24As an optically transparent model organism with an endothelial blood-brain barrier (BBB), 25 zebrafish offer a powerful tool to study the vertebrate BBB. However, the precise developmental 26 profile of functional zebrafish BBB acquisition and the subcellular and molecular mechanisms 27 governing the zebrafish BBB remain poorly characterized. Here we find a spatiotemporal gradient 28 of barrier acquisition. Moreover, we capture the dynamics of developmental BBB leakage using 29 live imaging, revealing a combination of steady accumulation in the parenchyma and sporadic 30 bursts of tracer leakage. Electron microscopy studies further reveal that this steady accumulation 31 results from high levels of transcytosis that are eventually suppressed, sealing the BBB. Finally,
32we demonstrate a key mammalian BBB regulator Mfsd2a, which inhibits transcytosis, plays a 33 conserved role in zebrafish. Mfsd2aa mutants display increased larval and adult BBB permeability 34 due to increased transcytosis. Our findings indicate a conserved developmental program of 35 barrier acquisition between zebrafish and mice.36 109 leakage of low molecular weight tracers 1 kDa and below into the brain parenchyma (Nitta et al.,110 2003; Campbell et al. 2008; Sohet et al., 2015; Yanagida et al. 2017). At 3 dpf, we observed a 111 sealed barrier in the hindbrain as previously described (Jeong et al., 2008), with only a few of the 112 parenchymal cells taking up the injected tracer (average of 2 ± 0.3 cells/embryo with NHS and 2 113 ± 0.4 cells/embryo with Dextran; Figure 1 -Supplement 1), which we quantify as a proxy of tracer 114 6 leakage into the brain. However, in the midbrain we observed an increased number of 115 parenchymal cells that accumulated the circulating tracers (average of 24 ± 1 cells/embryo with 116 NHS and 24 ± 1 cells/embryo with Dextran; Figure 1C and 1D), indicating that the tracers leaked 117 out of the blood vessels into the brain and that the midbrain barrier is not yet functional. In addition 118 to the use of exogenous injected fluorescent tracers, we also assayed BBB permeability with an 119 endogenous transgenic serum DBP-EGFP fusion protein (Tg(l-fabp:DBP-EGFP)) to account for 120 injection artifacts (Xie et al., 2010). At 3 dpf, we observed similar leakage patterns with the 121 transgenic serum protein as we did with the injected tracers (average of 24 ± 1 cells/embryo in 122 the midbrain and average of 2 ± 0.4 cells/embryo in the hindbrain; Figure 1C and D; Figure 1 -123 Supplement 1). At 4 dpf, the BBB in the hindbrain is completely functional with few tracer-filled 124 parenchymal cells (average of 3 ± 0.4 cells/embryo with NHS, 2 ± 0.4 cells/embryo with Dextran 125 and DBP-EGFP; Figure 1 -Supplement 1). However, the midbrain BBB remains leaky (average 126 of 23 ± 1 cells/embryo with NHS and DBP-EGFP and 24 ± 1 cells/embryo with Dextran; p=0.68 127 compared to 3 dpf, one-way ANOVA; Figure 1C and 1D). However, at 5 dpf the number of 128 midbrain parenchymal cells that uptake the tracers is dramatically reduced ...