Eight-week-old C57BL/6J wild-type (WT) mice were obtained from GemPharmatech Co., Ltd (Jiangsu, China). FGL2 knockout (FGL2^-/-) C57BL/6J mice were constructed by Shanghai Model Organisms Center, Inc. (Shanghai, China) via CRISPR/Cas9 technology targeting the first and second exons of the mouse FGL2 gene (17). Heterozygous mice (FGL2^+/-) were generated by crossing FGL2^-/- mice and their wild-type littermates. The mice were identified by PCR genotyping of genomic DNA from toe biopsies. Genotyping of mice was performed with two pairs of primers: P1/P2, F 5’-AATGCGCCCGCCCTTTTCTGG-3’ and R 5’-CTGCCGCGGGAGCTGGATGGT-3’, resulting in a 302bp product; P3/P4, F 5’-TATGGTATCTTTTGGGCACTGGTA-3’ and R 5’-TTCTTGGGCCTAATCATCATCTTG-3’, resulting in a 468bp product. The expressions of FGL2 in the brain of WT and FGL2^-/- mice were confirmed by RT-qPCR and western blotting. All mice were housed in the SPF animal experiment center of Tongji Hospital and pregnant mice were housed separately per cage. All the animal procedures were in accordance with Chinese guidelines and regulations for experimental animals, and approved by the Tongji Hospital Animal Care and Use Committee (number: TJH-201901018).

At Embryonic day (E) 17, pregnant dams were weighted and randomly intraperitoneally injected with 200 μg/kg lipopolysaccharide (Escherichia coli, serotype 0111: B4, L2630; Sigma-Aldrich, St. Louis, MO, USA) or 100 μl phosphate buffer saline (PBS) in accordance with the previous studies (19, 22). Dams were sacrificed at E18 or allowed to deliver at term (E19-20), while pups were sacrificed one or seven days after birth. The placenta was washed in PBS solution to remove the maternal blood in tissue, and the brain was rolled carefully on dried filter paper to clean the surface.

The BBB permeability was assessed as described previously using sodium fluorescein (NaF) with slight modifications (23). All the operations below were avoided from light. 5 μl of 0.1 g/ml NaF (Solarbio, Beijing, China) was injected intraperitoneally. One hour after injection, brain tissues and serum were collected. Brain tissues were weighted and homogenized in cold-PBS and centrifugated at 14000 g for 5 min at 4°C. The supernatants were mixed with an equal volume of 15% trichloroacetic acid followed by centrifugation at 1000 g for 10 min. Serum was also mixed with ten times the serum volume of 15% trichloroacetic acid before centrifugation. 25 μl of 5 M NaOH were added in 100 μl of the reaction solutions above before measurement. The fluorescence was measured (excitation 485/20 nm, emission 528/20 nm; Synergy HT, BioTek, Vermont, USA), and the fluorescence intensity was calculated (the fluorescence intensity = brain fluorescence/serum fluorescence).

According to our previous study, brain tissues were fixed in 4% paraformaldehyde, dehydrated by gradient ethanol, embedded in paraffin, and sectioned at 6 μm (19). The slides were then deparaffinized with xylene, rehydrated through graded ethanol, antigen retrieved by boiled citrate buffer under high temperature and high pressure for 2 min, and blocked in 5% goat serum (19). The cell slides were fixed in 4% paraformaldehyde directly and blocked in 5% goat serum. After being blocked, the slides were incubated with primary antibodies overnight at 4°C. After being incubated with the appropriate secondary antibodies (1:200, Life technologies, California, USA) at room temperature for 2 h, the sections were visualized and photographed using a fluorescence microscope (IX71, OLYMPUS, Tokyo, Japan). DAPI (Servicebio, Wuhan, China) was used for nuclear staining. The primary antibodies used were as follows: FGL2 (1:500, Abnova Cat# H00010875-M01), NOX2 (1:100, Proteintech Cat# 19013-1-AP), iNOS (1:100, Proteintech Cat# 18985-1-AP), vWF (1:100, Abcam Cat#ab11713), CD31 (1:2000, Abcam Cat# ab182981; 1:400, Abcam Cat# ab56299). For quantification, the fluorescence intensity or the percentage of at least 3 neonatal brains in each group was analyzed by the ImageJ software, and 3-5 different visual fields per animal were chosen for analysis. The fluorescence intensity of FGL2 or iNOS in CD31+ vessels was analyzed through selecting the CD31+ vessels as our region of interest (ROI) followed by “% Area” function of ImageJ in the ROI of each section. Meanwhile, the number of CD31+ and NOX2+/CD31+ cells in the cortex was counted manually using the “Multi-point” tool of Image J and the percentage of NOX2+ cells in CD31+ cells was calculated according to the reference (24).

Primary brain microvascular endothelial cells (BMECs) were prepared from 10-day-old C57BL/6J mice according to Liu’s protocol with minimal modifications (25). Briefly, the gray matters were isolated from the brain and minced in cold D-Hanks’ solution. The homogenate was supplied with 25% BSA and centrifugated at 600 g for 10 min at 4°C. The deposition in the lowest layer was collected. After three times of centrifugation, all microvessel pellets were further digested with 0.1% collage type II (Sigma-Aldrich, St. Louis, MO, USA) for 20 min at 37°C and finally cultured in DMEM/F12 medium supplemented with 20% FBS, 100 U/ml penicillin-streptomycin at 37°C, 5% CO₂. Culture medium was changed the next day and renewed every 2–3 days. When the cells reached 90% confluency, they were washed with D-hanks’ solution and digested by trypsin-EDTA for 3–4 min at 37°C until the majority of long spindle-shape cells contracted to round. After being suspended with fresh full medium, the cells were seeded into culture flasks at an appropriate density. The characterizations of BMECs were observed by morphology and immunofluorescent staining.

The levels of reactive oxygen species (ROS) in brain or cultured cells were assessed by dihydroethidium (DHE; Beyotime, Shanghai, China), according to manufacturers’ instructions and the previous study (26). All the operations below were avoided from light. Fresh brain tissues were sectioned to frozen sections at 6 μm, incubated with DHE for 40 min at 37°C, and directly observed using a fluorescence microscope (IX71, OLYMPUS, Tokyo, Japan) after nuclear staining, further quantification was used “% Area” function of the ImageJ software and 3-5 different visual fields per animal were chosen for analysis. BMECs were seeded in 12-well plates or 96-well black plates. After treatment, living cells were incubated directly with DHE for 30 min. The fluorescence intensity was observed in 12-well plates using the fluorescence microscope (IX71, OLYMPUS, Tokyo, Japan) or measured in 96-well plates by a fluorescence spectrophotometer (excitation 360/40 nm, emission 460/40 nm; Synergy HT, BioTek, Vermont, USA).

The concentrations of NO in brain or culture supernatant were assessed by Griess reagent in the Total Nitric Oxide Assay Kit (Beyotime, Shanghai, China) according to manufacturers’ instructions and the previous study (27). The absorbance was measured at 540 nm using a microplate spectrophotometer (Synergy HT, BioTek, Vermont, USA), and NO levels were calculated using a standard curve prepared by standard nitrite solutions.

To suppress the FGL2 expression, the plasmid containing double-stranded miRNA templates for FGL2 (NM_008013) was constructed into adenovirus vector (pADM-CMV-GFP) by Vigene Biosciences Co., Ltd. (Shandong, China). The oligonucleotides encoding FGL2-miRNA sequences were the same as the previous study (28). BMECs were infected at a multiplicity of infection (MOI) of 250 and supplied with 5 μg/ml polybrene. To overexpress the PI3K expression, the flag-tagged PI3K overexpressed plasmid (NM_001077495) was constructed into lentivirus vector (pHBLV-CMV-MCS-ZsGreen-T2A-PURO) by Hanbio Biotechnology Co., Ltd. (Shanghai, China). BMECs were infected at a MOI of 40 and supplied with 5 μg/ml polybrene. 48 h after infection, the fluorescence intensity of cells was observed under a fluorescence microscope (IX71, OLYMPUS, Tokyo, Japan). The expressions of FGL2 and flag were detected by western blotting.

Proteins from tissues and cells were extracted using RIPA lysis buffer (BOSTER, Wuhan, China) and quantified by BCA Protein Assay kit (BOSTER, Wuhan, China). The samples were then separated by 10% SDS/PAGE gel and transferred onto nitrocellulose membranes. The membranes were subsequently blocked with 5% skim milk for 1h at room temperature and incubated with the primary antibodies ( Table S1 ) overnight at 4°C. The next day, the membranes were incubated with HRP-conjugated secondary antibodies (1:5000, Servicebio, Wuhan, China) for 1 h at 37°C. The images were further detected with an enhanced chemiluminescence system (ChemiDoc XRS+, Bio-Rad, California, USA) and analyzed using ImageLab software. GAPDH or ACTIN was used to normalize the expressions of protein.

Total RNA from tissues and cells was extracted with RNAiso Plus (Takara, Kyoto, Japan). The concentration and quality of RNA were detected with a NanoDrop (Thermo Fisher Scientific, Waltham, USA). cDNA was prepared using PrimeScript RT Master Mix Kit (Takara, Kyoto, Japan). Quantitative real-time PCR was performed on CFX Connect Real-Time System (Bio-Rad, California, USA) using TB Green Premix EX Taq (Takara, Kyoto, Japan). The data was analyzed by 2^-ΔΔCt and normalized to GAPDH. The primer sequences were listed in Table S2 .

All statistical analyses were performed using GraphPad Prism^® 8.0 (GraphPad Software, Inc., California, USA) and date were expressed as the means ± standard error of the mean (SEM). Unpaired two-tailed Student’s t test was used for comparisons between two groups and one-way analysis of variance (ANOVA) with Tukey’s post-hoc test was for multiple groups. P value of <0.05 was considered statistically significant. The number of animals or samples in each group was indicated in the figure legends and the data was the representative of at least three independent experiments.

Wild-type dams were intraperitoneally injected with 200 µg/kg LPS or PBS at E17. The schematic of experimental design was shown in Figure 1A . 24h after injection, the expressions of pro-inflammatory cytokines in the placenta were all significantly increased, including TNF-α, IL-1β, and IL-6, while the anti-inflammatory IL-10 expression was decreased ( Figure S1A ). It suggested the reliable construction of maternal inflammation mouse model. At the dose used, LPS decreased the maternal weight 24h post injection but had no significant effect on time of birth and litter size compared with controls ( Figures S1B–D ). Occludin, Claudin-5, and ZO-1 were the representative tight junction (TJ) proteins, performing a crucial role in maintaining BBB integrity (29). Therefore, the expressions of TJ proteins in the brain of postnatal pups were further assessed. Compared with controls, LPS-exposed pups showed significantly decreased cerebral expressions of Claudin-5, Occludin, and ZO-1 at postnatal day 1 (P1) ( Figure 1B ). Additionally, the expressions of Claudin-5 and Occludin were still lower than controls at postnatal day 7 (P7) ( Figure 1B ). These results indicated the tight junctions were reduced in the brain of neonatal mice with maternal inflammation, implying the impaired BBB integrity.

To explore the role of FGL2 in BBB integrity under the condition of maternal inflammation, we firstly analyzed the expression of FGL2 in LPS-exposed pups. Compared with controls, cerebral mRNA and protein expressions of FGL2 were remarkably increased in LPS-exposed pups at both P1 and P7 ( Figures 1C, D ). Moreover, FGL2 expression in cerebral vessels was precisely investigated by immunofluorescence staining. It showed that the fluorescence intensity of FGL2 in CD31-positive vessels was also significantly elevated in LPS-exposed pups ( Figure 1E ). We then took advantage of LPS-treated primary BMECs to further confirm the link between endothelial FGL2 and TJ proteins. As previously (25), primary BMECs presented a swirl monolayer and expressed von Willebrand factor (vWF), an endothelial specific marker ( Figure S2 ). Moreover, the results showed that the mRNA and protein expressions of FGL2 in LPS-treated BMECs were markedly increased ( Figures 1F, G ), while the mRNA levels of TJ proteins, including Claudin-5, Occludin, and ZO-1, were significantly decreased ( Figure 1H ). These data strongly suggested that the increased endothelial expression of FGL2 in vessels was associated with the BBB integrity in maternal inflammation-exposed neonatal mice.

To further investigate the role of FGL2 in BBB function, we established FGL2^-/- mice and confirmed the cerebral knockout of FGL2 by RT-qPCR and western blotting ( Figure S3 ). In order to exclude the influence of maternal FGL2 levels on the offspring, we treated FGL2^+/- dams with LPS and explored the effect of FGL2 on FGL2^+/+ pups and their FGL2^-/- littermates ( Figure 2A ). Compared with controls, the reductions in both body weight and brain weight were found in LPS-exposed FGL2^+/+ pups at P1 ( Figures S4A–D ), while FGL2^-/- littermates were resistant to the LPS-induced loss of brain weight at P1 ( Figure S4B ). However, there was no significant difference in brain-to-body weight ratio among the four groups at both P1 and P7 ( Figures S4E, F ). Moreover, we found the induction of pro-inflammatory cytokines in the placenta from FGL2^+/- dams after LPS treatment ( Figure S5A ), as well as the increased cerebral FGL2 expression in their FGL2^+/+ pups ( Figures S5B, C ).

Next, we compared the BBB permeability in FGL2^+/+ and FGL2^-/- neonatal mice. After LPS exposure, the relative fluorescence intensities of NaF in the brain of FGL2^+/+ pups were obviously higher than PBS-treated controls at both P1 and P7 ( Figures 2B, C ), suggesting the hyperpermeability of BBB in conditions of maternal inflammation. In contrast with FGL2^+/+ pups, the hyperpermeability of BBB was notably attenuated in FGL2^-/- littermates at both P1 and P7, indicated by the reduced NaF fluorescence intensity in the brain ( Figures 2B, C ). These results pointed out that FGL2 depletion restored the BBB permeability during maternal inflammation. Tight junctions are the important components determining the BBB permeability (30). Consistent with the increased permeability, the expressions of TJ proteins were significantly reduced in LPS-treated FGL2^+/+ pups ( Figures 2D, E ). However, in FGL2^-/- pups the LPS-induced reduction of TJ proteins was significantly mitigated ( Figures 2D, E ). Moreover, when FGL2 expression was suppressed by adenovirus in vitro ( Figure 2F ), we also found that the reduction of mRNA expressions of Claudin-5, Occludin, and ZO-1 were all alleviated in LPS-treated BMECs ( Figure 2G ). These data highly suggested FGL2 deficiency restored the function of BBB by upregulating the expressions of TJ proteins and recovering the permeability of BBB in maternal inflammation-exposed pups.

Furthermore, we evaluated the mRNA levels of cytokines in brains of FGL2^+/+ and FGL2^-/- offspring. Consistent with our previous study in BALB/c mice (19), we observed elevated cytokines, including TNF-α, IL-6, and IL-10, in LPS-exposed FGL2^+/+ pups at P1 ( Figure 2H ). However, the eruption of pro-inflammatory cytokines was strongly inhibited in FGL2^-/- littermates at P1 ( Figure 2H ), and the protection against inflammation was still significant on TNF-α at P7 ( Figure 2I ). Therefore, FGL2 deficiency showed the remarkable protection on neonatal BBB disruption, which might further decrease the levels of pro-inflammatory cytokines in the brain.

Oxidative stress, due to uncontrolled accumulation of intracellular reactive oxygen species (ROS) and nitrogen species, is regarded as a crucial mechanism of BBB damage (9, 31). Previous study has observed that the imbalance of oxidation and antioxidant was concomitant with the BBB disruption in LPS-exposed fetal brain (15). To clarify the mechanisms of FGL2-mediated BBB damage, we further investigated oxidative stress in the brain of FGL2^+/+ and FGL2^-/- pups. We found LPS-exposed FGL2^+/+ brain exhibited a general increase in ROS levels ex vivo, indicated by the increased fluorescence intensity of DHE staining at P7 ( Figure 3A ). The contents of cerebral nitric oxide (NO) were also increased at P1 and P7 ( Figures 3B, C ). In contrast, FGL2 deficiency strikingly suppressed the inductions of ROS and NO ( Figures 3A–C ), suggesting the beneficial effects of FGL2 depletion on maternal inflammation-induced cerebral oxidative stress. Primary BMECs isolated from FGL2^+/+ and FGL2^-/- mice were subsequently used to evaluate the endothelial production of ROS and NO. Similarly, the production of LPS-induced ROS and NO were markedly increased in FGL2^+/+ BMECs but significantly inhibited in FGL2^-/- cells ( Figures 3D, E ).

In the context of BBB impairment, NADPH oxidase 2 (NOX2) and inducible nitric oxide synthase (iNOS) are responsible for the major production of ROS and NO, respectively (32–34). We further explored the regulation of FGL2 on NOX2 and iNOS expressions. The cerebral expressions of iNOS in pups were induced by LPS exposure and markedly inhibited upon FGL2 deficiency at both P1 and P7 ( Figure 4A ). Meanwhile, we observed the induction of NOX2 expression in LPS-exposed pups at P1, which was also inhibited by FGL2 deletion ( Figure 4B ). To precisely investigate the expressions of iNOS and NOX2 in vivo, we used immunofluorescence staining to analyze their vascular expressions in cortex at P7. After LPS exposure, the percentage of NOX2-positive cells in CD31-positive vessels was increased in FGL2^+/+ pups and was obviously decreased in FGL2^-/- littermates ( Figure 4C ). Simultaneously, the fluorescence intensity of iNOS in CD31-positive vessels was also elevated by LPS and suppressed by FGL2 deletion ( Figure 4D ). Likewise, the endothelial expressions of NOX2 and iNOS were both increased in LPS-treated FGL2^+/+ BMECs ( Figure 4E ). FGL2 deletion evidently downregulated the expressions of both proteins in vitro ( Figure 4E ). Together, these data pointed out that FGL2 deficiency dramatically decreased the vascular endothelial oxidative stress in the brain of neonatal mice upon maternal inflammation exposure.

Previous studies have revealed that phosphatidylinositol 3-kinase (PI3K)-dependent pathway mediates the FGL2-induced chemokine production (35). More importantly, the regulatory subunit of PI3K, PI3K-p85, has been reported to regulate endothelial oxidative stress (36). To explore the molecular mechanisms underlying FGL2 facilitated endothelial oxidative stress, we detected the expression of PI3K-p85 in pups upon maternal inflammation exposure. Compared with controls, the cerebral expression of PI3K-p85 was elevated in LPS-exposed FGL2^+/+ pups at both P1 and P7 ( Figure 5A ). In contrast, the increase of PI3K-p85 was significantly suppressed in FGL2^-/- littermates ( Figure 5A ). Nuclear factor kappa B (NF-κB) is an important downstream of PI3K and has been widely reported to be involved in oxidative stress (32). We found the activation of NF-κB p65 was enhanced in LPS-exposed FGL2^+/+ pups at P1 and was strikingly inhibited when FGL2 was deleted ( Figure 5B ). In vitro, the activation of PI3K/NF-κB pathway was also increased in LPS-treated BMECs and further inhibited by FGL2 deletion ( Figures 5C–E ). These data suggested the PI3K/NF-κB pathway could be the molecular mechanism underlying FGL2-regulated oxidative stress.

To confirm the role of PI3K/NF-κB pathway in FGL2-regulated oxidative stress in BMECs, we firstly pretreated FGL2^+/+ BMECs with a specific PI3K pathway inhibitor, LY294002, before LPS exposure. Similar to the findings in FGL2-deficient cells, the LPS-induced expressions of iNOS and NOX2, as well as intracellular ROS and NO levels were all downregulated by PI3K inhibitor ( Figures 5F–H ). We further overexpressed PI3K in FGL2^+/+ and FGL2^-/- BMECs by lentivirus. PI3K overexpression had an increasing trend on the expressions of NOX2 and iNOS in PBS-treated FGL2^+/+ BMECs, and these effects were enhanced by LPS ( Figure S6 ). More importantly, the inhibitory effects of FGL2 depletion on NOX2 and iNOS expressions were both blocked by PI3K overexpression in LPS-treated FGL2^-/- BMECs ( Figure 5I ). These findings strengthened the regulatory role of PI3K in FGL2-mediated oxidative stress. Additionally, the LPS-induced activation of NF-κB was inhibited in PI3K-inhibited BMECs ( Figure 5J ), but re-activated when PI3K was overexpressed in FGL2^-/- BMECs ( Figure 5I ). These results indicated NF-κB was the downstream of PI3K in our experimental systems. Collectively, all of these results strongly supported that PI3K/NF-κB were the key factors in FGL2-mediated endothelial oxidative stress.