Stromal-Like Cells and Retinal Pigment Epithelium Modulate Choroidal Sprouting Through Galectin-1-Dependent and Independent Pathways.

In our previous work, we reported a crucial interaction between EC and stromal cells in the RPE/choroid, underscoring its significant role in maintaining choroidal homeostasis, with potential implications for the pathophysiology of ocular disorders ( 9 ). Interestingly, the analysis of our reported single cell RNAseq expression profile data ( 9 ) showed that Gal1 was the most abundant galectin found in the RPE/choroid ( Fig. 1a , purple column). Indeed, while Gal1 mRNA expression is detected in a minority of other cell types as well, this analysis revealed that Gal1 is enriched in all stromal cell clusters ( Fig. 1a , red columns), suggesting that stromal cell populations are an important source of Gal1 in mouse choroidal tissue. To validate these observations, immunofluorescence staining of histological sections of RPE/choroid from WT adult mice was performed. Confocal microscopy showed that, while Gal1 was observed in other cell types such as RPE, it was mainly localized in both perivascular and non-perivascular stromal-like cell populations, according to analysis from single-cell RNA-seq expression profiles ( Fig. 1b ). We previously established that PDGFRβ and NG2 are expressed exclusively in stromal cells and smooth muscle cells (SMCs), while αSMA is specific to SMCs of mouse RPE/choroid. Additionally, we identified different stromal cell clusters expressing PDGFRβ, NG2, and Sca1, but not αSMA ( 9 ). Consistent with these findings, confocal microscopy revealed higher colocalization of Gal1 with PDGFRβ, NG2, and Sca1, but not with αSMA or Isolectin B4 ( Fig. 1c ). These findings suggest that Gal1 is predominantly associated with stromal cell populations rather than SMCs or EC of RPE/choroid from WT adult mice. Since Gal1 was also found to be expressed in the RPE, we then studied Gal1 secretion using polarized hfRPE cells in culture. Our findings revealed that approximately 75 % of Gal1 protein was secreted toward the apical side ( Fig. 1d ), suggesting that Gal1 released by the RPE may not be a major physiological source reaching the choroid. Thus, stromal cell populations appear to be the major source of Gal1 in mouse choroidal tissue in physiological settings. Functional and structural studies in adult mice lacking Gal1 ( Lgals1 −/− ) were performed to assess the role of Gal1 in visual pathophysiology. ERG measurements indicated normal phototransduction (a-wave) and synaptic transmission (b-wave), with no significant differences in scotopic or photopic responses among Lgals1 −/− , Lgals1 +/- , and WT mice ( Fig. S1 ), consistent with previous findings ( 33 ). Additionally, SD-OCT analysis revealed no significant changes in retinal thickness or layer organization ( Fig. 2a ). These results suggest that lack of Gal1 in knockout mice does not impair the function or structure of neural retina. Using the OMR test, a more sensitive and non-invasive method ( 47 , 48 ), we observed a slight but statistically significant reduction in spatial frequency, suggesting altered behavioral parameters associated with spatial vision, acuity and contrast sensitivity ( 34 ) in Lgals1 −/− respect to WT mice ( Fig. 2b ). Moreover, RNAseq data analyses from RPE/choroid showed a down regulation of genes linked to visual or light stimulus detection, photoreceptor physiology and outer retinal function ( Fig. 2c , right) with an enrichment in gene expression profiles associated with immune response in Lgals1 −/− respect to WT mice ( Fig. 2c , left). All together these results suggest that the lack of Gal1 in Lgals1 −/− mice does not significantly impact neural retina function, but affects gene expression profiles linked to eye homeostasis and altered spatial vision parameters. Based on the critical role of Gal1 in the pathophysiology of several tissues, electron microscopy (EM) of RPE/choroid from WT and Lgals1 −/− adult mice was performed. Stromal-like cells exhibited a lengthwise shape in WT mice ( Fig. 3a , left panel), but adopted an altered morphology with larger nuclei in Lgals1 −/− mice ( Fig. 3a , right panel). Furthermore, choroidal vessels were considerably more tortuous and chaotic in Lgals1 −/− versus WT mice, while the morphology of the RPE was similar between those phenotypes ( Fig. 3a ). Flat mounts of mouse RPE/choroid were generated to further analyze RPE morphology. Phalloidin staining did not show any difference in RPE shape compatible with RPE damage in Lgals1 −/− versus WT adult mice ( Fig. 3b ). Immunofluorescence using an anti-ZO1 antibody and Hoechst 33342 staining showed similar results (data not shown). To better understand the molecular bases of the observed alterations in the morphology of stromal-like cells and choroidal vessels, bulk RNAseq of RPE/choroid from WT and Lgals1 −/− mice was performed. GO enriched terms in Cellular Component category revealed a downregulation of genes associated with ocular extracellular matrix, detection of light stimuli involved in sensory perception and cellular anatomical entity ( Fig. 3c , left) in Lgals1 −/− versus WT mice. Besides, GO enriched terms in the Molecular Function category showed a reduction in ion transmembrane transport and glycosaminoglycan binding in these samples ( Fig. 3c , right). Our results indicate that the Gal1 deficiency impairs the structure of stromal-like cells and blood vessels. This effect is associated with altered gene expression profiles linked to choroidal homeostasis. These findings suggest an essential role for Gal1 in maintaining the structural integrity of choroidal tissue. The role of Gal1 in angiogenesis has been well established in a variety of pathological conditions ( 15 , 16 , 26 , 29 ). We set up an ex vivo model of microvascular angiogenesis ( 46 ) that preserves the interaction of chECs with other cellular components, including macrophages and pericytes, to analyze the contribution of Gal1 to choroidal sprouting. After several days of culture, we observed that choroidal explants obtained from Lgals1 −/− mice developed less sprouting compared to those obtained from WT mice ( Fig. 4a ). Further, choroidal sprouting was much lower in explants from WT mice exposed to an anti-Gal1 neutralizing monoclonal antibody (Anti-Gal1 mAb) ( 22 ) ( Fig. 4b ). These results demonstrate the contribution of Gal1 to choroidal angiogenesis. Given the strong evidence supporting Gal1's involvement in hypoxia-driven NV ( 23 , 26 , 29 ), we also conducted ex vivo choroid sprouting assays under hypoxic conditions. Choroidal explants from WT mice treated with recombinant Gal1 showed a significant increase in angiogenesis, while those exposed to anti-Gal1 mAb exhibited reduced growth ( Fig. 4c ). These results suggest that Gal1 plays a crucial role in hypoxia-driven CNV, as previously reported in an in vivo mouse model ( 33 ). To understand the potential role of Gal1 in the communication between stromal cells and choroidal vessels, we analyzed cellular communication networks through ligand–receptor analysis using LIANA ( 43 ). The interaction heatmap revealed a high frequency of ligand interactions among stromal cells, RPE, and choroidal ECs ( Fig. 5a ). The analysis of cell–cell ligand–receptor interactions revealed that Gal1 expressed by stromal cells specifically and strongly interacted with ECs via vascular endothelial growth factor receptor-2 (VEGFR-2)/(kinase-insert domain receptor (KDR) ( Fig. 5b ), highlighting the critical role of this lectin as a mediator of stromal cells-EC communication. Although other minor cell populations in the RPE/choroid have been shown to also interact through Gal1-dependent mechanisms, it should be noted that under basal conditions, the RPE itself does not appear to be involved. To further investigate the communication between stromal cells and choroidal vessels in vitro , we analyzed whether factors released by stromal cells could modulate EC tube formation. For this purpose, we employed previously characterized Gli1+ choroidal stromal cell primary cultures ( 9 ) as an in vitro model of choroidal stroma-like cells. After 3 days of culture, supernatant was obtained and employed as conditioned cell medium. E4-HUVEC tube formation assays were performed in conditioned or non-conditioned media from the stroma-like cell cultures. Results showed altered tubulogenesis in E4-HUVEC cells with respect to non-conditioned choroidal stroma-like cell medium. Quantitative analysis revealed that conditioned media significantly decreased the total segment length and induced greater inhibition of tube formation with respect to control in E4-HUVEC cells in the studied conditions ( Fig. 5c ). In agreement with our single cell RNAseq data from mouse adult RPE/choroid ( 9 ), qRT-PCR results showed that Gli1+ choroidal stroma mesenchymal cells (MSC)-like cell primary cultures express NG2, PDGFRβ, collagen type 1 alpha 1 chain (COL1a1) and Sca1, but not the EC marker CD31 showing no contamination with ECs ( Fig. S2 ). Moreover, single cell RNAseq analysis showed that, in addition to Gal1, stromal cells were enriched in insulin-like growth factor 1 (IGF-1) and pigment epithelium-derived factor (PEDF or Serpinf1) and expressed VEGF and bFGF (FGF2 or FGF-β) ( 9 ). Among them, the stromal-enriched anti-angiogenic molecule, PEDF , was the most highly expressed transcript in the whole mouse RPE/choroid ( Fig. 5d , upper panel). Besides, receptors of these molecules were mainly expressed in the different EC clusters ( Fig. 5d , lower panel), suggesting a possible crosstalk between these cell populations. The analysis of cell–cell ligand–receptor interactions revealed that Serpinf1 expressed by stromal cells specifically and powerfully interacted with ECs via VEGFR-2/KDR ( Fig. 5e ). These results suggest that cell communication between stromal cells and chECs proceeds through Serpinf1-Kdr signaling. Altogether, these findings indicate that stromal cells communicate with ECs to maintain choroidal tissue homeostasis through both Gal1-dependent and -independent pathways. To investigate whether Gal1-enriched stromal cells could represent a choroidal cell population involved in pathological settings, we studied the in vitro phenotype of our previously characterized Gli1+ choroidal stromal MSC-like cell primary cultures ( 9 ) in response to hypoxic or inflammatory conditions. For this purpose, primary cultures were exposed to pro- or anti-inflammatory cytokines, proliferative stimuli (bFGF) or hypoxia during 24 or 40 h. The expression of VEGF, Gal1 and PEDF (Serpinf1) was measured by qRT-PCR and both Gal1 and VEGF secretion were determined in culture supernatants by ELISA. No signs of cell death were observed at any of the conditions analyzed (data not shown). Choroidal stroma-like cells exposed to the anti-inflammatory cytokine IL-10 did not alter the expression or secretion of pro- or anti-angiogenic molecules; however, IL-1β and hypoxia treatments induced higher VEGF expression and secretion by choroidal stroma-like cells ( Fig. 6a and 6b ). Furthermore, IL-1β as well as IFN-γ lowered PEDF mRNA expression and did not alter Gal1 expression or secretion ( Fig. 6a and 6b ). Gal1 mRNA levels were slightly but significantly increased in choroidal stroma-like cells exposed to hypoxia during 24 h and were highly up-regulated in bFGF-treated cells ( Fig. 6a ). Although Gal1 secretion was not altered by any of the treatments applied ( Fig. 6b ), exposure to hypoxia or bFGF significantly modified the reticular pattern of Gal1 protein expression in choroidal stroma-like cells ( Fig. 6c ). Given the existing evidence highlighting a pathological role for Gal1 during CNV associated with RPE ( 33 ), we analyzed if Gal1 secretion was modulated by hypoxia in hfRPE-polarized cultures. Results showed that Gal1 secretion increased considerably after exposure to hypoxia and this up-regulated secretion was even more pronounced at the basolateral side of the polarized hfRPE cultures ( Fig. 6d ). These findings suggest that choroidal stroma-like cells are highly sensitive to in vitro pathological conditions, which modulate production of angiogenesis-related molecules. This effect underscores their essential role as a distinct choroidal cell population that may contribute to pathological CNV. While Gal1 expression was not strongly modulated in stromal-like cells, its secretion was markedly increased in hfRPE cultures exposed to hypoxia, indicating that this lectin is differentially regulated in stromal-like cells and RPE under physiological o pathological conditions.

Wild type (WT) and Gal1-deficient ( Lgals1 −/− ) mice on C57BL/6 background were handled according to ARVO Statement guidelines for the Use of Animals in Ophthalmic and Vision Research. Lgals1 −/− mice were originally provided by Dr. Francoise Poirier (Jacques Monod Institut, Paris, France). Animal procedures were reviewed and approved by the Institutional Animal Care and Use Committee at Weill Cornell Medicine. All efforts were made to reduce the number of experimental mice. Scotopic and photopic electroretinography (ERG) was performed in WT, Lgals1 ₊/- and Lgals1 −/− adult mice as previously described ( 8 ). Briefly, mice were dark-adapted overnight and under dim red illumination and were anesthetized with 100 mg/kg ketamine and ~10 mg/kg xylazine solution per hour. Anesthetized animals were placed on a stage with a heater to keep the appropriate body temperature. The pupils were dilated with one drop of 1 % Tropicamide / 2.5 % Phenylephrine solution and one drop Gonak (Akorn) was applied along with gold wire corneal electrodes. Retinal function was measured using a Ganzfeld ColorDome Diagnosys LLC (Littleton, MA, USA). For scotopic ERG, single-flash responses were recorded at various stimuli intensities varying from −4 log cd·s/m 2 to 2 log cd·s/m 2 , with inter-stimuli intervals of 10 to 60 s. For photopic ERG, mice were light-adapted under background light of 1.48 log cd·s/m 2 for approximately 5–10 mins. Single flash photopic responses were employed from −1 log cd·s/m 2 to 2 log cd·s/m 2 , presented on the 1.48 log cd·s/m 2 background. Both eyes were subject of scotopic and photopic ERG recordings. ERG amplitudes for scotopic a- and b-wave and photopic a- and b-wave were analyzed with software Diagnosys Espion and graphs were generated with Prism. To assess visual acuity, we performed the OptoMotor Response (OMR) test ( 34 ). For that purpose, visual thresholds of adult WT and Lgals1 −/− mice were measured by evaluating optokinetic tracking in a virtual optokinetic system as described previously ( 35 ). The optomotor response (OMR) assesses spatial frequency (similar to visual acuity) and contrast sensitivity by monitoring characteristic reflexive head tracking movements to gauge visual thresholds for each eye ( 36 ). Spatial frequency refers to the thickness or fineness of the bars, and contrast sensitivity refers to how much contrast there is between the bars and the background. Adult WT, Lgals1 ₊/- and Lgals1 −/− mouse retinas were examined in situ by fundoscopy and optical coherence tomography (OCT, Phoenix MICRON IV system, Phoenix Research Lab, Pleasanton, CA, USA). Briefly, mice were anesthetized with ketamine/xylazine solution and transferred to a heated water pad to regulate their body temperature. Pupils were dilated with a 1% tropicamide solution and corneas were kept moist with 1X PBS, pH 7.4 to prevent corneal desiccation. Images were acquired from both eyes using a 30° lens and Heidelberg Retina Angiograph-optical coherence tomography (Spectralis Heidelberg, Germany) with a 488-nm argon blue laser and standard 500-nm long-pass filter. The mice were transferred to a recovery heat pad and monitored until fully conscious. We further analyzed the structure of RPE/choroid by transmission electron microscopy (EM). Briefly, adult mice were sacrificed, and eyes were fixed in 2 % paraformaldehyde / 2 % glutaraldehyde solution. A hole was made into the cornea and eyes maintained in the same buffer at 4 °C. The following day, eyes were enucleated and any extraocular tissue attached to the sclera was carefully removed. Eye cups were further processed for EM as previously described ( 8 ). Briefly, eye cups were embedded in Spurr’s resin (Electron Microscopy Sciences, PA, USA) and 70 nm thin sections were cut, placed on copper grids, and counterstained with 2% uranyl acetate and 3.5% lead citrate (Ted Pella, CA, USA). Samples were imaged at the EM Core Facility at Weill Cornell Medicine. We performed bulk RNA sequencing (RNAseq) of choroidal tissue from adult WT and Lgals1 −/− mice as previously described ( 9 ). Three mice per group were euthanized and their eyes were enucleated and any remaining extraocular tissue attached to the sclera was removed. The neural retina was carefully detached from the optic nerve and RPE/choroid and discarded RPE/choroid tissue was mechanically dissected from the sclera using a scalpel and immediately used for RNA extraction (RNeasy Mini Kit, Qiagen, Hilden, Germany). Sample quality control, library preparation and sequencing were performed at the Weill Cornell Medicine Epigenomics Core. Briefly, cDNA libraries were prepared with the TruSeq RNA Sample Preparation Kit (Illumina) and sequenced on an Illumina HiSeq2000 platform ( 9 ). The RNAseq analysis was conducted following a standard pipeline. Briefly, high-quality sequencing reads were obtained after adapter trimming and quality filtering of raw reads using Trim Galore (version 0.6.5, www.bioinformatics.babraham.ac.uk/projects/trim_galore/ ). The cleaned reads were then aligned to the mouse reference genome assembly GRCm38.p6 (ENSEMBL release 97) using the Subread aligner (version 2.0.1) ( 37 ). Gene-level counts were quantified from the mapped reads using featureCounts (version 1.6.4) ( 38 ). The Bioconductor package edgeR (version 3.28.0) ( 39 ) was utilized for normalization and differential expression analysis, leveraging empirical Bayes methods to estimate gene-specific biological variation ( 40 ). Genes expressed at low levels were filtered out prior to normalization by retaining genes with worthwhile counts in a minimum number of samples. Trimmed mean of M-values (TMM) normalization was applied to account for compositional differences between libraries ( 40 ). Negative binomial generalized linear models were fitted to identify differentially expressed genes (DEGs). DEGs were defined as having |log2(fold change)| > 2 and p-value < 0.05. Volcano plots were generated using the EnhancedVolcano (version 1.14.0) packages. Gene Ontology (GO) enrichment analysis was performed using g:Profiler ( 41 ) and functional enrichments were visualized using Cytoscape (version 3.9.1) and plugins EnrichmentMap (version 3.3.1), ClusterMaker2, and AutoAnnotate (version 1.3.3) ( 42 ). The Single cell analysis Seurat object from Lehmann et.al. ( 9 , 43 ), has been downloaded from GEO accession GSE135167 . Interactome analysis was performed using the LIANA R package ( 43 ) with default parameterization and using the LIANA consensus database transformed to a mouse model and manually adding the Lgals1-Kdr interaction. The cell type annotation was used as a grouping variable. In addition, we manually added known PEDF interactions ( 44 , 45 ) because of their relevance to pathological angiogenesis. Adult WT, Lgals1 ₊/- and Lgals1 −/− mice were dark-adapted overnight and euthanized the following day. The eyes were enucleated (removal of cornea, lens and retina) and fixed in 4 % paraformaldehyde in phosphate-buffered saline (PBS) buffer for 2 h at room temperature. After two brief washes in PBS, RPE/choroid tissue was blocked and permeated in PBS containing 2 % of Bovine Serum Albumin (BSA) and 0.1 % of Triton-X-100 during 1 h at room temperature. The tissue was incubated with primary antibodies Phalloidin-488 (1:200) and anti-Zonula occludens-1 (ZO-1, 1:100) ON at 4 °C in humidifier chamber, washed three times in PBS for 10 min each, and incubated for 1–2 h at room temperature with secondary antibody solution (anti-Rabbit Alexa 568, 1:800; Hoechst, 1:100). Finally, stained tissue was washed three times in PBS for 10 min each, mounted onto slides in Fluoromount G (Invitrogen, MA, USA) and coverslipped. Images were taken with Zeiss Axio Observer spinning disk confocal microscope (Zeiss, CA, USA) equipped with a Yokogawa scanner unit, Hamamatsu Evolve electron-multiplying charge-coupled device cameras. Image analysis and processing was performed with ImageJ and Zeiss’s Zen 2 software. A characterized and well-defined ex vivo model of microvascular angiogenesis ( 46 ) was employed. Briefly, euthanized WT or Lgals1 −/− adult mice were immediately enucleated and eyes were kept in ice-cold complete medium [HyClone Medium 199 with 50 µg/ml of Endothelial Cell Growth Supplement (ECGS), 10 mM of HEPES, 50 µg/ml heparin, 5 % of fetal bovine serum (FBS), 100 IU/ml penicillin, 100 µg/ml streptomycin, 0.25 µg/ml amphotericin and GlutaMAX] before dissection. RPE/choroid was obtained, and peripheral tissue was cut into approximately 1 mm x 1 mm. Each piece of tissue was placed into a 30 µl Matrigel (Basement Membrane Matrix Growth Factor Reduced, Corning, NY, USA) drop per well of ice-cold 24-well plate. After 10 min in the 37 °C incubator for the Matrigel solidify, 500 µl of complete culture medium per well was added and plates were returned to 37 °C and 5 % CO 2 culture incubator. After 2 days, cells were subjected to overnight starvation with reduced-serum (1 %) complete medium before treatments with recombinant Gal1 (rGal1) purified as described ( 26 ) or function-blocking anti-Gal1 antibody (Anti-Gal1) generated and characterized as described ( 22 , 25 ) under normoxia (37 °C, 5 % CO 2 and atmospheric O 2 ) or hypoxia (37 °C, 5 % CO 2 and 1 % O 2 ). The culture medium was renewed every two days. Photos were taken after 4 and 7 days of treatments with a Zeiss Axio ObserverZ.1 widefield microscope (Zeiss, CA, USA) equipped a Plan-Apochromat 20X/0.8 M27 objective and Hamamatsu Flash4.0 V2 sCMOS camera and the sprouting area relative to control conditions was calculated using ImageJ Software as was described ( 46 ). Positive control included 250 ng/ml of recombinant mouse VEGF, (Peprotech, NJ, USA) and negative control included 15 µM of D,L-sulforaphane (Sigma, MO, USA) (data not shown). We employed Gli1+ choroidal stromal MSC-like cell primary cultures obtained and characterized for the first time as a choroidal stromal-like cell in vitro model. Gli1GFP/+ cells were isolated from RPE/choroid tissue by sorting experiments at the Weill Cornell Medicine Flow Cytometry Core ( 9 ). MSC-like cells were unfrozen and cultured in complete medium (Sigma M4526 MEM supplemented with 20 % of FBS, 100 IU/ml penicillin, 100 µg/ml streptomycin and GlutaMAX) and maintained in 37 °C and 5 % CO 2 incubators. Each vial employed for experiments had less than three passages. MSC-like cells were plated and, after overnight starvation with serum-reduced (0.5 %) complete medium, cells were exposed to 10 ng/ml recombinant interleukin-1β (IL-1β), 100 ng/mL recombinant interferon-γ (IFN-γ), 100 ng/ml recombinant interleukin 10 (IL-10), 10 ng/ml recombinant basic fibroblast growth factor (bFGF) or hypoxia (1 % O 2 ) during 24 or 40 h. Culture medium was collected, centrifuged at 400 x g during 5 min and supernatants were used for determinations of Gal1 and VEGF secretion by ELISA according to manufacturing protocols (R&D Systems, MO, USA). Cells were taken in RLT Buffer for RNA extraction with the RNeasy Mini Kit (Qiagen, Hilden, Germany. Cover glasses were added to plate wells for immunofluorescence assays. Human fetal (hf) RPE cells from donors at 16 to 18 weeks of gestation were kindly provided by S. Miller (National Eye Institute, National Institutes of Health, Bethesda, Maryland) and cultured on Transwells with RPE medium as previously described ( 8 ). Cells were used in passage 1 or 2. Cultures were maintained at 37 °C, 5 % CO 2 and atmospheric O 2 (normoxia) or at 37 °C, 5 % CO 2 and 1 % O 2 (hypoxia) during 72 h. Culture medium from the apical and basal side of transwells was collected and employed as samples for determination of Gal1 secretion by ELISA according to the manufacturer's recommended protocol (R&D Systems, MO, USA). Tube formation assays were performed employing immortalized human umbilical vein endothelial cells (E4-HUVEC) as an EC model for tube assays. E4-HUVECs were cultured in complete medium (HyClone Medium 199 supplemented with 50 µg/ml ECGS, 10 mM HEPES, 50 µg/ml heparin, 20 % FBS, 100 IU/ml penicillin, 100 µg/ml streptomycin and GlutaMAX) in plates coated with 0.1 % gelatin and maintained at 37 °C and 5 % CO 2 . Typical cell passaging protocol using Accutase was employed. Tube assays were performed in 96-well plates coated with 80 µl Matrigel per well. Briefly, after Matrigel solidified, 15000 cells in 100 µl culture medium were plated per well. After 1 h, plates were maintained on the Zeiss Axio ObserverZ.1 widefield microscope (Zeiss, CA, USA) equipped a Plan-Apochromat 10X/0.45 M27 objective and Hamamatsu Flash4.0 V2 sCMOS camera 37 °C and 5 % CO 2 and photos were taken each 30 min during approximately 16 h. Positive control included treatment with 20 ng/ml recombinant mouse VEGF (Peprotech,, NJ, USA) and negative control was performed by adding 15 µM D,L-sulforaphane (Sigma, MO, USA). Each condition was performed by duplicate or triplicate. The tubular structures were quantified, and the percentages of inhibition (I%) were calculated as follows: I% = [1 − (total tube length treatment/total tube length control)] X 100. Eyes from adult mice were enucleated, washed in PBS and a hole was made in the cornea with a 22G needle. For fixation and storage, the eyes were placed in tubes with 2 ml Excalibuŕs Alcoholic Z-Fix (solution was provided by Excalibur Pathology, Inc, Norman, OK) and stored at room temperature. Samples were shipped to Excalibur Pathology, Inc. for paraffin section preparation. Paraffin-embedded retinal cross-sections were incubated in xylene (4 times, 5 min each), ethanol 100 % (3 times, 5 min each), ethanol 95 % (10 min) and washed in distilled water (10 min) and PBS (5 min). For immunostaining, sections were blocked in PBS containing 2 % BSA and 0.1 % Triton X-100 for 1 h at room temperature. The sections were incubated with primary antibodies overnight at 4 °C. The following primary antibodies were used: goat anti-Gal1 (1:50, cat. AF1245, R&D Systems, MO, USA), rabbit anti-NG2 (1:50, cat AB5320, Millipore, MA, USA), mouse anti-α smooth muscle actin (α-SMA, 1:30, cat. 2547, Sigma, MO, USA), rat-anti-PDGFRβ (1:30, cat. 14-1402-81, Invitrogen, MA, USA), rat anti-Sca1 (1:30, cat. AB51317, Abcam, MA, USA). Samples were also stained by Hoechst 33342 (1:100, 10 µM, Thermo Fisher Scientific, MO, USA). To detect ECs, Alexa Fluor 647-conjugated IsolectinB4 (1:100) was added to the primary antibody solution. After three washes in PBS for 10 min each, retinal sections were incubated for 2 h at room temperature with appropriate secondary antibodies conjugated with Alexa Fluor 488, or 568 (1:500; Invitrogen, Life Technologies, CA, USA). Sections were washed three times in PBS for 10 min each and once in water, dried at room temperature and mounted on slides in Fluoromount G (Invitrogen, MA, USA). Immunofluorescence was also performed in MSC-like cell primary cultures. Briefly, glass covers with cells were fixed with 4 % PFA in PBS, washed with PBS and blocked in PBS containing 2 % BSA and 0.1 % Triton X-100, followed by staining with anti-Gal1 polyclonal antibody and incubation with appropriate secondary antibody as described above. Images were taken of the Zeiss Axio Observer spinning disk confocal microscope (Zeiss, CA, USA) equipped with a Yokogawa scanner unit, Hamamatsu Evolve electron-multiplying charge-coupled device cameras and Zeiss Plan Apochromat 63×/1.46 oil immersion objective. RNA extraction was performed according to the protocol of Qiagen RNeasy Mini Kit and cDNA samples were obtained by the High-Capacity cDNA Reverse Transcription Kit (Life Technologies, CA, USA) according to the manufacturer's protocols. Real time (q) RT-PCR analysis was carried out in a StepOnePlus Real Time PCR System (Life Technologies, CA, USA) using SYBR Select Master Mix (Life Technologies, CA, USA) and the mouse forward and reverse primers (Integrated DNA Technologies, IA, USA) listed in Table 1 . The cycling conditions included a hot start at 95 °C for 10 min, followed by 40 cycles at 95 °C for 15 sec and 60 °C for 1 min. Specificity was verified by melting curve analysis. Results were normalized to β-actin. Relative gene expression was calculated according to the 2 −ΔΔCt method. Each sample was analyzed by triplicate. No amplification was observed using water as a template. Results are representative of at least three independent experiments. Statistical analysis using the GraphPad Prism 5.0 software was performed. F or Barlettś test was employed for equal variances. Accordingly, two-tailed unpaired t test (analysis between two different groups) and one-way analysis of variance (ANOVA) followed by Dunnetťs multiple comparison post-test or Kruskal-Wallis followed by Dunn's multiple comparison post-test (analysis among more than two groups) were employed to determine statistical significance. Two-way ANOVA and Bonferronís post-tests were employed to analyze interaction between the two independent variables on the dependent variable. Mean ± standard error (SEM) or median with interquartile range are shown in graphs accordingly if parametric or nonparametric tests were employed. A p-value <0.05 was considered statistically significant.

To unravel the physiologic relevance of Gal1 in visual function, we performed structural and functional studies in Lgals1 −/− mice. It was shown that lack of Gal1 did not alter the physiologic vascular and neural architecture in the retina during development and no morphologic difference in retinal histology was observed in adult Lgasl1 −/− mice by histological analysis and immunofluorescence microscopy ( 33 ). In this regard, we did not find alterations in the thickness or layer organization of Lgasl1 −/− retinas through SD-OCT ( Fig. 2a ). In accordance with the reported results ( 33 ), we observed equivalent scotopic and photopic responses between WT and Lgasl1 −/− mice by ERG ( Fig. S1 ), suggesting that Gal1 is not essential for the electrical responses of different cell types in the retina ( 33 , 49 ). Moreover, no differences in gene expression profiles of other galectins or related pathways were evident in bioinformatic analysis, which indicates that the absence of Gal1 does not likely trigger compensatory mechanisms ( Table. S1 ). Here, we also performed the OMR assay, a more sensitive and clinically relevant test that evaluates visual function, which is based on stereotyped head gestures in response to movement in the surrounding environment, serving to stabilize the visual image of tracking rotating stripes in the retina to allow for high resolution vision ( 35 ). Although mice proficiently tracked a rotating grating, the average of grating acuity was close to 0.37 c/d in WT and 0.33 c/d in Lgasl1 −/− mice ( Fig. 2b ). Our previous reported studies based on single-cell profiling, uncovered the molecular identity of the major choroidal cell types being 13 transcriptionally distinct clusters ( 9 ). This transcriptional signature provided new insights to understand the pathophysiologic mechanisms underlying RPE/choroid in physiologic and pathologic conditions. Here, through reanalysis of our single-cell RNAseq expression profiling, we identified the elevated expression of Gal1 in the RPE/choroid of adult mice ( Fig. 1a ). Since the functions of Gal1 in the outer retina are less understood, we focused on its involvement in the pathophysiology of RPE/choroid. We demonstrated that Gal1 is mainly expressed by perivascular and non-perivascular stromal-like cells ( Fig. 1b , c ), which are categorized into four different clusters in the mouse RPE/choroid ( 9 ). Interestingly, previous works demonstrated the constitutive presence of Gal1 in stromal cells in other tissues ( 50 – 52 ). In line with these findings EM studies showed alterations in the shape of stromal-like cells in the RPE/choroid and the morphology of choroidal vessels in Lgasl1 −/− mice ( Fig. 3a ). RNAseq data from RPE/choroid of Lgasl1 −/− mice revealed altered expression of genes associated to ocular extracellular matrix, ion transmembrane transport and glycosaminoglycan recognition, suggesting that Gal1 participates in stromal cell communication with chECs. Based on these findings, we employed an ex vivo mouse model of choroidal microvascular angiogenesis that maintains the local interactions of ECs with neighboring cells ( 46 ) to analyze Gal1 involvement in this process. RPE/choroidal explants from Lgals1 −/− mice triggered less sprouting than WT mice ( Fig. 4a ) suggesting that Gal1 is involved in choroidal angiogenesis. Similarly, choroidal sprouting was greatly reduced in explants from WT mice exposed to function-blocking anti-Gal1 mAb ( Fig. 4b ). Taking these results into account we hypothesized that Gal1 may participate in the communication between stromal cells and chECs modulating choroidal sprouting. Using an interactome analysis of scRNA sequencing we report here the communication between these cells via Gal1 signaling ( Fig. 5b ). Moreover, some populations of choroidal stromal cells express markers shared by pericytes such as PDGFRβ and NG2 ( Fig. 1c and S2 ), highlighting the involvement of Gal1 in vessel pericyte coverage as previously observed in tumor microenvironments ( 53 , 54 ). Pericytes can both stabilize and promote angiogenesis, and their importance is evident when considering cases of PDR and tumor growth, in which lack of pericytes is associated with aberrant angiogenesis ( 55 ). Despite the intense debate with regards to pericytes given the lack of cell type-specific markers, a number of studies suggested that these cells as well as stromal cells may share a common lineage and possess phenotypic plasticity which can be modulated by different environmental factors which modify their angioregulatory activity ( 55 ). On top of the physical stabilizing support rendered by pericytes, PDGF stimulates these cells to upregulate VEGF which promotes EC survival and recruits more pericytes to the neovasculature for microvessel maturation ( 56 ). Indeed, newly recruited pericytes form a protective barrier reducing the effect of VEGF inhibitors and explaining, at least in part, the plateau phase in long term anti-VEGF treatments ( 56 ). PDGF inhibitors were proposed in combination therapy with current anti-VEGF monotherapy for wet AMD, however, their clinical failure suggests the need to shift efforts to target other VEGF-independent pathways ( 56 ). Thus, similar to the effects in cancer settings, targeting the Gal1-glycan axis could represent an attractive approach to complement anti-VEGF blockade programs in wet AMD. We have recently characterized a perivascular stromal Gli1 + MSC-like cell population in the RPE/choroid target of the EC-secreted Indian Hedgehog. This cellular pathway is critical for survival of choroidal mast cells and inflammatory responses after tissue damage ( 9 ). Using primary cultures of choroidal stromal Gli+MSC-like cells, we observed that these cells secrete molecules which impair EC tube formation and angiogenesis ( Fig. 5c ). Indeed, single cell RNAseq profiling showed that choroidal stromal cells express VEGF, IGF-1, bFGF and are enriched in the anti-angiogenic molecule PEDF (Serpinf1). Moreover, the intercellular communication between stromal cells and chECs mainly via Serpinf1-Kdr signaling was also demonstrated ( Fig. 5e ). This is particularly interesting since most reports published up to date suggested that RPE expresses the highest levels of PEDF transcripts among ocular tissues and secretes this regulatory factor into the interphotoreceptor matrix to prevent pathological NV ( 57 ). Decreased levels of PEDF have been linked to several retinal diseases, such as AMD and diabetic retinopathy although PEDF-mediated mechanisms of neuroprotection and angiogenesis inhibition remain unknown ( 57 ). Our results unveil a key role of stromal-like cells in the control of angiogenesis. We found that exposure of Gli1 + MSC-like cells to IL-1β or hypoxia resulted in high VEGF expression and secretion ( Fig. 6a – b ), suggesting that this cytokine could couple angiogenesis to immunoregulation and tissue inflammation. Interestingly, IL-1β as well as IFN-γ attenuated PEDF mRNA production. Our results are in line with the hypothesis that choroidal stromal cells could have an important role during hypoxia and inflammation recalibrating the production of pro-angiogenic and anti-angiogenic mediators. Interestingly, we found an increased transcription of Lgals1 gene in stromal-like cells exposed to hypoxia and bFGF ( Fig. 6a ). Nevertheless, we could not detect an increase in Gal1 secretion in any of the conditions tested ( Fig. 6b ), even though the intracellular reticular pattern of Gal1 protein was modified in stromal-like cells following exposure to hypoxia and bFGF ( Fig. 6c ). Based on these findings, the involvement of Gal1 in stromal cell responses under pathologic conditions needs to be further explored in the choroid. In this regard, Gal1 has recently been proposed to be a critical soluble factor capable of regulating tumor–stroma crosstalk, proliferation, angiogenesis, and inflammatory responses, leading to enhanced tumor progression and metastasis ( 58 , 59 ). Indeed, Gal1 expression by stromal cells has been described during endometriosis ( 60 ) and tumor progression ( 58 , 61 , 62 ). MSCs are one of the most important stromal cells in the tumor microenvironment playing a major role in immunosuppression and tumor progression given their ability to produce a wide range of cytokines and growth factors capable of influencing tumor progression through secretion of a variety of growth factors including galectins ( 63 , 64 , 65 ). However, a recent report that compared MSCs present in stage II breast cancer and normal breast adipose tissue showed that tumor-associated MSCs exhibit higher secretion of VEGF compared with MSCs associated to normal tissues with no significant differences in Gal1 secretion ( 63 ). Moreover, Gal1 actively participates in the cross-talk between MSCs and other cell types, including basophils/mast cells, megakariocytes, myeloid progenitors and inflammatory fibroblasts during myelofibrosis progression ( 65 ). Thus, the impact of stromal-derived Gal1 in different pathologic settings may depend on the coordinated action of different cytokines and growth factors in different tissue microenvironments. Research on AMD has mainly focused on the deficit of RPE as it has been shown to be the final pathological change leading to vision loss. However, mechanisms of EC dysfunction in CNV remain elusive even though chECs are lost even before the occurrence of RPE dysfunction. In fact, patients with early vascular abnormalities have been found to acquire AMD and recent studies have found a significant involvement of EC pathology in CNV, highlighting the vascular etiology of this disease ( 56 ). The resultant loss of vascular support to RPE triggers pro-angiogenic signals which stimulate abnormal intravasation of choroidal vessels into subretinal layers, an effect which has been observed in cases of nonexudative AMD which progress to wet AMD ( 56 ). It is well known that VEGF released from the basal side of the RPE monolayer is required for the formation and maintenance of choriocapillaris. However, in pathological conditions like CNV, increased VEGF secretion from the hypoxic retina appears to be the key driver of pathological angiogenesis. In this model, activation of the HIF-1 pathway in RPE cells has been shown to foster inflammatory and vascular programs in CNV ( 56 ). Although Gal1 is highly expressed in RPE ( Fig. 1 ), we observed that this lectin is preferentially secreted toward the apical side in polarized hfRPE primary cultures ( Fig. 1d ). This result supports the hypothesis that RPE may not be the main physiologic source of the choroidal Gal1 pool. In line with this finding, RPE morphology observed by flat mount staining and EM images were preserved in Lgals1 −/− mice ( Fig. 3a – b ). However, Wu et al. reported results supporting Gal1 upregulation in RPE during laser-induced CNV, triggering VEGFR2 signaling and inflammation-driven angiogenesis ( 33 ). Accordingly, we found that under hypoxia hfRPE cultures increased Gal1 secretion mainly toward the basolateral side ( Fig. 6d ), suggesting that RPE could be the main choroidal source of Gal1 under pathologic conditions. Moreover, we found that recombinant Gal1 triggered a significant increase in ex vivo choroid sprouting under hypoxic conditions ( Fig. 6 ), suggesting that hypoxia could modulate the glycan profile of chECs towards a permissive pattern for Gal1 binding as previously observed in tumor settings ( 23 , 26 ). Related to the involvement of Gal1 in pathological conditions, it has been recently observed that human recombinantGal1 suppresses PDGF-BB-mediated proliferation of RPE cells via attenuation of PDGF-BB-induced AKT signaling, suggesting that Gal1 may function to block an exaggerated wound-healing response, as observed in severe proliferative vitreoretinopathy ( 66 ). Laser-induced CNV has become a standard model for assessment of novel therapeutic strategies and study of potential mechanisms; however, laser-induced acute injury does not recapitulate the chronic onset of pathological ocular NV ( 56 ). As the etiologic mechanisms of AMD remain elusive, numerous experimental models have attempted to mirror the clinical features of the disease both in its early and late stages, but none of them successfully recapitulated the pathological features of the disease ( 56 ). This complexity is further compounded by the differences in the ocular anatomy and histopathology between animal models, cell systems and patient tissues ( 56 ). However, despite their limitations, different experimental models have been essential to uncover the role of vascular networks in wet AMD, particularly those assessing effectiveness of anti-VEGF therapies. In summary, in this study we demonstrated the up-regulated expression of Gal1, a pro-angiogenic β-galactoside-binding lectin, in recently described stromal cell populations. Our results support a model in which stromal cells can orchestrate choroidal homeostasis, including modulation of vascular and inflammatory programs. Our findings set the bases for studying the crosstalk between inflammation and angiogenesis in ocular disorders such as AMD. Additionally, our results are in line with the fact that RPE could be a major source of Gal1 during hypoxia-driven CNV, suggesting that this lectin could link immune and vascular programs mediated by these cells with potential implications in modulation of immune privilege mechanisms. Supporting the clinical relevance of these findings, increased aqueous levels of Gal1 in the eyes of wet AMD patients were recently reported correlating positively with VEGF-A levels. Moreover, after intravitreal injections of anti-VEGF agents, the levels of Gal1 and VEGF-A decreased significantly ( 67 ). Taken together, our findings suggest that Gal1 derived from both stromal cells and RPE may have distinctive roles in physiological and pathological choroidal sprouting. In addition, our findings highlight the relevance of choroidal stromal cells as a critical cellular population that may contribute significantly to the pathophysiology of AMD. In summary, our results identify new pathways that contribute to choroidal sprouting and suggest potential therapeutic strategies based on modulation of the galectin-glycan axis to control resolution of wet AMD.

The outer retinal physiology is extensively supported by two underlying tissues: the RPE and the choroid. The RPE is a polarized cell monolayer that resides on a pentalaminar extracellular matrix referred to as Bruch's membrane (BrM) which, together with the RPE, forms the outer blood-retinal barrier (BRB) to create an immune privileged ocular environment that separates the eye from the choroid vasculature ( 1 ). The outermost layer of BrM in-turn forms an integral part of the choriocapillaris; a dense network of blood vessels of the choroid providing oxygen/nutrients to the outer retina whilst removing metabolic waste ( 1 ). Different mechanisms underlying epithelial cell polarity of RPE and its involvement in the pathogenesis of age-related macular degeneration (AMD) have been identified ( 2 – 5 ) and RPE phenotype has been shown to be crucial for BRB maintenance ( 6 ). Growing evidence suggests that cross-signaling between basement membrane and perivascular cells may regulate the functional properties of endothelial cells (ECs) and epithelial cells in organized organ–blood barriers ( 7 ). Accordingly, we have described a novel regulatory pathway where ECs secrete factors that remodel RPE basement membrane and modulate RPE tight junctions enhancing RPE barrier function ( 8 ). We have also reported a new endothelial-mediated immunomodulatory pathway in the choroid and characterized the tissue-specific transcriptomic features of choroidal ECs (chEC) ( 9 ). Our studies revealed the cellular and molecular landscapes of adult mouse RPE/choroid, uncovering the molecular identity of 13 transcriptionally distinct clusters within RPE/choroidal cells categorized as RPE, smooth muscle cells (SMC), melanocytes, hematopoietic cells, two different populations of Schwann cells, three clusters of ECs and four clusters of stromal cells ( 9 ). Interestingly, stromal cells, a group that had not been previously explored, constitute more than 40 % of the total cells in the adult mouse RPE/choroid ( 9 ). RPE and choroid act as a functional unit, and the dysfunction of their supportive roles can cause retinal degeneration and blinding diseases such as AMD. This incurable disease that affects 8.7 % of the population worldwide and 25 % people > 80 yr old, is characterized by photoreceptor loss secondary to dysfunction or death of the RPE and chEC ( 10 – 13 ). The etiology of AMD remains largely unknown, in part due to the limited information about the communication and signaling of different cell types that populate the choroid and the intercellular networks that maintain RPE/choroid tissue in a healthy and functional state. Galectin-1 (Gal1), a member of a highly conserved family of β-galactoside-binding lectins, has been implicated in several pathologies including autoimmune inflammation, neoplastic transformation, retinal neovascularization (NV) and neurodegeneration ( 14 – 22 ). Both Gal1 and its glycosylated ligands are upregulated by hypoxia ( 23 , 24 ) and control EC signaling ( 15 ) through mechanisms that are dependent or independent of hypoxia-inducible factor 1 (HIF-1α)- and VEGF- pathways ( 16 – 19 , 23 – 26 ). Previous studies demonstrated that a dynamic remodeling of the EC glycome may influence Gal1 binding to VEGFR 2 and trigger VEGF-like signaling preserving angiogenesis even in anti-VEGF refractory tumors ( 23 – 26 ). Similar to other galectins Gal1 has been shown to contribute to ocular angiogenesis ( 27 ), emerging as a potential new therapeutic target in neovascular retinopathies ( 28 ). In this regard, we have previously demonstrated that Gal1 is increased in the aqueous humor of patients with proliferative diabetic retinopathy (PDR) or neovascular glaucoma ( 29 ) and alterations in the underlying N-glycome of vascular pathologic processes have also been reported ( 30 ), suggesting the involvement of galectin-glycan lattices in ocular disorders. Accordingly, we found that Gal1 can imprint a neurovascular phenotype in a mouse model of oxygen-induced retinopathy (OIR) and delineate responses to anti-VEGF similar to those observed in tumors ( 29 ). Moreover, disruption of Gal1-glycan lattices improved retinal function in the OIR model ( 29 , 31 , 32 ). Although Gal1's role in the outer retina has not been studied in detail, this lectin has been suggested to promote choroidal NV (CNV) and subretinal fibrosis ( 33 ). These findings suggested that Gal1 could play a critical regulatory mechanism in the RPE/choroid with critical implications in AMD pathogenesis. Here, we demonstrate that stromal cells and RPE are the major sources of Gal1 in the RPE/choroid from adult mice and illustrate the crucial role of this lectin in the communication between these cells and choroidal vessels to maintain choroidal tissue homeostasis. Moreover, this study examines the differential effects of Gal1 from stromal cells and RPE cells under normoxic and hypoxic conditions with implications in both physiologic and pathologic processes.