Complexity of PEDF multifunctional signaling in retinal development and diseases

Abstract

PEDF has neuroprotective, anti-inflammatory, anti-oxidative stress, and anti-angiogenic actions, but its multifunctional mechanisms remain unclear. We map the developmental distribution of the PEDF receptors LR, LRP6, and ATGL in naïve and Pedf^−/− retinas, and define receptor-specific signaling using siRNA and pharmacological inhibition. These receptors are broadly expressed in retinoblasts and acquire cell-type–specific patterns during retinal maturation via PEDF-independent mechanisms. LR and LRP6 also colocalize at the outer limiting membrane (OLM). Functionally, PEDF activates AKT, STAT3, ERK, and p38 and suppresses β-catenin in both HUVEC and ARPE-19 cells, while mTOR activation remains receptor-independent. LR and LRP6 couple PEDF to STAT3 and β-catenin in HUVECs, whereas in ARPE-19 cells LR links PEDF to AKT/STAT3/ERK and LRP6 to AKT/β-catenin. ATGL inhibition abolishes PEDF-induced AKT/STAT3/ERK/p38 signaling in both cell types. A 17-mer PEDF mimetic (Ppx) recapitulates these effects. These findings help define a framework for PEDF pleiotropy and highlight the OLM as a therapeutic interface.

PEDF exerts pleiotropic actions in the retina through the differential distribution of its receptors and receptor-specific coupling to distinct signaling pathways.

Introduction

The onset and progression of many neurodegenerative diseases is influenced by activation of membrane receptors and their coupled signals. Understanding the ways in which receptor ligands influence these signals both increases knowledge of the disease processes and offers the possibility of manipulation to mitigate disease. Cell death, inflammation and oxidative stress are three interlinked processes common to many neurodegenerative diseases^1–4. Although inhibition of any one of these three may reduce degeneration, blocking all three is likely to be far more effective. There is strong evidence that one of the few factors that can inhibit all three processes is Pigment Epithelium-Derived Factor (PEDF), a 418 amino acid secreted glycoprotein of the serpin superfamily of serine protease inhibitors. PEDF was first identified in the supernatant of cultured human fetal retinal pigment epithelial cells and the interphotoreceptor matrix^5–7, but has since been shown to be widely expressed throughout the body, especially in the nervous system and retina, where it is a potent protective factor for neurons^8–10. Further studies provide evidence that in addition to its neuroprotective effects, PEDF has strong anti-inflammatory and antioxidant activities and potent endogenous anti-angiogenic properties that counter neovascularization induced by vascular endothelial growth factor (VEGF)^11–15. Modulations of PEDF expression levels are linked to a range of diseases including obesity and metabolic disorders, cardiovascular disease, diabetic eye and kidney complications, liver diseases, and reproductive system disorders¹⁶, and its loss with osteogenesis imperfecta type VI and other more subtle bone deficits¹⁷.

The diversity in functional specificity of PEDF in a range of biological systems raises the question of whether these actions are achieved through distinct receptors and independent downstream signals or through multiple converging pathways. To address this question, we studied expression and PEDF-activation of three well-characterized PEDF receptors in the retina: non-integrin 37/67-kDa laminin receptor/ribosomal protein(LR)¹⁸, low-density lipoprotein receptor-related protein-6 (LRP6)¹⁹, and adipose triglyceride lipase (ATGL), also referred to as patatin-phospholipase domain containing 2 (PNPLA2)²⁰.

Many LR studies have focused on cancer biology where PEDF binding affects tumor growth, apoptosis and metastasis²¹. In the nervous system, however, LR mediates cell adhesion to basement membranes and regulates the blood brain barrier, while PEDF-LR receptor interactions control neovascularization²². Expression of LR in the retina has been detected on cultured retinal ganglion cells, where it induces pro-survival mechanisms, and in the developing retinal vasculature^23,24. LRP6, in conjunction with the Frizzled receptor, is an important component of the Wnt signaling pathway linked to intracellular mechanisms that regulate multiple cellular processes. PEDF binds to the LRP6 receptor with high affinity and can attenuate the response to Wnt²⁵. Elevated LRP6 levels are detected in retinal samples from age-related macular degeneration (AMD) patients²⁶. PEDF modulates levels of ATGL/PNPLA2, a key lipase predominantly localized to the cytoplasm, and their interaction activates of a number of downstream signal pathways^20,27–31. Ablation of ATGL/ PNPLA2 leads to altered phospholipid composition in the retina and malformation of photoreceptors³². The findings make these receptors key players in several retinal pathologies.

Much of the information on PEDF receptors comes from binding studies using tissue homogenates or cell lines, and there is currently sparse information about the distribution of these receptors among cell types in a given tissue or the signaling mechanisms coupled to them. This article provides new data on the expression and distribution of these three major PEDF receptors in the retina where PEDF is synthesized and secreted throughout normal development and where it has beneficial effects on many retinal cell types. We present evidence that these receptors show cell type–dependent distribution patterns and elicit both shared and receptor-biased signaling responses. Our findings suggest that PEDF exhibits pleiotropic effects on cell survival, neovascularization, and angiogenesis, largely through a combination of cell-type specific receptors and downstream cascades. These provide a set of new therapeutic targets for some early retinal diseases. LR and LRP6 are also detected in the outer limiting membrane (OLM) which begs the question of their function in the OLM and the results of their interactions with PEDF in this retinal barrier structure.

Results

Developmental expression of mRNAs for the PEDF receptors LR, LRP6 and ATGL in naïve and Pedf null mouse retinas

Expression of all three major PEDF receptors was detected in RNA isolated from mouse retinas at postnatal (PN) day 1, PN7 and adult (8 weeks) (Fig. 1a). Using Polr2f as the reference gene, LR mRNA exhibited a mild non-significant decrease with age. In contrast, LRP6 expression was significantly higher in adult retinas compared with PN1 and PN7, while ATGL expression increased at PN7 and remained elevated into adulthood. Since expression levels of many receptors are regulated by their respective ligands, we also examined whether expression of these receptors was modulated when the PEDF gene is deleted. In adult Pedf^-/- mouse retinas, both LRP6 and ATGL mRNA levels were significantly increased relative to age-matched naïve wild-type controls, whereas LR showed a modest but non-significant increase (Fig. 1b). In addition, we compared expression levels among these receptors in adult mouse retinas, and found that LR was most abundantly expressed, followed by ATGL, whereas LRP6 was present at relatively low levels (Fig. 1C).

Fig. 1. Expression levels of LR, LRP6, and ATGL in mouse retinas, ARPE-19 cells, and HUVECs.

LR, LRP6, and ATGL are expressed in distinct patterns in mouse retinas, HUVECs, and ARPE-19 cells. a Developmental expression of LR, LRP6, and ATGL mRNA in mouse retinas at PN1, PN7, and adult stages. Data were normalized to Polr2f and expressed as 2^(-ΔΔCT)relative to PN1 for each target gene, with the mean value of the PN1 group set to 1 (PN1 as calibrator). Comparisons are intended within each gene across developmental stages, not between different genes (n = 6). b Relative expression of LR, LRP6, and ATGL mRNA in retinas from WT and Pedf^−/− mice (8 weeks). Data were normalized to Polr2f and expressed as 2^(-ΔΔCT) relative to the WT group for each target gene, with the mean value of the WT group set to 1 (WT as calibrator). Comparisons are intended within each gene between genotypes, not between different genes (n = 6). Relative mRNA abundance of LR, LRP6 and ATGL in mouse retinas (c), ARPE-19 cells (e) and HUVECs (g). Retina data were normalized to Polr2f, whereas ARPE-19 and HUVEC data were normalized to GAPDH, and presented as 2^(-ΔCT) (arbitrary units) (retina n = 6; ARPE-19 n = 3; HUVEC n = 3). Representative western blots of LR, LRP6, and ATGL in adult (8-week) C57BL/6J WT mouse retinas (d) and in ARPE-19 cells (f) or HUVECs (h) at passages 3–5. HSP90 was used as the loading control. Data are presented as mean ± SD. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.

Expression of mRNA levels of PEDF receptors LR, LRP6, and ATGL in two human cell line models of retinal vascular barrier function

Similarly, we found comparable RNA expression levels for all three receptors in human umbilical vein endothelial cells (HUVECs) and retinal pigmented epithelial cells (ARPE19), two human cell lines widely used as models of retinal vascular barrier function (Fig. 1e, g). When normalized to GAPDH and presented as 2^(−ΔCt), LR showed the highest relative transcript abundance in both cell lines, ATGL exhibited intermediate levels, whereas LRP6 was expressed at the lowest levels among the three receptors, consistent with the ordering observed in whole-retina samples. While cross-target comparisons using different primer sets remain semi-quantitative, primer efficiencies were validated, and assays were performed under consistent conditions to minimize bias (see Methods).

Protein expression levels of all three receptors in adult retina, ARPE-19, and HUVEC cells

Western blot analyses confirmed protein expression levels of these receptors in adult retina, ARPE-19, and HUVECs, with LR consistently produced the strongest signal among the three receptors under our assay conditions, whereas LRP6 and ATGL were lower but clearly detectable (Fig. 1d, f, h). Although we focused this study on examining LR, LRP6 and ATGL expression in the retina and cell lines, the other four known PEDF receptors, ATP synthase subunit beta (ATP5F1b), plexin domain-containing protein-1 (PLXDC1), plexin domain-containing protein-2 (PLXDC2) and vascular endothelial growth factor receptor 2 (VEGFR2), were also expressed in the retina, HUVECs and ARPE-19 cells, albeit at different levels (Supplementary Fig. 1). Studies of ATP5F1b are difficult to analyze because the vast majority of this protein is localized to the mitochondria and resolving the cell surface expression is challenging. PLXDC1 and PLXDC2 were not studied in further detail here because available antibodies are poorly characterized. PEDF interactions with VEGFR2 have been described previously^11,33,34.

Cellular distribution of PEDF receptors in the mouse retina by immunolabeling

PN1, PN7, and adult mouse retinal sections were immunolabeled using antibodies that recognize each of the three major PEDF receptors (Figs. 2–4), and the results are summarized in Table 1 and in the schematic of the graphical abstract.

Fig. 2. Distribution of LR in mouse retinas.

Identification of retinal cell types expressing LR. Representative immunofluorescence images of LR and LR with RBPMS, NfL, GFAP, GS, Calbindin or PKC-α in C57BL/6 wildtype mouse retinas. a–c Distribution of LR (green) in PN1, PN7, and adult (8 weeks) C57BL/6 wildtype mouse retinas. d–i Double labeling for LR with RBPMS, NfL, GFAP, GS, Calbindin or PKC-α (red) in adult (8 weeks) C57BL/6 wildtype mouse retinas. Nuclei are labeled with DAPI (blue). The region enclosed by the dashed square is further zoomed in. (Scale bars: 20 μM).

Fig. 4. Distribution of ATGL in mouse retinas.

Identification of retinal cell types expressing ATGL. Representative immunofluorescence images of ATGL and ATGL with RBPMS, NfL, GFAP, GS, Calbindin, or PKC-α in C57BL/6 wildtype mouse retinas. a–c Distribution of ATGL (green) in PN1, PN7, and adult (8 weeks) C57BL/6 wildtype mouse retinas. d–i Double labeling for ATGL with RBPMS, NfL, GFAP, GS, Calbindin or PKC-α (red) in adult (8 weeks) C57BL/6 wildtype mouse retinas. Nuclei are labeled with DAPI (blue). The region enclosed by the dashed square is further zoomed in. (Scale bars: 20 μM).

Table 1.

Partial layers of the adult mouse retina and the main distribution of PEDF receptors

Layer’s Name		LR	LRP6	ATGL
Ganglion cell layer (GCL)	Ganglion cells	++	+	+
	Astrocytes	+	+	++
Inner plexiform layer (IPL)		-	+	+
Inner nuclear layer (INL)	Amacrine cells	++	+/-	+
	Müller cells	++	+/-	++
	Bipolar cells	++	++	-
	Horizontal cells	++	+	++
Outer plexiform layer (OPL)		+	++	++
Outer nuclear layer (ONL)		+/-	+/-	+
Outer limiting membrane (OLM)		++	+	-
Photoreceptor outer segments		+	++	++

Summary of the localization of PEDF receptors in different retinal layers and cells. ++, extensive overlap. +, limited overlap. -, lack of overlap.

LR was localized to all retinal cell types at PN1, with the greatest labeling intensity in the developing outer retina (Fig. 2a). By PN7, the inner plexiform layer (IPL) layer was present in the retina but devoid of LR immunolabel (Fig. 2b). In the adult retinas, essentially all cells in the ganglion cell layer (GCL) and inner nuclear layer (INL) showed intense fluorescence for LR expression, while labeling was not detected in the IPL (Fig. 2c). The outer nuclear layer (ONL) had low-level punctate labeling throughout its thickness, and a diffuse band of labeling was observed toward the outer edge, most likely in the photoreceptor inner and outer segments. In a series of double-labeling studies aimed at identifying cell types expressing LR, we found that essentially all of the cells in the GCL, identified by the ganglion cell-specific marker RNA-binding protein with multiple splicing (RBPMS), also positively colabeled for LR (Fig. 2d). Other LR-positive cells in the GCL not expressing RBPMS were likely to be displaced amacrine cells. To study LR sublocalization in retinal ganglion cells in more detail, we co-labeled with an antibody against the light chain of the neurofilaments (NfL) protein (Fig. 2e). While the ganglion cell bodies showed areas of co-incident labeling, the nerve fiber bundles and dendrites did not, suggesting the LR expression is restricted to ganglion cell bodies. The neurofilament-positive horizontal cell processes in the outer plexiform layer (OPL) were also not co-labeled with LR antibodies. When co-labeled with glial fibrillary acidic protein (GFAP) antibodies, there were only small regions of overlap, suggesting limited or no expression of LR in astrocytes, the major cell type expressing GFAP in the normal tissue (Fig. 2f). Co-labeling with the more general glial marker glutamine synthetase (GS), confirmed the minimal overlap in the astrocyte layer (Fig. 2g). In addition, Muller fibers coursing through the IPL and the ONL showed no sign of LR expression. Muller cell processes ramifying the OPL also showed minimal overlap with LR. We did, however, observe occasional cells in the middle of the INL that appeared to be double-labeled. These were in the position expected for Muller cell bodies, suggesting that LR is expressed by Muller cells, but in a highly localized manner. We used calbindin labeling to identify horizontal cells and their processes and detected calbindin-positive horizontal cell bodies clearly expressing LR (Fig. 2h). Some of the calbindin-positive processes proximal to the cell bodies showed positive LR labeling, but those further into the OPL did not. The results are consistent with LR expression on horizontal cell bodies and proximal processes but not on more distal processes. A similar subcellular distribution was seen in bipolar cells, identified using antibodies against protein kinase C-alpha (PKC-α). There was clear double labeling of cell bodies, but very little of either dendrites in the OPL or axon terminals in the IPL. The inner half of the INL is almost entirely composed of amacrine cells. The extensive labeling in this region, but lack of labeling in the IPL, indicates that amacrine cells also express LR on their cell bodies but not their processes (Fig. 2i). Although occasional photoreceptor cell bodies at the outer edge of the ONL were immunolabeled, Most of the ONL showed no labeling. There was light labeling of the outer segment regions and a brighter band of label at the location of the outer limiting membrane.

These immunolabeling studies provide evidence that the LR receptor is expressed by most neuronal cell types in the adult retina, though most expression is confined to cell soma regions.

LRP6 expression was also present throughout the entire retina at PN1, with the brightest fluorescence detected in the more developed GCL and inner edge of the INL (Fig. 3a). At PN7, expression was visible in all three nuclear layers, with lower levels in the ONL and strong fluorescence in the developing OPL and the outer limiting membrane (OLM) (Fig. 3b). In the adult retinas, labeling was strongest in the GCL, the inner edge of the INL, the OPL, and the OLM, with weaker expression in the remainder of the INL (Fig. 3c). Labeling of photoreceptor inner/outer segments, but not cell bodies in the ONL, was also apparent. Double labeling with RBPMS antibodies indicated that the LRP6 labeling was coincident with ganglion cells (Fig. 3d). The displaced amacrine cells in the GCL (RBPMS-negative cells) were also LRP6-positive. Double labeling with neurofilament antibodies indicated that there was co-incident labeling of many cell bodies in the GCL, but no clear overlap of labeling of axons in the nerve fiber layer or dendrites in the IPL (Fig. 3e). GFAP labeling showed very little overlap with LRP6 in the astrocyte layer (Fig. 3f). This pattern was confirmed with GS double labeling (Fig. 3g), where no clear overlap was found between LRP6 and GS labeling in the astrocyte layer and the Muller fibers in the IPL and the ONL. There was, however, an overlap between LRP6 and GS at the level of the OLM. Calbindin-positive horizontal cell bodies did not label with LRP6 antibodies (Fig. 3h). There was strong apparent overlap between calbindin and LRP6 labeling in the OPL, but at this level of resolution, it was not possible to discern whether the yellow color was due to true double labeling or close apposition of fine horizontal and bipolar cell dendrites. Though less bright, calbindin-positive cells in the INL and GCL were also LRP6-positive. Bipolar cell bodies and proximal dendrites, detected with antibodies against PKC-α, expressed LRP6, but axons passing through the IPL did not (Fig. 3i). Our findings provide evidence that the LRP6 receptor is expressed by specific subregions of many neuronal cell types in the adult retina.

Fig. 3. Distribution of LRP6 in mouse retinas.

Identification of retinal cell types expressing LRP6. Representative immunofluorescence images of LRP6 and LRP6 with RBPMS, NfL, GFAP, GS, Calbindin or PKC-α in C57BL/6 wildtype mouse retinas. a–c Distribution of LRP6 (green) in PN1, PN7, and adult (8 weeks) C57BL/6 wildtype mouse retinas. d–i Double labeling for LRP6 with RBPMS, NfL, GFAP, GS, Calbindin, or PKC-α (red) in adult (8 weeks) C57BL/6 wildtype mouse retinas. Nuclei are labeled with DAPI (blue). White arrows point to the outer limiting membrane. The region enclosed by the dashed square is further zoomed in. (Scale bars: 20 μM).

ATGL labeling was diffuse and present as small puncta at all ages with no clear outlining of cell bodies. At PN1, the label was distributed across the retina, with the heaviest concentration in the GCL (Fig. 4a). Expression of this receptor was also seen at PN7, but it was stronger in the OPL (Fig. 4b) whereas in adults, ATGL expression was detected in the GCL, IPL, INL (with stronger labeling at the inner edge), OPL and ONL (Fig. 4c). Puncta were present over and around ganglion cells labeled with RBPMS antibodies (Fig. 4d). Only limited expression of ATGL coincided with neurofilament labeling suggesting that this receptor was confined to the ganglion cell bodies with less on the dendrites and axons (Fig. 4e). Double label with GFAP indicated expression of ATGL by astrocytes in the GCL (Fig. 4f). GS antibody labeling overlapped with ATGL antibody labeling in parts of the astrocyte layer but Muller endfeet and Muller fibers in the IPL, OPL, ONL and OLM did not show double labeling. Double labeling with calbindin antibodies suggested that positive cells in the GCL, the inner edge of the INL, and horizontal cell bodies were all ATGL-positive (Fig. 4h). There was also some double labeling seen with ATGL and calbindin antibodies in the OPL. This may explain the limited double label seen with bipolar cells as detected with PKC-α antibodies. No double label was detected on either the bipolar cell bodies or axon terminals. Because bipolar dendrites and horizontal cell processes are very closely apposed in the triad synapses of rod spherules, it is possible that the punctate double label seen in the OPL is due to this close proximity of processes from the two cell types rather than co-incident expression (Fig. 4i). ATGL labeling was also detected in the photoreceptor outer segments, but not the cell bodies (Fig. 4). There was extensive ATGL label in the OPL that was not accounted for by either Bipolar or horizontal cells. It is possible that this was due to expression in photoreceptor terminals. Overall, ATGL shows a more restricted expression than LR or LRP6 and differs in that it is also expressed in astrocytes.

Given the importance of the retinal vascular system in PEDF activity, we performed retinal whole-mount staining, using isolectin B4 (IB4) to label retinal blood vessels and antibodies against the three receptors for double immunostaining. The results showed that LR strongly labeled numerous cells in the inner retina and exhibited extensive colocalization with retinal vessels in the middle and outer retinal layers, whereas LRP6 and ATGL colocalized with vessels throughout the full thickness of the retina (Fig. 5a-c). These findings indicate that all three receptors are expressed in endothelial cells. The vascular expression was confirmed on a non-retinal endothelial cell line, HUVEC cells, where all three receptors showed diffuse and somewhat punctate expression consistent with membrane localization, and also substantial cytoplasmic labeling, particularly in perinuclear regions (Fig. 5d). To determine whether this pattern was conserved in retinal pigment epithelial cells, we further examined receptor localization in ARPE-19 cells. Compared with HUVECs, ARPE-19 cells showed weaker overall immunoreactivity for all three receptors. LR and ATGL displayed predominantly punctate cytoplasmic labeling, whereas LRP6 demonstrated considerable overlap with nuclear staining, indicating a distinct subcellular distribution pattern in these epithelial cells (Fig. 5e). From these findings, we suggest that there is extensive receptor dynamics with variable amounts expressed on the cell surface from a larger intracellular pool.

Fig. 5. Immunolocalization of LR, LRP6 and ATGL in IB4-labeled mouse retinal vasculature and in HUVECs and ARPE-19 cells.

LR, LRP6 and ATGL are expressed by the retinal vasculature. a–c Representative double immunofluorescence images of LR, LRP6 and ATGL (red) with IB4 (green) in adult (8 weeks) C57BL/6 wildtype mouse retinas. (Scale bars: 20 μM). d Representative immunofluorescence images of LR, LRP6, and ATGL (green) in HUVEC cells. (Scale bars: 20 μM). e Representative immunofluorescence images of LR, LRP6, and ATGL (green) in ARPE-19 cells. (Scale bars: 20 μm) The region enclosed by the dashed square is further zoomed in.

Overall, the expression studies show that there is considerable overlap in the expression of the receptors by both neuronal and non-neuronal elements of the retina.

PEDF activation of LR is specifically coupled to STAT3 activation, whereas PEDF activation of LRP6 is specifically coupled to β-catenin inhibition in HUVECs

PEDF is known to activate STAT3 in a time and concentration-dependent manner in various cell types and tissues^35,36. Here, we confirmed this PEDF action using HUVECs.The peak response to PEDF occurred at 10 min with 200 ng/ml (Supplementary Fig. 3a–d). These conditions were used in subsequent follow-on experiments. siRNA knockdown of LR almost completely blocked the STAT3 response to PEDF (Fig. 6a, b) while control siRNAs had no effect. On the other hand, knockdown of LRP6 did not block PEDF activation of STAT3 (Fig. 6E, F) suggest specificity of PEDF coupling to STAT3 via LR. Other studies show that PEDF binds to the co-receptor LRP6 in the Wnt/β-catenin pathway and inhibits Wnt signal transduction in liver, retina, and other tissues^19,25,37. We also demonstrated this response to PEDF in the human HUVEC vascular cells. Our results show that the effect of PEDF in inhibiting β-catenin activation also occurred at 200 ng/ml PEDF and 10 minutes of cell stimulation (Supplementary Fig. 3e–h). Further study found that after knocking down LRP6, basal levels of active β-catenin decreased, and after adding PEDF, these levels returned to normal expression (Fig. 6g). However, knockdown of LR had no significant effect on the inhibition of β-catenin activation by PEDF (Fig. 6c). These results suggest that these two receptors are coupled to different signaling pathways in promoting PEDF downstream activities.

Fig. 6. LR is coupled to active STAT3, whereas LRP6 is coupled to inhibit β-catenin.

PEDF activates STAT3 phosphorylation through LR but not LRP6, whereas PEDF inhibits β-catenin activation through LRP6 but not LR in HUVECs. We transfected HUVECs with LR siRNA, LRP6 siRNA, or scrambled siRNA. Cells were stimulated with 200 ng/ml PEDF for 10 min. a–d Representative Western blots and densitometric measurements of LR, P-STAT3, STAT3, active β-catenin, and β-catenin in HUVECs. Compared with the control and scrambled siRNA groups, the expression level of LR in the LR siRNA group decreased significantly, and the phospho-STAT3 level was inhibited after PEDF stimulation, whereas the inhibition of β-catenin was not affected. e–h Representative Western blots and densitometric measurements of LRP6 P-STAT3, STAT3, active β-catenin, and β-catenin in HUVECs. Compared with the control and scrambled siRNA groups, the expression level of LRP6 in the LRP6 siRNA group decreased significantly, and the inhibition of β-catenin was affected, whereas the STAT3 level was still activated after PEDF stimulation. HSP90 was used as the control for total protein loaded. Data are expressed as the mean ± SD. (n = 3/group. *P < 0.05; **P < 0.01).

To further examine signaling mechanisms coupled to LR and LRP6, we screened PEDF-induced activation of a broader range of signal transduction enzymes in LR and LRP6 knockdown cells. As shown in supplementary Fig. 4, neither knockdown blocked PEDF activation of AKT, mTOR, ERK or p38, suggesting that these PEDF actions may be due, in part, to ATGL or another PEDF receptor.

PEDF activation of ATGL regulates multiple signal pathways in HUVECs

The other major PEDF receptor detected in the retina is ATGL. Although we tried multiple siRNAs that reduced ATGL mRNA levels, we were unable to reduce ATGL protein levels, suggesting a long mRNA half-life or a unique nuclear or vesicular compartmentalization of the ATGL protein. To examine possible signaling linked to the ATGL receptor in HUVECs, we used the inhibitor Atglistatin (ATGLi). Treatment with ATGLi almost completely removed the phosphorylation of AKT, STAT3, and p38 in response to PEDF without significantly changing the basal level of phosphorylation (Fig. 7a, b, d, and f). ATGLi was not, however, able to block the PEDF-induced phosphorylation of mTOR, nor the PEDF-induced decrease in β-catenin activation (Fig. 7c, g). ATGLi decreased the basal level of ERK phosphorylation and blocked its PEDF-induced increase (Fig. 7e).

Fig. 7. ATGL is coupled to AKT, STAT3, ERK, and p38 pathways.

ATGL is coupled to AKT, STAT3, ERK, and p38 pathways. We treated HUVECs with ATGL inhibitor or DMSO at the same concentration as the inhibitor solvent, then the cells were stimulated with 200 ng/ml PEDF for 10 min. a Representative Western blots of different signaling pathways in HUVECs. b–g Densitometric measurements of different signaling pathways in HUVECs. Compared with control and DMSO treatment groups, the ATGL inhibitor blocked the PEDF-induced phosphorylation of AKT, STAT3, ERK, and p38. Data are expressed as the mean ± SD. (n = 3/group. *P < 0.05; **P < 0.01; ***P < 0.001).

Receptor-specific and cell-type–dependent coupling of PEDF signaling in ARPE-19 cells

We next examined whether PEDF engages similar signaling pathways and receptor mechanisms in ARPE-19 cells. Under the same stimulation conditions used in HUVECs (10 min), PEDF induced dose-dependent increases in phosphorylated AKT, mTOR, STAT3, ERK, and p38, while active β-catenin levels decreased, recapitulating the overall trends observed in HUVECs (Supplementary Fig. 6). Based on these results, receptor knockdown and pharmacological inhibition assays were performed to determine whether receptor-mediated coupling differs between cell types. LR knockdown abolished PEDF-induced phosphorylation of AKT, STAT3, and ERK, while LRP6 knockdown prevented AKT phosphorylation and impaired β-catenin suppression (Fig. 8a–h). Notably, LRP6 knockdown reduced basal active β-catenin in both cell types; however, PEDF restored β-catenin levels only in HUVECs, but not in ARPE-19 cells, indicating cell-type–specific control of β-catenin by LRP6 (Fig. 8i–p). Inhibition of ATGL eliminated PEDF-induced phosphorylation of AKT, STAT3, ERK, and p38, matching the ATGL dependency observed in HUVECs (Fig. 8q–w). Taken together, these findings indicate that ATGL broadly governs PEDF-activated signaling in both cell types, that LR-STAT3 and LRP6-β-catenin coupling is shared across cell types, and that in ARPE-19 cells, LR and LRP6 additionally regulate AKT and ERK pathways, suggesting cell-type-specific functional diversification of PEDF receptor signaling in the retina.

Fig. 8. Shared and receptor-biased PEDF signaling through LR, LRP6 and ATGL in ARPE-19 cells.

Receptor-specific coupling of PEDF signaling pathways in ARPE-19 cells. ARPE-19 cells were transfected with LR siRNA, LRP6 siRNA, or scrambled siRNA, or treated with an ATGL inhibitor or vehicle control (DMSO), and then stimulated with 200 ng/mL PEDF for 10 min. a–h Representative Western blots and densitometric measurements of LR and downstream signaling proteins in ARPE-19 cells following LR knockdown. i–p Representative Western blots and densitometric measurements of LRP6 and downstream signaling proteins in ARPE-19 cells following LRP6 knockdown. q–w Representative Western blots and densitometric measurements of downstream signaling proteins in ARPE-19 cells following ATGL inhibition. HSP90 was used as the loading control. Data are expressed as mean ± SD (n = 3/group). *P < 0.05; **P < 0.01; ***P < 0.001.

Ppx, a modified PEDF peptide mimetic, interacts with multiple PEDF receptors, resulting in actions similar to the full-length PEDF protein

PEDF is a large 418 amino acid glycoprotein, raising the possibility that the multiple actions of this protein are mediated by different portions of the molecule. Several PEDF peptide fragments are shown to be biologically active^30,38–40. The PEDF mimetic, Ppx, is a 17-amino acid fragment that is a modified peptide with improved biological activity to its native counterpart and can reverse a number of ocular pathological events⁴¹. We analyzed its ability to activate several signal transduction enzymes that are regulated by PEDF and showed that both PEDF and Ppx activate AKT, mTOR, STAT3, ERK, and p38 and inhibit β-catenin activation in HUVECs (Fig. 9), suggesting that this peptide interacts with multiple PEDF receptors and can serve as a readily available and less costly pharmacological tool to dissect the multifunctional nature of PEDF with possible implications for retinal vascular diseases. In addition, the results suggest that all the downstream signaling activities measured in this study are due to one small region of the PEDF molecule that can be modified to improve bioactivity.

Fig. 9. Effect of PEDF and Ppx on HUVECs signaling pathways.

Both PEDF and Ppx can activate AKT, mTOR, STAT3, ERK, and p38 phosphorylation and inhibit β-catenin activation in HUVECs. a, b Representative Western blot of AKT, mTOR, STAT3, ERK, p38, and β-catenin protein levels and their activation after stimulation of HUVECs with different concentrations of PEDF and Ppx for 10 min. c–n Densitometric analysis of the blots. Data are expressed as the mean ± SD. (n = 3/group).

Our findings highlight downstream signaling events of PEDF and the importance of HUVEC and RPE cell lines as important retinal cellular models for knockdown pharmacological studies of receptor function and response to PEDF in pathologically challenged environments.

Discussion

Most retinal diseases involve multiple pathological events, including some of the most challenging, such as inflammation, oxidative stress, neurodegeneration, and neovascularization. There is strong evidence that PEDF, as an endogenous multifunctional protective factor, can mitigate these events. PEDF protects many types of neurons from ischemic injury, oxidative stress, and other neurodegenerative damage^5,6,42–44. It exerts anti-inflammatory effects, partly by inhibiting the Wnt signaling pathway⁴⁵. Importantly, PEDF maintains microvascular homeostasis and resists neovascularization in several systems^13,46–48. However, it is not yet established whether this serpin acts on multiple cell types and whether the response can be segregated among cells through different receptors.

The three PEDF receptors, LRP6, LR, and ATGL, studied here are strongly associated with retinal diseases, including AMD and DR. For example, deletion of ATGL in mice causes photoreceptor and retinal pigment epithelium dysfunction, with RPE cells exhibiting accelerated cellular aging^32,49. Similarly, activation of LRP6 and increased LRP6 protein levels are linked to AMD-related changes in both human and mouse retinas, as well as to proliferative diabetic retinopathy and choroidal neovascularization^26,50,51. The 67-kDa laminin receptor (67LR) also shows a strong correlation with developing neovascular disease. This receptor is preferentially expressed by proliferating retinal endothelial cells during developmental angiogenesis and in models of ischemic or proliferative retinopathy. Its expression increases during active vessel growth and declines as vessels become quiescent. These findings identify 67LR as a key mediator of endothelial activity in retinal vascular disease^24,52. Collectively, these studies highlight all three receptors as potential therapeutic targets for retinal disease. We provide a clearer understanding of their distribution and signaling in the retina, RPE cells, and HUVECs, which is essential to support the refinement of targeted therapeutic strategies.

The retina, which contains a diverse range of cell types, expresses all seven known PEDF receptors. We show that the expression of several receptors studied here is also detected in human endothelial and RPE cell lines, two important cell types in the retina that govern neuroprotective and neovascularization outcomes (Fig. 1 and Supplementary Fig. 1). The three major receptors studied here in detail, and our previous studies of the VEGFR2/PEDF receptor, showed developmental changes in PEDF receptor expression, with widespread expression in early postnatal retinoblasts followed by segregation among various retinal cell types as the tissue develops, which may indicate segregation of specific functions as the retinal cell types mature. For example, previous studies show that the expression of the 67 kD LR is closely linked to retinal vascular development in the neonatal mouse²⁴. LR is highly expressed from PN1 to PN7, a period when the retinal capillary network matures, and endothelial cells become established. It then decreases with the maturation of the vasculature and the immobility of blood vessels and endothelial cells. In our dataset, LR expression at PN7 and in adult retinas showed only a mild reduction without statistical significance, indicating that LR expression may persist beyond the period of active vascular remodeling. Such differences may reflect strain-specific effects, methodological differences (e.g., qPCR vs. protein detection), or cell-type heterogeneity in the whole-retina samples. However, the distribution and continued expression in postmitotic retinal neurons strongly suggest additional functions for LR. Similarly, studies of tumor cell lines have implicated LR in cell proliferation, but this is not likely relevant for the mature retina⁵³. LR has previously been described on isolated mature ganglion cell cultures, and our studies of its expression in vivo confirm this²³. In addition, we found expression of this receptor on all three major neuronal cell types in the INL (Table 1). Lack of labeling in the NFL, IPL, and OPL indicates that LR expression is restricted to cell body regions of inner retinal neurons. Interestingly, in the outer retina, the reverse was found. We could not detect LR on the photoreceptor cell bodies but observed labeling in the outer segments. Several laminins have been described in the interphotoreceptor matrix, a region that is also rich in functionally active PEDF⁵⁴.

LRP6 is a member of the low-density lipoprotein (LDL) receptor superfamily of cell-surface receptors. Together with Frizzled, it plays an essential role in the activation of the canonical Wnt/β-catenin signaling pathway^55,56, which is involved in many biological activities, including cell proliferation, migration, and development. Activation of LRP6 leads to inhibition of β-catenin phosphorylation and the eventual activation of Wnt target genes⁵⁷. LRP6 is widely expressed in tissues and is present on multiple cell types in the retina (Table 1). Knockdown of LRP6 does not broadly affect the major signal pathways but alters activation of β-catenin (nuclear localization), supporting previous evidence that PEDF decreases β-catenin activation^19,25. Our finding is that LRP6 knockdown reduced basal active β-catenin in both HUVECs and ARPE-19 cells. However, PEDF restored β-catenin activation only in HUVECs, not in ARPE-19 cells, revealing a cell-type-specific dependency on LRP6 for β-catenin regulation. Wnt signaling through the LRP6 receptor is important in embryonic development of many tissues where PEDF is expressed. Therefore, it is possible that PEDF regulates development, in part, through interaction with LRP6^7,58. In ARPE-19 cells, LRP6 knockdown additionally impaired PEDF-induced AKT phosphorylation. Previous studies have shown extensive crosstalk between Wnt signaling and the PI3K–AKT–mTOR axis, and our findings raise the possibility that LRP6 integrates PEDF inputs into both β-catenin and AKT pathways⁵⁹.

Previous studies have also shown that PEDF is a potent modulator of the MAP kinase pathway^60–63. In our study, inhibition of ATGL blocked PEDF-induced phosphorylation of AKT, STAT3, ERK, and p38 in both HUVECs and ARPE-19 cells, placing ATGL as a proximal node that couples PEDF to multiple intracellular pathways. In contrast, LR knockdown selectively impaired STAT3 activation in HUVECs, whereas in ARPE-19 cells, LR knockdown additionally suppressed PEDF-induced ERK and AKT phosphorylation. Together with the differential effects of LRP6 knockdown on β-catenin and AKT signaling in the two cell types, these findings argue that PEDF signaling is not executed through a single linear cascade, but rather through receptor-specific modules that confer cell-type–dependent tuning of downstream signaling responses.

While PEDF activates mTOR, none of the three receptors we studied appears to be involved in its phosphorylation-induced activation. In previous studies, we found that mTOR is activated by the enzyme PDK1, which, in turn, is activated by PI3kinase⁶⁴ through a number of growth factor receptors, including IGF1 and PDGF^65,66. Thus, it is highly likely that other PEDF receptor interactions remain to be identified.

Since PEDF is a potent neuroprotective factor, it is not surprising that many of the pathways we found coupled to PEDF receptors are implicated in neuroprotective events. For example, active STAT3 protects retinal ganglion cells and neurons in a number of other neurodegenerative diseases^67–70. Similarly, in a neurotoxic prion peptide model, LRP6-mediated canonical Wnt/β-catenin signaling supports neuronal survival⁷¹. Moreover, prior studies have reported that multiple downstream pathways initiated by LR signaling are associated with neuroprotective outcomes^72,73. AKT is a common target of growth factors and can protect neurons both through regulating the cell death machinery and by regulating nuclear proteins⁷⁴. Similarly, ERK can protect neurons in culture from excitotoxic damage and the hippocampus from hypoxic injury in vivo^75,76. Unlike the others, activation of p38 has often been associated with neuronal apoptosis^77,78. On the other hand, activation of both p38 and ERK is found to be neuroprotective in a model of glutamate excitotoxicity⁷⁹. Thus, further downstream studies are essential to follow pathways activated by PEDF binding to its receptors in retinal development, homeostatic control, and diseases.

Both LR and ATGL have some cytoplasmic/nuclear localization in addition to their presence on the plasma membrane. Previous studies have shown that endogenous ATGL is localized to the external surface of lipid droplets in adipocytes' cytoplasm, but its precise subcellular localization in endothelial cells is unknown. The LR receptor is also detected on both the cell surface and in the nucleus, suggesting that it may be linked to processes that carry external signals into the nucleus⁸⁰. By contrast, LRP6 localized to the plasma membrane, cytoplasm, and nucleus in HUVECs, whereas it appeared more prominently nuclear in ARPE-19 cells. Although LRP6 is classically described as a plasma-membrane Wnt co-receptor, prior work has documented nuclear pools of LRP6 in differentiating cells and demonstrated that the LRP6 intracellular domain can translocate to the nucleus^81,82. The function of intracellular pools of PEDF receptors is not clear, but is interesting to note that this serpin itself has been found in the nucleus of several cell types^7,58,83–85. It is possible that PEDF is transported to the nucleus via one or more of its receptors or by its putative nuclear localization signal⁸⁵. Since PEDF has a nuclear localization motif, it is also possible that PEDF is shuttling LR and ATGL via its nuclear localization motif^85,86.

LR and LRP6 were also detected in the OLM and may engage in barrier function. This structure lies adjacent to the interphotoreceptor matrix that we have previously shown contains an abundance of RPE-secreted PEDF. How PEDF activates these receptors in the OLM and their signaling in barrier function is still unclear from our results, but suggests that the OLM may be a key target for early therapeutics in early developing retinal diseases.

One interesting common feature of molecules identified as PEDF receptors is that many of them have strong interactions with other ligands associated with development, neuroprotection, inflammation, and angiogenesis. For example, LRP6 is a receptor for Wnt and plays a major role in transducing Wnt signals during development and in the adult. PEDF interaction with LRP6 inhibits Wnt responses; thus, this serpin could be thought of as a receptor modulator^25,87. In previous studies, we also found that PEDF interacts with the VEGFR2 receptor and, while it had no effect on the normally inactive receptor, it was able to inhibit VEGF-induced signals by blocking VEGFR2 autophosphorylation, possibly by preventing receptor dimerization³³. LR binds the extracellular matrix protein laminin as well as PEDF^88,89. How binding to one affects the other has yet to be determined.

PEDF is a multifunctional protein, and its neuroprotective and anti-angiogenic functions have been preferentially localized to different peptide regions. The neuroprotective region has been further dissected, and a 17mer modified peptide (Ppx) carries much of this activity^{30,41,90–92}. The ability of Ppx to activate AKT, p38, and ERK all suggests it can act through the ATGL receptor in HUVECs. Ppx can, however, activate mTOR and β-catenin, functions linked to LRP6 and an as-yet-undefined receptor. Because ATGL can influence STAT3 activation, it is unclear whether the action of Ppx on this protein is due to ATGL or to LR.

There is extensive evidence that PEDF plays multiple roles both in normal retinal homeostasis and in retinal disease progression, but less is known about how PEDF signals are transmitted through receptors to various retinal cell types. In this study, we found that each of the three PEDF receptors studied showed a characteristic developmental pattern of expression in the retina, an expression that was increased in the absence of PEDF. PEDF exerts its multiple actions both through differential distribution of receptors and by coupling to different signal pathways, as summarized in the graphical abstract. The study also provides a set of key cellular targets for early-stage retinal diseases, including AMD and diabetic retinopathy, for which there are currently poor early management strategies. There are clearly other PEDF binding partners in the retina, as this protein activates mTOR but not through any of the receptors studied here. PEDF is secreted by RPE cells into the interphotoreceptor matrix, and this allows it to regulate both photoreceptor survival and function as well as cells in the inner retina through the Muller cell microvilli making up the OLM. The identification of a small PEDF peptide mimetic that mimics actions of the holoprotein on multiple signal pathways suggests that this mimetic has significant potential for pharmacological intervention in the many retinal diseases in which PEDF has been implicated

As we gain a better understanding of other PEDF receptors, we will obtain a more comprehensive understanding of how this multifunctional protein alters cell physiology directly or by modulating responses to other factors.

Methods

Materials

Unless otherwise stated, all materials were from Thermo Fisher. PEDF was collected from the culture supernatant of HEK-293T cells transiently transfected with a CAG promoter–driven expression plasmid encoding full-length human PEDF (SERPINF1; NM_002615.7), as described by Zhang et al.⁹³. PEDF concentrations were measured using a commercial ELISA kit (R&D Systems, Minneapolis, MN, USA, No. DY117705). Ppx (Tombran-Tink et al., “Functional Peptide Analogs of PEDF”; United States Patent No.: US 9,611,314 B2; Date Issued: Apr. 4, 2017) is a modified 17-amino acid peptide containing the neuroprotective region of PEDF (90–106)⁴¹ and was a gift from Skyran Biologics, Inc. (Harrisburg, PA, USA). Atglistatin, an ATGL inhibitor, was obtained from Millipore Sigma.

Animals

C57BL/6J male and female mice were purchased from Vital River (Beijing, China). Pedf^-/- mice have been described previously and were maintained on a C57BL/6 background⁹⁴. We have complied with all relevant ethical regulations for animal use. Animal care and experimental procedures were carried out according to the ARVO Guidelines for the Use of Animals in Ophthalmic and Vision Research and were approved by the Laboratory Animal Management Committee, Tianjin Medical University Eye Hospital (Approval No. TJYY2023120232). Mice were housed and bred at the Laboratory Animal Center of the Tianjin Medical University Eye Institute, Tianjin Medical University Eye Hospital (Tianjin, China). All mice were housed in the same facility under identical husbandry conditions, in stable social groups with appropriate bedding and environmental enrichment, and maintained on a 12-h light/dark cycle with standard chow and water available ad libitum. Animals were handled gently by trained personnel to reduce handling-related stress, and all efforts were made to minimize pain and distress. For terminal tissue collection, mice were euthanized by rapid cervical dislocation performed by trained personnel in accordance with institutional and international guidelines, after which eyes were immediately enucleated. The study comprised two prespecified comparisons: a developmental series using wild-type (WT) mice at postnatal day 1 (PN1), postnatal day 7 (PN7), and adult (8 weeks), and a genotype comparison at 8 weeks between age-matched WT (control) and Pedf^-/- mice to determine the effects of PEDF deletion on PEDF receptor expression. Group sizes were set on the basis of relevant literature and the ethical imperative to reduce the number of animals used. Each group included n = 6 biological replicates (experimental unit = single mouse) and at least 3 technical replicates per mouse. No additional inclusion/exclusion criteria (e.g., sex, weight, litter, glucose levels) were set, and no animals, experimental units, or data points were excluded from analysis. Within each group, animals selected for retinal collection were chosen at random. The timing of euthanasia/tissue collection was based only on the prespecified age stage and genotype (PN1, PN7, WT 8 weeks, Pedf^-/- 8 weeks). Group allocation by age/genotype was necessarily known during allocation and conduct.

Cell culture and reagents

HUVEC and ARPE19 cells were acquired from American Type Culture Collection (ATCC, Manassas, VA, USA). In this study, both cell lines were used at passage 3–5 for all studies and grown to ~70-80% confluence to ensure consistent growth kinetics and responsiveness to stimulation. HUVECs were routinely cultured in Endothelial Cell Medium (ECM, ScienCell, Cat. No. 1001) containing 5% fetal bovine serum (FBS, ScienCell, Cat. No.0025), 1% endothelial cell growth supplement (ECGS, ScienCell, Cat. No.1052), and 1% antibiotic solution (P/S, ScienCell, Cat. No.0503) at 37 °C and 5% CO₂. ARPE-19 cells were cultured in DMEM/F12 minimal medium (Gibco, Cat. No.11320033) containing 10% FBS (Gibco, Cat. No.A5256701), 1% P/S (Gibco, Cat. No.15140122) solution at 37 °C and 5% CO₂. ARPE-19 cells contained pigment granules, as shown in Supplementary Fig. 5. All cells were routinely tested for mycoplasma infection. Both cell lines are also routinely used as models to study retinal disease mechanisms^95,96.

Quantitative real-time PCR

All PCR processes were carried out according to the manufacturer’s recommendations for PCR reagents. Total RNA was extracted from freshly isolated mouse retinas, or cultured cells (HUVECs and ARPE-19) using the EZ-press RNA Purification Kit (EZBioscience,Cat.No.EZB-RN4). The concentration and quality of RNA were examined spectrometrically (Nanodrop 2000; ThermoFisher), and 1 µg RNA was used to synthesize cDNA using the Color Reverse Transcription Kit (EZBioscience, Cat. No. A0010CGQ). The resulting cDNA concentrations were measured spectrophotometrically, and quantitative RT-PCR (qPCR) was performed in triplicate for each sample using SYBR Green (EZBioscience, Cat. No. A0012-R1).

PCR parameters used were as follows: cDNA: 500 ng; Primer: 4 ng; Cycle number: 45 after a Hot Start at 95°C for 10 minutes. For mouse retinal samples, Polr2f was used as the internal reference gene for normalization; for cultured HUVECs and ARPE-19 cells, GAPDH was used as the internal reference. CT values from technical replicates were first averaged for each biological sample, and target expression was normalized to the reference gene using ΔCT = CT_target − CT_ reference. For comparisons within the same target across experimental groups (e.g., developmental stage or genotype), relative expression was calculated using the comparative CT method and presented as 2^(−ΔΔCT) with the mean value of the control group set to 1 to facilitate visualization of fold changes. In addition, to enable comparison of relative transcript abundance across targets within the same sample type, data were also presented as 2^(−ΔCT) (reference gene -normalized) where indicated. Gene-specific primers were designed using the NCBI Primer-Blast primer design tool (Primer3web version 4.1.0) (Supplementary Data 1). Amplification efficiency for each primer set was determined from standard curves generated using serial (limiting-dilution) cDNA dilutions. The slopes of the standard curves ranged from –3.58 to –3.10, corresponding to amplification efficiencies of approximately 90–110%, with correlation coefficients (R²) > 0.99, meeting MIQE quality criteria⁹⁷ (Supplementary Data 2). Both melting curves and gel electrophoretic analyses were used to confirm the DNA quality and gene expression levels. PCR products were resolved and verified by 2% agarose gel electrophoresis containing 10 μg/mL ethidium bromide.

Western Blot

Mouse retinas, ARPE-19 cells, and HUVECs were harvested and lysed in RIPA buffer (Solarbio, Cat. No.R0020) containing PMSF (Solarbio, Cat. No.P0100) and phosphatase inhibitor (CST, Cat. No.5870) for 20 minutes, and total protein concentrations were measured using Bicinchoninic Acid protein assay (Solarbio, Cat. No: PC0020). Twenty micrograms of protein for each sample or 3 microliters of Multicolor Prestained Protein Ladder (Epizyme, Cat. No. WJ103) were separated by 10% sodium dodecyl sulfate polyacrylamide gel electrophoresis (EpiZyme, Cat. No. PG112) for two hours using Tris-Glycine buffer (25 and 250 mM respectively) and gels subsequently transblotted to a polyvinylidene fluoride membrane (Millipore, Cat. No.03010040001) using Tris-glycine buffer containing 20% methanol. Membranes were subsequently blocked using 5% nonfat dry milk (BD Difco, Cat. No.232100) for two hours at room temperature, then incubated overnight at 4°C with a primary antibody. Antibodies used in the study are listed in Supplementary Data 3. Membranes were then washed and incubated in an appropriate secondary antibody for two hours at room temperature. The processed blots were developed using Immobilon ECL reagent (GE, Cat. No.RPN2232) and imaged using a transilluminator (Tanon, Shanghai, China). After labeling with antibodies against phosphorylated/activated proteins, membranes were stripped in Stripping Buffer (Cwbiotech, Cat. No.CW0056), and reprobed with antibodies against the unphosphorylated parent protein. All incubations were carried out using optimized antibody concentrations and dilution volumes to avoid potential variation due to antibody affinity differences or uneven blot coverage. Protein band densities were calculated using NIH ImageJ software (Version 2.1.0, National Institutes of Health, Bethesda, MD, USA) and phosphorylated signals were normalized to the total protein.

Immunolabeling

Eyes of euthanized mice were removed, fixed at room temperature in 4% paraformaldehyde (PFA, Sigma, Cat. No.158127) for 2 hours, before removing the anterior segment. Eyecups were placed in 20% sucrose (Sigma, Cat. No.S0389) solution in PFA overnight at 4°C, then embedded in OCT (Sakura Tissue-Tek, Cat. No.A4583), and frozen with liquid nitrogen. The samples were sectioned at a thickness of 10 µm along the vertical meridian of the eyeball through the optic nerve head. All sections were fixed again in 4% PFA for 15 min, rinsed with PBS, and then placed in PBS block solution containing 0.3% Triton X-100 (Solarbio, Cat. No.T8200), 0.2% bovine serum albumin (Beyotime, Cat. No.ST2249), and 5% goat serum at 4 °C for 30 min. The. After blocking, frozen sections were incubated with primary antibodies at 4 °C overnight with constant shaking and then, after washubg, with an appropriate secondary antibodies (anti-rabbit or anti-mouse IgG) for 1 hour at room temperature. Antibodies used in the study are described in Supplementary Data 3. For double-labeling studies, we used well-characterized antibodies against RNA Binding Protein with Multiple Splicing (RBPMS) (ganglion cell bodies), neurofilament light chain (NfL) (ganglion cell axons and dendrites), glial fibrillary acidic protein (GFAP) (primarily astrocytes in normal retina), glutamine synthetase (GS) (Muller glia and astrocytes), calbindin (horizontal cells), and Protein kinase C-alpha (PKC-α) (bipolar cells). Slides containing the labeled retinal sections were mounted with 4′,6′-diamidino-2-phenylindole dihydrochloride (DAPI) (Solarbio, Cat. No.D8200) for 10 min, in fluorescent mounting medium (Solarbio, Cat. No.S2100) prior to imaging.

For retinal flat-mount labeling, intact eyes were fixed in 4% PFA for 15 min at room temperature. Whole retinas were carefully dissected and rinsed three times with PBS. After blocking with a block solution for 2 hours at 4 °C, tissues were incubated with the appropriate primary antibodies at 4 °C overnight. Following this, the samples were washed three times with 0.3% PBST for 1 hour each, and then incubated with the corresponding secondary antibodies and/or isolectin B4 (IB4) at 4 °C overnight. This was followed by three additional washes using 0.3% PBST for 1 hour each before samples were flat-mounted onto glass slides.

For immunolabeling of live cells, HUVECs and ARPE-19 cells were routinely grown in 25 cm² flasks (as described for cell culture above). Cells were physically scraped off the flask, transferred to 1.5 ml Eppendorf tubes, and centrifuged briefly. The supernatant was discarded, and the cells were resuspended and washed in serum-free medium. Cells were centrifuged again, and pellets resuspended and incubated with an appropriate primary antibody for 30 min on ice. After washing, cells were incubated with fluorescent secondary antibodies for another 30 min on ice. During incubation, cells were shaken every 10 minutes. Subsequently, cell nuclei were stained with Hoechst 33342 for 10 min, and cell suspensions were dropped onto a glass slide and coverslipped, without mounting medium, for imaging.

A Zeiss confocal microscope was used to capture images of immunolabeled samples using identical parameters with gain, light intensity (laser power), and slice thickness (Z-intervals) held constant for reliable and reproducible comparisons between controls and experimental samples. Retinal sections and HUVECs were imaged using 5 optical slices at 3 μm intervals, while retinal flat mounts were captured using 11 optical slices at 10 μm intervals. All images were taken with a 5x, 20x, or 63x objective.

siRNA-mediated knockdown

siRNAs for Laminin receptor (Santa Cruz, Lot: sc-35789), LRP6 (Shanghai GenePharma Co., Ltd., Shanghai, China), and control siRNA-A (Santa Cruz, Lot: sc-37007) were used for gene knockdown experiments using the transfection reagent, Lipofectamine 3000 (Invitrogen, Cat. No.L3000150). Sequence information is listed in Supplementary Data 1. All siRNAs were stored until use at -20 °C as a 10 μM stock solution using RNase-free water. Log phase cultured cells were plated in a 6-well plate at 200,000 cells/well, and the medium was immediately changed to serum-free and antibiotic-free medium for 24 h. When the cell density reached 70%, the medium was again replaced with 1 ml/well serum-free Opti-MEM™ medium for 1 h. 10 μL/well of a PEDF receptor siRNA or control scrambled siRNA-A was added to 1.5 μL/well Lipo3000 and mixed gently in Opti-MEM™ medium. After standing at room temperature for 15 min, the mixture was added to the 6-well plate and, 6 h later, the medium was replaced with complete growth medium containing 5% FBS and antibiotics and cultured further for 36 h. Cultures were then divided, with half replated into 6-well plates and the other half used for RNA extraction. Knockdown of PEDF receptors was verified by qPCR and Western Blot analyses. All knockdown experiments were carried out at least three times, and each was assayed in triplicate.

Statistics and Reproducibility

Statistical comparisons were all conducted using Prism 9 software (Version 9.3.1, GraphPad Software, La Jolla, CA, USA). Data are represented as the mean ± standard deviation of at least three independent experiments, with triplicate measurements for each experiment, unless otherwise noted. Statistical analysis was performed using a t-test for two groups and an ordinary one-way ANOVA followed by a Tukey post-hoc test for multiple groups. Where applicable, effect sizes are reported as mean differences with 95% confidence intervals. Differences were considered significant at p < 0.05.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

Z.Y. and M.Z: conceptualization, methodology, investigation, formal analysis, data interpretation, and writing—original draft; C.R.: formal analysis; X.Z.: conceptualization and methodology; C.J.B.: conceptualization, methodology, data interpretation, and writing—original draft; J.T.-T.: conceptualization, methodology, data interpretation, and writing—original draft; X.L.: conceptualization, methodology, and supervision. All authors read and approved the final manuscript.

Peer review

Peer review information

Communications Biology thanks Thomas Stax Jakobsen, Brian McKay, and the other anonymous reviewer(s) for their contribution to the peer review of this work. Primary Handling Editors: Eliana Scemes and Dario Ummarino. A peer review file is available.

Data availability

The raw data for all charts and summary statistics are provided as Supplementary Data 4. Supplementary Figs. 1 to 6 and uncropped whole blot images (Supplementary Fig. 7) are provided in the Supplementary information file. Other data presented in this study are available on request from the corresponding author.

Competing interests

The authors declare the following competing interests: J.T.-T. and C.J.B. have equity interests that relate to PEDF and PEDF mimetics. All other authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s42003-026-10175-2.