Stimulation of Insulin Signaling Reverses Inammation-Induced Cone Death in Retinal Detachment

Background Rhegmatogenous retinal detachment (RD) involving the macula is a major cause of visual impairment despite high surgical success rate, mainly because of cone death. We and others have shown that cone loss occurs secondary to chronic inammation in many retinal diseases, including age-related macular degeneration and retinitis pigmentosa. We here investigated the yet unknown mechanisms of inammation-induced cone death in RD. Methods Vitreous samples from patients with RD and from control patients with macular hole were analyzed to characterize the inammatory response to RD. A mouse model of RD and retinal explants culture were then used to explore the mechanisms leading to cone death. Results Analysis of vitreous samples shows that RD induces a marked inammatory response with increased cytokine and chemokine expression in humans, which is closely mimicked by experimental murine RD. In this model, we demonstrate that myeloid cells and T-lymphocytes directly contribute to cone loss, as the inhibition of their accumulation by Thrombospondin 1 increased cone survival. We show that cones are highly dependent on glucose and insulin signaling for survival in vitro and that insulin, and the insulin sensitizers rosiglitazone and metformin, prevent RD-induced cone loss in vivo, despite the persistence of inammation. Conclusion Our results describe a new mechanism by which inammation, likely through a combination of competition for glucose and the inhibition of insulin signaling, promotes cone death in RD. Therapeutic inhibition of inammation and stimulation of insulin signaling might prevent RD-associated cone death until the RD can be surgically repaired and improve visual outcome after RD. cells and T-lymphocytes directly contribute to cone loss in RD. We show that cones are highly dependent on insulin signaling for survival in vitro and that insulin and the insulin sensitizers rosiglitazone and metformin prevent RD-induced cone loss in vivo, despite the persistence of subretinal MP accumulation. Taken together, our results suggest that inltrating MPs contribute to cone death by reducing the availability of glucose and inhibiting insulin signaling. Improving insulin signaling in cones might represent a therapeutic target for delaying RD-associated cone degeneration and subsequent vision loss. adaptater molecule-1; ONL:outer nuclear layer; Th-1 cells:T helper type 1 cells; MCs:microglial cells; Mφs:monocyte derived-macrophages; Mos:monocytes.


Abstract
Background Rhegmatogenous retinal detachment (RD) involving the macula is a major cause of visual impairment despite high surgical success rate, mainly because of cone death. We and others have shown that cone loss occurs secondary to chronic in ammation in many retinal diseases, including age-related macular degeneration and retinitis pigmentosa. We here investigated the yet unknown mechanisms of in ammation-induced cone death in RD.
Methods Vitreous samples from patients with RD and from control patients with macular hole were analyzed to characterize the in ammatory response to RD. A mouse model of RD and retinal explants culture were then used to explore the mechanisms leading to cone death.
Results Analysis of vitreous samples shows that RD induces a marked in ammatory response with increased cytokine and chemokine expression in humans, which is closely mimicked by experimental murine RD. In this model, we demonstrate that myeloid cells and T-lymphocytes directly contribute to cone loss, as the inhibition of their accumulation by Thrombospondin 1 increased cone survival. We show that cones are highly dependent on glucose and insulin signaling for survival in vitro and that insulin, and the insulin sensitizers rosiglitazone and metformin, prevent RD-induced cone loss in vivo, despite the persistence of in ammation.
Conclusion Our results describe a new mechanism by which in ammation, likely through a combination of competition for glucose and the inhibition of insulin signaling, promotes cone death in RD. Therapeutic inhibition of in ammation and stimulation of insulin signaling might prevent RD-associated cone death until the RD can be surgically repaired and improve visual outcome after RD.
Trial registration: ClinicalTrials.gov Identi er NCT03318588 Background Rhegmatogenous retinal detachment (RD) is a sight-threatening condition with an annual incidence of 10.5 per 100,000 people [1]. Advances in surgical techniques over recent decades have greatly improved the anatomical results with a primary success rate currently up to 80% [2]. However, despite a successful retinal reattachment, visual recovery may still be disappointing, especially in cases involving the cone-rich macula [3] and this loss of vision is primarily due to photoreceptor cell death [4,5]. Several pathogenic mechanisms have been identi ed but accumulating evidence suggests that in ammation plays a key role in the pathogenesis of RD-induced rod-photoreceptor cell death. Human studies have thus reported elevated levels of cytokines and chemokines in the vitreous of patients with RD [6][7][8][9]. Furthermore, experimental models have demonstrated that cytokines from in ltrating mononuclear phagocytes (MP) induce the death of rods following retinal detachment [10][11][12][13].
Although cones represent only 5% of all photoreceptor cells in humans, they are responsible for daylight, high-acuity and color vision. In retinal detachment, cone density decreases despite successful surgery and there is a strong correlation between post-operative cone density and visual acuity [14]. It is therefore important to identify the yet unknown mechanisms of detachment-associated cone death to establish therapeutic targets for preventing visual impairment.
Cones are highly metabolically active cells which particularly depend on glucose for function and longterm survival [15,16]. Contrary to muscle and fat (in which glucose absorption is mediated by the insulindependent Glucose transporter 4 (GLUT4)), glucose uptake in adult neurons depends mainly on the insulin-independent GLUT3 [17]. Interestingly, cone glucose uptake relies on GLUT1, which is regulated by the Rod-derived cone viability factor (RdCVF) [15,16], and the Insulin/mTOR pathway [18,19]. Indeed, cones have their own endogenous insulin receptor signaling pathway including the phosphoinositide 3kinase (PI3K) and m-TOR and cone-speci c deletion of PI3K (p85) is su cient to induce age-related cone degeneration [20,21]. In retinitis pigmentosa, where RdCVF levels are extinguished due to the primary loss of rods, systemic administration of insulin improved glucose uptake in cones and delayed their death, despite the absence of RdCVF [18,19]. Together these studies reveal the crucial role of insulin signaling in cone viability.
In ammatory cells, are also very reliant on glucose for metabolic activity, in particular when activated [22]. Their glucose uptake is mainly insulin-independent and mediated by GLUT1 and GLUT3 [17]. On the other hand, these leukocytes can produce cytokines such as IL-1ß, IL-6, and IFN-γ, which inhibit insulin signaling in adjacent stromal cells, an important mechanism of type 2 diabetes [23]. In RD, it is not clear whether and to what degree in ltrating in ammatory cells disturbe insulin signaling and cone glucose availability.
We here demonstrate, using human vitreous samples and a mouse model of RD, that RD causes a severe in ammatory response characterized by increased cytokine expression. In experimental RD, we demonstrate that the recruitment of myeloid cells and T-lymphocytes directly contribute to cone loss in RD. We show that cones are highly dependent on insulin signaling for survival in vitro and that insulin and the insulin sensitizers rosiglitazone and metformin prevent RD-induced cone loss in vivo, despite the persistence of subretinal MP accumulation. Taken together, our results suggest that in ltrating MPs contribute to cone death by reducing the availability of glucose and inhibiting insulin signaling. Improving insulin signaling in cones might represent a therapeutic target for delaying RD-associated cone degeneration and subsequent vision loss.

Patients
We conducted a nonrandomized clinical study at Nancy University Hospital from November 2017 to August 2018. Forty-one patients with primary RD requiring vitrectomy and 33 control patients undergoing vitrectomy for vitreomacular traction (VMT) or macular hole (MH) were included in this study.
Exclusion criteria were any history of vitreoretinal surgery on the eye studied, diabetic retinopathy or uveitis.
All patients underwent a detailed ophthalmologic examination before surgery, including best-corrected visual acuity measured with projected-light Snellen charts, axial length measurement using IOLMaster (Carl Zeiss Meditec, Dublin, CA), biomicroscopy with anterior segment evaluation, fundus and careful peripheral retina examination. For the RD group, an Amsler-Dubois scheme was systematically established for each patient, specifying the extent of the RD, number, type and location of retinal breaks, existence of vitreous hemorrhage and preoperative proliferative vitreoretinopathy (PVR) grading according to Machemer et al [24].
All patients underwent a three-port 23-or 25-gauge pars plana vitrectomy. At the beginning of vitrectomy, air perfusion was set to open and undiluted vitreous uid samples (1 mL) were collected from each eye with 3 mL syringe. Samples were sent to the Biological Resource Center (Centre de Ressources biologiques, Nancy, France) within 30 min, cooled on ice and transferred into microfuge tubes. Each sample was centrifuged at 10000 g for 5 min and the supernatant was then collected and frozen at -80 °C before analysis.

Mouse model of RD
Wild-type (C57BL/6J) mice were purchased from Janvier Labs at the age of 8 weeks. Mice were housed in the animal facility under speci c pathogen-free condition, in a 12 h/12 h light/dark (100-500 lux) cycle with water and normal diet food available ad libitum. All experimental protocols and procedures were approved by the local animal care ethics committee (N°APAFIS#5201-20160427103344).
RD was induced with a previously described method [11][12][13]. Brie y, mice were anesthetized with an intraperitoneal injection of xylazine hydrochloride (10 mg/kg) and ketamine hydrochloride (100 mg/kg) and pupils were dilated with topical phenylephrine (5%) and tropicamide (0.5%). A 30-gauge needle was rst used to create two sclerotomies 1.5 mm posterior to the limbus. A glass needle (with a 80-gauge manually beveled tip) connected to a Hamilton syringe lled with diluted sodium hyaluronate (Healon GV®, Alcon) was then introduced into the vitreous cavity through one of the sclerotomy. The tip of the needle was nally inserted into the subretinal space through a peripheral retinotomy and 4 µl of diluted sodium hyaluronate containing or not recombinant human TSP-1 (thrombospondin-1) (100 µg/ml, Biotechne), human insulin (2 IU/ml, Umuline NPH), recombinant IGF-1 (insulin growth factor-1) (200 ng/ml, Biotechne) or metformin (50 mg/ml, Merck Millipore), was gently injected, detaching approximately two-third of the retina from the underlying RPE.
For treatment with rosiglitazone, mice received intraperitoneal injections of 10 mg/kg rosiglitazone or vehicle (5% DMSO) 3 days before and 4 to 7 days after RD induction.
Eyes with subretinal hemorrhage were excluded from analysis. Mice were sacri ed from 1 to 10 days As insulin has been demonstrated to be quite unstable in media containing cysteine [25], and porcine/human insulin (identical except one amino acid at the C-terminus of the beta chain) is signi cantly less effective in rodents [26], we used supraphysiological doses of insulin (5 mIU/ml) in our model. For this reason, HNMPA was used at a concentration 10-fold higher than the IC50 (1 mM). Each culture medium was renewed every 36 hours and after 5 days, the retinal explants were carefully removed. Immunohistochemistry and cones quanti cation were then performed as described for retina below.

Immunohistochemistry of retinal atmounts
Eyes were enucleated, xed in 4% paraformaldehyde for 1 hour at room temperature and sectioned at the limbus; the cornea and lens were discarded. The retinas were peeled from the RPE/choroid/sclera and incubated overnight at 4 °C in PBS-1% triton with the following primary antibodies: peanut agglutinin (PNA) Alexa uor® 594 (Thermo Fisher Scienti c; 1/100), rabbit polyclonal anti-human cone arrestin (CAR) antibody (LUMIF-hCAR; 1:10000) and goat polyclonal anti-IBA1 (ionized calcium-binding adaptater molecule-1) (Abcam; 1:100). After few washes, the retinas were incubated for 2 hours at room temperature with appropriate Alexa Fluor® conjugated secondary antibodies (Thermo Fisher Scienti c; 1:500) in PBS-1% triton and nuclei were counterstained with Hoechst (1:1000, Sigma Aldrich). The retinas were atmounted and viewed with a uorescence microscope (DM5500, Leica). Images centered on the area with the lowest number of PNA + cone arrestin + cells were captured with a confocal laser-scanning microscope (FV1000, Olympus) using a 40X lens. Each cell population was manually counted in a masked fashion. IBA-1 + cells were quanti ed on atmounts on the outer segment side of the detached retina while PNA + cone arrestin + cells were counted on confocal microscopy Z-stacks using ImageJ software.
Reverse transcription and real-time quantitative polymerase chain reaction Total RNA was extracted from mouse retina with the Nucleospin RNAII extraction kit according to the manufacturer's protocol (Macherey Nagel). Single-stranded cDNA was synthetized from total mRNA (pretreated with DNase) using oligo-dT as primer and superscript II reverse transcriptase (Thermo Fisher Scienti c). Subsequent RT-PCR was performed using cDNA, PowerSYBR Green PCR Master Mix (Applied Biosystems) and primers (IDT technology) available upon request. qPCR was performed using StepOne Plus real-time PCR systems (Applied Biosystems) with the following pro le: 45 cycles of 15 s at 95 °C, 45 s at 60 °C. Results were normalized using house-keeping gene RPS26.

Statistical analysis
Graph Pad Prism 7 (GraphPad Software) was used for data analysis and graphic representation. All values are reported as mean ± SEM. Statistical analyses were performed by one-way Anova analysis of variance, Student t-test or Mann-Whitney U test for comparison among means depending on the experimental design. The p values are indicated in the gure legends.

Results
RD causes a marked in ammatory response with increased cytokine and chemokine expression in both human and experimental models To characterize the in ammatory response to RD, we rst analyzed the expression pro le of cytokines in vitreous samples from 41 patients with RD and from 34 control patients with macular hole. The mean extent of detachment in the RD group was 2.1 ± 0.8 quadrants with a macular involvement in 34.1% of cases and some degree of retinal wrinkling and folding (grade B or C proliferative vitreoretinopathy) in 36.6% of cases. The mean duration of symptoms before surgery was 7.7 ± 7.3 days with a median of 4.5 days . Using a Human Cytokine 27-plex Assay we showed that the cytokines IL-1ra, IL-6, IL-7, IL-8, IFN-γ (Fig. 1A), the chemokines CCL2, CCL3, CCL4, CXCL10 and CCL11 (Fig. 1B) and the growth factor G-CSF (Fig. 1C) were signi cantly increased in the vitreous from RD patients. In contrast, the levels of IL-10, IL-13 and VEGF were not statistically different between the two groups ( Fig. 1A-C). The remaining cytokines, such as IL-4, and IL-17 of the assay were not detectable.
We next evaluated the expression of the cytokines by RT-qPCR of mouse retinal control tissues and retinas harvested after four days of experimental RD. This time point was chosen for analysis as it was similar to the median duration of symptoms in our clinical study and because photoreceptor cell death peaks at around three days after RD in both experimental models and human samples [4,[27][28][29]. The transcription levels of nine out of eleven mediators found to be elevated in human vitreous from RD patients (except for Ccl11 and Il-8 that does not exist in mice) were signi cantly upregulated in detached mouse retinas compared to controls ( Fig. 1D-F).
In summary, our ndings con rm that RD induces a marked in ammatory response in human patients, which is closely mimicked by experimental murine retinal detachment. The cytokine pro l is suggestive of an in ltration of mainly myeloid cells, but the increased levels of IFN-γ might be suggestive of T helper type 1 (Th1) cells recruitment.

RD-associated leukocyte in ltration is associated with cone loss
In the healthy retina, microglial cells (MCs) populate the inner retina, but the photoreceptor cell layer and subretinal space are devoid of any immune cells [30]. In RD, MPs have been shown to accumulate in the detached area and are highly associated with TUNEL-positive nuclei in the inner aspects of the outer nuclear layer (ONL), where rod nuclei are located [10,12,13]. Using ow cytometry and a gating strategy ( Fig. 2A) adapted from O'Koren et al. [31], we here analyzed the leukocyte population in healthy and detached (day 1, day 3, day 7) mouse retinas. In healthy retinas, we only detected MCs (CD11b + CD45 low ) that steadily increased to quadruple their numbers at the end of the observation period (Fig. 2B). The myeloid cell population (CD11b + CD45 high ) sharply increased after RD, peaking at 24 h and remained strongly elevated throughout (Fig. 2B). Interestingly, we also found a sizeable population of T-cells (CD45 + CD3 + ) that in ltrated the retina at day 3 and remained elevated, although to a lesser extent, at day 7 (Fig. 2B). A more detailed analysis of the myeloid population reveals a rapid, but short-lived recruitment of neutrophils (CD11b + CD45 high Ly6G + ) and monocytes (CD11b + CD45 high Ly6G neg Ly6C high ) at day 1 (Fig. 2C). The number of macrophages (CD11b + CD45 high Ly6G neg Ly6C low ) mainly increased at day 3 and stayed elevated, likely re ecting the differentiation of monocytes (Mos) into macrophages (Fig. 2C).
Together, these results demonstrate that RD induces the in ltration of T-cells, monocyte derived-macrophages (Mφs) and accumulation of MCs, and that MPs, MCs and Mφs, represent the main accumulating immune cells.
Next, we quanti ed the presence of MPs (MCs and Mφs) and the cone population on IBA-1 (MP marker)-, peanut agglutinin (PNA cone outer segment marker)-, cone arrestin (CAR, cone marker) triple-stained retinal at-mounts (Fig. 2D). Confocal microscopy con rmed that subretinal MPs are not observed in normal mice and highlight the elongated shape of the cone outer segments (Fig. 2D). Interestingly, despite the peak of in ltrating myeloid cells measured in the whole retina at day 1, IBA-1 + MPs only accumulated in the subretinal space by day 3 and continued to rise to reach a plateau at day 7 (Fig. 2E).
Although cone outer segments seemed shortened at day 1 (Fig. 2D), their number only decreased signi cantly in the following days and was reduced by approximately 50% at day 7 (Fig. 2D, F and G) mirroring the subretinal MPs accumulation.
Taken together, our results demonstrate that RD leads to a rapid in ltration of myeloid cells, followed by T-cells and a protracted increase of the numbers of MCs and Mφs that started accumulating in the subretinal space by day 3. We also showed that this accumulation was strongly associated with cone death.

TSP-1 inhibits RD-induced subretinal MPs in ltration and associated cone loss
Thrombospondin-1 (TSP-1) is an extracellular matrix molecule that is produced by a wide variety of cell types, notably the RPE, in ammatory and resident macrophages [30]. We and others have shown that it physiologically prevents age-related subretinal MP accumulation and inhibits excessive subretinal MP in ltration and choroidal neovascularization in the context of age-related macular degeneration [32][33][34] and controls T-cell response [35]. To further explore the role of in ltrating MPs in RD-associated cone loss, we induced RD with diluted sodium hyaluronate that contained or not recombinant TSP-1 (100 µg/ml). Using ow cytometry and the same gating strategy as in Fig. 2 (Fig. 3A), we found that recombinant TSP-1 signi cantly reduced the number of T-cells, Mos, and Mφs but had no effect on the MC population (Fig. 3B). Accordingly, RT-qPCR analysis showed that recombinant TSP-1 signi cantly reduced the transcription levels of Il-6, Ifn-γ, Ccl2, Ccl3 and Ccl4 (Fig. 3C). Quanti cation of IBA-1 + MPs and PNA + CAR + cones on triple-stained retinal at-mounts at day 7 (Fig. 3D) showed that recombinant TSP-1 also signi cantly decreased subretinal MP accumulation (Fig. 3E) and increased cone survival in detached retinas compared with PBS controls (Fig. 3F and 3G).
Together, these ndings show that the pharmacologically induced reduction of the population of in ltrating MPs and T-cells and cytokines expression signi cantly protects against RD-induced cone loss, suggesting that in ltrating T-cells, Mos, and Mφs directly contribute to cone loss in RD.

Insulin is essential for cone survival in vitro and delays RD-induced cone loss in vivo
Cones are highly dependent on glucose for metabolic activity and long term survival [15,16]. It has been shown that insulin signaling pathways play a key role in cone glucose uptake [20,21] and that activation of these pathways by systemic injection of insulin promotes cone survival in a mouse model of RP [18,19].
In RD, cone glucose availability could become critical as highly metabolic active immune cells compete for fuel in the photoreceptor cell layer [22]. Additionally, in ammatory cytokines such as IL-6, and IFN-γ might impede with insulin signaling, similar to the mechanism of type 2 diabetes [23].
To investigate whether diminished insulin signaling could contribute to RD-associated cone loss in adult retinas, we rst examined the effect of insulin on 5 day retinal explants with or without human insulin and/or the speci c insulin receptor inhibitor HNMPA (Fig. 4A). Quanti cation of PNA + CAR + cones showed that cone survival was signi cantly increased in the presence of insulin in the medium culture compared to the control condition ( Fig. 4A and 4B). The addition of an insulin receptor inhibitor HNMPA that blocks insulin receptor autophosphorylation, but not insulin growth factor 1 (IGF-1) receptor activation [36], resulted in a severe loss of PNA + CAR + cones, not observed with its vehicle (DMSO) (Fig. 4A and 4B). Our results indicate that insulin signaling promotes cone survival in retinal explants con rming previous results in a mouse retinal degeneration model [18,19].
Next, we induced RD in vivo with subretinal injection of diluted sodium hyaluronate containing (or not) human insulin (2 IU/ml). Quanti cation on immuno-stained retinal at-mounts at day 7 (Fig. 4C) revealed that insulin treatment did not alter the numbers of subretinal IBA1 + MPs, but very signi cantly increased the number of PNA + CAR + cones compared to PBS controls (Fig. 4D-F). Comparatively, addition of IGF-1 to the detachment inducing gel (at a concentration 100-fold higher than IGF-1 s ED50, which has been shown to reverse hypoalgesia in diabetic mice [37]) had no effect on the numbers of IBA1 + MPs or PNA + CAR + cones, quanti ed on day 7 immuno-stained retinas (Fig. 4G-J).
Taken together, our results showed that insulin and insulin receptor signaling were essential for cone survival ex vivo of adult retinas and that insulin treatment very signi cantly inhibited RD-induced cone loss despite the unchanged MPs in ltration in vivo. This effect was not due to insulin-induced IGF-1R signaling, which can activate anti-apoptotic IGF-1 receptor signaling [38,39], as IGF-1 had no comparable effect.
The insulin sensitizer rosiglitazone and metformin prevent RD-induced cone loss IL-6 and IFN-γ, which we show are increased in RD, can inhibit insulin signaling in type 2 diabetes [23]. This insulin resistance can at least in part be reversed by insulin sensitizers, such as rosaglitazone [40,41]. To explore whether pharmacological improvement of insulin-signaling could reduce in ammationinduced cone degeneration in RD, we next examined whether rosiglitazone, could prevent cone loss in our mouse model of RD.
Mice received daily intraperitoneal injections of rosiglitazone or vehicle (DMSO 5%) 3 days before and throughout the 7 days of RD. The treatment also did not alter the increased levels of Il-6 and Ifn-γ mRNA levels in whole retinal mRNA and increased Ccl2 in four day RD samples (Fig. 5A). Quanti cation of IBA1-, PNA-, CAR triple-stained retinal atmounts (Fig. 5B) also showed that rosiglitazone had no effect on the number of in ltration of subretinal IBA-1 + MPs at day 7 (Fig. 5C). Despite this lack of an antiin ammatory effect, quanti cation of PNA + CAR + cones, revealed that rosiglitazone signi cantly protected retinas against cone loss at day 7 ( Fig. 5D and E). Interestingly, subretinal injection of metformin, a commonly used insulin sensitizer that also exerts an independent anti-in ammatory effect [42,43], signi cantly increased cone survival and decreased subretinal MP accumulation in detached retinas compared with PBS controls (Fig. 5G-J).
In summary, our results show that the well-established insulin sensitizers rosiglitazone and metformin signi cantly curb cone loss in RD. The fact that we observed increased cone survival under rosiglitazoneand insulin-treatment in the absence of an anti-in ammatory effect strongly suggests that restored insulin signaling was the likely mode of action.

Discussion
Despite successful surgical repair, visual recovery remains incomplete in many eyes that have suffered from macula-off RD, mainly because of photoreceptor cell death [4,5,27], in particular cone loss [14].
Although evidence from both human and experimental studies show that in ammation is strongly associated with RD-induced photoreceptor cell death [7,[10][11][12][13], little is known about the mechanisms that speci cally lead to cone death.
Our analysis of vitreous samples con rms that RD in humans is associated with increased expression of cytokines, chemokines and growth factors [6,7,9,44]. Interestingly, our experimental model revealed a very similar induction in RD in mice, making it therefore well-suited for studying the effect of these mediators on RD-associated cone loss. We con rm previous reports of RD-induced MP accumulation [7,10,12,[45][46][47]. Our ow cytometric analysis showed a more nuanced picture of a rapid in ltration of neutrophils and Mos (day 1), followed by T-lymphocytes (day 3) and a protracted increase of the numbers of MCs and Mφs that were immunohistochemically detected in the subretinal space from day 3 following RD.
RD-induced MP accumulation has previously been shown to be highly associated with TUNEL-positive nuclei in the inner aspects of the ONL, where the rod nuclei are located [10,12,13]. We here show that the accumulation of immune cells was strongly associated with the loss of more than 50% of the cones. This loss was demonstrated by the disappearance of PNA + cells (which might have suggested the loss of cone outer segments only), but also CAR + cells, a marker that is found throughout the cone cytoplasme, demonstrating the loss of the cells. Recombinant TSP-1 severely reduced the numbers of accumulating in ltrating T-cells, Mos and Mφs, but interestingly not MCs, and very signi cantly reduced the transcription level of cytokines at day 4, which strongly suggests they were expressed by the leukocytes. Importantly, TSP-1 treatment also signi cantly prevented the loss of cones at day 7, suggesting that in ltrating T-cells, Mos, and Mφs directly contribute to cone loss in RD, similarly to previous results regarding rods [10,12,13].
Cones are particularly reliant on glucose and insulin signaling for survival [15,16], evidenced by progressive cone loss in mice with cone-speci c deletion of PI3K (p85, indispensible for insulin signaling) [20,21] and the protective effect of insulin on secondary cone-loss in retinitis pigmentosa [18,19]. Using insulin and a speci c insulin receptor inhibitor on mouse retinal explants, we con rmed that insulin receptor signaling was essential for cone survival ex vivo.
During RD the retina is in ltrated by activated leukocytes. Leukocyte activation induces a metabolic switch in the immune cells to aerobic glycolysis, which dramatically increases their glucose consumption and makes them very reliant on surrounding glucose concentrations for survival and function [22]. In RD the in ltrating activated leukocytes therefore likely compete for glucose, which might become critical for cone survival. Upon activation of the leukocytes, GLUT1 (lymphocytes) and GLUT3 (monocytes) are recruited to the plasma membrane mediating the massive glucose uptake [17,48]. Insulin only plays a minor role in the increased uptake of glucose in in ammatory cells, as it does not in uence GLUT translocation to the plasma membrane in neutrophils and T-lymphocytes and only marginally increases glucose consumption in macrophages [48,49]. Insulin supplementation could therefore redirect glucose uptake preferentially to cones and save cones from starvation. To make matters worse, the in ammatory cytokines, such as IL-6 and IFN-γ, that are secreted by the in ltrating leukocytes, can inhibit insulin signaling, which inhibits insulins trophic effects and further reduces glucose uptake [23]. In ammationinduced reduced insulin signaling and glucose uptake is increasingly recognized to play an important role in insulin-resistance in the adipose tissue of type 2 diabetes [23], but has also been suggested to play an important role in neuronal death in uveitis, a blinding auto-immune disease of the retina and choroid [50].
To evaluate if glucose could be redirected to cones and trophic signals restored, we treated the experimental animals with insulin or the insulin sensitizer rosiglitazone. Our data demonstrates that insulin supplementation, but also treatment with rosiglitazone, signi cantly prevented RD-induced cone loss, despite having no effect on the level of in ammation. In contrast, IGF-1 supplementation did not increase cone survival in RD, underlining the speci city of insulin signaling for this effect.
Mechanistically, insulin receptor signaling has been described to increase the GLUT1-dependent glucose uptake via the activation of the mammalian target of rapamycin complex 1 (mTORC1) [18,19]. In RD, the insulin supplementation might thereby increase glucose-uptake to cones without affecting the insulinindependent glucose uptake by in ltrating leukocyte. This redirection of glucose to the cones could prevent cone starvation in RD. Additionally, insulin and insulin sensitizers might restore the in ammation-induced reduction in insulin receptor signaling and boost the trophic, anti-apoptotic signals via mTORC2 [18,19] Conclusion In summary, our results describe a new mechanism by which in ammation induces cone death in RD. This mechanism might not be speci c to RD but could also be involved in other retinal diseases characterized by chronic in ammation and cone loss such as retinitis pigmentosa and AMD [32,51]. Therapeutic inhibition of in ammation and restoration of insulin signaling and glucose availability to cones might prevent cone death in these diseases. Indeed, our study demonstrates that Metformin, an insulin sensitizer with known anti-in ammatory effects [42,43], signi cantly decreased in ammation and increased cone survival. Metformin or similar agents might prevent RD-associated cone death until the RD can be surgically repaired in the future.

Declarations
Ethics approval and consent to participate This study was approved by the regional Institutional Ethics Committee (Comité de Protection des Personnes CPP17-059/2017-A02195-48) and adhered to the Declaration of Helsinki. This study was also registered in the database of the National Institutes of Health at clinicaltrials.gov (identi cation number NCT03318588). All patients had complete information about the study and the risks and bene ts of the surgical procedure and gave their written consent for participation before surgery.

Consent for publication
Not applicable Availability of data and materials All data generated or analyzed during this study are included in this published article. The datasets used and/or analyzed during the current study are also available from the corresponding author on reasonable request.

Competing interests
The authors declare that they have no competing interests.