The dynamic balance between energy intake and consumption is critical for maintaining the integrity of an organism. When calorie intake exceeds energy expenditure, the remaining energy is converted into fatty acids (FAs) and stored as triglycerides (TGs). However, when energy expenditure (EE) surpasses caloric intake, TG is hydrolyzed and FAs are released (80). Dysregulated lipolysis leads to an overabundance of circulating FAs, which leads to several destructive, lipotoxic pathologic changes, such as insulin resistance, type 2 diabetes, fatty liver, nonalcoholic fatty liver disease, and inflammation (44–47). In 2004, three independent institutes identified ATGL, the main enzyme critical for the initial step of TG degradation, and named it ATGL (81), iPLA2ζ (82), and desnutrin (83). To prevent confusion, we use ATGL in the following discussion. The human gene encoding ATGL is located on chromosome 11p15.5 and includes 10 exons encoding a 504 amino acid protein (NCBI Reference Sequence: NG_023394.1). As a key participant in lipid catabolism, ATGL is highly expressed in brown adipose tissue (BAT) and white adipose tissue (WAT). However, in nonadipose tissues such as skeletal muscle (84), heart (85), liver (86), lung (87), and ocular tissues, and especially in the neural retina and RPE cells (ARPE-19 and hTERT) (39), ATGL mRNA is highly expressed, although at a lower level than in adipose tissues. In addition, researchers found that ATGL was also expressed in human Y-79 cells (11), mouse photoreceptor (661W) cells (88), a rat photoreceptor precursor (R28) cell line (39) and bovine RPE cells (89). ATGL is mainly localized in TG-rich intracellular lipid droplets (LDs), and the catalytic site in human and murine ATGL comprises a dyad consisting of serine 47 and aspartate 166 at the N-terminus of the protein (81, 90). Notably, the following factors may affect ATGL localization: the C-terminal portion includes a hydrophobic LD-binding region, and mutation in the ATGL C-terminus can negatively affect AGTL-LD binding (44). In addition, ATGL binds LC3, an autophagosome marker that promotes ATGL recruitment to LDs (47).

ATGL and hormone-sensitive lipase (HSL) are the major lipases in adipocytes in humans and mice. Several factors are involved in the regulation of ATGL activity. During catecholamine-stimulated lipolysis, for example, PKA phosphorylates α/β hydrolase domain-containing 5 (ABHD5) and promotes ATGL release from the LD surface. The direct interaction of ABHD5 with ATGL promotes the hydrolysis of triacylglycerols (TAG) (91) and broadens region-specificity of ATGL, because as it no longer exclusively engages in sn-2 reactions and engages in more sn-1 and sn-2 reactions (92). Notably, in adipocytes, the combination of an adipocyte- or FA-binding protein with ABHD5 induces ATGL activity, probably by preventing the inhibition of ATGL protease activity (93). In nonadipocytes, ubiquitin regulatory X domain-containing protein 8 (UBXD8) interacts with ATGL and causes the dissociation of CGI-58, resulting in a decrease in ATGL activity (94). ATGL activity is activated by ABHD5 and inhibited by the G0S2 protein in a noncompetitive manner (95). Specifically, G0S2 engages in direct protein-protein interactions and thus impedes substrate accessibility (96). Hypoxia-inducible gene 2 (HIG2), a HIF-1 target, has recently been identified as an inhibitor of ATGL that directly interacts with the AT GL patatin-like phospholipase domain (97, 98).

PEDF is also involved in regulating ATGL (see Figure 1 ). As a secreted protein, PEDF binds to ATGL on the RPE cell membrane, mediating the first step in its biological effects (33). PEDF regulates lipid metabolic functions through a concomitant elevation in the scavenger receptor CD36. The crucial roles played by PEDF are thought to be mediated through ATGL activation (99). For example, PEDF enhances the nuclear import of the predominantly cytosolic ATGL protein to promote its subsequent proteasomal degradation in the nucleus, thus diminishing ATGL protein stability in a COP1-dependent manner (99). During the development of inflammation-associated hepatic steatosis, endogenous PEDF directly interacted with ATGL and was neutralized by TNF-α, possibly in a PPARα-dependent manner (100). Endothelial damage is critical to vascular leakage and the subsequent shock in sepsis patients. Elevated serum PEDF levels in patients with sepsis induced increased extravasation, which was induced by the abnormal expression of ATGL (101). In addition, both PEDF and the PEDF-derived peptide 44mer promoted cardiac triglyceride degeneration by binding to ATGL (33).

ATGL is regulated by many factors at the transcriptional and post-transcriptional levels. In mice, ATGL controls energy supply and systemically hydrolyzes FAs, thereby influencing mitochondrial β-oxidation and the activation of peroxisome proliferator-activated receptors (PPARs) (102). ATGL decreases PPAR-γ signaling in the brain and suppresses appetite. Additionally, ATGL-deficient humans typically have obesity (103). At the transcriptional level, tumor necrosis factor-α (TNF-α) and insulin decrease ATGL expression, respectively. In this process, PPAR-γ interacts with the ATGL promoter and induces ATGL mRNA expression (104). Under starvation conditions, forkhead box protein O1 (FoxO1) showed increased affinity for the ATGL promoter (105), and this affinity was decreased by an increase in insulin level (106). Regarding the posttranscriptional phosphorylation of ATGL, studies with mice have shown that ATGL phosphorylation at Ser406 either by protein kinase A (PKA) or adenosine 5’-monophosphate-activated protein7 kinase (AMPK) increased ATGL activity (107, 108). In addition, fat-specific protein 27 (Fsp27) reduced ATGL transcript levels by regulating the affinity of early growth response proteins for the ATGL promoter in human adipocytes and thus decreased ATGL-mediated lipolysis (109, 110) (see Figure 1 ).

Lipoprotein receptor-related protein 6 (LRP6) was first isolated by Sheryl D. Brown et al., who reported the chromosomal location of LRP6 on human chromosome 12 in 1998 (111). The low-density lipoprotein receptor (LDLR) family comprises cell surface receptors, and several members are involved in numerous signaling pathways and are expressed in various target organs (112). LRP6 is a unique member of the LDL receptor gene family, as indicated by sequence comparisons, and exhibits the closest relationship to the LRP5 gene, a possible therapeutic target for type 1 diabetes (113). Both LRP5 and LRP6 function as coreceptors in the Wnt/β-catenin pathway (111); nevertheless, studies have indicated that LRP5 and LRP6 act distinctly owing to differences in their tissue distribution and affinity for individual Wnt ligands (114). LRP5 and LRP6 are structurally related, exhibiting approximately 71% homology at the nucleotide level, and are broadly expressed in humans, including the hippocampus, renal tubular cells, hepatocytes, intestinal epithelial cells, osteoblasts, and osteoclasts (115). Although LRP5 and LRP6 are functionally and structurally similar, LRP6 is more closely related to glucose and lipid metabolism signaling pathways and plays a more crucial role in developmental processes than LRP5 (116–118).

The human LRP6 (hLRP6) gene is located on chromosome 12p13.2, is 150 kb long and consists of 23 exons. hLRP6 is highly evolutionarily conserved, with negligible differences observed among Drosophila, Xenopus and Mus (119). LRP6 is highly expressed in the heart, brain, placenta, spleen, testes and ovaries but is expressed at low levels in the liver, skeletal muscle, prostate and colon mucosa (111). Mice show LRP6 expression levels in the heart, brain, lungs and kidneys similar to those in humans (111). According to a βgeo reporter activity assay, LRP6 is an endogenous gene and has been detected in all types of developing embryo cells (120).

In mice, LRP6 is essential to the canonical and noncanonical Wnt signaling pathways. When the LRP6 gene carries an insertion mutation and an individual Wnt gene is mutated, Lrp6^-/- homozygous embryos present developmental defects (120). Mice that lack LRP6 genes typically exhibit mid-/hindbrain defects, posterior limb truncations, and abnormal limb patterning, similar to Wnt7a-mutant mice (48). In addition, LRP6 plays an important role in the mouse forebrain (121). In Alzheimer’s disease (AD), a neurodegenerative disorder characterized by deficient cognitive processes and the accumulation of amyloid precursor protein (APP) and amyloid-β, LRP6 mutations are considered to be pathogenic factors. Notably, the conditional deletion of the LRP6 gene in mouse forebrain neuronal cells led to the destruction of synaptic integrity and memory loss in an age-dependent manner, potentially due to increased processing of APP in to amyloid-β (51). In addition, De Ferrari GV et al. showed that in two brain bank data series, a single nucleotide polymorphism (SNP) in exon 18 of LRP6 was associated with Alzheimer’s disease (49).

Recently, a reduction in Wnt-mediated transcription resulting from mutations in LRP6 has been confirmed to lead to several features of metabolic syndrome, including atherosclerosis (50), diabetes (55), hyperlipidemia (50, 53), and hypertension (54), but not obesity. Coronary artery disease (CAD) is usually caused by several risk factors of metabolic syndrome. In 2007, Mani et al. reported on a family with autosomal dominant early CAD, characterized by the features of hyperlipidemia, hypertension, and osteoporosis, which were found to be associated with a short piece of chromosome 12p; specifically, the group identified a mutation in LPR6 that encoded a coreceptor in the Wnt signaling pathway (52). The identified mutation damaged the Wnt signaling pathway in vitro because an arginine residue in the epidermal growth factor-like (EGF-like) domain was replaced with a cysteine residue. This finding encouraged the investigators to focus on the relationships between LRP6 and various human diseases, especially atherosclerosis (51).

After the binding of the Wnt ligand, the Fzd/LRP6 receptor and the scaffold protein Dishevelled (Dvl) form a signaling body that is endocytosed. The signaling body induced glycogen synthase kinase 3 (GSK3)-mediated phosphorylation of LRP6 on the PPPSP motif, which initiated the phosphorylation of adjacent S/T clusters through casein kinase 1γ (CK1γ) (122, 123). In addition to GSK3, other kinases phosphorylate the LRP6-PPPSP motif in a Wnt-independent manner. For example, due to the initiation of phosphorylation of cyclin-dependent kinase 14 (CDK14/PFTK1) and its regulators, mitotic cyclin Y and cyclin Y-like 1 (CCNY/CCNYL1), the ability of LRP6 to respond to Wnt ligands is greatest during the G2/M period (123–127). Phosphorylated LRP6 recruits a polyprotein called the “destruction complex”, which includes the scaffold protein Axin and Adenomatous Polyposis Coli(APC) tumor suppressors, the kinases CK1α and GSK3, and the protein E3 ubiquitin ligase β-transduction protein repeat sequence (β-TrCP). In the absence of Wnt ligands, CK1α and GSK3 in the destruction complex phosphorylate β-catenin, inducing β-TrCP-mediated degradation by the proteasome. Wnt/LRP6 signaling inhibits destruction complex activity through at least two mechanisms. First, GSK3 is the product of inhibited LRP6 phosphorylation. Second, the signaling body matures into a multivesicular body, in which the destruction complex is isolated with LRP6. Wnt/LRP6 signaling is enhanced by members of the R-sponge protein (RSPO) family of secreted proteins, which bind to the LGR4/5 receptor and the E3 protein ligase RNF43/ZNRF3 to stabilize the Wnt receptor on the cell surface (54, 128–133).

The canonical Wnt signaling pathway is a conservative intracellular signaling pathway. When the receptor complex on the cell surface, specifically, the Wnt-Fzd-LRP5/6 complex, binds to a Wnt ligand, the canonical Wnt pathway is rapidly activated. Then, Wnt phosphorylation of the cytoplasmic tail of the Fzd-LRP5/6 complex triggers the recruitment and phosphorylation of disheveled protein (also known as Dishevelled, Dvl) in the cytoplasm (134, 135). Subsequently, Axin is recruited to the receptor complex to inhibit its decomposition and the phosphorylation of β-catenin. The stable nuclear translocation of soluble β-catenin in the cytoplasm promotes the activation of the downstream Wnt/β-catenin signaling pathway via the transcription and translation of target genes (134, 135). In the absence of Wnt ligands or in the presence of Wnt receptor inhibitors, β-catenin is bound by Axin, Axin and casein kinase Ia (CKIa) and glycogen synthase kinase 3β (GSK3β); moreover, two APC protein kinases combine to form a destruction complex (134). CKIa and GSK3β mediate the phosphorylation of β-catenin in the cytoplasm. Phosphorylated β-catenin is recognized by b-TrCP and induces ubiquitination, which ultimately leads to the degradation of β-catenin (136). Many inhibitors of the Wnt signaling pathway, including Dickkopf WNT Signaling Pathway Inhibitor 1, DKK-1 (137), R-spondins (138), Norrin (139), endostatin (140), PEDF (40), and very low-density lipoprotein receptor, have been identified (141, 142). These negative regulatory molecules of the Wnt signaling pathway ultimately lead to the transcription of target genes, inducing cell proliferation, differentiation, or apoptosis (see Figure 2 ).

Plexin domain-containing 1 (PLXDC1), also known as tumor endothelial marker 7, is a member of the tumor vascular endothelial marker family (143). The genes encoding PLXDC1 are highly expressed during tumor angiogenesis and enriched in many types of tumor endothelial cells (144), including glioblastoma endothelium (57, 63), ovarian cancer (62), gastric cancer (60), renal cell carcinoma (61), colorectal cancer (64), lung cancer (58), breast cancer (56), and osteosarcoma (59). PLXDC1 has been found to be highly expressed in the endothelial cells in the ocular disease DR context and is highly specific to diseased blood vessels (145). PLXDC1 has also been shown to be expressed in rodent retinal ganglion cells (RGCs) (145). Transmission electron microscopy has revealed that PLXDC1 was mainly expressed at tight junctions between endothelial cells, and some PLXDC1 expression was detected on the surface of the blood vessel lumen (146).

PLXDC1 is a novel LRP1-regulated cell-signaling protein. Regulation of PLXDC1 activity by LRP1 was confirmed in CHO cells, and in extracts of RAW264.7 cells and mouse liver, PLXDC1 coimmunoprecipitated with LRP1 (147). The Nidogen/PLXDC1 interaction enhanced cell spreading in PLXDC1-transfected 293T cells, suggesting that Nidogen is a candidate ligand for PLXDC1 (148). Akash Nanda et al. identified cortactin as a protein candidate for binding to the extracellular region of both PLXDC1 and its closest homolog, which were all expressed in the tumor endothelium (146).

Plexin domain-containing 2 (PLXDC2) is a largely uncharacterized transmembrane protein with a homologous nidogen region and a plexin repeat in its extracellular domain. In mice, PLXDC2-betageo expression is prominent in the brain, including in the cortical hem, midbrain-hindbrain boundary, and midbrain floorplate, during midembryonic stages (E9.5-E11.5). On E15.5, PLXDC2 expression was apparent in a large number of discrete nuclei and structures throughout the brain, including the glial wedge and derivatives of the cortical hem. PLXDC2 was also found to be expressed in other tissues, most notably the limbs, lung buds and developing heart, as well as in the spinal cord and dorsal root ganglia (149). In humans, PLXDC2 has been found to be expressed in endothelial cells of the tumor stroma, neural progenitor cells, and pluripotent stem cells and in human hepatocellular carcinoma (HCC) tissues, including HCC cells, tumor vascular endothelial cells, and some infiltrating cells (150).

The PLXDC2 level is elevated during tumor angiogenesis (143). PLXDC2, a mitogen for neural progenitors, has been shown to be a novel regulator of the development of different brain regions involved in the coordinated control of cell proliferation and cell fate along and across the neuraxis (151). PLXDC2 expression increases during cellular senescence and is a marker of adult stem cells (152). In addition, PLXDC2 has been associated with CAD (72), ovarian cancer (73), diabetes mellitus (71), DR (68), testicular germ cell tumors (74), Sotos syndrome (66), ischemic stroke (70), laryngeal squamous cell carcinoma (69), breast cancer (56), and colorectal tumors (65, 67).

In 2014, Guo Cheng et al. identified PEDF binding to PLXDC1 and PLXDC2 on the surface of 293T cells (see Figure 3 ). PEDF interacted with PLXDC1 to promote the secretion of interleukin-10 by macrophages, which exerted neuroprotective effects on rat RGCs, and PEDF combined with PLXDC2 promoted apoptosis in endothelial cells (41). With few reports on the role or mechanism of PEDF and PLXDC in DR or neovascular fundus diseases, PLXDC1 has been shown to inhibit cell proliferation and invasion in tumor cells, promote endothelial cell separation in mice. With a protective effect on RGCs, PLXDC1 is likely to be an important receptor for PEDF, particularly in fundus diseases. In addition to PEDF, other molecules are ligands of PLXDC1 or PLXDC2. Moreover, members of the Wnt family (Wnt3a, Wnt5a and Wnt8b) show a striking overlap with PLXDC2 expression in certain areas (69). Cortactin can bind to the extracellular terminus of PLXDC2 (152). PLXDC2 has also been shown to be an activating ligand for ADGRD1 on cumulus cells (153).

The laminin receptor (LR) is a laminin-binding protein with a molecular weight of approximately 67 kDa and is a nonintegrin cell surface receptor of the extracellular matrix (154). LR contributes to laminin binding, ribosome biosynthesis, cytoskeletal organization, and nuclear function, and it controls key cellular processes, including cell adhesion, migration, proliferation and survival, protein synthesis, development, and differentiation (76). Research on LR is mainly focused on its role on the development of tumors, specifically its promotion of tumor cell adhesion to the basement membrane, which is the first step in the tumor cell invasion and metastasis. Therefore, compared with normal cells, LR in tumor cells is overexpressed, and its overexpression is considered to be a molecular marker of cancer metastasis and aggressiveness, including breast cancer, lung cancer, ovarian cancer, prostate cancer, gastric cancer, thyroid cancer, leukemia and lymphoma (75, 76).

In vitro yeast two-hybrid screening experiments confirmed that LR is a cell membrane surface receptor of PEDF. Researchers have also determined that the PEDF isoform that interacts with LR consists of 34 peptides. Surface plasmon resonance analysis showed that PEDF interacts with LR, and PEDF also interacts with endothelial cells. LR-plasma membrane binding induces endothelial cell apoptosis and inhibits endothelial cell migration and the formation of vascular networks in vitro (25). The combination of the PEDF 34mer and LR may also mediate endothelial cell apoptosis through the LR-JNK-PPAR-γ-FasL-caspase 8 pathway, thereby exerting an anti-NV effect in the retina. In apoptotic cells, the PEDF 34mer binds to LR to phosphorylate JNK/MAPK, which then activates the transcription factor PPARγ through PPARγ ligand binding and nuclear translocation. In the nucleus, PPAR-γ and RXR combine in the form of heterodimers and recruit specific nuclear coactivators that activate the transcription of target FasL genes in a ligand-dependent manner. Both PPAR-γ and RXR promote the transcription mechanism initiated at FasL gene promoters, ultimately inducing the expression of FasL protein (155) (see Figure 4 ).

Three VEGFR subtypes are encoded by different genes: VEGFR1 (Flt-1 in mice) and VEGFR2 (Flk-1; KDR) are structurally similar, while VEGFR3 (Flt-4) carries proteolytically cleaved extracellular domains (156–158). VEGF receptors are highly expressed in endothelial cells, hematopoietic cells, macrophages, and smooth muscle cells (159–161). VEGF-A isoforms drive physiological and pathophysiological angiogenesis via VEGFR1 and VEGFR2 signaling, while lymphatic angiogenesis is mediated by VEGF-C/D isoforms via VEGFR3 (162). VEGFR2 is a large membrane protein consisting of seven extracellular immunoglobulin (Ig)-like domains, a single transmembrane helix, and a split intracellular kinase domain. VEGF-A is an endogenous agonist of VEGFR2 that binds to ligand binding sites in Ig-like domain 2 (D2) and D3 on the basis of the stoichiometry of one VEGF-A dimer to one VEGFR dimer (163, 164).

VEGF-A, VEGFR1, and VEGFR2 play important roles in physiological and pathological angiogenesis, including tumor angiogenesis. Although VEGF-A isoforms exhibit similar VEGFR2-binding properties, activation of VEGFR2 is a complex multistep process. The VEGFR2 signaling pathway is critical for controlling vascular endothelial cell activity, migration, and permeability (165, 166). In cancer cells, transcription factors induced by hypoxia-inducible factors rapidly increase gene transcription rates, resulting in elevated levels of VEGF gene expression while enhancing the stability of VEGF mRNA (78, 79). The direct result of VEGF-VEGFR binding is endothelial cell budding, increased vascular permeability, and increased expression of tissue matrix metalloproteinases, resulting in extracellular matrix digestion. These processes lead to the movement of endothelial and pericyte cells, eventually leading to vasodilation and the formation of pathological vascular networks.

In 2006, Jianxing Ma demonstrated that PEDF inhibited VEGFR2 (167). PEDF inhibits VEGF expression at the transcriptional level by inhibiting hypoxia-induced increases in VEGF promoter activity, HIF-1 nuclear translocation, and mitogen-activated protein kinase phosphorylation. Other investigators have reported that PEDF exerts the anti-VEGF actions by binding to its key angiogenic receptor, VEGF-R2 (43). This binding inhibits receptor activation and ultimately reverses critical downstream VEGF-A signaling by reducing VEGF-R2 tyrosine kinase activity and inhibiting two phosphorylation sites in VEGF-induced HUVECs (Y951 is critical for cell migration and pathological angiogenesis, and Y1175 phosphorylates downstream signaling). Inhibition of VEGF angiogenesis signaling molecules such as PI3K, PLC-γ, FAK and Src may regulate many aspects of endothelial cell biology, including cell proliferation, migration, and survival. Behzad Shahbazi et al. also showed that PEDF binds to the VEGFR2, as well as laminin receptor much stronger than ATP synthase β-subunit, VEGFR1, and LRP6 using molecular docking, molecular dynamics (MD) simulation, and Molecular mechanics/Poisson-Boltzmann surface area (MM/PBSA) approach. Besides, the N-terminal of PEDF owns the most important enactment in the interaction with its receptors (168).

F-ATP is an ATP synthase in the bacterial plasma membrane, the eukaryotic mitochondrial inner membrane and chloroplast thylakoid membrane. F-ATPase leverages the membrane proton gradient to drive ATP synthesis, passing the passive proton flux through the membrane along its electrochemical gradient, and releasing newly formed ATP from the active site of F-ATP synthase using the energy released by transport reaction (169). F-ATP synthase is composed of two domains, F0-ATP synthase and F1-ATP synthase. In previous studies on angiogenesis, lipoprotein metabolism, innate immunity, hypertension or food intake regulation, F1-ATP synthase was found to be expressed only in cell mitochondria and was used to synthesize ATP that was needed by cells(77). Luigi Notari et al. demonstrated for the first time that PEDF binds F1 ATP synthase b subunit on the endothelial cell surface and inhibits endothelial extracellular ATP synthesis (170). In addition, F1-ATP synthase can inhibit cell growth, proliferation, and migration and directly induce multiple forms of cell death (171).