Sortilin Inhibition Protects Neurons From Degeneration in the Diabetic Retina

Abstract

Purpose: To investigate the level and localization of the multifunctional receptor sortilin in the diabetic retina, as well as the effect of sortilin inhibition on retinal neurodegeneration in experimental diabetes.

Methods: The localization of sortilin and colocalization with the p75 neurotrophin receptor (p75^NTR) and Müller cell (MC) markers were determined using immunofluorescence on retinal sections from human patients with diabetes and streptozotocin-induced diabetic C57BL/6J male mice. In the diabetic mice, levels were further quantified using Western blot and quantitative PCR. Therapeutic studies were performed on diabetic mice using intravitreally injected anti-sortilin antibodies. Neuroprotection was evaluated in vivo by optical coherence tomography and by quantification of retinal ganglion cells (RGCs) in flat mounts.

Results: Increased levels of sortilin were observed in human and murine diabetic retinas compared with nondiabetic control retinas. Sortilin was highly localized to retinal MCs, and, notably, colocalization with p75^NTR was only seen in diabetic retinas. A remarkable protective effect of sortilin inhibition on inner retinal cells was observed in diabetic mice. At eight weeks after diabetes induction, inner retinal thickness was reduced by 9.7% (−12.7%, −6.6%; P < 0.0001; n = 11−12) in the PBS-injected control group compared with the anti-sortilin injected group. Similarly, the count of RGCs was reduced by 20.5% (−30.8%, −10.2%; P = 0.0009) in the PBS-injected control group compared with the anti-sortilin–injected group.

Conclusions: Sortilin is upregulated in the diabetic retina, and sortilin inhibition effectively protects against neuronal loss. Thus sortilin emerges as a novel pharmacological target in diabetic retinal neurodegeneration—an important early event in the pathogenesis of diabetic retinopathy.

Diabetic retinopathy (DR) is a leading cause of visual impairment in the Western world.¹ The underlying pathology is complex² and remains to be elucidated in detail. The hallmark of DR is changes in the microvasculature with loss of pericytes and formation of microaneurysms as early events.^3,4 However, already decades ago, it was compellingly argued also to consider DR as a neurosensory disorder.⁵ Mounting evidence now supports that degeneration and dysfunction of retinal neurons occur in the course of DR and contribute to its development.⁶ Interestingly, dysfunction and loss of neuronal cells seem to be early events that may precede the vascular lesions characterizing the disease.^7,8 

Sortilin⁹ is a multifunctional receptor belonging to the Vps10p domain receptor family. In addition to its important role as a sorting receptor, sortilin has emerged as an essential regulator of neurotrophin signaling.¹⁰ Neurotrophins are growth factors essential for neuronal development, integrity, and functionality. These growth factors are commonly secreted as pro-neurotrophins that as opposed to their mature counterparts are proapoptotic.¹¹ As an example, the mature nerve growth factor (NGF) targets the tyrosine receptor kinase A in conjunction with the p75 neurotrophin receptor (p75^NTR) for trophic signaling, whereas the immature pro-nerve growth factor (proNGF) binds with high affinity to a receptor complex comprising sortilin and p75^NTR for induction of proapoptotic¹⁰ or proinflammatory¹² pathways. Without the presence of sortilin, the proNGF/p75^NTR interaction is of low affinity, and there is no proapoptotic signaling because the p75^NTR/sortilin complex is not formed.^9,10 ProNGF-induced apoptotic signaling through the p75^NTR/sortilin complex has been shown to play a role in several neurodegenerative pathologies.^13,14 In addition, the pleiotropic effect of sortilin as a sorting receptor has implicated it in diverse pathologies such as cardiovascular disease¹⁵ and carcinogenesis.¹⁶ 

The potential role of sortilin in DR has not been investigated. However, diabetes mellitus (DM) has been shown to cause an imbalance of proNGF and NGF in the human retina,¹⁷ and animal studies have implicated disturbed neurotrophin signaling in retinal ganglion cell (RGC) degeneration in DR,^18,19 optic nerve lesions and glaucoma²⁰ through proNGF/p75^NTR-mediated mechanisms. In addition, proNGF itself induces RGC death.¹² Furthermore, postmitotic RGCs express sortilin^13,21 and p75^NTR13,22 when programmed RGC death peaks in the developing mouse retina. 

Thus we hypothesized that sortilin is upregulated in the diabetic retina and involved in the diabetes-induced loss of RGCs: The localization of sortilin and p75^NTR was studied by immunofluorescence (IF) in retinal tissue obtained postmortem from six human donors with DR and five controls. Subsequently, sortilin expression and localization were evaluated in the retina of streptozotocin (STZ)-induced diabetic mice. Therapeutic studies using intravitreally (IVT)-injected anti-sortilin polyclonal blocking antibodies in diabetic mice were used to evaluate the neuroprotective effect of sortilin inhibition. 

Methods

Human Tissues

Six eyes from six patients with a verified history of DR and five eyes from five normal controls were studied (Table). The eyes had been selected among eyes obtained by autopsy between 1991 and 1998 to study vascular changes in DR.²³ The retinal vascular system had been cast using the technique previously described.^24,25 The eyes had been divided by horizontal section through the fovea and the optic disk, and one half had been embedded in paraffin as previously described.²⁶ Clinical data at the time of casting are shown in the Table. The study using the previously collected tissue material was approved by the Regional Committee for Scientific Ethics, approval 1-10-72-299-12. 

Table.

 

Relevant Data of the Eyes Included in This Study Obtained at the Time of Examination

 

Immunofluorescence in Human Retinal Sections

Tissue sections were cut 2 µm thick from the equator region of the eye and were incubated at 60°C for 60 minutes. Sections were deparaffinized in xylene overnight (ON) before rehydration with graded ethanol washes. Antigens were retrieved by incubation in Tris-EDTA buffer in a microwave oven for three minutes (maximum temperature 60°C), left to cool for >30 minutes at room temperature on a rocking table, and washed in PBS. Unspecific binding was reduced by incubation in 2% BSA in PBS for 10 minutes. The sections were incubated with the primary antibodies (see Supplementary Table S1) in 1% BSA in PBS ON at 4°C. The following day, sections were washed in PBS before incubation with the secondary antibodies (see Supplementary Table S1) in 1% BSA in PBS at room temperature for 30 minutes. Cell nuclei were stained with DAPI (Sigma-Aldrich, St. Louis, MO, USA), washed in water, dipped in 70% ethanol, and left to dry. Cover glasses were mounted using ProLong Gold Antifade mounting medium (ThermoFisher Scientific, Waltham, MA, USA). A control stain for smooth muscle actin (SMA), a protein found in retinal vessels, was used to assess the quality of the tissue (Supplementary Fig. S3). Negative controls were performed, where the primary antibody was omitted (Supplementary Fig. S4A). 

Fluorescence staining and colocalization were examined using a Zeiss LSM 710 confocal laser scanning microscope (Zeiss, Jena, Germany) fitted with a Plan-Apochromat 20 × 63 × 1.4 objective. Excitation and emission wavelengths for Alexa Fluor 488 were λ_ex = 488 nm and λ_em 525 nm. Correspondingly for TRITC: λ_ex = 532 nm and λ_em = 576 nm. SMA staining was visualized by fluorescence microscopy (Leitz DM RB; Leica Microsystems GmbH, Wetzlar, Germany) with a 40 × 1.0 objective, and images were captured with a Leica DFC 360 FX camera and associated software (Leica Application Suite, version 3). Images were captured using the same settings for Alexa Fluor 488 and Alexa Fluor 568; fluorescence intensities can be compared among images in the same figure. 

Quantification of Sortilin and p75^NTR Levels in the Human Retina

Vertical densitometric profile plots representing the mean fluorescent signal intensity across the retinas in each group were generated using the Plot Profile function of ImageJ²⁷ for both sortilin and p75^NTR using the DAPI nuclear stain to determine the location in the retina. Three linear plot profiles were analyzed for each retinal image, and the mean fluorescent signal intensity was calculated. In each of the retinal images, eight different retinal layers were identified: Nerve fiber layer (NFL), ganglion cell layer (GCL), inner plexiform layer (IPL), inner nuclear layer, outer plexiform layer, outer nuclear layer, outer limiting membrane (OLM), and inner and outer segments of the photoreceptors. For each layer, the mean fluorescent signal intensity was calculated. Total fluorescence intensity of sortilin and p75^NTR across the retina was calculated as the area under the linear plot profiles of the respective fluorescence intensities. For the SMA-stained sections, the mean fluorescent signal intensity from the NFL to the OLM were similarly determined using the Plot Profile function in ImageJ. 

Animals

The use of animals was approved by the Danish Animal Inspectorate (authorization no. 2012-15-2934-00113 and 2016-15-0201-00947). For diabetes induction, eight-week-old C57BL/6J male mice were purchased from Janvier Labs (Le Genest-Saint-Isle, France). Animals were maintained under a 12-/12-hour light/dark cycle at the Animal Facilities at the Department of Biomedicine, Aarhus University, Denmark with free access to autoclaved tap water and standard chow (Altromin, Brogaarden, Denmark). The animals were handled in accordance with the Association for Research in Vision and Ophthalmology Statement for the Use of Animals in Ophthalmic and Vision Research. 

Before intravitreal injections or optical coherence tomography (OCT), mice were anesthetized with an intraperitoneal injection of a mixture of ketamine and medetomidine hydrochloride (Ketador 60–100 mg/kg [Richter Pharma AG, Wels, Austria] and Cepetor/Domitor 0.5–1 mg/kg [ScanVet Animal Health A/S, Fredensborg, Denmark]). Pupils were dilated with a drop of 1% tropicamide (Mydriacyl; Alcon Nordic A/S, Copenhagen, Denmark). During anesthesia, the eyes were lubricated with carbomer eye gel (Viscotears 2 mg/mL, Alcon Nordic). Immediately after procedures, anesthesia was reversed by atipamezole 0.5–1 mg/kg (Antisedan, Orion Pharma, Copenhagen, Denmark), and animals were placed on a warming plate until they moved spontaneously. 

Induction of Diabetes

Diabetes was induced in eight-week-old C57BL/6J male mice by five intraperitoneal STZ injections on five consecutive days using doses of 55 mg/kg body weight in 10 mM citrate buffer.²⁸ Nondiabetic control mice were injected with citrate buffer only. Animals were fasted for four to six hours before the STZ injection. Body weight and blood glucose were measured weekly throughout the study. Blood glucose was measured in tail-capillary blood by Contour Next (Ascensia Diabetes Care, Copenhagen, Denmark). Animals were considered diabetic when blood glucose was >15 mmol/L. Mice were reinjected with two doses of 55 mg STZ/kg body weight intraperitoneally until diabetes was achieved. When blood glucose reached >15 mmol/L for all animals in the diabetic groups, the experimental period started. Animals with signs of illness or ketonuria (i.e., fast weight loss) were excluded. 

Intravitreal Injections of Anti-Sortilin Polyclonal Antibodies

At 2.5 weeks (T_2.5) after the onset of diabetes, intravitreal injections were performed under direct visualization with an operating microscope (OPMI 1 FR PRO; Zeiss). A 30G disposable needle was used to puncture the sclera near the limbus, and a 33G blunt-ended needle of a Hamilton syringe (Hamilton Company, Reno, NV, USA) was then inserted into the opening and used to inject either antibody or buffer solution: The left eye of 20 diabetic mice was injected intravitreally with 2 µL of 1 mg/mL goat anti-sortilin polyclonal antibody (AF2934, R&D Systems, Minneapolis, MN, USA) and the right eye with 2 µL PBS. Twenty eyes from 10 age-matched nondiabetic mice were used as noninjected controls. 

OCT Imaging

At 2.5, four, six, and eight weeks after the onset of DM, the Micron IV image-guided OCT 2 system (Phoenix Research Laboratories, Pleasanton, CA, USA) was used to acquire B-scans in the central retina at an equal distance from the optic disc. The average thickness of the NGI (NFL, GCL, and IPL) complex was determined by loading the OCT images into InSight software (Phoenix Research Laboratories) and manually segmenting the retinal layers. 

Western Blotting

After sacrifice at 20 weeks of diabetes, eyes were extracted, and the retinas were processed for Western blotting as described previously.^29,30 Membranes were incubated at 4°C ON with an anti-sortilin antibody or an anti-neuronal growth factor receptor antibody, and with an anti-glyceraldehyde-3-phosphate dehydrogenase antibody as loading control (see Supplementary Table S1). Visualization and quantification were done as described previously.²⁹ Complete Western blots are provided as supplementary material (Supplementary Fig. S6). 

RT-qPCR Analyses

After sacrifice at 20 weeks of diabetes, eyes were enucleated and dropped in ice-cold Hanks’ balanced salt solution buffer. Periocular tissue was removed, and the anterior segment and lens were discarded. The neuroretina was dropped in ice-cold lysis buffer (buffer RLT + 40 mM Dithiothreitol). The tissue lysate was loaded onto a QIAshredder for homogenization, and, subsequently, RNA was purified using RNeasy plus kit, according to protocol (QIAGEN, Hilden, Germany). The iScript cDNA synthesis kit (Bio-Rad, Hercules, CA, USA) was used for first-strand cDNA synthesis, with a total input RNA of 50 ng. PCR reactions were performed in triplicate on 1:5 diluted cDNA using Taqman Gene expression assays and analyzed on a LightCyclerW480 (Roche Diagnostics, Basel, Switzerland) according to Taqman protocols. Relative gene expression of sortilin (Mm00490905; ThermoFisher Scientific) and p75^NTR (Mm00446296; ThermoFisher Scientific) was calculated using the standard curve method and related relative to an endogenous control, the housekeeping gene hypoxanthine-guanine phosphoribosyl transferase (Mm00446968; ThermoFisher Scientific). 

Immunofluorescence and TUNEL Staining in Murine Retinal Sections

Eyes were enucleated and fixed in 4% paraformaldehyde ON at 4°C. Paraffin embedding and sectioning were performed at the Department of Pathology, Aarhus University Hospital. For immunofluorescence, sagittal sections of 3 µm were prepared, processed, and stained as described above. However, for anti-glial fibrillary acidic protein (GFAP) staining enzymatic retrieval by 10 minutes’ incubation with 20 µg/mL proteinase K solution (Promega, Madison, WI, USA) was performed according to manufacturer's instruction. The antibodies used are listed in Supplementary Table S1. Negative controls were performed, where the primary antibody was omitted (Supplementary Fig. S4B). TUNEL staining was performed using the DeadEnd Fluorometric TUNEL System (Promega). Pretreatment of the paraffin-embedded sections and TUNEL staining were performed according to the supplied protocol. 

Fluorescence signals were visualized by fluorescence microscopy (Leitz DM RB; Leica Microsystems GmbH) with a 16 × 0.5 or 40 × 1.0 objective and images captured with a Leica DFC 360 FX camera and associated software (Leica Application Suite, version 3). Images were captured using the same settings for Alexa Fluor 488 and Alexa Fluor 568; fluorescence intensities can be compared among images in the same figure. 

Quantification of Retinal Ganglion Cells in Flat Mounts

Eight weeks after the onset of diabetes, mice were sacrificed and eyes were enucleated and fixed in 4% paraformaldehyde for two hours at room temperature. The anterior segment and lens were removed and the neuroretina was carefully peeled off the retinal pigment epithelium (RPE)/choroid and immersed in PBS buffer. Retinas were permeabilized and blocked in 0.3% BSA and 1% Triton X-100 in PBS ON at 4°C on a shaker. Retinas were washed and stained with anti-Brn3a dissolved in 1% Triton X-100 in PBS buffer with 1 mM MgCl₂, 1 mM CaCl₂, 0.1 mM MnCl₂ ON at 4°C on a shaker before being stained with secondary antibody dissolved in 0.3% BSA and 1% Triton X-100 in PBS for two hours at RT. Retinas were washed and stained with DAPI. Flat mounts were transferred to Super-FrostPlus glass slides and mounted with the RGC layer facing upward using ProLong Gold antifade reagent (ThermoFisher Scientific). Images for assessment of RGC densities were acquired 1 mm peripheral to the OD by fluorescence microscopy (Leitz DM RB; Leica Microsystems) and captured with a Leica DFC 360 FX camera and associated software (Leica Application Suite version 3, Leica Microsystems). Unfortunately, the quality of the Brn3A stain (Supplementary Fig. S4C) did not allow for quantification but validated the layer in which the RGCs were quantified. The number of RGCs was counted manually as neuronal cell nuclei in the GCL. RGCs were counted in three images from each retina, and a mean was calculated. 

Cell Culture and Immunofluorescence

Mouse neuroblastoma N2A (American Type Culture Collection, Manassas, VA, USA), HEK293 (American Type Culture Collection) and SH-SY5Y (American Type Culture Collection) cells were cultured in Dulbecco's modified Eagle's medium with GlutaMAX (ThermoFisher Scientific) supplemented with 10% fetal calf serum (Sigma-Aldrich) and 1% penicillin/streptomycin (ThermoFisher Scientific). The human Müller cell line Moorfields/Institute of Ophthalmology-Müller 1 (MIO-M1)³¹ was obtained from the UCL Institute of Ophthalmology, London, UK and cultured in Dulbecco's modified Eagle's medium with high glucose, GlutaMAX and pyruvate (Gibco, ThermoFisher Scientific) supplemented with 10% fetal calf serum (Sigma-Aldrich) and 1% penicillin/streptomycin (ThermoFisher Scientific). All cell lines were maintained in T-75 cell culture flasks (Sarstedt, Nümbrecht, Germany) at 37°C and 5% CO₂. At near confluence, cells were seeded into SlideFlasks (ThermoFisher Scientific) and allowed to grow for 24 hours until 50% confluence and fixed using 10% neutral formalin. For transfection, after 24 hours in SlideFlasks, MIO-M1 cells were transfected with 1.5 µg plasmid DNA encoding sortilin (pAAV/apoE-hAAT-Sortilin) using XtremeGENE 9 (Roche) according to manufacturer's protocol and after 72 hours fixed using 10% neutral formalin. For the protein-blocking experiment, slides were washed in PBS followed by incubation for one hour at room temperature with primary antibody that had been preincubated for 30 minutes with or without recombinant murine sortilin (2934-ST; R&D Systems) in a 1:5 ratio. For sortilin immunofluorescence in HEK-293, MIO-M1 and SH-SY5Y, fixed slides were incubated in 70% ethanol followed by incubation with primary antibody for one hour at room temperature. Slides were washed in PBS and incubated with secondary antibody for one hour at room temperature. Finally, the slides were counterstained with DAPI before being mounted with cover glasses. Images were acquired by fluorescence microscopy (Leitz DM RB; Leica Microsystems) and captured with a Leica DFC 360 FX camera and associated software (Leica Application Suite version 3, Leica Microsystems). Fluorescence intensities can be compared among images in the same figure unless exposure times are stated. 

Single-Cell RNA-Seq Analysis

Processed single cell RNA sequencing (scRNA-seq) data from human³² and mouse³³ retina samples was downloaded from, respectively, the Human Cell Atlas Data Portal (WongAdultRetina) and GEO (GSM4649095). Analyses were performed using Automated Single Cell Analysis Platform VI.³⁴ The data were normalized using Seurat³⁵ and dimensional reduction performed with UMAP.³⁶ Clusters of cells were identified using Seurat³⁷ and manually annotated using cell-type marker genes from Lukowski et al.³² 

Statistical Analysis

Measurements are presented as mean ± SD. Data were evaluated for normality. Statistical differences between two group means were evaluated using a two-tailed Student t-test. Differences between multiple means were evaluated using ANOVA followed by post hoc comparisons using the Tukey method. A mean of the measurements from the two eyes of an animal were used if included in the same group. A significance level of α = 0.05 was used. Confidence intervals are presented accordingly. Statistical analyses were performed using the R statistical software package, version R-4.0.5 (https://www.r-project.org/). 

Results

Elevated Sortilin and p75^NTR Levels in Patients With Diabetic Retinopathy

Sortilin and p75^NTR were detectable in all human retinas examined using IF (Fig. 1A, Supplementary Fig. S1). Densitometric profile analysis of the fluorescent signals revealed significantly higher levels of sortilin and p75^NTR in the retinas of patients with DM compared with controls (Fig. 1B). In human diabetic retinas, the fluorescent sortilin signal was strongest in the NFL, whereas it predominated in the GCL in nondiabetic control retinas (Fig. 1C). The highest p75^NTR signal was observed in the OLM in diabetic retinas and in the NFL in nondiabetic control retinas (Fig. 1C). Notably, the sortilin and p75^NTR signal covaried across the different layers of the diabetic retinas, with particular high intensities in inner retinal layers and corresponding to the OLM. 

Figure 1.

 

Increased levels and colocalization of sortilin and p75^NTR in the human diabetic retina. Sortilin and p75^NTR levels in retinas from patients with DR compared with control retinas determined by immunofluorescence and confocal microscopy analysis. (A) Paraffin sections from patients with DR D1, D4, and D6, and controls C1, C2, and C5 labeled with anti-sortilin (red) and anti-p75^NTR (green). Colocalization is shown in the right column (yellow). Nuclei are stained with DAPI (blue). Scale bars: 20 µm. (B) Bar plots with mean ± SD of total fluorescence intensity of sortilin (red) and p75^NTR (green) in retinas from patients with DR and controls. (C) Mean fluorescence intensities of sortilin (red) and p75^NTR (green) in different retinal layers in eyes from patients with DR (D) compared with controls (C). Mean ± SD is shown. (D) Detail of the inner retina from patients D6 and C5. White arrows (D6) show sortilin at cell boundaries and protrusions, and gray arrows (C5) show sortilin in intracellular vesicles. Scale bars: 20 µm. INL, inner nuclear layer; OPL, outer plexiform layer; ONL, outer nuclear layer; PR, inner and outer segments of the photoreceptors; D, patients with DM; C, controls; AU, arbitrary units. Significance levels: * [0.01, 0.05]; ** [0.001, 0.01]; *** [0.0001, 0.001]; **** [0, 0.0001].

Sortilin Colocalizes With p75^NTR and GFAP in Patients With Diabetic Retinopathy

In controls without DM, sortilin was predominately localized in the inner retina but was observed across retinal layers (Fig. 1A, Supplementary Fig. S1). Characteristically, sortilin showed a staining pattern compatible with localization to intracellular vesicles in retinal cells (Fig. 1D). In most cases, the p75^NTR signal was low, and no colocalization was observed in any part of the nondiabetic retinas. In eyes from patients with DM, a clear shift in sortilin and p75^NTR signal was observed. Both the sortilin and p75^NTR signal were increased, and the signal was colocalized to thread-like structures spanning the retina (Fig. 1A, Supplementary Fig. S1). The sortilin signal was no longer limited to intracellular structures but was predominantly localized at cell boundaries and protrusions (Fig. 1D). 

The localization pattern of sortilin and p75^NTR in the diabetic retinas was highly suggestive of expression in retinal Müller cells (MC). Co-immunostaining for a marker of reactive MCs, GFAP, and sortilin revealed a high degree of colocalization in the diabetic retinas (Fig. 2A, Supplementary Fig. S2). Noticeably, the colocalization signal spanned across the retina, but was most pronounced in the inner retina (Figs. 2A, 2B, Supplementary Fig. S2). The GFAP signal is expectedly lower in the nondiabetic retina. Thus additional co-immunostaining of sortilin and the MC-marker glutamine synthetase (GS) was performed in sections from controls without DM. Limited colocalization was observed in the inner retina (Fig. 2C). Thus sortilin levels increase in MCs in parallel with their development of a reactive phenotype during the development of DR. 

Figure 2.

 

Colocalization of sortilin and a Müller cell marker in human diabetic retinas. Localization of GFAP and sortilin in retinas from patients with DR compared with control retinas. (A) Paraffin sections of retinas from patients with DR (D1 and D6) and a control (C5) co-labeled with anti-sortilin (red), anti-GFAP (green), and DAPI (blue). Colocalization is shown in the right column (yellow). (B) High magnification of the inner retina from patients D6 and C5. Scale bars: 20 µm. (C) Colocalization of sortilin and GS in control retina. Paraffin section from control C1 labeled with anti-sortilin (red), anti-GS (green), and DAPI. Colocalization is shown in the right column (yellow); example is marked with white arrowhead. Scale bars: 20 µm. INL, inner nuclear layer; ONL, outer nuclear layer.

SMA was used as a positive control stain in all diabetic (D1-D6) and control (C1-C5) human retinas (Supplementary Fig. S3). No significant difference in background intensity was observed between the two groups (Supplementary Fig. S3). The negative controls revealed an autofluorescence signal corresponding to the RPE (Supplementary Fig. S4A). IF staining for sortilin was performed in cultured cells (Supplementary Fig. S5). As expected, sortilin was identified across the examined cell types including the human MC-line MIO-M1. 

Sortilin and p75^NTR are Upregulated in the Retinas of STZ-Induced Diabetic Mice

Sortilin and p75^NTR protein and mRNA levels were determined at 20 weeks after diabetes induction with STZ. Total protein levels of sortilin were increased by 58% (12%, 103%; P = 0.02; n = 5-6), and the total protein levels of p75^NTR were increased by 138% (87%, 188%; P = 0.0004; n = 5-6) in diabetic murine retinas compared with nondiabetic retinas (Figs. 3A–D, Supplementary Fig. S6). A similar pattern was found on the mRNA level, where sortilin and p75^NTR mRNA expression were increased by 84% (26%, 142%; P = 0.012; n = 4) and 194% (78%, 309%; P = 0.009; n = 4), respectively (Figs. 3E, 3F). Increased levels of sortilin and p75^NTR were also observed using IF in retinal sections at 20 weeks (Fig. 4A). 

Figure 3.

 

Sortilin and p75^NTR mRNA and protein levels are elevated in the eyes of diabetic mice. (A, B) Western blots of retinas from STZ-induced diabetic mice 20 weeks after the onset of DM and nondiabetic control mice. Bands corresponding to sortilin and p75^NTR compared with GAPDH are shown. (C, D) Corresponding bar plots with mean ± SD sortilin and p75^NTR protein levels compared with GAPDH from retinas of diabetic mice and control mice. (E) Bar plot with mean ± SD sortilin mRNA levels relative to HPRT evaluated by RT-qPCR. (F) Bar plot with mean ± SD p75^NTR mRNA levels relative to HPRT evaluated by RT-qPCR. DM, Diabetes mellitus; GAPDH, glyceraldehyde-3-phosphate dehydrogenase; HPRT, hypoxanthine-guanine phosphoribosyl transferase. Significance levels: * [0.01, 0.05]; ** [0.001, 0.01]; *** [0.0001, 0.001]; **** [0, 0.0001].

Figure 4.

 

Elevated levels of sortilin colocalize with p75^NTR and a Müller cell marker in the retinas of diabetic mice. (A) Sortilin levels are increased and colocalize with p75^NTR in the inner retina of diabetic mice. Paraffin sections from STZ-induced diabetic mice 20 weeks after the onset of diabetes and nondiabetic controls labeled with anti-sortilin (red) and anti-p75^NTR (green). (B) Sortilin colocalizes with the Müller cell marker GS in diabetic retinas. Paraffin sections from mice after 20 weeks of diabetes and control mice labeled with anti-sortilin (red) and anti-GS (green). Colocalization is shown in the second rightmost column (yellow). Inserts to the right show magnification of the merged images. Examples of colocalization are indicated by white arrowheads. Scale bars: 20 µm. INL, inner nuclear layer; ONL, outer nuclear layer.

Sortilin is Expressed in MC in the Diabetic Murine Retina and Colocalizes With p75^NTR

The localization of sortilin and p75^NTR was evaluated by IF in retinal sections (Fig. 4A) from mice sacrificed at 20 weeks after diabetes induction. Sortilin and p75^NTR were detectable in all murine retinas examined. Sortilin immunostaining was most prominent in the inner retina but spanned the entire retina (Fig. 4A). This pattern is comparable to the pattern observed in the human sections. Prominent sortilin and p75^NTR colocalization was observed in the retina of diabetic mice, particularly in the inner retina (Fig. 4A, insert). The MC marker GS colocalized with sortilin. The colocalization was most evident in diabetic retinas with increased expression of sortilin (Fig. 4B, insert). The negative controls revealed an intense autofluorescence signal in photoreceptor outer segments and the RPE (Supplementary Fig. S4B). 

NGFR/Ngfr are Differentially Expressed Across Retinal Cell Populations in the Normal Human and Murine Retina

Analysis of scRNA-seq data from a transcriptome atlas of the adult human retina³² revealed almost uniform expression levels of sortilin (SORT1) across retinal cell populations. In contrast, p75^NTR (NGFR) showed more varying expression with particular high expression levels in the MC population (Supplementary Fig. S7A). Similarly, analysis of scRNA-seq from the adult murine retina³³ revealed expression levels of sortilin (Sort1) across retinal cell populations whereas p75^NTR (Ngfr) showed high and relative uniform expression in MCs and to a lesser extent subpopulations of rods and bipolar cells (Supplementary Fig. S7B). 

Intravitreal Administration of Anti-Sortilin Antibodies Preserves Retinal Structure and Reduces RGC Death in Diabetic Mice

The effect of sortilin inhibition on diabetic retinal neurodegeneration was subsequently evaluated in STZ-induced diabetic mice (blood glucose and weight are provided in Supplementary Fig. S8C). Anti-sortilin antibodies or PBS were injected IVT 2.5 weeks (T_2.5) after induction of DM (Fig. 5A). The ability of the antibody to neutralize sortilin was validated by a protein-blocking experiment in N2A cells; incubation of the anti-sortilin antibody with recombinant murine sortilin completely blocked sortilin immunostaining (Supplementary Fig. S9). No obvious group differences in sortilin levels using IF (Supplementary Fig. S10) were evident at eight weeks (T₈). This suggests the primary action to result from ligand blocking rather than a general reduction of sortilin levels. 

As a common marker of glial reactivity, GFAP immunostaining was performed on retinal sections (n = 4) from eyes harvested at T₈. An innermost retinal signal corresponding to retinal astrocytes was observed in all sections. However, an increased GFAP signal in the vertically oriented MC was evident in PBS-injected diabetic mice compared with both the anti-sortilin injected diabetic and the nondiabetic mice (Fig. 5B). 

Figure 5.

 

Intravitreal delivery of anti-sortilin antibody preserves retinal structure and retinal ganglion cells in diabetic mice. (A) Timeline of the experiment. DM was induced with five intraperitoneal injections of STZ. At 2.5 weeks after onset of DM (blood glucose > 15 mmol/L) mice were injected IVT with anti-sortilin polyclonal antibodies and PBS in opposite eyes. (B) Representative images of GFAP (red) and DAPI (blue) labelled paraffin-embedded retinal sections at eight weeks after onset of DM. Scale bar: 20 µm. (C) Representative OCT scans from nondiabetic control mice, IVT PBS-injected, and IVT anti-sortilin antibody-injected eyes at different time points. The white bar marks the combined thickness of the nerve fiber, retinal ganglion cell, and inner plexiform layer (NGI-thickness). (D) Box and dot plots of NGI-thickness; gray lines connect paired data points (i.e., eyes from the same animal). Significant comparisons between the groups are presented. (E) Representative images of DAPI stained retinal flat mounts. Scale bar: 50 µm. (F) Bar plots with mean ± SD of RGC density in retinal flat mounts from control mice, IVT PBS-injected and IVT anti-sortilin pAb-injected eyes. pAb, polyclonal antibody; INL, inner nuclear layer; ONL, outer nuclear layer. Significance levels: * [0.01, 0.05]; ** [0.001,0.01]; *** [0.0001, 0.001]; **** [0, 0.0001].

Mice were non-invasively monitored with OCT until sacrifice at T₈ (Fig. 5A). Four weeks after onset of diabetes, inner retinal thickness as quantified by the combined thickness of the NFL, GCL, and inner IPL (i.e., the NGI) (Figs. 5C, 5D) was significantly reduced by 5.0 µm (−7.7, −2.3; P = 0.00053; n = 8) in the eyes of diabetic mice injected IVT with PBS compared with the eyes injected with anti-sortilin antibodies. At six weeks, a reduction by 6.7 µm (−10.0, −3.4; P = 0.00055; n = 4-5) was observed. The protective effect of anti-sortilin antibodies on NGI thickness was maintained until sacrifice at T₈ (i.e., for at least 5.5 weeks). At this point, the NGI of diabetic anti-sortilin antibody IVT-injected mice measured 66.6 ± 1.66 µm compared with 60.1 ± 2.61 µm in diabetic PBS IVT-injected eyes, corresponding to a reduction by 6.4 µm (−8.5, −4.4; P < 0.0001; n = 11−12). From four weeks, highly significant differences comparable to those reported in the literature^19,38,39 were observed between the PBS-injected group and the nondiabetic group. On the other hand, the NGI thickness in anti-sortilin antibody-injected eyes was comparable to nondiabetic mice with no significant difference at any time point further demonstrating the degree of rescue. IVT injection of anti-sortilin antibodies was well tolerated, as demonstrated further by fundus imaging (Supplementary Fig. S8A). No difference in outer nuclear layer thickness was observed between the treated group and the control group (P = 0.524; n = 8) (Supplementary Fig. S8B). 

After sacrifice at T₈ (i.e., 5.5 weeks after treatment), RGCs were counted on retinal flat mounts (Figs. 5E, 5F). A reduction in the number of RGCs of 23.9% (−33.8%, −14.0%; P = 0.0002; n = 4) was found in the PBS-injected diabetic eyes compared with nondiabetic control mice, whereas the 4.3% (−14.2%, 5.6%) reduction in the anti-sortilin antibody-injected eyes was insignificant compared with nondiabetic control mice (P = 0.47; n = 4). The increased loss of RGCs in PBS-injected diabetic eyes compared with anti-sortilin injected diabetic eyes amounted to 20.5% (−30.8%, −10.2%; P = 0.0009). TUNEL-stained retinal sections of eyes harvested at T8 contained very limited number of positive cells (Supplementary Fig. S11), suggesting that the apoptotic process occurred at an earlier stage. 

Discussion

The multifunctional receptor sortilin has been investigated comprehensively, especially in the nervous system. Sortilin is believed to play a pivotal role in major human diseases involving neurodegeneration, but its role in the human eye remains elusive. In this study, immunostaining for sortilin in six patients with DR and five controls revealed (1) higher levels of sortilin in the diabetic retina, (2) sortilin colocalization with GFAP, (3) sortilin levels covarying with GFAP and p75^NTR, and (4) sortilin colocalization with p75^NTR in diabetic but not in nondiabetic retinas. Similar findings were demonstrated in diabetic murine retinas. Interestingly, the delivery of anti-sortilin antibodies rescued STZ-induced diabetic mice from inner retinal neurodegeneration. The almost-complete preservation of inner retinal structure with concomitant reduction of RGC death after sortilin-inhibition is highly suggestive of an indispensable role of sortilin in the neurodegenerative pathology of DR. 

P75^NTR and its co-receptor sortilin are abundantly expressed in the brain and retina during development, but the expression is limited in healthy, adult tissue.¹³ In the nondiabetic human and murine retina, we similarly found low levels of both receptors compared with the diabetic retina. The intracellular localization of sortilin without colocalization with p75^NTR is consistent with findings in the postnatal mouse retina where sortilin was detected in the Golgi apparatus, and no double-labeling with p75^NTR was seen at postnatal day 10.²¹ Moreover, cell studies have shown that the majority of sortilin (∼90%) is found in the Golgi apparatus and vesicles, leaving only a minor fraction at the surface of the cells.⁴⁰ 

Sortilin has been implicated in light-induced photoreceptor death in adult rodents⁴¹ and retinal neuron death following elevated intraocular pressure-induced ischemia in rats.⁴² In these studies, retinal sortilin and p75^NTR levels were both increased. However, in a study with STZ-induced diabetic rats, increased p75^NTR expression was not associated with significantly different sortilin levels after 4-5 weeks of diabetes.⁴³ In contrast, we demonstrated significantly increased retinal sortilin and p75^NTR levels in both human patients with DM and STZ-induced diabetic mice at 20 weeks of diabetes. This difference could be explained by retinal sortilin accumulation during the course of DM. 

The retinal localization pattern of neurotrophins and their receptors is not completely delineated.⁴⁴ In the adult rodent retina, p75^NTR is predominately expressed in MCs,^{12, 45-47} as well as in photoreceptor cells⁴¹ and vascular cells.⁴⁸ Notably, the expression in RGCs is debated.⁴⁴ Sortilin expression has been reported in RGC,^41,42,46 astrocytes,^41,46 MCs⁴⁶ and photoreceptor cells.⁴¹ In our study, sortilin similarly localized intracellularly across cell types in control retinas, while p75^NTR showed a much more restricted distribution. Interestingly, similar findings were observed at the mRNA level when analyzing retinal scRNA-seq data of human and murine origin. Both sortilin and p75^NTR localized to GFAP/GS positive cells corresponding to MCs in both the human and murine retina, and the diabetes-induced accumulation of sortilin and p75^NTR was most pronounced in MC projections enveloping RGCs. Interestingly, the levels covaried with GFAP which is increased in activated MCs developing a gliotic phenotype in response to retinal insults.^49,50 

In retinal insult models, p75^NTR is overexpressed in MCs and not the cells that undergo apoptosis. Accordingly, RGC death through paracrine mechanisms has been demonstrated after proNGF stimulation of retinal glial cells in the ischemic retina,⁴⁶ models of ocular proNGF overexpression,¹² optic nerve lesions and glaucoma,²⁰ and in animal models of STZ-induced diabetes.³⁹ As an example, IVT injections of proNGF induced robust MC expression of TNFα in the murine eye, and the consequent RGC death was abolished in p75^NTR and sortilin knockout mice.¹² The observed colocalization of sortilin and p75^NTR on MC projections suggests that the effects of sortilin are partly indirect by regulating signaling through p75^NTR on MCs. This is further substantiated by the observation of reduced glial reactivity evaluated by GFAP IF after sortilin inhibition. The loss of pericytes and endothelial cells in DR may involve similar paracrine mechanisms or direct actions of proneurotrophins on the p75^NTR/sortilin complex.⁴⁸ MCs are essential for maintaining retinal homeostasis,⁴⁹ and the dysfunction of MCs that are intercalated between the vasculature and neurons probably forms an important link between neuronal dysfunction and microvascular changes in the progression of DR.⁵¹ Indeed, the vascular occlusions in DR have been suggested to occur because of the invasion of vessels by gliotic MCs based on serial histological sections in our cohort of patients with DM.^25,26 Although proneurotrophin signaling through the p75^NTR/sortilin complex is the most obvious cause of the observed pathology, other sortilin-dependent mechanism might be involved. The cellular sorting of neurotrophic and neurotoxic factors and their receptors may be affected. As examples, sortilin is known to regulate the release of cytokines⁵² and exosomes⁵³ in certain contexts. 

We observed a remarkable neuroprotective effect after IVT injection of anti-sortilin antibodies in diabetic mice. Targeting proNGF and p75^NTR has similarly enabled the rescue of RGCs in rodent models of DM,^19,39,54 optic nerve lesions,²⁰ and ocular hypertension.²⁰ Our findings suggest sortilin as a target for retinal neuroprotection and support the evidence for disturbed neurotrophin signaling in retinal disease, in particular as an early event in DR. Sortilin inhibition would protect retinal cells from proneurotrophin signaling through the p75^NTR/sortilin complex while enabling unbound proneurotrophins to be converted into mature forms for trophic support. Importantly, this will not be the case when blocking p75^NTR, which is crucial for trophic support by interacting with specific tyrosine receptor kinase-receptors.⁵⁵ The rodent model used only partially recapitulates the clinical entity of DR and within a limited time frame. Thus the further elucidation of sortilin action in the diabetic retina requires an integrated view of the interactions between vascular and nonvascular cells, as well as the neuroglial interactions. Nonetheless, our findings strongly suggest that sortilin and its interaction with p75^NTR is a target in DR and potentially other neurodegenerative pathologies of the retina. 

Acknowledgments

The authors thank Tina Hindkjær for expert technical assistance and Anne Kruse Hollensen for kindly providing the sortilin expression plasmid. Single cell RNA-seq analysis was performed by the Bioinformatics Core Facility at the Department of Biomedicine, Aarhus University, and the authors thank the staff at the core facility for their excellent support. 

Supported by the VELUX Foundation (ALA., Grant No. 17739), Gene Therapy Initiative Aarhus (GTI-Aarhus) funded by the Lundbeck Foundation (T.J.C., Grant No. R126-2012–12456), the Lundbeck Foundation (A.N., Grant Nos. R248-2017-431 and R315-2018-3066), Independent Research Fund Denmark (T.J.C., Grant No. 2034-00036B), Fight for Sight Denmark (T.S.J.), the Synoptik Foundation (T.S.J.), and Insusense Therapeutics. 

Disclosure: T.S. Jakobsen, None; J.A. Østergaard, None; M. Kjolby, Biogen (R); E.L. Birch, None; T. Bek, None; A. Nykjaer, None; T.J. Corydon, None; A.L. Askou, None