Latrophilin-2 is a novel receptor of LRG1 that rescues vascular and neurological abnormalities and restores diabetic erectile function

Diabetes mellitus (DM) is a chronic metabolic disorder characterized by inappropriate hyperglycemia, which causes endothelial dysfunction and peripheral neuropathy, ultimately leading to multiple complications. One prevalent complication is diabetic erectile dysfunction (ED), which is more severe and more resistant to treatment than nondiabetic ED. The serum glycoprotein leucine-rich ɑ-2-glycoprotein 1 (LRG1) is a modulator of TGF-β-mediated angiogenesis and has been proposed as a biomarker for a variety of diseases, including DM. Here, we found that the adhesion GPCR latrophilin-2 (LPHN2) is a TGF-β-independent receptor of LRG1. By interacting with LPHN2, LRG1 promotes both angiogenic and neurotrophic processes in mouse tissue explants under hyperglycemic conditions. Preclinical studies in a diabetic ED mouse model showed that LRG1 administration into the penile tissue, which exhibits significantly increased LPHN2 expression, fully restores erectile function by rescuing vascular and neurological abnormalities. Further investigations revealed that PI3K, AKT, and NF-κB p65 constitute the key intracellular signaling pathway of the LRG1/LPHN2 axis, providing important mechanistic insights into LRG1-mediated angiogenesis and nerve regeneration in DM. Our findings suggest that LRG1 can be a potential new therapeutic option for treating aberrant peripheral blood vessels and neuropathy associated with diabetic complications, such as diabetic ED.


INTRODUCTION
Diabetes mellitus (DM) is a chronic metabolic disorder that involves endothelial dysfunction and neuropathy and can lead to multiple complications, including cardiovascular disease, stroke, chronic kidney disease, foot ulcers, and retinopathy 1 . DM is also a major cause of erectile dysfunction (ED), which is a manifestation of microangiopathy and neuropathy 1 . Approximately 50-75% of male diabetic patients have ED 2 . Phosphodiesterase type 5 (PDE5) inhibitors are the most commonly used first-line treatment options for ED; however, these agents are ineffective in~30% of patients and ultimately cannot rescue angiopathy and neuropathy in diabetic ED patients 3 . Tested alternative therapeutic options for ED include various angiogenic or neurotrophic factors, such as COMP-Ang1, vascular endothelial growth factor (VEGF), dickkopf2, neurotrophin-3 (NT3), and brain-derived neurotrophic factor (BDNF) [4][5][6][7][8] ; however, these treatments have shown limited success in clinical trials.
Leucine-rich ɑ-2-glycoprotein 1 (LRG1) is a secreted glycoprotein that is abundant in serum (10-50 μg/mL) 9 . LRG1 has received attention as a prognostic/diagnostic biomarker for many diseases, including inflammatory disease 10,11 , heart failure 12 , neurodegenerative disease 13,14 , and several types of cancer [15][16][17][18] . Moreover, it has also been reported that LRG1 is associated with several types of diabetic complications [19][20][21] . However, little is known about the relationship between LRG1 and diabetic ED, and the molecular mechanisms for the multifunctional role of LRG1 in patients with DM complications remain to be established. Conversely, local administration of recombinant LRG1 was reported to promote corneal reepithelialization and nerve regeneration in a diabetic mouse model 22 and to accelerate keratinocyte migration and skin wound repair in normal mice 23 . Moreover, LRG1 overexpression reportedly improves glucose homeostasis and delays the diabetic phenotype in obese mice 24 . LRG1 promotes angiogenesis by modulating endothelial transforming growth factor (TGF)-β signaling 25 . In the presence of TGF-β, LRG1 directly binds to the TGF-β accessory receptor endoglin (ENG), thereby switching the angiostatic TGF-β signaling (ALK5-Smad2/3 pathway) toward pro-angiogenic TGF-β signaling (ALK1-Smad 1/5/8 pathway). However, compared with typical receptor-ligand interactions, LRG1 shows relatively weak binding affinity for ENG (~2.9 μM), raising questions of whether an additional unknown component is required for TGF-β receptor activation or if LRG1 signals through another mechanism.
Here, we used ligand-based receptor capture technology on living cells and discovered that the adhesion G-protein-coupled receptor latrophilin-2 (LPHN2) is a TGF-β-independent receptor of LRG1. We found that LRG1 acts through LPHN2 to induce both angiogenic and neurotrophic processes in hyperglycemia. Moreover, LPHN2 expression is significantly increased in the penile tissues of a streptozotocin (STZ)-induced diabetic model mice, as well as diabetic ED patients. Accordingly, intra-cavernous administration of recombinant LRG1 ameliorates vascular and neurological abnormalities, thus effectively restoring erectile function. These findings provide insight into the TGF-β-independent mechanism of LRG1 as an angioneurin in hyperglycemia and its potential therapeutic application against various diabetic complications associated with endothelial dysfunction and neuropathy.

MATERIALS AND METHODS Cell culture
Human umbilical vein endothelial cells (HUVECs; Cat# CC-2519, Lonza) and human embryonic kidney 293 T cells (HEK293T; Cat# CRL-3216, ATCC) were authenticated according to ATCC guidelines. Cells were maintained at 37°C under humidified 5% CO 2 conditions. Cells between passages 2 and 7 were used for all experiments. Primary mouse cavernous endothelial cells (MCECs) were prepared and maintained in complete M199 as described previously [26][27][28] . Diabetes-induced angiopathy was mimicked by serumstarving the cells overnight and exposing them to high-glucose (30 mM glucose) conditions for 72 h at 37°C in a humidified 5% CO 2 atmosphere 29 . Normal glucose (5 mM glucose, Cat# G7021, Sigma-Aldrich) conditions were used as a control.

Animals
Eight-week-old male C57BL/6 mice (Orient Bio, Korea) and LRG1-Tg male mice were used in this study. Transgenic mice with whole-body overexpression of LRG1 were generated by Macrogen (Seoul, Korea) using a construct in which the full LRG1 ORF was subcloned between the RsrII and Notl restriction sites of the pCMV6-Kan/Neo vector (#PCMV6KN, OriGene Technologies Inc.). All manipulations were conducted with the approval of Macrogen's Institutional Animal Care and Use Committee. The incorporation of the gene in transgenic offspring was screened by PCR analysis of genomic DNA using genotyping primers (CMV_LRG1_F: 5'-ACCATGGTG ATGCGGTTTTG-3'; CMV_LRG1_R: 5'-CACCGACAGCTGGACAGTGT-3'). In all experiments, mice were age-matched, and wild-type littermates were used as controls. Experiments were conducted with the approval of the Inha University Animal Care and Use Committee (Assurance Number: INHA 180523-570). Diabetes was induced by intraperitoneal injection of streptozotocin (STZ; 50 mg/kg body weight) for five consecutive days as described previously 30 . Eight weeks after diabetes induction, the mice were anesthetized with intramuscular injections of ketamine (100 mg/kg) and xylazine (5 mg/kg).

Tube formation assay and cell migration assay
Tube formation assays were performed as described previously 28 . Tube formation was monitored for 12-16 h under a phase-contrast microscope and quantified by counting the number of master junctions in four separate experiments in a blinded manner using ImageJ software. For Transwell migration assays, 1 × 10 5 HUVECs in 200 μl of serum-free M199 were seeded into the upper inserts (8-μm pores) of 12-well plates (Cat# 353182, Corning), and 600 μl of M199 supplemented with 20% FBS was added to the lower chamber. HUVECs or MCECs were then treated with 1 μg/ml PBS (control), 1 μg/ml LRG1, 5 ng/ml TGF-β1 (Cat# 240-B, R&D Systems), 10 μg/ml anti-TGF-β1 antibody (Cat# MAB1835, R&D Systems), and 5 ng/ml TGF-β1 + 10 μg/ml anti-TGF-β1 antibody (premixed overnight for neutralization). After 18 h of culture at 37°C, non-migrated cells on the upper surface of the insert were removed with a cotton swab. Migrated cells were fixed and stained with crystal violet (Cat# V5265, Sigma) and then eluted with 10% acetic acid. The number of migrated or invaded cells was counted in four random fields.

TriCEPS-based ligand-receptor capture (LRC-TriCEPS)
To identify the putative receptor of LRG1, TriCEPS-based ligand-receptor capture (LRC-TriCEPS; cat# P05203, Dualsystems Biotech) was used according to the manufacturer's directions. Briefly, 300 μg of recombinant LRG1 protein or transferrin (control ligand) was dissolved in 150 μl of 25 mM HEPES buffer (pH 8.2), after which 1.5 μl of the TriCEPS reagent was added to each sample and incubated for 90 min at 22°C with constant agitation. HEK293T cells (1.2 × 10 8 ) in PBS (pH 6.5) were mildly oxidized by incubation with 1.5 mM NaIO 4 at 4°C in the dark for 15 min with gentle rotation (Supplementary Fig. 1a). Cells were then washed twice at 300 × g for 5 min and resuspended in two separate tubes containing 18 ml of PBS with 1% FBS (pH 6.5), after which 150 μl of TriCEPS-coupled transferrin or TriCEPS-coupled LRG1 was added to each tube and incubated for 90 min at 4°C with constant, gentle agitation. After the coupling reaction was completed, the cells were collected, and the cell pellets were sent to dual systems for LC-MS/MS analysis.

Lentiviral shRNA delivery
After the animals were anesthetized with intramuscular injections of ketamine (100 mg/kg) and xylazine (5 mg/kg), the scrambled control shRNA (shCon) lentivirus particles (Cat# SC-108080, Santa Cruz) or SMART vector mouse lentivirus containing shRNA targeting LPHN2 (shLPHN2; Cat# V3SM7603-234963095, Dharmacon) were administered by intramuscular injection at a dose of 1 × 10 5 transforming units (TU) per mouse. For in vitro (HUVECs, HEK293T cells and MCECs) and ex vivo (MPGs and DEGs) cell culture experiments, shCon or shLPHN2 lentivirus particles were added to the culture medium at a concentration of 5 × 10 4 TU/ml. The shRNA sequence targeting murine LPHN2 was ACGGCTTATGGGTCATTTA, and the shRNA sequence targeting human LPHN2 was TCAACTGCTAGGTCGATAT. Experiments were performed 3 days after lentiviral infection.

Cell surface binding assay
HEK293T cells and HUVECs were plated on 8-well Lab-Tek chamber slides in standard growth medium. The cells were incubated with LRG1-YFP or YFP for 1 h at 37°C in a humidified 5% CO 2 environment and then washed twice with PBS and fixed with 4% formaldehyde in PBS for 15 min at room temperature. The wells were washed twice with PBS, and then the slides were mounted with ProLonged Antifade mounting solution (Cat# P10144, Molecular Probes). Fluorescently labeled LRG1 was observed under a confocal laser-scanning microscope (Zeiss LSM 780).
Orbitrap XL mass spectrometer (Thermo Fisher) equipped with a nanoelectrospray source 31 (Yonsei Proteome Research Center, Seoul, Republic of Korea). The Mascot algorithm (Matrix Science, USA) was used to identify peptide sequences present in a protein sequence database. The peptides were filtered with a significance threshold of P < 0.05.

Ex vivo endothelial cell sprouting assay
Mouse corpus cavernosum tissue was cut into two or three pieces and then plated on Matrigel stock solution-coated 6-well cell culture dishes. The Matrigel was polymerized by incubating for 15 min at 37°C, and 3 ml of normal-glucose (5 mM) or high-glucose (30 mM) conditioned complete M199, with or without 1 µg/ml LRG1 and shRNA (shCon or shLPHN2; 5 × 10 4 TU/ml culture medium), was added to the cell culture dish. The cells were then incubated at 37°C in a humidified 5% CO 2 environment, and the conditioned complete M199 was changed every 2 days. Images were obtained with a phase-contrast microscope after 7 days, and the sprouting cell density was analyzed using ImageJ 1.34.

Measurement of erectile function
Erectile function was measured as described previously 30 . The fasting and postprandial blood glucose concentrations in streptozotocin (STZ)-treated WT or LRG1-Tg diabetic mice were significantly higher than those in control or LRG1-Tg mice. In addition, body weight and serum insulin levels were significantly lower in STZ-induced diabetic mice than in controls or LRG1-Tg mice. However, the body weight and blood glucose levels of diabetic mice did not differ significantly regardless of treatment (Supplementary Tables 1-4).

BrdU labeling
Mice from each group were intraperitoneally injected with BrdU (50 mg/kg body weight; Cat# 19-160, Sigma-Aldrich) once daily for three consecutive days and sacrificed 1 day after the final BrdU injection. An additional antigen retrieval step was performed with 5'-bromo-2'-deoxyuridine antibody (BrdU; Cat. MCA2060, AbD Serotec; 1:50). The number of BrdUpositive endothelial cells was counted in four different regions. The values are expressed per high-power field.

TUNEL assay
TUNEL (terminal deoxynucleotidyl transferase-mediated deoxyuridine triphosphate nick-end labeling) assays were used to evaluate cell death in cavernous tissue from control and STZ-induced diabetic mice. The ApopTag Fluorescein In Situ Apoptosis Detection Kit (Cat# S7160, Chemicon) was used according to the manufacturer's instructions. Digital images and the numbers of apoptotic cells were determined using a confocal fluorescence microscope. For in vitro studies, the number and percentage of TUNEL-positive cells were evaluated.
Cignal Finder GPCR signaling 10-pathway reporter array and phospho array For Cignal Finder GPCR signaling 10-pathway reporter array analyses, HUVECs (5 × 10 3 cells/well) diluted in 80 µl of Opti-MEM supplemented with 5% FBS and 0.1 mM nonessential amino acids were transfected by mixing with the preincubation transfection mixture and then incubated for 24 h at 37°C in a 5% CO 2 environment. For the pre-incubation transfection mixture, 0.5 µl of Lipofectamine 2000 (Invitrogen) and reporter DNA constructs (Cignal Finder Reporter Array for GPCR signaling 10-Pathway; CCA-109L-2, Qiagen) in 50 μl Opti-MEM were incubated for 30 min at room temperature. After serum starvation for 4 h, reporter DNA-transfected HUVECs were treated with 1 µg/ ml LRG1 for 6 h. The 10 signaling pathway reporters in cells were then analyzed by quantifying luminescence signals using the Cignal 10-Pathway Reporter Array Kit according to the manufacturer's instructions.
For the phospho array, HUVECs were serum-starved for 8 h and then incubated with anti-TGF-β1 antibody for 1 h, after which HUVECs were stimulated with 1 µg/ml LRG1 for 30 min. Whole-cell lysates of HUVECs were prepared, and aliquots of lysates containing 50 µg of protein were used for Phospho Explorer antibody assays using an antibody array assay kit (Cat# KAS02, Fullmoon Biosystems, Sunnyvale, CA, USA) according to the manufacturer's instructions. The Phospho Explorer antibody array (Full Moon Biosystems, Inc.) consisted of 1318 antibodies, and each antibody was present as two replicates printed on a coated glass microscope slide, as well as multiple positive and negative controls. Phospho Explorer antibody array experiments and analyses were performed as a custom service by E-Biogen (Ebiogen Inc.)

RESULTS
The Adhesion GPCR Latrophilin-2 is a Novel TGFβ-independent Receptor of LRG1 LRG1 has been previously reported to promote angiogenesis in the presence of TGF-β by modulating the endothelial pro-angiogenic Smad1/5/8 signaling pathway 25 . However, we found that LRG1 alone was sufficient to induce tube formation and migration of human umbilical vein endothelial cells (HUVECs), which were cultured under high-glucose conditions to mimic diabetic hyperglycemia (Fig. 1a, b). Furthermore, LRG1-mediated tube formation and migration were unaffected by the presence of TGF-β1-blocking antibody. These findings led us to hypothesize that LRG1 may bind to an unknown receptor and/or activate TGF-β1-independent signaling to induce angiogenesis under hyperglycemic conditions. We next used a proteomics approach to identify the TGF-β1independent LRG1 receptor. First, we confirmed that LRG1 bound to the cell surface of the HUVECs and HEK293T cells, indicating that both cell lines expressed an unknown cell surface receptor of LRG1 (Fig. 1c). We then performed ligand-based receptor capture (LRC) in live HEK293T cells, which are amenable to large-scale culture. To this end, we used TriCEPS, a chemoproteomic reagent with three moieties for ligand conjugation, receptor capture, and receptor purification, respectively, which enabled the identification of a glycosylated target receptor 34 . Briefly, HEK293T cells were oxidized with sodium metaperiodate and incubated with TriCEPS-coupled LRG1, and then the captured glycoproteins were affinity purified and identified by liquid chromatography-tandem mass spectrometry (LC-MS/MS) analysis (Fig. 1d, Supplementary Fig. 1a). Based on our selection criteria (enrichment factor > 4 and false discovery rate [FDR]-adjusted P value < 0.05), we identified three prospective LRG1 binding partners: galectin-3 (LEG3), NADH dehydrogenase 1 alpha subcomplex subunit 5 (NDUA5), and latrophilin-2 (LPHN2) (Fig. 1e). Of these three proteins, LPHN2 was the only cell membrane protein and was thus the strongest candidate for an LRG1 receptor. Indeed, LPHN2 expression was confirmed in both HUVECs and HEK293T cells (Fig. 1f). LPHN2, also known as calcium-independent alpha-latrotoxin receptor 2 (CIRL2), is an adhesion-type G-protein-coupled receptor (GPCR) that is widely expressed in various tissues; however, its cellular functions are poorly understood 35,36 .
Lentiviral-mediated LPHN2 knockdown in HUVECs and HEK293T cells exhibited significantly decreased cell-surface binding of LRG1-YFP compared to cells treated with scrambled small hairpin RNA (shRNA) control lentivirus (shControl) (Fig. 1f, h). Moreover, LC-MS/MS analysis of HUVECs treated with soluble LRG1 revealed that LPHN2 are co-immunoprecipitated with LRG1, suggesting a specific and direct association of LRG1 with LPHN2 on HUVEC membranes (Fig. 1g, Supplementary Fig. 1b, c). Solidphase binding assays using LRG1 and the extracellular domains of LPHN2, including lectin (Lec), olfactomedin-like (Olf), and GPCR autoproteolysis-inducing/GPCR proteolysis site (GAIN/GPS) domains, revealed that the Olf domain of LPHN2 was the minimal binding domain for LRG1 (Kd = 920 nM for Olf domain, 890 nM for Lec/Olf domain) ( Supplementary Fig. 1d, e). Unlike the Olf domain alone and the Lec/Olf domain, recombinant ecto-full LPHN2 was unable to bind with LRG1. This finding suggested that proper proteolytic processing of the LPHN2 GPS motif at the cellular membrane might also be critical for LRG1 binding and LPHN2mediated cellular functions 37 . Interestingly, LRG1 could not bind to the Lec/Olf domain of LPHN1 or LPHN3, despite their 87% sequence similarity to the Lec/Olf domain of LPHN2 (Supplementary Fig. 1d, e). These findings demonstrated that LPHN2 is a novel TGF-β-independent receptor of LRG1.
The LRG1/LPHN2 Axis Induces Microvessel Sprouting and Neurite Outgrowth in a TGF-β-independent Manner Under High-Glucose Conditions Endothelial dysfunction in DM manifests as impaired proliferation and migration of endothelial cells (ECs) 1 . As expected, we found that high-glucose conditions impaired in vitro tube formation of mouse cavernous endothelial cells (MCECs) and decreased microvessel sprouting from aortic rings and penile cavernosum tissues (Fig. 2a, b, top; Supplementary Fig. 2a). LRG1 treatment completely rescued all of these high-glucose-induced phenotypes, and lentiviral-mediated LPHN2 knockdown significantly reduced these effects of LRG1 (Fig. 2a, b; Supplementary Fig. 2a, b). Consistent with previous reports 38 , we found slightly increased levels of secreted TGF-β1 in high-glucose culture medium conditioned by MCECs, aortic rings, and penile cavernosum tissues ( Supplementary Fig. 2c). However, even in the presence of TGF-β1-blocking antibody, LRG1 effectively promoted microvessel sprouting under hyperglycemic conditions (Fig. 2a, b).
Notably, under high-glucose culture conditions, MCECs showed significantly increased LPHN2 expression (Supplementary Fig. 2d), as did major pelvic ganglion (MPG; mixed parasympathetic and sympathetic ganglion that innervates the pelvic organs) and dorsal root ganglion (DRG; a cluster of cell bodies that transmit sensory information to the brain). Additionally, it has been reported that hyperglycemia leads to impairments of Schwann cell proliferation and migration and of axon regeneration from DRG neurons, which are broadly associated with diabetic peripheral neuropathy 39 . Therefore, we next tested whether LRG1 affects neurite outgrowth in MPG and DRG under hyperglycemic conditions. To this end, we exposed MPG and DRG mouse explants to hyperglycemia and analyzed neuronal tissues using βIII-tubulin immunofluorescence staining. In line with a previous report, high-glucose conditions significantly reduced neurite sprouting in both MPG and DRG explants (Fig. 2c, d, top). Importantly, LRG1 treatment stimulated neurite outgrowth from MPG and DRG explants under hyperglycemic conditions, and this effect was abolished by lentiviral-mediated LPHN2 knockdown (Fig. 2c, d). Although high-glucose conditions increased TGF-β1 levels in both MPG and DRG explant cultures, the TGF-β1-blocking antibody did not impact LRG1-mediated neurotrophic effects (Fig. 2c, d, Supplementary Fig. 2c). Taken together, our data showed that TGF-β-independent binding of LRG1 to LPHN2 promotes micro-vessel sprouting and neurite outgrowth under hyperglycemic conditions.
We previously reported that intra-cavernous delivery of Ninjurin1-neutralizing antibody or Dickkopf2 restored erectile function in diabetic mice through enhanced penile angiogenesis and neural regeneration 6,40 . Therefore, we investigated the therapeutic relevance of the LRG1/LPHN2 axis by evaluating erectile function in diabetic mice after intracavernous injection of LRG1. Upon electrical stimulation of the cavernous nerve, diabetic mice showed significantly decreased ratios of maximal intracavernous pressure (ICP) or total ICP to mean systolic blood pressure (MSBP) compared to those of control mice. LRG1 injection restored erection parameters in diabetic mice to almost 92% of control values (Fig. 3f). The rescue effects of LRG1 injection were significantly abrogated by shLPHN2 lentiviral infection, but not by shControl lentiviral infection (Supplementary Fig. 3a). The effects of LRG1 on erection were dose-dependent and maintained for at least 4 weeks (Supplementary Fig. 3b, c). MSBP did not differ among the experimental groups ( Supplementary Tables 1-4). We also found that the LRG1-mediated restoration of erectile function was slightly attenuated in the presence of a TGF-β1-blocking antibody and was significantly decreased by further infection with shLPHN2 lentivirus (Fig. 3f). Consistently, erectile function remained intact in LRG1-Tg mice, even after multiple STZ injections, and this effect was abolished by shRNA-mediated LPHN2 knockdown (Supplementary Fig. 3d). These results indicate that intracavernous administration of recombinant LRG1 effectively ameliorates erectile dysfunction in STZ-induced diabetic mice through the LPHN2 receptor.  4)). *P < 0.05; **P < 0.01; ***P < 0.001 (Student's t test). N.S., not significant. The relative ratio of the NG group was defined as 1.

The LRG1/LPHN2 Axis Exerts Dual Angiogenic and Neurotrophic Effects on the Recovery of Erectile Function in Diabetic Mice
We performed immunofluorescence staining for platelet/endothelial cell adhesion molecule (PECAM)-1 and the pericyte marker neuron-glial antigen 2 (NG2) in cavernosum tissue. The results confirmed that LRG1 treatment rescued the decreased numbers of cavernous endothelial cells and pericytes in diabetic mice, and this effect was abolished when LRG1 was co-administered with shLPHN2 lentivirus (LRG1 + shLPHN2 group) (Fig. 4a). Expression levels of PECAM-1 and NG-2 were comparable between shControl and shLPHN2 lentivirus-infected wild-type mice, regardless of the presence of LRG1. In contrast, in diabetic mice, LRG1 treatment increased the phosphorylation of endothelial NO synthase (eNOS) (Fig. 4b, c), the endothelial cell proliferation (BrdU incorporation assay) (Fig. 4d, h), and the expression of the endothelial cell-cell junction protein claudin-5 (Fig. 4e, h), but decreased cavernous endothelial permeability (extravasation of oxidized LDL) (Fig. 4f, h) and apoptosis (TUNEL assay) (Fig. 4g, h). Similar to the findings regarding erectile function, cavernous endothelial cell levels and permeability were preserved in LRG1-Tg mice, even after multiple STZ injections, but were abrogated in the LPHN2-knockdown condition ( Supplementary Fig. 4a). Interestingly, in STZ-induced diabetic mice, intracavernous injection of LRG1 significantly restored penile neuronal levels (βIII-tubulin) and neuronal NOS (nNOS)-containing nerve fibers, and these effects were abolished by LPHN2 knockdown (Fig. 4i, j). LRG1-Tg mice under STZ-induced diabetic conditions exhibited maintained nerve content in DNBs, similar to the cavernous endothelial cell content ( Supplementary  Fig. 4b). Taken together, our results indicate that exogenous LRG1 treatment profoundly restores the decreased numbers of endothelial cells in cavernosum tissue and neurofilaments at DNBs, thereby ameliorating erectile dysfunction in STZ-induced diabetic mice.
The LRG1-LPHN2 Signaling Pathway is Responsible for Angiogenic and Neurotrophic Effects Previous proteomic screens have identified two tyrosine residues in LPHN2 (Tyr1406 and Tyr1421) as candidate phosphorylation sites, although the significance of LPHN2 phosphorylation is unknown 41 . Therefore, we next examined whether LPHN2 is phosphorylated upon LRG1 binding under high-glucose conditions. Immunoprecipitation of LPHN2 and subsequent immunoblotting with anti-phospho-tyrosine antibody demonstrated that LRG1 treatment stimulated LPHN2 phosphorylation within 10 min. These results suggest that LRG1 binding can induce LPHN2mediated intracellular signaling (Fig. 5a).
To examine the LRG1/LPHN2-mediated intracellular signaling pathway responsible for promoting angiogenic and neurotrophic effects, we first used a Cignal Finder GPCR Signaling 10-Pathway Reporter Array in HUVECs upon LRG1 treatment. Of the 10 pathways tested, only nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB) was activated by LRG1 treatment under hyperglycemic culture conditions (Fig. 5b). To systematically investigate the signaling components affected by LRG1, we performed phosphorylation profiling of 363 key signaling proteins in HUVECs after LRG1 treatment. Among the 688 phosphorylation sites on the array, 102 exhibited phosphorylation increases (51 sites) or decreases (51 sites) of >25% upon LRG1 treatment ( Supplementary Fig. 5). We next built a signal network model describing the interactions between these proteins (Fig. 5c). This network model predicted that LRG1 can induce phosphorylation of NF-κB p65 and its upstream kinase AKT, and this prediction was confirmed by Western blot analysis of HUVECs (Fig. 5d). Western blot analysis also revealed phosphorylation of the upstream AKT kinase PI3K, which was not detected by the phosphor-array. Phosphorylation levels of PI3K, AKT, and NF-κB p65 increased 10-30 min after LRG1 treatment under high-glucose conditions and declined after 60 min (Fig. 5d). HUVECs with shRNA-mediated LPHN2 knockdown exhibited reduced phosphorylation of PI3K, AKT, and NF-κB p65, even upon LRG1 treatment (Fig. 5d). Moreover, LRG1-induced phosphorylation of Lyn was markedly decreased by PP2, a Lyn kinase (Src kinase family) inhibitor, and both PP2 and LY294002 (PI3K inhibitor) substantially suppressed PI3K, AKT and NF-κB p65 phosphorylation even in the presence of LRG1 (Fig. 5e). Consequently, both PP2 and LY294002 inhibited tube formation in HUVECs and axonal sprouting from DRG explants, even in the presence of LRG1 (Fig. 5f, g). Finally, we investigated whether LRG1 exerts its angiogenic effects directly or indirectly. The protein levels of typical angiogenic factors were analyzed, and there were no differences in the levels of VEGFA, angiopoietin-1 (ANG1) or fibroblast growth factor 2 (FGF2) between untreated and LRG1-treated HUVECs (Fig. 5h), indicating that endothelial cell proliferation and migration mainly result from the direct effects of LRG1/LPHN2-mediated signaling through Lyn/ PI3K/AKT/NF-κB p65. Collectively, these results imply that PI3K, AKT, and NF-κB p65 signaling constitutes the key signaling pathway for LRG1/LPHN2-mediated angiogenesis and neurite outgrowth under hyperglycemic conditions.

DISCUSSION
In this study, we found that even in the absence of TGF-β, LRG1 can induce endothelial tube formation and migration. Therefore, we postulated that LRG1 may also mediate angiogenesis through an alternative receptor that is independent of TGF-β signaling. Using a ligand-receptor capture (LRC)-based proteomics approach and comprehensive biochemical and cellular analysis, we found that LPHN2 is a novel TGF-β-independent receptor of LRG1. Among LPHN family members (LPHN1-3), LPHN1 and LPHN3 are predominantly expressed in the brain and play critical roles in growth cone migration and synapse formation by binding to adhesion ligands, including teneurin, neurexins, and fibronectin leucine-rich repeat transmembrane proteins (FLRTs) 36 . On the other hand, LPHN2 is widely expressed in various tissues and reportedly participates in showing LPHN2 and LRG1 in mouse cavernous tissue from control and STZ-induced diabetic mice (top). Normalized band intensity values were quantified by ImageJ and are presented as the means ± SEM (n = 4). c and d LPHN2 and LRG1 expression in CC (c) and DNB (d) tissue from control and STZ-induced diabetic mice (DM). Immunostaining for LPHN2 (red, left), LRG1 (red, middle) with PECAM-1 (c, green) or βIII-Tubulin (d, green). Scale bars, 100 μm. Immunopositive areas of LPHN2 and LRG1 in mouse CC (c) and DNB (d) were quantified using ImageJ and are presented as the means ± SEM (n = 6, right). e LPHN2 and LRG1 expression in human CC tissue from control and diabetic patients with erectile dysfunction (DM-ED). Immunostaining for LPHN2 (left, red) and LRG1 (middle, red). Nuclei were stained with DAPI (blue). Scale bars, 100 μm. Quantification of LPHN2-and LRG1-immunopositive areas in human CC using ImageJ (means ± SEM (n = 6, right). f Representative intracavernous pressure (ICP) responses in control and STZ-induced diabetic mice at 2 weeks after repeated intracavernous injections of Fc (negative control, 5 µg/20 µl), LRG1-Fc (5 µg/20 µl), or LRG1-Fc (5 µg/20 µl) + anti-TGF-β1 antibodies (10 μg/ml) on Days 0 and 3 with scramble shRNA control (shCon) or LPHN2 knockdown shRNA (shLPHN2) conditions (5 × 10 5 TU/mouse). The cavernous nerve was stimulated with 5 V. The stimulus interval is indicated by a solid bar. The ratios of mean maximal ICP (f, middle) and total ICP (f, right) (area under the curve) to mean systolic blood pressure (MSBP) were calculated for each group (n = 5). *P < 0.05; **P < 0.01; ***P < 0.001 (Student's t test). N.S., not significant. The relative ratio of the control group was defined as 1.
synapse formation, epithelial-mesenchymal transition (EMT), cardiac development, and vascular morphogenesis 35,[42][43][44][45] . However, little is known about specific ligands of LPHN2 or the downstream signaling pathways and molecular mechanisms underlying the cellular functions of LPHN2. Our results showed that LRG1 specifically and directly binds to the LPHN2 Olf domain but not to LPHN1 or LPHN3.
neural and vascular networks. Accordingly, various factors known as angioneurins (e.g., VEGF, angiopoietin, NGF, BDNF, FGF2, and HGF) have dual roles in vascular and neuronal development, and their malfunction can contribute to both neurological and vascular disorders 47 . Our results demonstrated that LRG1 can function as an angioneurin by directly binding and activating LPHN2. The binding of LRG1 to LPHN2 promoted endothelial cell proliferation and tube formation, microvessel sprouting from aortic rings, and neurite outgrowth from the DRG and MPG through PI3K-AKT-NF-κB p65 phosphorylation. NFκB is a key regulator of innate and adaptive immune responses, which can be activated by the cytokines TNFα and IL-1 and by bacterial components, such as lipopolysaccharide 48 . Other angioneurins, such as VEGF, NT3, and BDNF, can also activate NF-κB to induce vascular endothelial responses and axonal and dendritic growth 49,50 . Although the NF-κB target gene associated with LRG1/LPHN2-mediated angiogenesis and neurite outgrowth remains to be identified, we determined that the PI3K-AKT-NF-κB p65 signaling pathway is entirely independent of TGF-β. In future studies, it will be interesting to determine whether LRG1 can simultaneously or exclusively activate LPHN2-mediated and TGF-β receptormediated signaling in endothelial and neuronal cells under pathophysiological conditions. Erectile dysfunction is common in patients with DM. Despite its substantial negative impacts on quality of life, ED is often undiagnosed in the clinic and remains undertreated. Among diabetic men, the main etiologies of ED are endothelial dysfunction, autonomic neuropathy, and smooth muscle dysfunction 1 . There is an unmet need for therapies that can cure both diabetic vascular and neuronal complications, highlighting the urgency of developing novel treatment options. Several angioneurins, including VEGF, dickkopf2, COMP-Ang1, neurotrophin-3 (NT3), and brain-derived neurotrophic factor (BDNF), have been investigated as potential treatment options for diabetic ED [4][5][6][7][8] . Although the detailed mechanisms are not yet clear, we detected increased LPHN2 expression in the penile tissues of STZ-induced diabetic mice and human diabetic patients. Importantly, intracavernous administration of LRG1 in diabetic mice elicited a wide range of beneficial therapeutic effects through its interaction with LPHN2, including increases in eNOS phosphorylation, endothelial proliferation, endothelial cell-cell junctions, and the proportion of nNOS-containing nerve fibers, as well as decreases in cavernous endothelial permeability and apoptosis (Figs. 3 and 4). These benefits of exogeneous LRG1 ameliorated vascular and neurological abnormalities in the penile tissue of diabetic mice, ultimately leading to the restoration of erectile function. Notably, LRG1 treatment in normal mice did not cause any vascular or neuronal alterations, regardless of the presence of LPHN2, suggesting that intracavernous injection of LRG1 will have limited potential side effects. This preclinical evidence indicates that local administration of LRG1 could be a promising therapeutic strategy for the treatment of diabetic vascular and neuronal complications, such as diabetic erectile dysfunction and foot ulcers. The STZ-induced type 1 diabetic mouse model has been widely used and has contributed to substantial advancements in the study of ED 51 . However, in the clinical setting, only 10% of DM cases are type 1 DM, while 90% of cases are type 2 DM 52 . Thus, there remains a need for further investigation of the therapeutic effects of LRG1 on diabetic ED using animal models of ED associated with type 1 and/ or 2 diabetes.
LRG1 is known to play important roles in diabetic diseases. Patients with type 2 diabetes exhibit elevated plasma LRG1 levels, which seems to be related to the occurrence and progression of diabetic nephropathy 53,54 and proliferative diabetic retinopathy 20,55 . However, several recent studies have shown that LRG1 may be a potential therapeutic option for chronic diabetic wounds, diabetic keratopathy, and obesity [22][23][24] . Overall, these findings suggest that LRG1 is a multifunctional protein with context-dependent effects. Indeed, LRG1 is reportedly involved in neovascularization, epithelial-mesenchymal transition (EMT), neutrophil activation, re-epithelialization, and nerve regeneration 22,23,25,56,57 . Additionally, LPHN2 expression is highly increased during the differentiation of pluripotent stem cells into cardiac progenitor cells and cardiomyocytes, and Lphn2 −/− embryos exhibit serious defects in right ventricle, right atrium, and outflow tract development 43 , suggesting that the LRG1/ LPHN2 signaling axis may also be involved in vascular and neuronal development.
Based on our current and previous findings, we propose a model suggesting that high circulating LRG1 in serum can only exert its cellular functions at specific sites with increased expression of LPHN2 and/or TGF-β, although the role of highly abundant serum LRG1 is not yet clear. Conversely, this model predicts that under pathological conditions, excess LRG1 expression or overactivation of the LRG1/LPHN2 axis may cause chronic inflammation via NF-κB activation 21 . Thus, LRG1 can be viewed as a double-edged sword, similar to growth factors in cancers and cytokines in sepsis. It remains to be clarified whether increased LRG1 under pathological conditions causes pathological abnormalities or is a consequence of compensatory effects during recovery in pathologically damaged tissues. Since LRG1 has long been considered a promising biomarker for various pathologies, including inflammation, neurodegenerative disease, heart failure, and several types of cancer [10][11][12][13][14][15][16][17][18] , further studies should investigate other roles of LRG1 in these diseases.

DATA AVAILABILITY
All study data are included in the article and/or Supplemental Information. The phospho array data have been deposited in Gene Expression Omnibus (GSE160005). Fig. 5 Identification of the LRG1-LPHN2 signaling pathway. a HUVECs were stimulated with LRG1 (1 μg/ml) for the indicated times under high-glucose conditions. Immunoprecipitants with an anti-LPHN2 antibody were analyzed by Western blotting using antibodies specific for pY-LPHN2 and LPHN2. b Cignal Finder GPCR signaling 10-pathway reporter array analysis of HUVECs after treatment with LRG1 (1 μg/ml). The results of dual-luciferase assays are presented as normalized relative luminescence signals (means ± SEM, n = 3). The relative ratio of the untreated group was defined as 1. **P < 0.01 (Student's t test). c Network model showing interactions between the proteins with increased phosphorylation in response to LRG1. Node color indicates the increase (red) in the phosphorylation level. The color bar indicates the gradient of the log2-fold-change of phosphorylation levels by LRG1 with respect to those in untreated control conditions. Circled P on a node indicates phosphorylation of the corresponding protein. Arrows, activation; inhibition symbols, inhibition; solid arrows, direct activation; dotted arrows, indirect activation; gray lines, protein-protein interactions; and green lines, plasma membrane. d Western blot analysis of HUVECs stimulated with LRG1 (5 μg/ml) using the indicated antibodies. e Western blot analysis of HUVECs stimulated with 1 µg/ml LRG1 with or without pretreatment with PP2 (10 μM) or LY294002 (10 μM). f and g HUVECs (f ) and mouse DRG explants (g) were treated with LRG1 (1 μg/ml), LY (10 μM), PP2 (10 μM), LRG1 (1 μg/ml) + LY (10 μM), or LRG1 (1 μg/ml) + PP2 (10 μM). Top: Representative images of HUVEC tube formation (f) and βIIItubulin staining in mouse DRG explants (g). The relative ratio of the untreated group was defined as 1. Scale bars, 100 µm. Bottom: Master junctions (n = 4) (f) and βIII-tubulin-immunopositive axon lengths in DRGs (g) were quantified using ImageJ, and the results are presented as the means ± SEM (n = 4). h Western blot analysis of HUVECs stimulated with 1 µg/ml LRG1 for 30 min or 7 hr using antibodies specific for angiogenic factors (VEGFA, angiopoietin-1, FGF2).