Sult2b1 deficiency decreases the CNV lesion area in vivo

To better understand the role of Sult2b1 in nAMD in vivo, we measured Sult2b1 mRNA levels in RPE–choroid tissue in a laser-induced mouse model of nAMD. During CNV lesion development, Sult2b1 mRNA levels were significantly higher (CNV3d, P < 0.001; CNV5d, P < 0.01; CNV7d, P < 0.05) in the laser-induced CNV group than in the control group (Fig S1), indicating that Sult2b1 may play an important role in CNV development.

Then, we performed laser photocoagulation of the ocular fundus in WT- and Sult2b1-deficient (Sult2b1^-/-) mice to assess whether Sult2b1 deficiency alters CNV, as represented by leakage or neovascularization areas. The fundus fluorescein angiography (FFA) results showed that 7 d after laser photocoagulation, the leakage area of the laser-induced lesions was significantly smaller (P < 0.0001) in Sult2b1^-/- mice than in WT mice (Fig 1A and B). The neovascularization area (labelled with isolectin) was also significantly smaller (P < 0.001) in Sult2b1^-/- mice than in WT mice (Fig 1C and D). Consistently, the mRNA levels of proangiogenic growth markers (Vegfa, Pdgfb, and Tek), an endothelial cell marker (Tek), and profibrotic factors (Tgfb1, Pdgfb, and Fgf2) were lower in the CNV lesions of Sult2b1^-/- mice than in those of WT mice (Fig 1E). Together, these results suggest that Sult2b1 facilitates the progression of laser-induced CNV.

SULT2B1⁺ macrophages are required for CNV progression in vivo

To determine which cell type is responsible for CNV development, we first examined the expression pattern of SULT2B1 in CNV lesions. Interestingly, SULT2B1 was specifically expressed in macrophages (F4/80⁺) in the CNV areas of WT mice but not of Sult2b1^-/- mice (Fig 2A). Moreover, we noted less macrophage infiltration in the CNV lesions of Sult2b1^-/- mice than in those of WT mice according to F4/80 staining of RPE–choroid flat mounts (Fig 2B and C), suggesting that SULT2B1⁺ macrophages are involved in the laser-induced CNV.

To verify the hypothesis, we depleted macrophages in WT and Sult2b1^-/- mice by intraperitoneal injection (IP) of a Csf1r and c-Kit inhibitor (pexidartinib, PLX3397) that selectively induces macrophage death (26). After a continuous 7-d treatment with PLX3397, we observed a significant reduction (P < 0.01) in macrophages in the spleen compared with the untreated group (Fig 2D and E). After PLX3397 treatment, there was no significant difference in CNV area between Sult2b1^-/- and WT mice (Fig 2F–I). Therefore, these results suggest that SULT2B1⁺ macrophages contribute to laser-induced CNV.

Sult2b1 drives M2 polarization to promote pathological angiogenesis

Given that alternatively activated M2 macrophages can promote pathological angiogenesis (16, 17), we investigated whether Sult2b1 expression affects M2 macrophage polarization in vivo and in vitro. Immunofluorescence analysis of CNV lesions showed that the percentage of M2 macrophages (indicated by YM1 or ARG1 positivity) was lower in the CNV lesions of Sult2b1^-/- mice than in those of WT mice (Fig 3A–D). Then, we applied BMDMs in vitro and employed IL-4 to induce M2 polarization (27). Consistently, the SULT2B1 protein level in BMDMs showed a marked increase after IL-4 stimulation (Figs 3E and S2), which indicated that Sult2b1 might play a role in M2 polarization. Moreover, YM1 and ARG1 protein levels were significantly lower in Sult2b1^-/- BMDMs than in WT BMDMs after the induction of M2 polarization (Fig 3F). Taken together, these results indicate that Sult2b1 deletion inhibits differentiation to M2 macrophages.

To test our hypothesis that Sult2b1 is related to M2-mediated vessel formation in CNV lesions, we measured the angiogenic activity of M2 BMDMs harvested from Sult2b1^-/- and WT mice. To this end, we incubated C166 cells and HUVECs with the conditioned medium of WT or Sult2b1^-/- M2 BMDMs and performed tube formation assays to assess vessel formation. The M2 BMDMs from Sult2b1^-/- mice formed significantly fewer vessels in the three-dimensional (3D) matrix than those from WT mice (Fig 3G and J), consistent with the in vivo CNV results. Thus, Sult2b1 deficiency negatively affects the ability of macrophages to regulate pathological angiogenesis by inhibiting M2 polarization.

Cholesterol efflux mediated by LXR–ABCA1/G1 signalling is increased in Sult2b1^-/- macrophages

Having pinpointed that Sult2b1 is involved in M2 polarization, we next sought to explore the mechanism by which Sult2b1 skews macrophage polarization towards the M2 phenotype during CNV development. LXR, a key regulator of cholesterol homoeostasis, can be antagonized by Sult2b1 through the sulfation of oxysterols (23). In addition, the induction of efficient cholesterol efflux from senescent macrophages via treatment with an LXR agonist can regulate M2 polarization and pathological angiogenesis (28, 29). Hence, we speculated that Sult2b1 regulates pathological angiogenesis through LXR signalling. To test this hypothesis, we carried out immunofluorescence analyses of macrophages from Sult2b1^-/- CNV lesions and found that LXRα and LXRβ protein levels were elevated (Fig 4A and B). Furthermore, the proteins encoded by LXR target genes, including the cholesterol reverse transporters ATP-binding cassette subfamily members A1 (ABCA1) and G1 (ABCG1), were at significantly higher levels in macrophages from Sult2b1^-/- CNV lesions than in those from WT lesions, as shown by immunofluorescence (Fig 4C and D). Consistently, LXRα, LXRβ, ABCA1, and ABCG1 protein levels were increased in Sult2b1^-/- M2 BMDMs (Fig 4E). Therefore, signalling through LXR is suppressed by Sult2b1 deficiency in macrophages.

Furthermore, we determined the total intracellular cholesterol content in macrophages by fluorometry and found that cholesterol accumulation was attenuated in BMDMs of Sult2b1^-/- mice (Fig 4F). Moreover, the cholesterol efflux rate of Sult2b1^-/- BMDMs was higher than that of WT BMDMs (Fig 4G). These results demonstrate that the loss of Sult2b1 leads to the activation of LXR–ABCA1/G1 signalling, which subsequently accelerates cholesterol transport out of macrophages.

LXR inhibition eliminates the capacity of Sult2b1^-/- macrophages to restrain pathological angiogenesis

Thus, to evaluate the role of LXRs in pathological angiogenesis in the context of Sult2b1^-/- macrophages, we investigated whether the application of LXR antagonist (inhibiting LXRα and LXRβ, GSK2033) reversed the phenotype of Sult2b1^-/- macrophages in vivo and in vitro. After generating CNV lesions by laser photocoagulation, Sult2b1^-/- and WT mice were treated by intravitreal injection of GSK2033. The results showed that the expression levels of the LXRs and their targets ABCA1 and ABCG1 were successfully reduced by GSK2033 treatment (Fig 5A–D). However, there was no significant difference in their expression between Sult2b1^−/− and WT macrophages treated with the LXR antagonist (Fig 5A–D). Furthermore, GSK2033 significantly increased the leakage area (P < 0.0001) and CNV lesion size (P < 0.0001), regardless of Sult2b1 deficiency (Fig 5E and F), indicating that LXR inhibition could protect against pathological angiogenesis. The leakage area and CNV size in treated Sult2b1^-/- mice were larger than those in untreated Sult2b1^-/- mice. Interestingly, CNV lesion sizes were comparable between the Sult2b1^-/- and WT mice upon treatment with GSK2033 (Fig 5E and F), suggesting that the effects of Sult2b1 deficiency depend mainly on LXRs.

Moreover, GSK2033 treatment resulted in decreased protein levels of LXRα, LXRβ, ABCA1, and ABCG1 in WT BMDMs (Fig 6A and B) and increased the protein levels of YM1 and ARG1 in WT M2 BMDMs (Fig 6C and D), indicating that LXR inhibition could prevent M2 polarization regardless of Sult2b1 expression status. Notably, the expression levels of LXRs, LXR targets, and M2 markers in Sult2b1^−/− macrophages treated with the LXR antagonist did not significantly differ from those in treated WT macrophages (Fig 6A–D). We also examined the effect of the LXR antagonist on the tube-formation ability of WT and Sult2b1^-/- macrophages in vitro. The result showed that there was no significant difference in forming vessels between WT and Sult2b1^-/- macrophages treated by GSK2033 (Figs 6E and F and S3A and B), which were consistent with those observed in vivo (Fig 5E and F). Further analysis of cholesterol content and cholesterol efflux revealed that treatment with the LXR antagonist reversed both the decrease in cholesterol content and the increase in cholesterol efflux in Sult2b1-deficient macrophages (Fig 6G and H). It may suggest that the effect of Sult2b1 on cholesterol efflux capacity relies on LXRs. Overall, our data suggest that the antiangiogenic characteristics of Sult2b1^-/- macrophages are most likely attributable to the activation of LXRs.

Sts deficiency aggravates CNV by promoting M2 macrophage differentiation

In contrast to Sult2b1, Sts catalyses the conversion of sulfated oxysterol to oxysterol (30, 31). Therefore, we sought to determine whether the loss of Sts plays an opposing role in the pathological angiogenesis of CNV. To accomplish this, we performed laser photocoagulation in WT- and Sts-deficient (Sts^-/-) mice (Fig S4) and examined CNV lesions and infiltrating macrophages. STS and SULT2B1 were specifically expressed in macrophages within the CNV lesion (Fig 7A). Strikingly, Sts^-/- mice had larger CNV lesions after laser injury than the WT mice (Fig 7B–E). The mRNA levels of proangiogenic growth markers (Vegfa, Pdgfb, and Tek), an endothelial cell marker (Tek), and profibrotic factors (Tgfb1and Pdgfb) were higher in the CNV lesions of Sts^-/- mice than in those of WT mice (Fig 7F). Moreover, the number of infiltrating macrophages was significantly increased in the CNV lesions of Sts^-/- mice, as determined by analysis of RPE–choroid flat mounts (Fig S5A and B). The percentage of M2 macrophages (indicated by YM1 or ARG1 positivity) in Sts^-/- CNV lesions was higher than that in WT lesions (Fig S5C–F). In addition, the expression levels of LXRs and their targets ABCA1 and ABCG1 in macrophages were lower in CNV lesions from Sts^-/- mice than in those from WT mice (Fig S6A–D). These results demonstrate that Sts protects against CNV development by activating LXR–ABCA1/G1 signalling to block M2 macrophage differentiation, indicating that its role is opposite to that of Sult2b1.

Animals

Animal procedures were approved by the Shanghai Jiao Tong University Institutional Review Board. All animal experiments conformed to the Association for Research in Vision and Ophthalmology (ARVO) Statement for the Use of Animals in Ophthalmic and Vision Research. The C57BL/6J mice were provided by the Laboratory Animal Center of Shanghai General Hospital. The Sult2b1^-/- mice were provided by Shanghai Institute of Immunology, as described in a previous study (40). The Sts^−/− mice were generated by deleting 91 bp in exon 2 of the Sts gene as described in Fig S4. The target ending with NGG near exon 2 was designed and cut under the action of CRISPR/Cas9. The sgRNA1 sequence of sgRNA1 was GAGGATGATGTCAGCGAGCGTGG, and the sgRNA2 sequence of sgRNA2 was AGCCGAGGTCACCGATGCCCAGG. Primer sequences for genotyping are presented in Table S1. Male mice aged between 6 and 8 wk and ∼20 g were included in the experiment.

Laser-induced CNV

Mice were anaesthetized with an intraperitoneal injection of 1% pentobarbital sodium (0.1 ml/10 g body weight) after the application of tropicamide (Santen) for pupil dilatation. CNV lesions were induced as described in a previous study (11). Briefly, four laser spots were administered around the optic nerve head by an argon laser (120 mW in intensity, 50 μm in size, and 100 ms in duration) (OcuLight Infrared Laser System 810 nm; Iridex Corp.) in each eye. CNV lesion area was determined by FFA and immunofluorescence in RPE–choroid flat mounts on day 7 after injury.

Intravitreal injection

Mice were anaesthetized as described above. After laser treatment, the mice were immediately divided into two groups for intravitreal injection. For local anaesthesia, benoxinate HCl 0.4% drops (Santen) were administered. Subsequently, 2 μl of either GSK2033 (LXR antagonist, 500 ng/μl) or vehicle (PBS) was delivered into the vitreous cavity using a microsyringe (Hamilton). The operation was conducted under a dissecting microscope (Olympus) in a sterilized environment. After the operation, 0.5% levofloxacin eye drops (Santen) were applied to prevent infection.

FFA

Mice were anaesthetized, and their pupils were dilated as described above. Then, fluorescein sodium (1%, 0.1 ml/kg; Alcon) was injected intraperitoneally, and late-phase angiograms were performed 5 min after the injection. Fluorescein angiography (FA) was performed by a Micron IV System (Phoenix Research Laboratories). The area of each CNV lesion with leakage was quantified using ImageJ software (National Institutes of Health).

Macrophage depletion

Pexidartinib (16 mg/kg, PLX3397; MedChemExpress) was injected intraperitoneally twice daily for seven consecutive days before laser photocoagulation. No obvious behavioural or health problems were observed in animals injected with PLX3397. Splenocytes were collected and stained with F4/80 (123108; BioLegend) on day 7 after PLX3397 treatment. Because cessation of PLX3397 treatment induces macrophage regeneration (41), we administered PLX3397 for 7 d after laser injury.

Cell culture

HUVECs, from ATCC were grown in humidified air (37°C, 5% CO₂) in an endothelial cell medium (1001; Solarbio) supplemented with 5% FBS, 1% penicillin–streptomycin antibiotics and 1% endothelial cell growth supplement. C166 cells (from ATCC) and BMDMs were cultured in DMEM/high glucose (SH30243.01; Hyclone) with 10% FBS and 1% penicillin–streptomycin antibiotics.

BMDM polarization

Bone marrow cells were harvested from the femur and tibia and then differentiated with 10 ng/ml recombinant mouse M-CSF (PMC2044; Thermo Fisher Scientific) for 7 d (24). For M2 polarization, the macrophages were stimulated with 10 ng/ml recombinant mouse IL-4 (574302; BioLegend) (42) for 24 h. The conditioned medium and cells were collected separately for the following use.

Endothelial cell tube formation assay

Matrigel matrix (354234; Corning) was added to a 96-well plate (50 μl/well) and allowed to polymerize for 30 min at 37°C. Then, 1 × 10⁴ C166 cells or HUVECs and conditioned medium from BMDMs (described above) were mixed and seeded on the Matrigel matrix. After 6 h of incubation at 37°C, tube formation was photographed using a light microscope at 10× magnification. Automated analysis was performed with the ImageJ software (National Institutes of Health, Bethesda, MD, USA). Three parameters (number [Nb] of master junctions, Nb of meshes, and total length) were taken for the quantification experiments.

Cholesterol quantification assay

The free-cholesterol content of M2 BMDMs was measured using the cholesterol/cholesteryl ester quantitation assay kit (ab65359; Abcam). In brief, the dried lipids (extracted from 1 × 10⁶ M2 BMDMs) were dissolved in 200 μl assay buffer (supplied in the kit), and then incubated with a cholesterol probe/cholesterol enzyme mix solution (supplied in the kit) at 37°C for 1 h protected from light. Finally, the fluorescence was detected by a fluorescent microplate reader (Ex/Em = 535/587 nm). The free-cholesterol content was quantified using a cholesterol standard curve.

Cholesterol efflux assay

The cholesterol efflux was measured using the commercial cholesterol efflux assay kit (cell-based) (ab196985; Abcam). In brief, M2 BMDMs (1 × 10⁵) were incubated with a labelling reagent for 1 h and then with the Equilibration Buffer overnight. The cells were treated with the desired cholesterol acceptor (serum) for 6 h in an incubator (37°C, 5% CO₂). The fluorescence intensity of the media and cell lysates was measured via a fluorescent microplate reader (Ex/Em = 482/515 nm). The cholesterol efflux was calculated by dividing the fluorescence intensity of the media by the total fluorescence intensity of the cell lysate and media.

Immunofluorescence

Eyeballs were harvested from mice that were transcardially perfused with 4% paraformaldehyde and then prepared for RPE–choroid flat mounts or cryosections. The antibodies used for staining were anti-isolectin (FL-1201, 1:500; Vector Labs), anti-F4/80 (ab6640, 1:500; Abcam), anti-SULT2B1 (ab254616, 1:500; Abcam), anti-YM1 (1404, 1:500; Stem Cell), anti-ARG1 (16001-1-AP, 1:500; Proteintech), anti-LXRα (ab176323, 1:100; Abcam), anti-LXRβ (ab28479, 1:200; Abcam), anti-ABCA1 (ab18180, 1:200; Abcam), anti-ABCG1 (ab52617, 1:100; Abcam), and anti-STS (17870-1-AP, 1:500; Proteintech). Images were visualized by a fluorescence microscope (Olympus). ImageJ (National Institutes of Health, Bethesda, MD, USA) was used for analysis and quantification.

RNA extraction and real-time PCR

Total RNA was extracted from RPE–choroidal tissues (with CNV lesions) using a total RNA extraction kit (Tiangen). RNA quantity and quality were assessed with a NanoDrop spectrophotometer (Thermo Fisher Scientific). cDNA was synthesized with a cDNA synthesis kit (Takara Bio), and gene expression was analysed by real-time PCR (Applied Biosystems). Primer sequences are presented in Table S2. The mRNA levels of the target genes were quantitatively normalized against those of Gapdh using the ΔΔCT method.

Western blot analysis

Harvested BMDMs described above were homogenized with lysis buffer. The samples were resolved by 10% SDS‒PAGE and transferred to PVDF membranes. After blocking with 5% nonfat dry milk in TBST (TBS containing Tween-20) for 1 h at room temperature, the membranes were incubated at 4°C overnight with primary antibodies against GAPDH (10494-1-AP, 1:1,000; Proteintech), β-actin (3779, 1:1,000; Prosci), YM1 (1404, 1:1,000; Stem Cell), ARG1 (16001-1-AP, 1:5,000; Proteintech), SULT2B1 (ab254616, 1:1,000; Abcam), LXRα (ab176323, 1:1,000; Abcam), LXRβ (ab28479, 1:1,000; Abcam), ABCA1 (ab18180, 1:500; Abcam) or ABCG1 (ab52617, 1:1,000; Abcam). The membranes were washed three times with TBST and then incubated with the corresponding horseradish peroxidase-conjugated secondary antibodies for 1 h at room temperature. Finally, the membranes were washed three times with TBST, and the bands were visualized by chemiluminescence using a molecular imaging system (Amersham Imager 600; GE Healthcare). GAPDH or β-actin was used as loading control. The experiment was replicated three times using independent biological samples.

Statistical analysis

Data are shown as the mean ± SD. Each experiment was repeated three times. Differences among groups were analysed using unpaired two-tailed t test (Prism v7.0; Graph-Pad Software, Inc). P < 0.05 was set as the criterion for significance.