assessed by quantifying the mean fluorescent intensity within each labeled microglial cell using Image J software.

Fluorescein angiography

Request a detailed protocol

Mice were injected intraperitoneally with Fluorescein AK-FLUOR (100 mg/mL; Akorn) at 100 μg/g (body weight). Fluorescein angiography (FA) of the retina was performed using a Phoenix Micron III retinal imaging system (Phoenix Research Labs) at various times following fluorescein injection. Bright-field and fluorescence images of the central fundus were captured during early and late transit phases.

In vivo laser model for choroidal neovascularization

Request a detailed protocol

Choroidal neovascularization was induced in vivo using a laser injury model as previously described (Campos et al., 2006). Experimental animals were anesthetized with an intraperitoneal injection of ketamine (90 mg/kg) and xylazine (8 mg/kg) and their pupils were dilated with 1% tropicamide (Akorn Inc, Buffalo Grove, IL, USA) and 2.5% phenylephrine (Alcon Laboratories Inc, Fort Worth, TX, USA). Corneal anesthesia was provided using topical 0.5% proparacaine (Alcon Laboratories Inc). Laser injury was applied to the retina using a slit-lamp-mounted, 532 nm wavelength, photocoagulation laser (Iridex, Mountain View, CA, USA) and a handheld focusing lens. Using laser settings to create burns that ruptured Bruch’s membrane (power, 50 mW; duration, 100 ms; spot size, 100 µm), four well-spaced laser burns were placed circumferentially approximately 375 µm from the optic nerve. Animals were sacrificed 7 days after laser injury and the size of choroidal neovascularization (CNV) complexes was evaluated in RPE-choroidal flat-mounts following immunohistochemical staining with DAPI, Iba1, Alex633-conjugated phalloidin (1:100), and Alex568-conjugated lectin IB4 (1:100). Microglial recruitment was evaluated by measuring the intensity of the Iba1+ signal, the area of RPE cell disruption was determined using phalloidin labeling, and the area of CNV determined by IB4 labeling and image analysis.

References

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
  7. 7
  8. 8
  9. 9
  10. 10
  11. 11
  12. 12
  13. 13
  14. 14
  15. 15
  16. 16
  17. 17
  18. 18
  19. 19
  20. 20
    TGF-beta signaling in cancer treatment
    1. I Fabregat
    2. J Fernando
    3. J Mainez
    4. P Sancho
    (2014)
    Current Pharmaceutical Design 20:2934–2947.
  21. 21
  22. 22
  23. 23
  24. 24
  25. 25
    Seven new loci associated with age-related macular degeneration
    1. LG Fritsche
    2. W Chen
    3. M Schu
    4. BL Yaspan
    5. Y Yu
    6. G Thorleifsson
    7. DJ Zack
    8. S Arakawa
    9. V Cipriani
    10. S Ripke
    11. RP Igo
    12. GH Buitendijk
    13. X Sim
    14. DE Weeks
    15. RH Guymer
    16. JE Merriam
    17. PJ Francis
    18. G Hannum
    19. A Agarwal
    20. AM Armbrecht
    21. I Audo
    22. T Aung
    23. GR Barile
    24. M Benchaboune
    25. AC Bird
    26. PN Bishop
    27. KE Branham
    28. M Brooks
    29. AJ Brucker
    30. WH Cade
    31. MS Cain
    32. PA Campochiaro
    33. CC Chan
    34. CY Cheng
    35. EY Chew
    36. KA Chin
    37. I Chowers
    38. DG Clayton
    39. R Cojocaru
    40. YP Conley
    41. BK Cornes
    42. MJ Daly
    43. B Dhillon
    44. AO Edwards
    45. E Evangelou
    46. J Fagerness
    47. HA Ferreyra
    48. JS Friedman
    49. A Geirsdottir
    50. RJ George
    51. C Gieger
    52. N Gupta
    53. SA Hagstrom
    54. SP Harding
    55. C Haritoglou
    56. JR Heckenlively
    57. FG Holz
    58. G Hughes
    59. JP Ioannidis
    60. T Ishibashi
    61. P Joseph
    62. G Jun
    63. Y Kamatani
    64. N Katsanis
    65. C N Keilhauer
    66. JC Khan
    67. IK Kim
    68. Y Kiyohara
    69. BE Klein
    70. R Klein
    71. JL Kovach
    72. I Kozak
    73. CJ Lee
    74. KE Lee
    75. P Lichtner
    76. AJ Lotery
    77. T Meitinger
    78. P Mitchell
    79. S Mohand-Saïd
    80. AT Moore
    81. DJ Morgan
    82. MA Morrison
    83. CE Myers
    84. AC Naj
    85. Y Nakamura
    86. Y Okada
    87. A Orlin
    88. MC Ortube
    89. MI Othman
    90. C Pappas
    91. KH Park
    92. GJ Pauer
    93. NS Peachey
    94. O Poch
    95. RR Priya
    96. R Reynolds
    97. AJ Richardson
    98. R Ripp
    99. G Rudolph
    100. E Ryu
    101. JA Sahel
    102. DA Schaumberg
    103. HP Scholl
    104. SG Schwartz
    105. WK Scott
    106. H Shahid
    107. H Sigurdsson
    108. G Silvestri
    109. TA Sivakumaran
    110. RT Smith
    111. L Sobrin
    112. EH Souied
    113. DE Stambolian
    114. H Stefansson
    115. GM Sturgill-Short
    116. A Takahashi
    117. N Tosakulwong
    118. BJ Truitt
    119. EE Tsironi
    120. AG Uitterlinden
    121. CM van Duijn
    122. L Vijaya
    123. JR Vingerling
    124. EN Vithana
    125. AR Webster
    126. HE Wichmann
    127. TW Winkler
    128. TY Wong
    129. AF Wright
    130. D Zelenika
    131. M Zhang
    132. L Zhao
    133. K Zhang
    134. ML Klein
    135. GS Hageman
    136. GM Lathrop
    137. K Stefansson
    138. R Allikmets
    139. PN Baird
    140. MB Gorin
    141. JJ Wang
    142. CC Klaver
    143. JM Seddon
    144. MA Pericak-Vance
    145. SK Iyengar
    146. JR Yates
    147. A Swaroop
    148. BH Weber
    149. M Kubo
    150. MM Deangelis
    151. T Léveillard
    152. U Thorsteinsdottir
    153. JL Haines
    154. LA Farrer
    155. IM Heid
    156. GR Abecasis
    157. AMD Gene Consortium
    (2013)
    Nature Genetics 45:433–439.
    https://doi.org/10.1038/ng.2578
  26. 26
  27. 27
  28. 28
  29. 29
  30. 30
  31. 31
  32. 32
  33. 33
  34. 34
  35. 35
  36. 36
  37. 37
  38. 38
  39. 39
  40. 40
  41. 41
  42. 42
  43. 43
  44. 44
  45. 45
    Heterogeneity in localization of isoforms of TGF-beta in human retina, vitreous, and choroid
    1. GA Lutty
    2. C Merges
    3. AB Threlkeld
    4. S Crone
    5. DS McLeod
    (1993)
    Investigative Ophthalmology & Visual Science 34:477–487.
  46. 46
  47. 47
  48. 48
  49. 49
  50. 50
  51. 51
  52. 52
  53. 53
    Light-induced migration of retinal microglia into the subretinal space
    1. TF Ng
    2. JW Streilein
    (2001)
    Investigative Ophthalmology & Visual Science 42:3301–3310.
  54. 54
  55. 55
  56. 56
  57. 57
  58. 58
  59. 59
  60. 60
  61. 61
  62. 62
  63. 63
  64. 64
  65. 65
  66. 66
  67. 67
  68. 68
  69. 69
  70. 70
  71. 71
  72. 72
  73. 73
  74. 74
  75. 75
  76. 76
  77. 77
  78. 78
  79. 79

Decision letter

  1. Constance L Cepko
    Reviewing Editor; Harvard Medical School, United States
  2. Gary L Westbrook
    Senior Editor; Vollum Institute, United States

In the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included.

Thank you for submitting your article "Absence of TGFβ signaling in retinal microglia induces retinal degeneration and promotes choroidal neovascularization" for consideration by eLife. Your article has been reviewed by three peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Gary Westbrook as the Senior Editor. The following individuals involved in review of your submission have agreed to reveal their identity: Pat D'Amoore (Reviewer #2); Lois Smith (Reviewer #3). The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

Summary:

The study, "Absence of TGFβ signaling in retinal microglia induces retinal degeneration and promotes choroidal neovascularization" addresses an interesting question in retinal degeneration, perhaps relevant to AMD and other degenerative diseases. Previous studies of the role of TGFb in retinal biology, and suggestions from human genetics, have led to the notion that microglia are signaled through this pathway. A role for microglia in retinal health has been addressed previously, including studies by this group, where they found that they play a role in rod degeneration in models of retinitis pigmentosa. Here, the authors deleted the TGFBR2 gene from mice using a conditional allele and a tamoxifen regulated Cre expressed from the regulatory region of Cx3CR1. They find that retinal neurons degenerate, there is a change in the response of microglia to laser injury, there is a reduction in the expression of microglial genes comprising their sensome, and overall a change in the microglial gene expression signature suggesting alterations to immune system modulation. The authors conclude that all of the observed phenotypes are due to effects within microglia following loss of the receptor.

Essential revisions:

As pointed out by the reviewers, the Cre allele that is used likely can lead to recombination in two other cell types, macrophages and monocytes. The authors should attempt to determine if recombination is solely within microglia by using antibodies to a microglial-specific antibody to Tmem119, and one to P2RY12. If they are unable to prove their statements regarding the sole recombination within microglia, they should modify their title and Discussion to reflect these ambiguities. However, the study is well done and will stand without this proof, but modifications of their claims will need to be made accordingly. Several other suggestions are made by the reviewers that should be addressed to improve the manuscript, e.g. a discussion of the source of the ligand and reference to the new work from Zoller et al. using a similar model to examine effects in the brain.

Original reviews:

The original reviews from the reviewers are below. We think addressing other issues that are not explicit in the Essential revisions above would also improve the manuscript so we hope you will consider them as you prepare your revisions.

Reviewer #1:

In the manuscript entitled "Absence of TGFβ signaling the retinal microglia induces retinal degeneration and promotes choroidal neovascularization", the authors identify that loss of TGFβR2 in retinal microglia influences microglial activation and neurodegeneration. Previous work has shown that CNS deletion of TGFβ1 results in loss of a homeostatic gene signature in microglia and a shift to a heightened inflammatory state (Butovksy et al., 2014). The authors take this a significant step further using conditional genetic deletion of TGFβR2 specifically in microglia (Cx3cr1CreERT2;Tgfbr2fl/fl mice) and show that this results in reduced expression of homeostatic genes and a heightened inflammatory state in the retina. Importantly, the authors further show that this ultimately results in structural and functional degeneration of neurons in vivo in the retina. Overall, the authors address an important question with rigor. The data are also high quality and largely support the conclusions. The following are points that require further attention:

1) The authors do not address the possibility that the effects are due to changes in perivascular macrophages vs. retinal microglia. It is highly likely that PVMs also recombine using the Cx3cr1CreERT2 mouse. It would be important to address the potential contribution of these cells given their high association with blood vessels.

2) The contribution of infiltrating monocytes to increased numbers of Iba-1/CD11b-positive cells in Cx3cr1CreERT2;Tgfbr2fl/fl mice has not been ruled out. Using tdTomato as a reporter is not sufficient (Figure 3—figure supplement 1C). While monocytes should turn over following tamoxifen treatment to eventually become tdTomato-negative (within ~7-14 days), tdTomato-positive monocytes could have infiltrated and taken up residence in the retina prior to turning over. One would not be able to distinguish these early infiltrates in the current experiments.

3) The authors likely submitted the current manuscript prior to the publication of the recent paper (October 1, 2018) by Zöller et al. in Nature Communications (Zöller et al., 2018). This group uses the same genetic strategy to assess TGFβ signaling in microglia (Cx3cr1CreERT2;Tgfbr2fl/fl mice), but assessments were perform in the brain or in vitro. Interestingly, while the two studies observe similar results in heightened microglial inflammatory state following ablation of TGFβR2, the study by Zöller et al. does not observe decreases in microglial homeostatic gene expression or neurodegeneration. These differences in the two studies should be addressed in the revised manuscript.

4) It is concluded that effects on microglial morphology are cell autonomous (subsection “Specific in vivo ablation of TGFBR2 in retinal microglia induces rapid morphological transformation and proliferation”, first paragraph). While certainly a possibility, it is also plausible that these effects are secondary. For example, loss of TGFβ signaling in microglia could result in changes in signaling to Müller glia. The Müller glia then respond by releasing factors that alter the microglial morphology. Such alternative possibilities should be addressed.

5) One of the major conclusions is that microglia downregulate their 'sensome' related genes following ablation of TGFβR2. However, in Figure 7, the mutant microglia appear to react more robustly to a laser ablation? Purinergic signaling, a key sensome-related pathway, is known to regulate this injury response in microglia and many of these genes appear to be downregulated in the gene expression data set presented in this paper. Therefore, it is curious that such a robust response is observed in Figure 7. This should be discussed.

6) It seems inappropriate to include graphs of published data from other groups (Figure 1B and 1C). Instead, the authors should simply refer to these data in the text. Also, note that the Barres lab RNAseq database is largely from neonate animals, not adult.

7) The same image is used for 2 figures (3 week timepoint in Figure 2 and Figure Supplementary Figure 1). This should be corrected.

8) The image of Iba1 staining in Figure 3—figure supplement 1A at 3 weeks post-tamoxifen treatment is inconsistent with the increased Iba1 and CD11b-positve cells shown in Figure 2 and Supplementary figure 1. This should be addressed.

9) In Figure 2E and G, the authors would be better served to quantify the total area of microglia vs. the area 'subtended' by microglia. It is difficult to make such a measurement in static images and is better suited for live imaging. For example, perhaps mutant microglia are more motile and still 'subtend' the same amount of retina, but this can't be captured by static imaging.

10) It appears that microglia no longer express Cx3cr1 and Tmem119 in Figure 3A? It would be helpful to show these data on a graph where the y axis is a log2 scale.

11) CD68-positive lysosomes should be observed in control conditions, albeit at lower amounts compared to more activated cells. It is unclear why CD68 is virtually absent in controls (e.g. Figure 4C and D).

Reviewer #2:

While the work is novel for the retina, it is very reminiscent of a recent study from Qin et al., 2018, which shows that lack of activated TGFβ leads to microglial activation and progressive loss of myelinated axons and death. In addition, a paper was just published (Zoller et al., 2018) showing a similar role for TGFβ in brain microglia as in retina, which the authors may want to cite (unlikely they were aware of it prior to submission of this paper).

CX3CR1 is not specific for microglia and is also expressed by circulating monocytes. There is, however, one marker Tmem119 that has shown to be specific, which could provide "cleaner" results. A discussion of this fact would be useful. It is surprising that in spite of the demonstrated upregulation of proinflammatory cytokines, such as interferons and IL1β, by the activated microglial, infiltrating macrophages were not detected. The authors might speculate on why this is.

The authors conclude that the "constitutive neuron-microglia interactions in the form of TGFβ are necessary in the maintenance of the orderly organization and trophic function of the microglial in the retina". What is the evidence that neurons are providing the TGFβ, how is it activated and which isoform of TGFβ is mediating the homeostatic suppression of microglial activation?

The title is misleading. The data show that the absence of TGFβ signaling in microglial leads to exacerbated choroidal neovascularization, but the title suggests that this initiates the vessel growth.

Reviewer #3:

Inflammation/immune changes are associated with age-related macular degeneration (AMD). The authors examine how TGFβ signaling in microglia influences retinal neuroinflammation. They found that ablation of the TGFβ receptor, TGFBR2, in retinal microglia of adult mice induced abnormal microglial numbers, distribution, morphology, and activation status, and promoted a change in microglial gene expression profile. TGFBR2-deficient retinal microglia induced secondary gliotic changes in Müller cells, neuronal apoptosis, and decreased light-evoked retinal function (ERG) reflecting abnormal synaptic transmission and increased laser-induced choroidal neovascularization. These results suggest that TGFβ-mediated microglial regulation can drive neuroinflammatory contributions to AMD-related neurodegeneration and neovascularization.

General comments:

There is a dearth of cell specific information about "inflammation" and "immune" contributions to AMD. Although there are very clear indications of the importance of inflammation and immunity from human genetic studies, the cellular and biochemical details need to be worked out experimentally. There are many intrinsic problems with extrapolation of mouse data to human as the immune systems in each are very different and mouse models of AMD are approximate as mice have no macula, and aging in mice and humans is very different. Nonetheless, pathway analysis can be helpful. It is likely that these mouse analyses of TGFβ in macrophages/microglia will overlap with human AMD.

In many ways this paper is quite descriptive as there is no explanation for the neuronal loss following macrophage loss of TGFbR2. Nonetheless, the paper adds important new general information about macrophage neuronal interactions and TGFb control of macrophage function and related neuronal survival. I expect that with the techniques now available to look at multiple pathways the need to prove in detail that one pathway is involved after a genetic change in a cell will become less important.

Specific comments:

Figure 1B: need to label y axis as Tgfbr2 and reference the control. Check all figures for more complete labeling with units and specific mRNAs etc.

Figure 1 convincingly shows that TGFBR2 is expressed preferentially in microglia and is ablated in Cx3Cr1CreER/+,Tgfbr2flox/flox mice with Tamoxifen.

Figure 2: TGFBR2 ablation in retinal microglia clearly induces abnormalities in microglial density, distribution, and morphology. Interestingly, despite increased density, TMX-treated TG retinas have a greater area without microglial processes.

Figure 3: Constitutive expression of microglial "sensome" genes is decreased with TGFBR2 ablation in retinal microglia. mRNA levels of Cx3cr1, P2yr12, Tmem119, and Siglech are all decreased in microglia in TG vs. control mice.

Figure 4: Expression of genes associated with microglial activation is increased with TGFBR2 ablation in retinal microglia. mRNA levels for H2-Aa (MHCII), Cd68, Cd74, Apoe, Ccl2, and CCl8 are all increased in microglia in TG vs. control mice.

Figure 5: TGFBR2 ablation in retinal microglia preserved lamination but progressively decreased both the inner and the outer retinal layer thickness (the inner plexiform layer (IPL), inner nuclear layer (ONL), outer plexiform layer (OPL), and outer nuclear layer (ONL)) 3 weeks post-TMX.

In 10 weeks post-TMX, TG animals had altered dark-adapted responses with a small decrease in a-wave amplitude and a marked decrease in b-wave amplitude. Light-adapted responses were similar for a-wave amplitude but significantly decreased in b-wave amplitude.

Figure 6: Changes in the mRNA expression of immune regulated genes in the retina following microglial TGFBR2 ablation using Nanostring-based profiling. Volcano plots at 2 and 8 weeks reflect the activation of neuroinflammatory pathways and immune cell activation and maturation pathways.

Figure 7: TGFBR2 ablation in retinal microglia increases laser-induced choroidal neovascularization. No abnormal leakage or vascular structure were detected. Normal morphologies and distributions following TGFBR2 ablation in TG mice 12 weeks TGFBR2-ablated TG animals demonstrated a higher recruitment of Iba1+ myeloid cells to the laser injury site, which was correlated with a larger laser lesion size and a larger CNV area.

https://doi.org/10.7554/eLife.42049.026

Author response

Essential revisions:

As pointed out by the reviewers, the Cre allele that is used likely can lead to recombination in two other cell types, macrophages and monocytes. The authors should attempt to determine if recombination is solely within microglia by using antibodies to a microglial-specific antibody to Tmem119, and one to P2RY12. If they are unable to prove their statements regarding the sole recombination within microglia, they should modify their title and Discussion to reflect these ambiguities. However, the study is well done and will stand without this proof, but modifications of their claims will need to be made accordingly.

In this revision, we have performed additional experiments to further support the notion that recombination (that enables TGFBR2 genetic ablation) occurs within retinal microglia by using microglia-specific antibodies as the reviewers had indicated. In our original manuscript, we had found that retinal transcripts for Tmem119 and P2ry12 declined to low levels 2 weeks following tamoxifen induction in transgenic animals. We had also found that TMEM119 immunopositivity in microglia is rapidly lost upon induction (as previously shown in Figure 3E, now Figure 4E). In this revision, we performed experiments with P2RY12 immunohistochemistry that demonstrated that during the first week following induction, all endogenous CD11b+, P2RY12+ retinal microglia demonstrated progressive deramification and morphological changes; many of these deramifying CD11b+, P2RY12+ cells also acquired Ki67+ immunopositivity. We did not observe any CD11b+, P2RY12-negative infiltrating macrophages entering the retina during this early phase. Although P2RY12 immunopositivity, like that for TMEM119, gradually declined and became undetectable after >7 days following induction, there is no evidence for a contribution of infiltrating monocytes during this early phase to the final population of morphologically transformed cells. Our cell-fate mapping experiments also ruled out the possibility for a later phase of monocyte infiltration as a contributing population. Because Ki67+ immunopositivity can also be detected in P2RY12+ microglia following induction, proliferation of endogenous microglia is a mechanism that is present and capable of contributing to the increased numbers of myeloid cells detected without the need to invoke a contribution from infiltrating macrophages.

These results, together with supporting data from Lund et al., 2018 (PMID: 29662171) and Buttgereit et al., 2016 (PMID: 27776109) (elaborated below in responses to reviewer #1), provide further justification for a microglial, rather than a monocytic, source for the morphologically transformed cells following TGFBR2 ablation.

This new data is included in Figure 3—figure supplement 1. We have added to the Results:

“In addition, we observed that the myeloid cells in the retina demonstrating progressive morphological change in the first week following tamoxifen administration were immunopositive for P2RY12, a marker for endogenous microglia, as well as for Ki67, a marker of proliferating cells (Figure 3—figure supplement 1D). Although P2RY12 immunpositivity was gradually lost after one week following TGFBR2 ablation, these findings indicated that the population of morphologically-transformed myeloid cells in the retina arose from the proliferation and modification of pre-existing endogenous retinal microglia.”

Several other suggestions are made by the reviewers that should be addressed to improve the manuscript, e.g. a discussion of the source of the ligand and reference to the new work from Zoller et al. using a similar model to examine effects in the brain.

We have in this revision: (1) added a new paragraph in the Discussion discussing the work of Zoller et al., (2) added further clarification regarding potential cellular sources of TGFB ligands in the retina.

Original reviews:

The original reviews from the reviewers are below. We think addressing other issues that are not explicit in the Essential revisions above would also improve the manuscript so we hope you will consider them as you prepare your revisions.

Reviewer #1:

[…] The following are points that require further attention:

1) The authors do not address the possibility that the effects are due to changes in perivascular macrophages vs. retinal microglia. It is highly likely that PVMs also recombine using the Cx3cr1CreERT2 mouse. It would be important to address the potential contribution of these cells given their high association with blood vessels.

The reviewer raises a valid point regarding the potential contribution of perivascular macrophages to the transformed population of myeloid cells within the retina following TGFBR2 ablation. While much remains unknown about perivascular macrophages, it is thought that they are a sparse population of perivascular cells located outside the retinal neural parenchyma (PMID: 19608545), that are long-lived and CX3CR1-expressing (PMID: 27135602). While we cannot rule out that there may be some transformed perivascular macrophages in the retina following TGFBR2 ablation, we consider it more likely that the majority of the effects described here are attributable to TGFBR2-ablated retinal microglia for the following reasons: (1) retinal microglia are much more numerous in the retina, (2) we observed progressive, time-dependent changes following tamoxifen administration in IBA1+ microglia located in the retinal neural parenchyma, which transitioned progressively from a ramified to an elongated morphology following TGFBR2 ablation. It is likely that CX3CR1-expressing perivascular macrophages undergo TGFBR2 ablation also, but we believe that the contribution to overall retinal pathology observed is likely small, given their sparse numbers at baseline. The oft-used marker for typical perivascular macrophages, CD206, is not helpful in making this distinction as retinal microglia, following TGFBR2 ablation, also become CD206-immunpositive. We performed additional experiments in which we conducted close follow-up of retinal microglia during the first week following tamoxifen ablation; we found that microglia demonstrated progressive morphological changes that were coincident with the gradual increase of CD206 immunopositivity in these cells. In order for true perivascular macrophages (which are CD206+ at baseline) to contribute substantially to the final population of myeloid cells in the TGFBR2-ablated retina, these cells will have to proliferate rapidly and migrate to distribute themselves across the retina – these features were not observed on close scrutiny of early events following TGFBR2 ablation.

We have added the following statement to the Discussion to include mention of perivascular macrophages:

“It is possible that long-lived CX3CR1-expressing perivascular macrophages resident within the retina may also contribute to the transformed population, but this is likely a smaller contribution, owing to their sparser numbers at baseline (Goldmann et al., 2016; Mendes-Jorge et al., 2009). […] However, prominent proliferation and migration of CD206+ perivascular macrophages present at baseline prior to TGFBR2 ablation were not detected, indicating that true perivascular macrophages are unlikely to contribute substantially to the final population of transformed cells (Figure 3—figure supplement 2).”

2) The contribution of infiltrating monocytes to increased numbers of Iba-1/CD11b-positive cells in Cx3cr1CreERT2;Tgfbr2fl/fl mice has not been ruled out. Using tdTomato as a reporter is not sufficient (Figure 3—figure supplement 1C). While monocytes should turn over following tamoxifen treatment to eventually become tdTomato-negative (within ~7-14 days), tdTomato-positive monocytes could have infiltrated and taken up residence in the retina prior to turning over. One would not be able to distinguish these early infiltrates in the current experiments.

We agree with the reviewer that if monocytes did indeed infiltrate the retina soon after tamoxifen administration, subsequently becoming long-lived cells, they may, like endogenous microglia, demonstrate tdTomato labeling in this experiment. Our experimental design would rule out ongoing monocytic infiltration occurring after 7-14 days but may miss early infiltration. In this revision, we have provided additional data that show that all retinal microglia demonstrating morphological changes during this early period were initially immunopositive for P2RY12, a marker for endogenous microglia, supporting the notion that resident retinal microglia undergo morphological transformation to form the population of myeloid cells characterized in the manuscript. We did not observe any early infiltration of P2RY12-negative monocytes into the retina during this period.

The absence of a contribution by infiltrating monocytes in this context is additionally supported by previously published work: (1) Lund et al., 2018 (PMID: 29662171) in which bone-marrow transplantation from a CD45.1 WT donor mouse into a CD45.2, Cx3cr1CreER/+, Tgbr2fl/fl recipient mouse demonstrated that all the CD11b+, F4/80+ myeloid cells located in the brain following TGFBR2 ablation were all CD45.2+, indicating their status as resident microglia (with little or no contribution from CD45.1+ circulating monocytes), (2) Buttgereit et al., 2016 (PMID: 27776109), in which tamoxifen administration in Sall1CreER/+, Tgbr2fl/fl, R26-YFP mice in resulted in TGFBR2 ablation and YFP expression specifically in only microglia and not circulating monocytes (Sall1 being specifically expressed in microglia but not other members of the mononuclear phagocyte family, such as monocytes); all the CD45+,F4/80+ cells in the brain following TGFBR2 were YFP+, indicating minimal contribution by systemic monocytes. We have in this revision also added new data showing that morphologically transformation myeloid cells in the retina following TGFBR2 ablation are immunopositive for P2RY12, a marker of endogenous microglia (see response to Essential revisions above).

The evidence that we have demonstrated in the retina, combined with other supporting information from previous studies in the brain, together indicate that monocytic contribution to the population of TGFBR2-ablated myeloid cells in the CNS, is unlikely to be significant.

3) The authors likely submitted the current manuscript prior to the publication of the recent paper (October 1, 2018) by Zöller et al. in Nature Communications (Zöller et al., 2018). This group uses the same genetic strategy to assess TGFβ signaling in microglia (Cx3cr1CreERT2;Tgfbr2fl/fl mice), but assessments were perform in the brain or in vitro. Interestingly, while the two studies observe similar results in heightened microglial inflammatory state following ablation of TGFβR2, the study by Zöller et al. does not observe decreases in microglial homeostatic gene expression or neurodegeneration. These differences in the two studies should be addressed in the revised manuscript.

We have now included mention of this recent paper which had been published after our manuscript’s submission in the Discussion. While there are similarities between the findings in Zöller et al., and those in our manuscript and the work of others, there were also observations in their report that were not found elsewhere. They did not find in their observations evidence of decreased microglial homeostatic gene expression or the onset of neurodegenerative changes; this contrasted with not only our findings, but also those reported previously. We feel that these differences may have arisen in differences in methodology, and we remain in the opinion that microglia in the retina require constitutive TGF-β signaling to maintain a microglia-specific gene signature and to prevent a transition to an activated phenotype that help drive retinal neurodegeneration. We have added the following text in the Discussion to address these differences:

“In a study published following the submission of our manuscript, Zöller et al., (Zoller et al., 2018) using a similar transgenic model, had induced the ablation of exon 2/3 of TGFBR2 in CX3CR1-expressing cells (Chytil et al., 2002) and described in the brain an upregulation of microglial activation markers, but had failed to detect alterations in microglial density, microglia-specific gene expression or neuronal survival. […] Combined with findings in vitro demonstrating a requirement for TGF-β for the expression of a microglia-specific gene signature, and those in vivo in showing that decreased TGF-β signaling to microglia resulted in the downregulation of microglia-specific gene expression and neurodegenerative changes (Butovsky et al., 2014; Qin et al., 2018), it is likely that microglia in the retina require constitutive TGF-β signaling to maintain a microglia-specific gene signature and to prevent a transition to an activated phenotype that helps drive retinal neurodegeneration.

4) It is concluded that effects on microglial morphology are cell autonomous (subsection “Specific in vivo ablation of TGFBR2 in retinal microglia induces rapid morphological transformation and proliferation”, first paragraph). While certainly a possibility, it is also plausible that these effects are secondary. For example, loss of TGFβ signaling in microglia could result in changes in signaling to Müller glia. The Müller glia then respond by releasing factors that alter the microglial morphology. Such alternative possibilities should be addressed.

We had suggested that neuronal degeneration observed in our model may have been contributed by changes in secreted signals from Müller glia. We add to this point now by mentioning how these signals can also feedback onto microglia to influence them. The section in the Discussion now reads:

“We found evidence for widespread secondary gliotic changes in Müller cells that were spatiotemporally-coincident with microglial changes. […] These changes in Müller cells may additionally feedback onto nearby microglia via secreted signals to influence their activation (Wang et al., 2011; Wang et al., 2014).”

5) One of the major conclusions is that microglia downregulate their 'sensome' related genes following ablation of TGFβR2. However, in Figure 7, the mutant microglia appear to react more robustly to a laser ablation? Purinergic signaling, a key sensome-related pathway, is known to regulate this injury response in microglia and many of these genes appear to be downregulated in the gene expression data set presented in this paper. Therefore, it is curious that such a robust response is observed in Figure 7. This should be discussed.

The functional concept of expression of “sensome” genes by microglia in the healthy CNS has been related to the ability of microglia to recognize endogenous ligands in order to carry out homeostatic functions under normal conditions. In the absence of TGF-β signaling, the decrease in microglia-specific gene expression, and the increase in microglial activation markers and microglial proliferation, have been correlated to increased numbers of microglia demonstrating abnormal physiologies and aberrant responses that fail to successfully re-establish homeostasis in the face of perturbations, driving increased pathological change in the brain and spinal cord (Lund et al., 2018a, Qin et al., 2018, and Taylor et al., 2017).

In the retina, we also find that constitutive TGFβ signaling negatively regulates microglial responses to injury triggers, as TGFBR2 ablation results in more microglial recruitment and increased neovascular changes. While the decreased expression of P2RY12, a microglial receptor and sensome gene, may negatively influence microglial recruitment in response to ATP release induced by tissue injury, this single factor, in and of itself, does not necessarily predict a less robust microglial response. On the contrary, in the absence of TGF-β signaling, the increased numbers of microglia expressing higher levels of multiple priming and activation markers, culminate in increased inflammation and greater pathological change. Consistent with this, increased numbers of activated, proinflammatory microglia have been associated with increased pathological neovascularization in the retina (PMID: 29654250, 29767277, 23977372) – in the current context, proinflammatory changes in microglia induced by TGFBR2 ablation appear more influential to exacerbation of laser-induced neovascularization, than the potential loss of P2RY12 function. Our Nanostring mRNA profiling experiments detected in TGFBR2-ablated microglia increased expression of multiple chemokine receptors (CCR3, CCR5, CCR6) and Toll-like receptors (TLR1, TLR4, TLR9) that suggest an enhanced ability to respond to other inflammatory cues.

6) It seems inappropriate to include graphs of published data from other groups (Figure 1B and 1C). Instead, the authors should simply refer to these data in the text. Also, note that the Barres lab RNAseq database is largely from neonate animals, not adult.

We appreciate the reviewer’s point of view but we feel that the graphical representations of this data (which is other located within the published database and needs to be manually retrieved) help to communicate our point to the reader better than simply referring to it in the text. We have made the sources of the information very clear in the manuscript (there is no doubt as to its attribution). We have also changed the heading in Figure 1C (data from the Barres lab) from “adult” to “P7-17”.

7) The same image is used for 2 figures (3 week timepoint in Figure 2 and Supplementary Figure 1). This should be corrected.

8) The image of Iba1 staining in Figure 3—figure supplement 1A at 3 weeks post-tamoxifen treatment is inconsistent with the increased Iba1 and CD11b-positve cells shown in Figure 2 and Supplementary figure 1. This should be addressed.

We apologize for inadvertently using the same image in two places in the manuscript. While these images are illustrative of the data and the point made in the manuscript, we have now provided alternative images in Supplementary Figure 1 (now Figure 2).

9) In Figure 2E and G, the authors would be better served to quantify the total area of microglia vs. the area 'subtended' by microglia. It is difficult to make such a measurement in static images and is better suited for live imaging. For example, perhaps mutant microglia are more motile and still 'subtend' the same amount of retina, but this can't be captured by static imaging.

In our analyses, we had aimed to quantitate morphological measures that provide a comprehensive and insightful description of morphological changes that microglia undergo on TGFBR2 ablation. One of the features that microglia have in the flat, laminated structure of the healthy adult retina, is in their two-dimensionally oriented, ramified morphologies that provide spatial coverage in the OPL and IPL. With TGFBR2 ablation, the ramified processes are shortened, the geometries of the cells altered, and the spaces between neighboring cells not covered by the processes enlarge. We had attempted to illustrate this feature by quantifying area subtended by the territory of each microglia’s dendritic tree. This difference in space occupation between cell morphologies with and without TGFBR2 ablation will not be highlighted by simply quantifying the total area of the microglial processes. This morphological analysis had been used in previous publications by us (Ma et al., 2009 and PMID: 21108733) and others (PMID: 25064005).

10) It appears that microglia no longer express Cx3cr1 and Tmem119 in Figure 3A? It would be helpful to show these data on a graph where the y axis is a log2 scale.

The main point of this figure is to illustrate that mRNA levels for microglial sensome genes are markedly downregulated following TGFBR2 ablation. We feel that the representation of the same data on a log2 scale may not reflect this main point as intuitively (see Author response image 1). In Figure 4A, we use the same graphical format to show the upregulation of microglial activation genes – in this situation, a matching log2 scale will be even less intuitive and evocative of the main point. This is our point of view concerning the effective exposition of our data and we are amenable to changing the format of the graph if there is consensus opinion among the editors and reviewers that a change is preferable.

11) CD68-positive lysosomes should be observed in control conditions, albeit at lower amounts compared to more activated cells. It is unclear why CD68 is virtually absent in controls (e.g. Figure 4C and D).

CD68-positive lysosomes are indeed visible in control microglia in the figure (now Figure 5C). At higher magnification, it can be observed that this signal is present in microglia in the expected locations (i.e. within the soma in a perinuclear location, close to the base of a primary process).

Reviewer #2:

While the work is novel for the retina, it is very reminiscent of a recent study from Qin et al., 2018, which shows that lack of activated TGFβ leads to microglial activation and progressive loss of myelinated axons and death. In addition, a paper was just published (Zoller et al., 2018) showing a similar role for TGFβ in brain microglia as in retina, which the authors may want to cite (unlikely they were aware of it prior to submission of this paper).

We thank the reviewer for pointing out the paper Zoller et al., – it had not been published upon our submission. In this revision we had included mention of the data in these two very recent publications, including a detailed description of the findings in Zoller et al., (please see Discussion and comments above).

CX3CR1 is not specific for microglia and is also expressed by circulating monocytes. There is, however, one marker Tmem119 that has shown to be specific, which could provide "cleaner" results. A discussion of this fact would be useful.

Tmem119 has been described as a marker that is expressed in parenchymal microglia in the steady state, and not by non-CNS macrophages or by circulating monocytes. However, with the ablation of microglial TGFBR2, there is a downregulation of Tmem119 expression in microglia on a mRNA and protein level that makes long-term tracking of these cells difficult in this context (now Figure 4E). To address the reviewer’s suggestion however, we have performed additional experiments with another microglia-specific marker, P2RY12 (experiments described under “Essential revisions”) as part of this revision that provide further support to the notion that altered IBA1+ cells in the retina following TGFBR2 ablation originate from resident retinal microglia (that express the specific microglial marker, P2RY12), rather than from circulating monocytes.

It is surprising that in spite of the demonstrated upregulation of proinflammatory cytokines, such as interferons and IL1β, by the activated microglial, infiltrating macrophages were not detected. The authors might speculate on why this is.

It is not completely clear why circulating monocytes following TGFBR2 ablation do not infiltrate into the retina despite the upregulation of microglial activation factors. This however appears to be a finding that is also observed in the brain and spinal cord. We do not know with certainty the reasons why this is the case but we are happy to speculate. One possibility that is that infiltration of monocytes in various contexts are accompanied by alterations/breakdown in the blood-retina (or blood-brain) barrier, a feature that not evident the situation here. Another factor is the availability of an empty myeloid cell niche within the retina. In our recently published data (Zhang et al., 2018), we find that microglial repopulation in the retina following depletion also involve little monocytic infiltration, as endogenous microglia appear to proliferate sufficiently and rapidly enough to fill up the vacated niches. In our other work (Ma et al., 2017), we found that rapid redistribution of microglia from the inner to the outer retina in response to RPE injury creates empty niches in the inner retina which are then filled by infiltrating monocytes. It appears that there is a strong tendency of the retina to maintain myeloid cell homeostasis – replacement cells may be drawn from infiltrating monocytes if there is (1) an empty niche that is insufficiently filled by endogenous microglial proliferation and migration (Lund et al., 2018b), and/or (2) a viable route of vascular entry (i.e. such as through a compromised blood-CNS barrier). In TGFBR2 ablation, endogenous microglia proliferate and increase in number without presenting an obvious empty niche and without accompanying blood-retinal barrier breakdown. These factors may have limited the ability of monocytes to infiltrate the retina in this context. We have added to the Discussion:

“The reasons for why monocytic infiltration did not occur in this context are unclear and may be related to the absence of evident breakdown of the blood-retinal barrier, despite increased microglial activation, and the absence of an empty myeloid cell niche ready to accommodate infiltrating monocytes (Lund et al., 2018b). “(Discussion, p17).

The authors conclude that the "constitutive neuron-microglia interactions in the form of TGFβ are necessary in the maintenance of the orderly organization and trophic function of the microglial in the retina". What is the evidence that neurons are providing the TGFβ, how is it activated and which isoform of TGFβ is mediating the homeostatic suppression of microglial activation?

We appreciate the reviewer’s insightful question. While we have given it deep consideration, we may not be able to give it a full discussion in our manuscript. Here are some of our thoughts on these points:

On the topic of TGFβ ligand expression in the retina on the mRNA level, mRNA profiling of sorted retinal cell types in the adult mouse retina (Siegert et al., 2012) reveals that among the 3 Tgfb ligands, Tgfb1 is primarily expressed by retinal microglia,Tgfb2 is primarily expressed by a variety of retinal neurons, including photoreceptors, bipolar cells, amacrine cells, and retinal ganglion cells, and Tgfb3 is expressed by amacrine neurons (see Author response image 2).

In addition, mRNA expression of Tgfb1 and Tgfb2 has been also attributed to Müller cells in culture (PMID: 9617551). On a protein level, TGFB2 has been found at higher levels in the monkey neural retina than TGFB1 on ELISA, with TGFB2 localizing to photoreceptors on immunohistochemistry (PMID: 7821377), but TGFB1 has also been localized to photoreceptors by immunohistochemistry in the human retina (Lutty et al., 1991).Immunohistochemical analyses of TGFB1, TGFB2, and TGFB3 have also been examined across different species (monkey, human, and cat) by Anderson et al., (PMID: Anderson et al., 1995) with overlapping patterns of immunopositivity in RPE cells, photoreceptors, Mueller cells, ganglion cells, hyalocytes, and vascular cells. In the context of glaucoma, TGFB2 has been located to astrocytes in the optic nerve head (PMID: 10396201) and TGFB1 to activated microglia there (PMID: 11391707).

Taken together, there is evidence that retinal neurons are a prominent source of TGFB2, and possibly some TGFB1 and TGFB3, while retinal microglia appear to express primarily TGFB1. While the details regarding the action of specific neuronal cell-type to microglia signaling via specific TGFb isoforms are not yet worked out, there is good evidence that retinal neurons are a prominent source of TGFb ligands (particularly TGFB2), with the possibility of microglia providing autocrine signaling via TGFB1. While we do not enter into these details in the manuscript, we have used the schematic in Figure 10 to display these likely relationships.

We have provided the following in the Discussion to reference the work examining the localization of TGFB ligands in the retina:

“TGFβ ligands, TGFβ1, TGFβ2, and TGFβ3, are expressed by multiple retinal cell types, including different classes of retinal neurons, endothelial cells, RPE cells, and retinal microglia (Anderson et al., 1995; Close et al., 2005; Lutty et al., 1993). In particular, Tgfb2 and Tgfb3 mRNA have been detected in amacrine, bipolar, and retinal ganglion cells (Siegert et al., 2012), and TGFB2 protein localized to photoreceptors (Lutty et al., 1991).”

The title is misleading. The data show that the absence of TGFβ signaling in microglial leads to exacerbated choroidal neovascularization, but the title suggests that this initiates the vessel growth.

This has been amended to read “Pathological transformation of retinal microglia in the absence of constitutive TGFβ signaling induces retinal degeneration and exacerbates choroidal neovascularization” as suggested by the reviewer.

Reviewer #3:

[…] In many ways this paper is quite descriptive as there is no explanation for the neuronal loss following macrophage loss of TGFbR2. Nonetheless, the paper adds important new general information about macrophage neuronal interactions and TGFb control of macrophage function and related neuronal survival. I expect that with the techniques now available to look at multiple pathways the need to prove in detail that one pathway is involved after a genetic change in a cell will become less important.

We appreciate the reviewer’s comments and encouragements. We concur that there are likely multiple pathways that can contribute to microglial transitions from a healthy supportive state to one that contributes to pathological neurodegeneration. Even so, examining the specific contribution of TGF-b signaling to this transition is instructive and potentially valuable; because TGF-b has been highlighted in GWAS study as a contributor to AMD risk and has been put forward as a pathway to be targeted for AMD treatment, it may be insightful to discover further how this pathway connects to the physiology of retinal microglia.

We do not yet fully understand how neuronal loss in the retina occurs following microglial TGFBR2 ablation. We do not think that neurodegeneration arises from an insufficiency of microglial function, as prolonged microglial depletion in the retina, while resulting in synaptic degeneration, does not result in retinal thinning and atrophy (PMID: 28235894). As such, we hypothesize that neurodegeneration results from the acquisition of aberrant microglial functions that then induce neuronal cell death. We speculate that the secondary induction of Muller cell gliosis, which acquire a neurotoxic A2-like astrocytic signature (Liddelow et al., 2017), may contribute to the death of neurons.

Specific comments:

Figure 1B: need to label y axis as Tgfbr2 and reference the control. Check all figures for more complete labeling with units and specific mRNAs etc.

The data in Figure 1B is referenced to Siegert et al., 2012 and additional information concerning the definition of expression level units and the linearity of the scale may be found in the originating paper. In the rest of the figures involving mRNA quantification that we have performed, we have checked to ensure that we consistently express it as a normalized value relative to the control group in the experiment.

https://doi.org/10.7554/eLife.42049.027

Article and author information

Author details

  1. Wenxin Ma

    Unit on Neuron-Glia Interactions in Retinal Disease, National Eye Institute, National Institutes of Health, Bethesda, United States
    Contribution
    Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Writing—original draft, Writing—review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-8396-6625
  2. Sean M Silverman

    Unit on Neuron-Glia Interactions in Retinal Disease, National Eye Institute, National Institutes of Health, Bethesda, United States
    Contribution
    Investigation, Writing—original draft, Writing—review and editing
    Competing interests
    No competing interests declared
  3. Lian Zhao

    Unit on Neuron-Glia Interactions in Retinal Disease, National Eye Institute, National Institutes of Health, Bethesda, United States
    Contribution
    Investigation, Writing—original draft, Writing—review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-0120-1969
  4. Rafael Villasmil

    Flow Cytometry Core Facility, National Eye Institute, National Institutes of Health, Bethesda, United States
    Contribution
    Investigation, Writing—original draft, Writing—review and editing
    Competing interests
    No competing interests declared
  5. Maria M Campos

    Section on Histopathology, National Eye Institute, National Institutes of Health, Bethesda, United States
    Contribution
    Investigation, Writing—original draft, Writing—review and editing
    Competing interests
    No competing interests declared
  6. Juan Amaral

    Unit on Ocular Stem Cell and Translational Research, National Eye Institute, National Institutes of Health, Bethesda, United States
    Contribution
    Investigation, Writing—original draft, Writing—review and editing
    Competing interests
    No competing interests declared
  7. Wai T Wong

    Unit on Neuron-Glia Interactions in Retinal Disease, National Eye Institute, National Institutes of Health, Bethesda, United States
    Contribution
    Conceptualization, Resources, Data curation, Formal analysis, Supervision, Funding acquisition, Methodology, Writing—original draft, Project administration, Writing—review and editing
    For correspondence
    wongw@nei.nih.gov
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-0681-4016

Funding

National Eye Institute (Intramural Research Program)

  • Wai T Wong

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

This study is supported by funds from the National Eye Institute Intramural Research Program.

Ethics

Animal experimentation: This study was performed in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. All of the animals were handled according to approved institutional animal care and use committee (IACUC) protocols (NEI-602, NEI-665) of the National Eye Institute.

Senior Editor

  1. Gary L Westbrook, Vollum Institute, United States

Reviewing Editor

  1. Constance L Cepko, Harvard Medical School, United States

Publication history

  1. Received: September 15, 2018
  2. Accepted: January 2, 2019
  3. Version of Record published: January 22, 2019 (version 1)

Copyright

This is an open-access article, free of all copyright, and may be freely reproduced, distributed, transmitted, modified, built upon, or otherwise used by anyone for any lawful purpose. The work is made available under the Creative Commons CC0 public domain dedication.

Metrics

  • 1,652
    Page views
  • 258
    Downloads
  • 4
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, Scopus, PubMed Central.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

Further reading

    1. Immunology and Inflammation
    David George Saliba et al.
    Research Article
    1. Immunology and Inflammation
    Eric E Irons et al.
    Research Article Updated