Hfe2 liver knock out alters BBB integrity
Using widefield imaging, we observed that Hfe2 knock-out mice demonstrated an intensive leakage pattern, indicative of BBB breakdown (Extended Data Fig. 1a-c). This was unexpected as Hfe2 is not expressed in the brain 5 (Extended Data Fig. 1d) and suggested that a non-brain source of Hfe2 regulates BBB integrity. Hfe2 is mainly expressed by the liver 14,15; hence we hypothesized that Hfe2 released by the liver into the systemic blood stream regulates BBB integrity. To investigate the role of liver-secreted Hfe2 on BBB integrity, we generated Hfe2ΔAlb−cre (Hfe2fl/fl; Alb-cre) transgenic mice to genetically ablate Hfe2 production, specifically in the liver. ELISA analysis of the serum demonstrated a 56±8 % reduction of Hfe2 levels in the serum, confirming that the liver is a major source of serum Hfe2 (Extended Data Fig. 1e).
To characterize BBB integrity in detail, we developed a method allowing for 3D representations of BBB-leakages at any given cortical depth (Fig. 1a). Leakages were defined by the extravasation of Lysine-fixable Texas Red-conjugated 70 kDa dextran (TR-dextran). Animals were perfused to remove residual TR-dextran in the vessels. By using CUBIC lipid-clearing, the optically transparent brains were imaged using light-sheet microscopy to map out BBB-leakage throughout the whole brain. In order to ensure that our perfusion protocol removed all traces of TR-dextran in the intact brain, while also allowing for visualization of BBB-leakages, we performed middle cerebral artery occlusion (MCAO), which is known to open the BBB of one half of the brain while leaving the other half unperturbed. As expected, leakage-staining was seen in the MCAO damaged hemisphere, while the healthy hemisphere remained devoid of any leakage staining (Extended Data Fig. 2). Having validated our perfusion protocol, we studied the role of liver Hfe2 on BBB integrity. In Hfe2ΔAlb−cre animals, we observed intense BBB-leakage in superficial as well as deep regions of the brain, whereas Hfe2fl/fl control littermates did not display any such leakage (Fig. 1b).
To ensure that the BBB-leakage observed in the Hfe2ΔAlb−cre animals did not result from a developmental defect, we ablated Hfe2 secretion specifically in adult mice. We infected a cohort of adult Hfe2fl/fl mice with a liver-directed adenovirus, AAV8-AlbCre 16, or a control adenovirus, AAV8-GFP. Three weeks post-infection, AAV8-AlbCre infected mice displayed significant reduction in Hfe2 serum-levels, which was comparable to that of Hfe2ΔAlb−cre mice (Extended Data Fig. 1e). Hfe2fl/fl mice infected with AAV8-AlbCre invariably displayed severe BBB-leakages, which were not observed in the AAV8-GFP infected controls (Fig. 1b). Both Hfe2 liver-specific knockout models exhibited significant BBB-leakage, as indicated by increased total fluorescence intensity of residual TR-dextran, indicating that Hfe2 plays a significant role on maintaining BBB integrity (Fig. 1b). To confirm this phenotype and validate our BBB-assessment method, we performed in-vivo multiphoton imaging of TR-dextran in Hfe2ΔAlb−cre and AAV8-AlbCre mice along with their respective controls. We confirmed that the observed leakage was occurring within the brain, by imaging the height of the dura relative to the observed blood vessels (Supplementary Video 1). Following this validation, we did time-lapse imaging over a 40-minute window to record the temporal dynamics of the leakage. As expected, both Hfe2 liver knock-out mouse models displayed obvious extravascular leakage (Fig. 1c and Supplementary Video 2). Time-series analysis revealed a significant increase in the total TR-dextran extravasation in the knockout animals compared to their respective controls (Fig. 1c) and this extravasation was observed only in larger blood vessels (>24 µm, Extended Data Fig. 1f). The leakage pattern seen in the in-vivo imaging supported what was observed in our light-sheet BBB assessments (Fig. 1b). We subsequently investigated whether BBB-breakdown in Hfe2-deficient mice led to accumulation of blood-born substances in the CNS. We performed staining for blood-born Fibrinogen, a well-established marker of BBB-breakdown 17. Immunostaining for fibrinogen and the endothelial marker isolectin revealed abundant fibrinogen deposition around the cerebral vasculature in Hfe2-deficient mice, which was significantly elevated compared to controls (Fig. 1d). Finally, we performed intravenous injection of TR-dextran and studied TR-dextran presence in sections (Fig. 1e and Extended Data Fig. 1g) 18. We observed TR-dextran deposition similar to that observed by others demonstrating BBB disruption in mice 18. In conclusion, using 4 different experimental techniques we demonstrate that levels of Hfe2 are a critical regulator of BBB-integrity.
Hfe2 muscle knock out alters BBB integrity
Hfe2 plays a major role in iron homeostasis, and Hfe2ΔAlb−cre mice display a 52±13 % increase in iron levels in the blood (Extended Data Fig. 3a)15. To test if altered iron levels were the cause of the BBB-leakage, we generated Hfe2ΔActa−cre (Hfe2fl/fl; Acta-cre) transgenic mice to genetically ablate Hfe2 production specifically in skeletal muscles. Analysis demonstrated that this procedure reduced Hfe2 levels by 32±10 % without altering iron levels in the serum (Extended Data Fig. 3a-b)15. In Hfe2ΔActacre animals, we nonetheless observed BBB-leakages using light sheet imaging and in-vivo multiphoton imaging, suggesting BBB-alteration is not the result of increased of iron levels (Fig. 1f and Extended Data Fig. 1h). Leakages appeared less pronounced than in Hfe2Δalb−cre animals, which may reflect the lower decrease in blood Hfe2 in Hfe2ΔActa−cre when compared to Hfe2Δalb−cre. Although we cannot totally exclude a role for iron in the leakage observed in Hfe2Δalb−cre animals, these results indicate that liver- and muscle-secreted Hfe2 may play a pivotal role in the maintenance of BBB-integrity.
Hfe2 muscle-and liver-knock out results in neuronal loss and behavioral deficits
BBB disruption in Hfe2ΔAlb−cre animals triggers the extravasation of fibrinogen, which can be toxic to brain neurons 19, hence, we surveyed the number of cortical neurons in Hfe2ΔAlb−cre, Hfe2ΔActa−cre, and AAV8-AlbCre animals. Interestingly, in both the liver- and muscle-specific Hfe2 knock outs, we observed a strong reduction of the number of cortical neurons (Fig. 2a). The observed reduction in the number of neurons in Hfe2ΔActa−cre animals fits data showing that a serum pool of Hfe2 serves to regulate BBB integrity. Following infection with AAV8-AlbCre, to ablate liver production of Hfe2, we also observed a progressive loss of cortical neurons over 3 to 4 weeks (Fig. 2a). Next, we performed a series of behavioral tests to determine whether neuronal loss correlated with functional deficits. Hfe2 knock out in the liver resulted in a strong reduction of functional scores in marble burying, open field, and water maze tests, indicating impaired brain functions (Fig. 2b-d). Similarly, Hfe2ΔActa−cre animals also displayed functional deficits, however in this case we cannot exclude the possibility that this is a result of muscle weakness.
RGMa and Hfe2 have opposite effects on Claudin-5 and PDGF-B expression
Pericytes are key to maintaining BBB-integrity 20,21, hence we studied pericyte coverage of endothelial cells in Hfe2 liver KO (Fig. 3a). Interestingly, both Hfe2Δalb−cre and AAV8-AlbCre animals displayed a significant reduction of pericyte coverage when compared to control animals, suggesting that this alteration of pericyte coverage may trigger BBB-breakdown (Fig. 3a). PDGF-B expression by endothelial cells is critical for maintaining pericyte coverage. Consequently, we used quantitative PCR on purified endothelial cells to determine whether knocking out Hfe2 affects mRNA levels for PDGF-B (Fig. 3b). PDGF-B mRNA levels were reduced by ~ 4.5±1.6× fold in Hfe2ΔAlb−cre when compared to Hfe2fl/fl mice (Fig. 3b), suggesting that Hfe2 contributes to maintenance of BBB integrity by promoting PDGF-B expression. To investigate the mechanisms whereby Hfe2 regulates PDGF-B expression, we studied PDGF-B mRNA levels in cultured endothelial cells. Surprisingly, addition of Hfe2 to endothelial cultures did not influence PDGF-B mRNA levels (Fig. 3c). We hypothesized that Hfe2 may counteract the effect of another factor that negatively regulates PDGF-B expression in endothelial cells (bEnd3 cells). In axonal growth experiments, we have observed that Hfe2 blocks the inhibitory activity of RGMa. The addition of RGMa to the laminin-substrate shortens axons when compared to laminin on its own, while the addition of Hfe2 to the medium restored axonal length to control levels, suggesting that Hfe2 can fully neutralize RGMa (Extended Data Fig. 4). We detected RGMa in human and murine blood serum by Western Blotting (Extended Data Fig. 5) and ELISA (~0.9 µg/mL), respectively, indicating that endothelial cells are in direct contact with both blood-borne RGMa and Hfe2. We addressed the possibility that RGMa and Hfe2 – similar to what was observed in growing axons – have opposite effect on endothelial cells. RGMa was added to the medium of cultured bEnd3 cells. This induced a 4.1±1.3× fold reduction in PDGF-B mRNA levels, which prompted us to test whether Hfe2 positively impacts PDGF-B expression by counteracting this RGMa effect (Fig. 3c). As predicted, when Hfe2 was added to RGMa treated cells, we observed an abrogation of the RGMa effect on PDGF-B mRNA levels (Fig. 3c). PDGF-B is present in the matrix of endothelial cells, hence, we evaluated the expression of this protein using immunohistochemistry. In agreement with our mRNA data, RGMa addition to the medium reduced PDGF-B, which was rescued by the addition of Hfe2 to the medium (Extended Data Fig. 6).
BBB integrity is also regulated by tight junctional molecules, hence, we asked whether the expression of tight junction proteins is altered upon ligand treatment in the cultured bEnd3 cells. Immunocytochemistry revealed obvious claudin-5 discontinuity upon RGMa treatment, and such disruption was prevented by Hfe2 co-treatment (Fig. 3d). Western blot further confirmed a significant reduction in claudin-5 expression in RGMa-treated bEnd3 cells compared to Hfe2 co-treatment and controls (Fig. 3e). RGMa did not appear to impact the expression of Occludin, another tight junction protein (not shown). Hence our results indicated that RGMa treatment on cerebral endothelial cells significantly alters both the expression of Claudin-5 and PDGF-B, which can be completely prevented by Hfe2.
RGMa and Hfe2 have opposite effects on endothelial cells and BBB integrity
To further assess the effect of RGMa and Hfe2 on BBB integrity, we performed a Transwell permeability assay to study the extravasation of horse radish peroxidase (HRP) through a monolayer of bEnd3 cells upon ligand treatment. We found that soluble RGMa significantly increases the monolayer permeability to HRP to an extent comparable with TNF-α, which was used as a positive control. The effect of RGMa on the bEnd3 monolayer was significantly attenuated by the addition of Hfe2 (Fig. 3f). Additional in-vitro assessment using trans-endothelial electrical resistance (TEER) confirmed that RGMa significantly reduced bEnd3 barrier function which was suppressed upon Hfe2 co-treatment (Fig. 3g).
The above presented data suggest that RGMa alters endothelial barrier function in vitro. To address this in vivo, we intravenously administered either soluble RGMa alone or with Hfe2 into wild type (WT) mice and assessed BBB-integrity 24h post-injection. Using the light-sheet BBB assessment tool, we observed intense accumulation of TR-dextran in the brains of RGMa-treated WT mice, indicating that RGMa-injection severely disrupts BBB function (Fig. 4a). Co-treatment of RGMa and Hfe2 significantly prevented RGMa-mediated BBB-breakdown (Fig. 4a and Extended Data Fig. 7). Multiphoton imaging further confirmed the disruptive role of RGMa, which displayed a significant increase in TR-dextran extravasation. BBB-integrity in WT animals co-treated with both RGMa and Hfe2 were comparable to the PBS treatment (Fig. 4b and Supplementary Video 3). RGMa-treated WT animals exhibited a significant amount of perivascular fibrinogen accumulation, which was completely abrogated by Hfe2 co-treatment (Fig. 4c). Furthermore, we looked at BBB-leakage using widefield imaging, confirming that RGMa induces leakage that is not observed when Hfe2 is co-injected (Extended Data Fig. 7). Together, these data identify RGMa and Hfe2 as novel regulators of PDGF-B and BBB integrity.
Neogenin is involved in RGMa mediated opening of the BBB
RGMa interacts with Neogenin to inhibit axonal growth 22. Although Hfe2 also interacts with Neogenin, Hfe2 does not inhibit outgrowth, but rather suppresses the RGMa inhibition of axonal growth (Extended Data Fig. 4). In the chick brain, Hfe2 did not lead to a displacement of the Neogenin receptor and it still localized to lipid rafts (data not shown), hence, these outgrowth data suggest that Hfe2 prevents the interaction between RGMa and Neogenin. To address this possible role for Hfe2, we developed an assay in which soluble Neogenin (AP-tagged) will interact with RGMa. We observed that soluble Neogenin-AP interacts with ELISA-plate coated RGMa (Fig. 5a). In a competitive binding assay, we show that Hfe2 significantly prevented the binding of Neogenin-AP to RGMa (Fig. 5a). Also, we show that Neogenin-AP binds to Hfe2 and that this interaction is blocked by RGMa (Fig. 5a). Next, we studied Neogenin expression in blood vessels, and show that it is strongly expressed in endothelial cells (Fig. 5b). Neogenin staining was observed in the lumen of human and murine blood vessels suggesting that it interacts with blood proteins (Fig. 5b). The presence of Neogenin in endothelial cells raised the possibility that Hfe2 restores BBB integrity by preventing RGMa binding to Neogenin. A hypomorphic allele with ~90% loss of Neogenin has been described previously, and shows that mice die around P28 23. Hence, to study the involvement of Neogenin in RGMa-induced BBB breakdown, we used NeoΔTie2−creERT2 (Neofl/fl; Tie2-creERT2) knockout animals in vivo, wherein Neogenin is genetically deleted from endothelial cells upon Tamoxifen induction. Immunoblotting on isolated cerebral endothelial cells of NeoΔTie2−creERT2 revealed complete deletion of Neogenin (Extended Data Fig. 8). Soluble RGMa treatment applied to tamoxifen-treated NeoΔTie2−creERT2 mice did not induce BBB-leakage as observed with WT mice, indicating that RGMa-mediated BBB breakdown is primarily mediated by Neogenin (Fig. 5c and Extended Data Fig. 8). Multiphoton imaging as well as fibrinogen staining in RGMa-treated NeoΔTie2−creERT2 mice also indicated that Neogenin ablation in cerebral endothelial cells significantly prevents RGMa-mediated BBB alterations (Fig. 5d-f and Supplementary Video 4). These results demonstrate the role of the Neogenin receptor in mediating the effect of RGMa on BBB integrity.
The RGM/Neogenin pathway modulates BBB integrity in a model of MS
RGMa is upregulated in MS, a disease manifesting severe BBB-disruption 12,13. Furthermore, inflammation has been shown to reduce Hfe2/RGMc expression 24. Therefore, we explored the possibility that the BBB-breakdown caused by RGMa/Hfe2 imbalance may be a critical component in MS pathology. To test this hypothesis, we used EAE, an animal model for MS 12. Animals received a MOG peptide that induces paralysis within 2-3 weeks. ELISA and Western Blotting showed that Hfe2-levels are strongly downregulated in the serum of EAE mice when compared to sham induction mice (Fig. 6a). Interestingly RGMa levels were increased in EAE animals (Fig. 6a) and we observed high levels of RGMa around blood vessels within the plaques found in the brain of MS patients (Extended Data Fig. 9). We reasoned that low Hfe2 combined with high RGMa levels open the BBB in EAE and that administrating Hfe2 should restore the RGMa/Hfe2 balance and prevent RGMa-mediated BBB opening. Treatment with soluble Hfe2 to EAE mice significantly delayed disease onset and significantly reduced disease severity (Fig. 6b). This effect was suppressed by co-injecting RGMa with Hfe2, which fits with the opposite activities of RGMa and Hfe2 shown earlier (Fig. 3&4). Because Hfe2-liver knock out reduces Hfe2 levels in the serum and alters BBB-integrity, Hfe2ΔAlb−cre mice should present an aggravation of EAE symptoms. Indeed, Hfe2ΔAlb−cre mice displayed earlier disease onset and significantly increased disease severity at early stages of the disease when compared to controls (Fig. 6c). These data suggest that circulating Hfe2 prevents functional impairment by blocking the detrimental effects of RGMa following EAE induction. Because Neogenin mediates the RGMa effect on BBB (Fig. 5), we tested whether Neogenin knock-out in endothelial cells prevents EAE symptoms. As expected, NeoΔTie2−creERT2 mice exhibited significant improvement following EAE induction (Extended Data Fig. 10), which fits with a model in which Neogenin mediates the RGMa/Hfe2 effect.
Examination of cell infiltrates within the spinal cord of the Hfe2-treated EAE mice showed significant reduction of cellular infiltrates as assessed using the H&E staining inflammatory index, where 0 is no infiltrates and 4 is cellular infiltration within the grey and white matter (Extended Data Fig. 10). Immune cells play a critical role in CNS damage in EAE25, hence we examined immune cells infiltrations in spinal cord. Hfe2-treated mice showed significant reduction in the pan T cell marker CD3+, the pan B cell marker B220+, as well as the myeloid marker CD11B+ when compared to vehicle-treated controls (Fig. 6d). To ensure that reduced immune cell infiltration did not result from an effect of Hfe2 on immune cell activation, we performed adoptive transfer of activated immune cells into the recipient mice (Fig. 6e). Hfe2 treatment once again prevented the infiltration of immune cells into the CNS indicating that the positive effect of Hfe2 on EAE progression was likely independent of any effects on immune cell priming (Fig. 6e). Treatment with Hfe2 had no effect on i) the adhesion properties of naïve T and B cells, ii) naïve antigen presenting cells, iii) naïve immune cell populations, iv) on activated immune cells, and v) antigen specific immune cells which also indicates that the Hfe2 effect is independent of immune cell priming (Extended Data Fig. 11a-e). This suggests that Hfe2 prevents immune cell infiltration by preventing BBB-alteration following EAE induction, which was further confirmed by a significant reduction in endogenous fibrinogen extravasation into the CNS of EAE mice treated with Hfe2 (Extended Data Fig. 11f). Taken together, our data validate a role for Hfe2, RGMa, and Neogenin in BBB-maintenance and highlight their importance in MS pathology.