Macrophages play a crucial role in vascular smooth muscle cell coverage

Angiogenesis involves the formation of new blood vessels from pre-existing vasculature. Whereas coordinated angiogenesis is essential for normal embryonic and postnatal development, disorganized angiogenesis causes either a reduction in blood vessels or excess neovascularization, and is related to various pathologies such as malignant tumors, diabetic retinopathy and inflammation. During angiogenesis, interactions between endothelial and mural cells, pericytes, and vascular smooth muscle cells (VSMCs) are essential for vascular integrity (Gaengel et al., 2009). Additionally, previous studies have suggested that mural cell loss or dysfunction triggers vessel-related pathological conditions such as diabetic retinopathy (Hammes et al., 2002). The distribution of pericytes and VSMCs differs greatly. Generally, pericytes are evenly distributed along small blood vessels and capillaries, whereas VSMCs cover large vessels and arterioles preferentially, but not venules and capillaries. Despite the importance of mural cells during angiogenesis, the molecular mechanisms that regulate the number and distribution of mural cells during in vivo angiogenesis remain unclear. Several molecular mechanisms underlying endothelial cell-pericyte communication in vascular homeostasis have been identified. In particular, the platelet-derived growth factor B (PDGF-B) homodimer (PDGF-BB) derived from endothelial cells plays a pivotal role in the recruitment, proliferation and survival of pericytes expressing PDGF receptor β (PDGFRβ) (Armulik et al., 2005; Hoch and Soriano, 2003). The systemic deletion of Pdgfb and Pdgfrb results in pericyte loss and severe vessel abnormalities, and is associated with embryonic lethality (Hellström et al., 1999). Moreover, a previous study has shown that the genetic ablation of Pdgfb in endothelial cells results in severe pericyte depletion and endothelial cell dysfunction in developing retinas (Park et al., 2017). These reports indicate that PDGF-B plays a key role in pericyte recruitment and vessel stabilization. Interestingly, another study showed that PDGFRβ⁺ NG2⁺ perivascular cells express mesenchymal stem cell markers and have multi-lineage potential, including myofiber, chondrocyte and adipocyte lineages (Crisan et al., 2008). Notably, it has been reported that a subset of pericytes can differentiate into VSMCs (Volz et al., 2015). Several in vitro pericyte culture studies have shown that transforming growth factor β (TGFβ) is associated with the acquisition of a contractile phenotype such as VSMCs (Armulik et al., 2005). However, the detailed molecular and cellular basis underlying pericyte-derived VSMC supply during angiogenesis remains unclear.

Forkhead transcription factor O1 (FOXO1) is a multi-functional transcription factor that promotes the transcription of genes involved in various physiological processes, such as cell cycle arrest, apoptosis and oxidative stress resistance (Accili and Arden, 2004; Eijkelenboom and Burgering, 2013; Huang and Tindall, 2007). Although FOXO1 is most abundant in adipose tissue, liver tissue and pancreatic β cells (Accili and Arden, 2004; Nakae et al., 2003), blood and lymphatic endothelial cells also express FOXO1 during development (Niimi et al., 2020; Wilhelm et al., 2016). Previously, we reported that the systemic deletion of murine Foxo1 resulted in embryonic lethality in the mid-gestational stage owing to large vessel abnormalities (Furuyama et al., 2004). Furthermore, we demonstrated that FOXO1 was preferentially expressed in tip cells (specifically, differentiated endothelial cells for distal migration) rather than in proximal stalk endothelial cells (Fukumoto et al., 2018). Endothelial cell-specific deletion of Foxo1 impaired tip cell behavior in the developing mouse retinal angiogenic front (Fukumoto et al., 2018). Another study showed that endothelial cell-specific deletion of Foxo1 increased retinal blood vessel density, whereas the overexpression of Foxo1 reduced vessel density by regulating endothelial cell proliferation (Wilhelm et al., 2016). We also demonstrated that endothelial cell-specific deletion of Foxo1 alters pericyte coverage to an aberrant morphology with disturbed cytoplasmic processes (Niimi et al., 2019). Despite these findings regarding Foxo1 function in endothelial cells, whether endothelial Foxo1 affects the pericyte-to-VSMC transition has not yet been investigated. In this study, we aimed to investigate the role of endothelial Foxo1 in pericyte-derived VSMCs during retinal angiogenesis, using transgenic mice expressing a constitutively active form of FOXO1.

To determine the role of Foxo1 in mural cell distribution, we used a mouse model with endothelial cell-specific tamoxifen-inducible overexpression of a constitutively active Foxo1 mutant (Cdh5-CreER^T2; R26-stop^f/f-Foxo13A; referred to as Foxo1^CAiEC). Tamoxifen was administered to the mice on postnatal days (P) 1, 2 and 3, and they were euthanized on P6 (Fig. 1A). Consistent with a previous study (Kim et al., 2019), anti-FOXO1 immunohistochemistry revealed that FOXO1 was abundant in tip endothelial cells, whereas stalk endothelial cells expressed low levels of FOXO1 in the retinas of P6 wild-type mice (Fig. 1B). An increase in nuclear FOXO1 expression in Foxo1^CAiEC retinal vasculature was confirmed using anti-FOXO1 immunohistochemistry (Fig. 1B-D). As previously reported (Wilhelm et al., 2016), FOXO1 overexpression decreased radial vessel expansion and endothelial cell density (Fig. 1E-H). In the angiogenic front region, typical sprouts consisting of a long stalk and tip endothelium were reduced in Foxo1^CAiEC retinas compared with wild-type retinas (Fig. 1F,I). As tip cells highly express FOXO1, ectopic expression of FOXO1 in all endothelial cells may result in a tip phenotype. To verify this hypothesis, we performed immunohistochemistry or fluorescence in situ hybridization to visualize the expression of several tip cell markers (Esm1, Pdgfb, Angpt2 and Dll4) in the angiogenic front region of both wild-type and Foxo1^CAiEC retinas. These tip cell markers were locally expressed in only the tip cells of wild-type retinas; however, proximal endothelial cells also expressed these markers in Foxo1^CAiEC retinas, except for DLL4 (Fig. 1J-Q). These results indicate that FOXO1 plays a role in determining tip cell phenotype during angiogenesis. Because this study focused on pericytes and VSMCs, the detailed mechanisms regarding endothelial cell abnormalities, such as reduced vessel sprouting in Foxo1^CAiEC retinas, were not further investigated.

In Foxo1^CAiEC retinas, Pdgfb expression was upregulated and not restricted to tip cells (Fig. 1N,O). As PDGF-B is a well-established PDGFRβ⁺ pericyte attractant, we subsequently investigated pericyte coverage using anti-NG2 and anti-PDGFRβ immunohistochemistry in wild-type and Foxo1^CAiEC retinas. Consistent with the upregulation of Pdgfb, both NG2⁺ and PDGFRβ⁺ pericyte coverage increased significantly in the angiogenic front of Foxo1^CAiEC retinas (Fig. 2A-D). Interestingly, a substantial number of ectopic α-smooth muscle actin (αSMA)-positive VSMCs were present in the angiogenic front of Foxo1^CAiEC retinas, whereas only a few VSMCs were present in the wild-type angiogenic front (Fig. 2E-G). Here, we defined αSMA-positive perivascular cells as VSMCs. As it was difficult to distinguish the αSMA expression in mural cells from that in endothelial cells in the Z-stacked images, we confirmed that αSMA was certainly upregulated in mural cells of Foxo1^CAiEC retinas by observing single slice images (Fig. S1A,B). VSMC coverage pattern in small vessels near the optic disc (high in arteriole and little in venule) was comparable between wild-type and Foxo1^CAiEC retinas (Fig. 2H). However, αSMA expression was also upregulated in the venule of Foxo1^CAiEC retinas (Fig. 2I). As previous reports indicated that a subset of pericytes has the potential to differentiate into VSMCs via the TGFβ signaling pathway (Papetti et al., 2003), we hypothesized that excess TGFβ signaling forces pericytes to preferentially differentiate into VSMCs in Foxo1^CAiEC retinas. To investigate TGFβ-producing cells in the retinal angiogenic front region, we performed a FISH study using a Tgfb1 probe. The major Tgfb1-producing cells were endothelial cells; however, the counts of Tgfb1 signals in the endothelium did not vary in wild-type and Foxo1^CAiEC retinas (Fig. 3A,B). Conversely, we observed that ∼80% of IBA1⁺ macrophages located in the perivascular area also expressed Tgfb1 (Fig. 3A,C), and the number of Tgfb1-producing macrophages was higher in Foxo1^CAiEC retinas than that in wild-type retinas (Fig. 3D). The total number of IBA1⁺ macrophages increased significantly in Foxo1^CAiEC retinas compared with that in wild-type retinas (Fig. 3E,F), without promoting macrophage proliferation or the inflammatory response (Fig. S2A-C). These IBA1⁺ macrophages were located near the NG2⁺ pericytes, particularly in the Foxo1^CAiEC retina (Fig. 3G,H). Moreover, phosphorylated SMAD3⁺ (downstream of TGFβ1 signaling) PDGFRβ⁺ pericytes increased significantly in Foxo1^CAiEC retinas, indicating that TGFβ signaling was active in the pericytes of Foxo1^CAiEC retinas (Fig. 3I,J). These results indicate that endothelial FOXO1 attracts macrophages to nearby blood vessels to differentiate pericytes into VSMCs via macrophage-derived TGFβ1 signaling.

Subsequently, we investigated endothelial cell-derived factors that attracted macrophages to the endothelial cells. C-C motif chemokine ligand 2 (CCL2), also known as monocyte chemoattractant protein 1, and C-X-C motif chemokine ligand 12 (CXCL12) are known to attract myeloid cells including macrophages (Vogel et al., 2014). However, we found that CCL2 and CXCL12 mRNA expression in constitutively active Foxo1 (Foxo1-3A)-transfected human umbilical vein endothelial cells (HUVECs) and empty vector-transfected HUVECs were comparable (Fig. S3A,B). The semaphorin 3 (SEMA3A, SEMA3B, SEMA3C, SEMA3D, SEMA3E, SEMA3F and SEMA3G) family includes secretory-type semaphorins related to the morphogenesis of various tissues, including those in the cardiovascular and nervous systems (Kruger et al., 2005; Zhou et al., 2008). A previous study showed that SEMA3A, SEMA3C, SEMA3E and SEMA3G function as attractants for macrophages (Shimizu et al., 2013). Another study suggested that SEMA3C expression is regulated by FOXO1 (Ramaswamy et al., 2002). Based on these findings, we investigated the types of semaphorin 3 expressed in HUVECs. Reverse transcription PCR (RT-PCR) analysis revealed that the HUVECs expressed SEMA3A, SEMA3C, SEMA3D and SEMA3F (Fig. 4A). The expression level of SEMA3C was lower than that of SEMA3A, SEMA3D and SEMA3F. Among the semaphorin 3 family expressed in HUVECs, only SEMA3C mRNA expression was upregulated by Foxo1-3A overexpression (Fig. 4B). Anti-SEMA3C western blot analysis of conditioned medium cultured with HUVECs (Foxo1-3A or empty) revealed that Foxo1-3A overexpression upregulated SEMA3C secretion into the culture medium (Fig. 4C,D). We performed a transwell cell migration assay to confirm that SEMA3C secreted from endothelial cells functions as a chemoattractant for macrophages. The conditioned medium cultured with HUVECs transfected with either Foxo1-3A or an empty vector was placed in the basal side of the transwell chamber, with or without anti-SEMA3C neutralizing antibodies. The macrophage cell line RAW264.7 was seeded into the apical side of the chamber and incubated to allow migration in response to the gradient concentration of humoral factors contained in the basal chamber. As a result, the HUVEC-conditioned medium transfected with Foxo1-3A attracted considerably more RAW264.7 cells to the basal side of the chamber, compared with the medium with empty-transfected HUVECs (Fig. 4E,F). Moreover, the anti-SEMA3C neutralizing antibody significantly reduced the number of attracted cells in the medium with HUVECs transfected with Foxo1-3A (Fig. 4E,F).

Next, we performed fluorescence in situ hybridization using a Sema3c probe to identify SEMA3C-producing cells during angiogenesis. In the angiogenic front region, most SEMA3C-producing cells were endothelial cells, and Sema3c mRNA signal intensity increased significantly in Foxo1^CAiEC retinas compared with that in wild-type retinas (Fig. 5A-C). Subsequently, we investigated the receptor expression of SEMA3C in macrophages in vivo. SEMA3C receptors include the neuropilin (Chen et al., 1997; He and Tessier-Lavigne, 1997; Kolodkin et al., 1997) and plexin families (Comeau et al., 1998; Winberg et al., 1998). A previous study showed that SEMA3C is a repulsive guidance cue for cortical axons when the receptor is a neuropilin 1 (Nrp1) homodimer. However, it is an attractive guidance cue when the receptor is an Nrp1/Nrp2 heterodimer or a Nrp2 homodimer (Ruediger et al., 2013). This indicates that Nrp2 is essential for the chemoattractant activity of SEMA3C. To investigate neuropilin expression in retinal macrophages, we performed a fluorescence in situ hybridization study using Nrp1 and Nrp2 probes. Approximately half of the IBA1⁺ cells near the front vasculature expressed the Nrp2 mRNA, whereas less than 20% of IBA1⁺ cells expressed the Nrp1 mRNA (Fig. 5D-F). In Foxo1^CAiEC retinas, Nrp2⁺ IBA1⁺ cells were enriched in the angiogenic front region (Fig. 5E,G). To confirm the role of SEMA3C in the phenotype of Foxo1^CAiEC retinas, we intravitreally administered an anti-SEMA3C neutralizing antibody in wild-type and Foxo1^CAiEC mice. Tamoxifen was intraperitoneally administered at P1, P2 and P3. Subsequently, normal rat IgG and anti-SEMA3C neutralizing antibodies were intravitreally injected at P4 and P5 in the left and right eyes, respectively (Fig. 6A). Vessel density was not altered by SEMA3C neutralization in wild-type and Foxo1^CAiEC mice (Fig. 6B,C). An anti-IBA1 immunohistochemistry study in P6 retinas revealed that SEMA3C neutralization reduced macrophage density in both wild-type and Foxo1^CAiEC retinas (Fig. 6D,E). These results indicate that SEMA3C attracts macrophages not only under Foxo1-abundant conditions but also under physiological conditions in normal developing retinas. Notably, excess VSMC coverage in Foxo1^CAiEC retinas was almost completely counteracted by SEMA3C neutralization (Fig. 6F,G). Indeed, increased phosphorylated SMAD3⁺ pericytes in Foxo1^CAiEC retinas reduced significantly in SEMA3C-neutralized Foxo1^CAiEC retinas, without changing the number of PDGFRβ cells (Fig. 6H-J).

To exclude the possibility that SEMA3C directly promotes the pericyte-to-VSMC transition independently of macrophages, we performed a drug-induced macrophage depletion experiment. Intraperitoneal administration of the colony-stimulating factor 1 receptor (CSF1R) inhibitor Ki20227 has been reported to reduce the number of retinal macrophages (Kubota et al., 2009). As Ki20227 also inhibits VEGFR2 and PDGFRβ activity at a high dose, we injected Ki20227 at a low dose (0.1 mg kg⁻¹ d⁻¹) every other day to avoid primary vessel abnormalities (Fig. 7A). Ki20227 administration effectively suppressed macrophage accumulation in P6 Foxo1^CAiEC retinas without affecting vessel density (Fig. 7B-E). Moreover, Ki20227 administration inhibited ectopic VSMC coverage in Foxo1^CAiEC retinas without affecting pericyte recruitment (Fig. 7F-I). These results suggest that SEMA3C-dependent TGFβ-producing macrophage attraction is essential for excess VSMC coverage in Foxo1^CAiEC retinas.

To investigate the physiological relevance of Foxo1 on VSMC coverage, we used endothelial cell-specific Foxo1 knockout mice (Cdh5-CreER^T2; Foxo1^flox/flox, referred to as Foxo1^ΔiEC) (Fig. S4A). Anti-FOXO1 immunohistochemistry revealed that Foxo1 was successfully knocked out in Foxo1^ΔiEC retinas (Fig. S4B). As previously reported (Kim et al., 2019; Park et al., 2017), vessel density was decreased and front vasculature was malformed in Foxo1^ΔiEC retinas, compared with those of wild-type retinas (Fig. S4B,C). These results indicate successful loss of function of FOXO1 in Foxo1^ΔiEC retinas. Unexpectedly, αSMA⁺ VSMC coverage was upregulated in the angiogenic front region of Foxo1^ΔiEC retinas, similar to Foxo1^CAiEC retinas (Fig. S4D,E). Interestingly, the number of PDGFRβ⁺ pericytes, the pSMAD3⁺ to PDGFRβ⁺ ratio and the recruitment of IBA1⁺ macrophages also increased in Foxo1^ΔiEC retinas, similar to Foxo1^CAiEC retinas (Fig. S4F-J). To determine the cause of this paradoxical phenotype, we performed quantitative RT-PCR to analyze SEMA3A, SEMA3C, SEMA3D and SEMA3F expressions in HUVECs transfected with FOXO1-specific (siFOXO1) or non-targeting (siNC) siRNA. Consistent with Foxo1-3A overexpression (Fig. 4B), SEMA3C was downregulated in siFOXO1-transfected HUVECs (Fig. S4K). Interestingly, SEMA3A and SEMA3F expression was significantly upregulated in siFOXO1-transfected HUVECs compared with that in siNC-transfected HUVECs (Fig. S4K). As SEMA3A has also been reported to be an attractant for macrophages (Shimizu et al., 2013), SEMA3A upregulation may surpass the effect of SEMA3C reduction. These results indicate that the normal expression level of FOXO1 in tip cells is tightly regulated; any dysregulation leading to lower or higher FOXO1 expression levels than normal results in a disorganized mural cell distribution. Last, we investigated the physiological significance of SEMA3C in mural cell coverage during retinal angiogenesis. As shown in Fig. 6F, SEMA3C neutralization had no effect on VSMC coverage in wild-type P6 retinas, except for a slight decrease in the number of IBA1⁺ cells. We hypothesized that a longer incubation period might be required for the VSMCs to respond to SEMA3C neutralization. Therefore, we injected SEMA3C-neutralizing antibodies at P2, P4 and P6 in the wild-type retina, with P14 retinas used for immunohistochemistry (Fig. 8A). Interestingly, anti-αSMA immunohistochemistry revealed a significant reduction in the length of VSMC-covered arteriole in SEMA3C-neutralized retina (Fig. 8B,C), without affecting PDGFRβ⁺ cell coverage (Fig. 8D). This indicates that SEMA3C physiologically contributes to the arterial VSMC supply during retinal vasculature development.

Our study proposes a multicellular mechanism for mural cell supply during angiogenesis (Fig. 8E). FOXO1 in tip endothelial cells promotes PDGF-B expression and attracts PDGFRβ⁺ pericyte progenitors to the front angiogenic vasculature. Concurrently, tip FOXO1 attracts retinal macrophages to the perivascular region via SEMA3C. The recruited macrophages adjoin pericytes, consequently activating macrophage-dependent TGFβ1 signaling, resulting in the initiation of pericyte-to-VSMC transition. A previous study has revealed that pericytes certainly express the TGFβ1 receptor ALK5, and TGFβ1/ALK5 contributes to vessel morphogenesis via Smad2/3 phosphorylation in pericytes (Dave et al., 2018). Considering that the endothelium, rather than macrophages, preferentially expresses Tgfb1 (Fig. 3A), endothelial cell-derived Tgfb1 may promote the pericyte-to-VSMC transition. Conversely, a previous immunohistochemistry study demonstrated higher expression of TGFβ1 at P3 in retinal perivascular IBA1⁺ cells than in endothelial cells (Biswas et al., 2017). This suggests that macrophages are the major source of Tgfb1 during the early developmental period. As TGFβ is secreted in its inactive latent form, it requires protease-dependent cleavage or integrin-dependent structural changes to be activated (Massagué and Chen, 2000). Macrophages synthesize and secrete the TGFβ activator thrombospondin1 (Jaffe et al., 1985). Recently, a study demonstrated that activated macrophages transiently express integrin αv, which can physically activate latent TGFβ (Li et al., 2023). These macrophage-derived factors might partially contribute to the activation of endothelium-derived latent TGFβ in the basement membrane.

As both Foxo1 deletion and overexpression upregulated macrophage-dependent VSMC coverage, it is speculated that the physiological expression level of FOXO1 in tip cells is optimized to execute temporally coordinated pericyte-to-VSMC differentiation. Determining FOXO1 expression in tip cells is interesting. Park et al. demonstrated that pericytes stabilize endothelial cells via the Ang1-(pericyte)/Tie2-(endothelial cell) dependent inactivation of endothelial FOXO1 (Park et al., 2017). Conversely, pericyte detachment activates FOXO1 and destabilizes endothelial cells, enabling them to behave as tip cells. The fine balance between pericyte attachment and detachment around the tip endothelial cells might determine FOXO1 activation and/or inactivation, and regulate FOXO1 levels in tip cells.

A recent genome-wide in vitro study has shown that FOXO1 contributes to the transcription of tip cell genes (Esm1, Angpt2 and Cxcr4) in endothelial cells (Miyamura et al., 2024). In addition to these tip-cell genes, we identified Sema3c as a tip-derived macrophage attractant that is transcriptionally regulated by FOXO1. Although a previous study found that pericytes abundantly expressed SEMA3C, compared with endothelial cells (Yang et al., 2015), we did not observe perivascular Sema3C-FISH signals in vivo. As the origin of pericytes is heterogeneous (Yamazaki and Mukouyama, 2018), SEMA3C expression in mural cells might vary among different tissues. To investigate the detailed function of endothelial SEMA3C in tip endothelial cells, an analysis of conditional Sema3C knockout phenotypes is necessary, because heterogeneous Sema3c mutant mice die of artery malformation within 24 h of birth (Feiner et al., 2001). As shown in Fig. 6B,C, SEMA3C neutralization did not affect retinal vessel density. A previous study has shown that intravitreal injection of SEMA3C-Fc inhibits retinal vessel formation by suppressing several angiogenic intracellular signaling pathways in endothelial cells (Yang et al., 2015). Considering these findings together with the comparatively low SEMA3C expression levels observed in the wild-type angiogenic front (Fig. 5A,B), the contribution of SEMA3C to endothelial intracellular angiogenic signaling might be small. SEMA3C has previously been identified as a FOXO1 target gene (Ramaswamy et al., 2002); however, whether its transcriptional regulation is direct or indirect has not yet been investigated. The public ChIP-seq database (ChIP-Atlas; SRX5548892) showed no reliable reads precipitated by the anti-FOXO1 antibody around the human SEMA3C promoter in HUVECs. This suggests that the regulation is indirect.

Our present study demonstrates that SEMA3C-dependent macrophage attraction to tip cells is required for arteriolar VSMC coverage during vasculature development in wild-type mice (Fig. 8). Previous reports have shown that ESM1⁺ tip cells are committed to becoming arteriole endothelial cells in mice retina (Xu et al., 2014). Pericytes around tip cells, which initiate differentiation into VSMCs, might follow the same trajectory as adjacent tip cells to the arteriole. In conclusion, these findings shed light on the molecular and cellular mechanisms regulating mural cell supply during angiogenesis, and may provide therapeutic targets for mural cell-associated pathology.

The Cdh5-CreER^T2 mouse strain (Okabe et al., 2014) was crossed with the Rosa26-stop^flox/flox-Foxo13A mouse strain (Iskandar et al., 2010) to obtain tamoxifen-inducible endothelial cell-specific overexpression of the constitutively active Foxo1 mutant. FOXO1 is phosphorylated by AKT at the three serine or threonine residues to enable its nuclear translocation (Brunet et al., 1999; Huang and Tindall, 2007); therefore, mutated Foxo1 with alanine substitutions at these three residues (Foxo1-3A) is always localized in the nucleus and acts as a constitutively active transcription factor (Furuyama et al., 2003). Tamoxifen (Sigma-Aldrich) was dissolved in peanut oil (10 mg ml⁻¹) and administered (50 μg g⁻¹ body weight) to the mice intraperitoneally at P1, P2 and P3. The mice were euthanized at P6. Tamoxifen-treated Cdh5-CreER^T2×Rosa26-stop^flox/flox-Foxo13A (Foxo1^CAiEC) and Cdh5-CreER^T2 mice with the wild-type Rosa26 locus were used for the experiments. To achieve endothelium-specific deletion of Foxo1, we used Cdh5-CreER^T2×Foxo1^flox/flox mice (Foxo1^ΔiEC) (Miyazaki et al., 2012; Okabe et al., 2014). The tamoxifen treatment regimen in Foxo1^ΔiEC mice was the same as that in Foxo1^CAiEC mice. The CSF1R inhibitor Ki20227 (Cayman Chemical) was administered (0.1 mg kg⁻¹ day⁻¹) intraperitoneally at P1, P3 and P5 to deplete retinal macrophages at P6. Anti-SEMA3C neutralizing antibodies (rat anti-SEMA3C monoclonal antibody; R&D Systems, MAB1728, AB_2301533) and isotype IgG were administered intravitreally in the right and left eyes of the same pup, respectively. This was performed using low-volume (10 µl) glass syringes and 34 G needles. The antibody was diluted in sterile PBS (200 ng µl⁻¹), and 1 µl of that diluted solution was injected in each mouse daily. We performed all animal experiments without sex distinction of the pups. All animal experiments were approved by the Committee of Animal Experiments of the Kagawa Prefectural College of Health Science. We minimized the number of animals used and their discomfort during the experiment.

The mouse retina is commonly used as a model for observing postnatal angiogenesis. Retinal neovascularization begins during the postnatal period and expands radially from the optic disc to the periphery, facilitating observation of the angiogenic front region. As retinal neovascularization reaches the retinal periphery around the P7–P8 period, P6 is a suitable time point for observing developmental angiogenesis. We isolated eyeballs from either P6 or P14 pups and made a hole in the cornea using a 27 G needle. The eyeballs were subsequently fixed in 4% paraformaldehyde for 5 (P6) or 10 min (P14). After washing, the sclera and choroid were peeled off, and the cornea, iris, vitreous and vitreous vessels were removed using forceps. The retinas were stored in methanol at −25°C.

Retinas were fixed in 4% paraformaldehyde and blocked with 0.5% blocking reagent (Perkin Elmer). Subsequently, retinas were incubated with diluted primary antibodies overnight at 4°C. The primary antibodies used included: rat anti-CD31 monoclonal antibody (BD Biosciences, 550274, AB_393571, 1:100); goat anti-collagen type IV polyclonal antibody (Merck Millipore, AB769, AB_92262, 1:1000); rabbit anti-ERG monoclonal antibody (Abcam, ab92513, AB_2630401, 1:1000); rat anti-NG2/MCSP monoclonal antibody (Novus Biologicals, MAB6689, AB_10890940, 1:100); rat anti-PDGFRβ monoclonal antibody (Thermo Fisher Scientific, 14-1402, AB_467493, 1:200); mouse anti-α smooth muscle actin FITC-conjugated monoclonal antibody (Sigma-Aldrich, F3777, AB_476977, 1:500); rabbit anti-FOXO1 monoclonal antibody (Cell Signaling Technology, 2880, AB_2106495, 1:200); rabbit anti-KI67 polyclonal antibody (Abcam, ab15580, AB_443209, 1:1000); rabbit anti-IBA1 monoclonal antibody (Cell Signaling Technology, 17198, AB_2820254, 1:500); goat anti-DLL4 polyclonal antibody (R&D Systems, AF1389, AB_354770, 1:100); rabbit anti-SMAD3 (phospho S423+S425) monoclonal antibody (Abcam, ab52903, AB_882596, 1:200); and goat anti-ESM1 polyclonal antibody (R&D Systems, AF1999, AB_2101810, 1:200). The retinas were washed in 0.5% TritonX-100, and further incubated with the appropriate fluorescent dye-conjugated secondary antibodies for 2 h at 20°C. After this, the retinas were washed and flat mounted. ABC signal amplification was needed to detect FOXO1. All fluorescent images were acquired using a FV10i confocal fluorescence microscope (Olympus). All images were captured in the depth of the retinal superficial plexus layer; the deeper layers were not captured for any images. The depth of each z-stacked image was approximately 5-15 µm.

The RNAscope Multiplex Fluorescent Reagent Kit (Advanced Cell Diagnostics) was used for conducting fluorescence in situ hybridization, according to the manufacturer's protocol, with some modifications. Briefly, the retinas were fixed in 4% paraformaldehyde overnight at 4°C. They were then subjected to hydroperoxide (20°C, 10 min) and protease treatments (40°C, 7 min). After washing, probes were applied to the retinas for hybridization (40°C, 3 h or overnight). Signal amplification was performed following the manufacturer's instructions. However, instead of using the kit component wash buffer, 1×PBST containing 1% bovine serum albumin was used for washing the retinas. The FISH-treated retinas were subsequently subjected to immunohistochemistry analysis. The target probes used in this study were obtained from Advanced Cell Diagnostics and the catalog codes are as follows: Mm-Pdgfb C1 probe (424651), Mm-Angpt2 C1 probe (406091), Mm-Tgfb1 C1 probe (407751), Mm-Sema3c C1 probe (441441), Mm-Nrp1 C1 probe (471621) and Mm-Nrp2 C1 probe (500661). The probe sequences were not provided by the manufacturer.

All image analyses, including vessel area measurement, cell number count and fluorescence intensity analysis, were performed using the ImageJ software (version 1.51j8) (Abràmofff et al., 2004; Schneider et al., 2012). The vessel area was measured using binarized CD31 immunohistochemistry images (×100) and divided by the vascularized area. Macrophage numbers were determined using automated cell counts in binarized IBA1 immunohistochemistry images (×200). The number of pericytes was counted manually (×600) and divided by the vessel area obtained from the binarized collagen IV immunohistochemistry images of the same microscopic fields. Because counting the overlapped pericytes in z-stacked images was difficult, we quantified the pericyte number in multiple single slice images while shifting focus. Because anti-NG2 and anti-PDGFRβ immunohistochemistry showed cytoplasmic signals with the absence of nuclear signals, we counted pericytes as the nuclei number. The number of pSMAD3⁺ pericytes was also counted manually (×600) and divided by the total PDGFRβ⁺ cell count of the same microscopic field. The IBA1⁺ macrophages contacting NG2⁺ pericytes were quantified manually in ×600 multiple slice images, while shifting the depth and then dividing by the count of total IBA1⁺ cells in the same images. We excluded from the count the macrophages that simply touched the pericytes via only the elongated cytoplasmic processes. Fluorescence intensity was measured in the original ×600 images where the vascularized area was the region of interest. The images acquired by fluorescence in situ hybridization were quantified by counting positive signals. To exclude the count of fluorescence in situ hybridization signals from non-endothelial cells, we configured region of interest exactly along the contour of CD31 immunostaining signals and quantified the fluorescence in situ hybridization signals only in the region of interest. Measurements were performed using multiple microscopic fields from one retina and averaged as one data point, except for low-magnification (×100) images. Data from at least three mice were analyzed in each experiment.

The HUVECs (pooled cells from three newborn donors; PromoCell, C-12208) were cultured in an endothelial cell basal medium supplemented with fetal calf serum, epidermal growth factor (EGF), basic fibroblast growth factor (bFGF), insulin-like growth factor (IGF), vascular endothelial growth factor (VEGF), ascorbic acid and hydrocortisone (PromoCell). HUVECs were authenticated and confirmed to have no microbiological contamination or infectious viruses by the provider. RAW264.7 cells were cultured in Dulbecco's modified Eagle medium (low glucose) containing 10% fetal bovine serum. Both cells were cultured in a 5% CO₂ incubator at 37°C, with a humidity of 95% or more. To overexpress Foxo1, HUVECs were transfected with the Foxo1-3A overexpression vector (pcDNA3.1-Foxo1-myc) using the Via-Fect Transfection Reagent (Promega) for 24 h. The empty pcDNA3.1 vector was used as the negative control. Construction of the Foxo1-3A overexpression vector has been previously reported (Furuyama et al., 2000). To reduce FOXO1 expression, HUVECs were transfected with a specific siRNA against FOXO1 mRNA (siGENOME Human FOXO1 siRNA, Horizon Discovery) using the Lipofectamine RNA iMAX reagent (Thermo Fisher Scientific) for 48 h. A randomized siRNA sequence (siNC) was used as the negative control.

The HUVECs were homogenized in the ISOGEN reagent (Nippon Gene), total RNA was isolated, and reverse transcription was performed using the ReverTra Ace reagent (TOYOBO). Quantitative mRNA expression analysis was performed using the THUNDERBIRD qPCR reagent (TOYOBO) and the Step One Plus real-time PCR system (Applied Biosystems). The primers used for the RT-PCR and qPCR analyses are listed in Table S1.

The HUVECs and the conditioned media cultured with HUVECs were lysed in Laemmli sample buffer and subjected to SDS-PAGE. Proteins were transferred to polyvinylidene difluoride membranes and blocked with the Blocking One reagent (Nacalai Tesque) for 30 min at room temperature. The membranes were further incubated with rat anti-SEMA3C monoclonal antibodies (R&D Systems, MAB1728, AB_2301533, 1:250) overnight, at 4°C. After washing, the membranes were incubated with the appropriate horseradish peroxidase-conjugated secondary antibodies. Next, the membranes were subjected to chemiluminescence-dependent antigen detection. After chemiluminescence detection, the membrane was stained with the Ponceau S staining solution as a loading control.

The Foxo1-overexpression vector or an empty vector was transfected in HUVECs for 1 days and the conditioned media were prepared in the basolateral side of the transwell chamber (Corning, 353097, 8 µm pore membrane) with/without anti-SEMA3C neutralizing antibodies. Normal rat IgG was used for the isotype control. RAW264.7 cells were seeded onto the apical side of the Transwell chamber at a concentration of 5.0×10⁴ cells/well. After 4 h of incubation, the basolateral side of the transwell membrane was fixed and stained with the Giemsa stain. Five microscopic fields (×200) per membrane were observed, and the average cell number was counted.

All experiments were performed at least thrice. We used an unpaired Student's t-test (for homoscedasticity) and Welch's t-test (for heteroscedasticity) to compare the average values. Detailed information, including the P-value for each statistical analysis, is provided in the figure legends. Throughout the study, P<0.05 was considered statistically significant. For multiple comparisons, the P-values were corrected using Shaffer's method. We used Microsoft Excel LTSC for the statistical analysis.

We thank Editage (https://www.editage.com/) for editing and checking this manuscript for English language.

The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.203080.reviewer-comments.pdf