Main text

Neurovascular endothelial cells (ECs) have unique functional and structural properties to regulate substance exchange between blood and brain, establishing the blood-brain barrier (BBB)^1,2. Loss of normal BBB function is associated with neuropathological states, including neurodegeneration, infection and cancer^1,2. Mural cells are vascular support cells that include pericytes and vascular smooth muscle cells (vSMCs) and control vascular development, function and BBB permeability^3–9. PDGFB/PDGFRβ signalling is required for the development of mural cells^3,4,10–12 and their depletion in PdgfB or Pdgfrβ deficient rodents causes altered angiogenesis, aneurysm and vessel leakiness^3–6,13. Strikingly, rodent models of pericyte loss have an open BBB, with free transport of tracer dyes from the blood stream into the brain parenchyma but an otherwise intact vasculature^5,6. These and other observations, have led to a current model whereby pericytes control BBB function by suppressing EC adsorptive transcytosis. This prominent mechanism holds significant translational promise to safely open up the BBB in therapeutic settings, by discovering ways to disrupt pericyte control of ECs^5,14.

Zebrafish have a highly conserved BBB^15–19 with the brain endothelium expressing conserved molecular markers of a functional BBB¹⁸ and showing conserved cellular interactions with pericytes, smooth muscle cells and glial cells^20,21. The zebrafish BBB becomes functional by 5 days post fertilisation (dpf)¹⁹ and as in mice, pdgfrb mutants lack brain pericytes and exhibit loss of vSMCs^13,20. Thus, the zebrafish is accepted as a highly accessible model to study BBB function and mural cell control of the BBB¹⁷. To investigate the influence of mural cells upon the neurovasculature, we generated a pdgfrb mutant (pdgfrb^uq30bh), which possesses an early deletion leading to a frameshift, predicted to cause a premature stop codon and a putative null allele. Using the TgBAC(pdgfrb:egfp)^ncv22 line, we observed that pericytes were lost in the cerebral central arteries of pdgfrb mutants (Fig. 1a), as in previous studies^13,20. To assess how brain angiogenesis occurs without pericytes, we imaged, traced and quantified blood vessel branching from the middle mesencephalic central arteries in the midbrain (Fig. 1a-c; Extended Data Fig. 1a). At 7 dpf, mutants displayed significantly reduced vascular complexity compared to siblings, with reduced vessel length and branch points (Fig. 1d-e). This was more severe at 14 dpf, demonstrating the phenotype is progressive (Fig. 1d-e). Timecourse imaging revealed the reduction in vascular complexity was apparent from 5 dpf (Extended Data Fig. 1a-b). These results suggest that pericytes (the vSMCs only begin to develop after 5 dpf²²) regulate angiogenesis from early in larval development.

Fig. 1:

We next tested BBB function in mutants at 7 and 14 dpf. We performed intravenous injection of fluorescently conjugated tracers of various molecular weights (1 kDa NHS, 10 kDa and 70 kDa Dextran), and live imaged fish at 2 hours after injection (as in ¹⁹) (Fig. 1f; Extended Data Fig, 1c). 2000 kDa Dextran was coinjected to normalize quantification of vascular tracer dye delivery (see methods for details). We found no evidence of 10 kDa or 70 kDa Dextran tracer extravasation into the brain parenchyma of pdgfrb^uq30bhmutants at either timepoint (Fig. 1g; Extended Data Fig. 1c-e). 1 kDa tracer showed no difference between genotypes but was observable in the parenchyma in both sibling and mutants, indicating the smaller cargo can pass the BBB (Extended Data Fig. 1c & e). Together, these results show that neither pericytes nor vSMCs play a role in establishment of BBB function in larval zebrafish.

To investigate how mural cells control cerebral vasculature at later stages of life, we tissue-cleared brains of adult stage pdgfrb^uq30bhmutants and siblings²³ and performed whole brain imaging. We confirmed loss of pericytes and vSMCs at this stage (Fig. 2a). We examined a consistent region of the left optic tectum (mid brain) and detected decreased branching of capillaries and increased diameter of major vessels (Fig. 1b-d). This vessel dilation was restricted to larger calibre arteries that would normally be invested with vSMCs and not seen in capillaries (Fig. 1c-d). This suggests vessel dilation in mutants is due to loss of vSMCs and reduced branching of capillaries to loss of pericytes.

Fig. 2:

We assessed BBB function in juvenile 2-month-old zebrafish by intravenous tracer injection (as in ²⁴) and using a mixture of 70- and 2000 kDa Dextran. We euthanized fish 2 hours post-injection and tissue-cleared the brains. The whole brain imaging revealed that mutants displayed widespread aneurysm of major vessels (Fig. 2e-f, Extended Data Fig. 2a-a′). To quantify potential BBB leakage, we imaged a consistent region with high capillary density (Fig. 2e′), and quantified parenchymal 70 kDa Dextran intensity relative to vascular 2000 kDa Dextran intensity. Both tracers remained restricted to the vascular lumen regardless of the severity of the aneurysms (Fig. 2g). This correlates with our findings at larval stages and demonstrated that mural cells are dispensable for the establishment and early maturation of the BBB in zebrafish up to 2 months of age.

To investigate BBB integrity at later stages, we injected 3-month-old adult zebrafish with 10 or 70 kDa Dextran. Brightfield imaging of vibratome sectioned brains revealed severe aneurysms across large calibre vessels (Fig. 3a) with blood pooling that was potentially consistent with vessel ruptures (Fig. 3c). Furthermore, we detected parenchymal accumulation of tracer dye closely associated with large calibre vessel aneurysms (Fig. 3b-c; Extended Data Fig. 3a-á). To quantify the distribution of parenchymal tracer dye from deep (around major artery aneurysms) to superficial (in capillary beds) brain regions, we measured fluorescence intensity in 10 brain regions 100 microns apart from medial to lateral locations. This demonstrated accumulation of tracer dye in medial regions associated with aneurysm in mutants, with no evidence of major leakage in capillary beds (Fig. 3d-e). To better understand the extravasated tracer accumulation, we conducted whole brain imaging after 70 kDa Dextran injection. Interestingly, we observed highly localised tracer accumulated in regions that could be considered leakage "hotspots" at large calibre vessel aneurysms (Fig. 4a; Extended Data Fig. 4a; Supplementary Videos 1 and 2). Quantification of three independent whole cleared brains throughout the optic tecta revealed on average 11 hotspots occurring at major vessels in each mutant animal (Fig. 4b). To further understand the nature of these hotspots, we examined red blood cells in pdgfrb mutant adults (Crispants, validated in Extended Data Fig. 4b) using the Tg(gata1:DsRed)^sd2 transgenic line together with tracer injection. These animals showed that tracer accumulation occurred concomitant with extravasated red blood cells at hotspots, identifying them as focal haemorrhages (Fig. 4c). Notably, these sites were not detected at capillaries (Fig. 4d). Taken together, this shows that in older mutants, loss of BBB integrity is associated with focal haemorrhage along aneurysmal vessels.

Fig. 3:

Fig. 4:

Pericytes are proposed to regulate the BBB by suppressing adsorptive transcytosis through brain capillary ECs^5,6. This model is built on the observation by electron microscopy of the induction of caveolae in leaky pericyte deficient vessels^5,6. To investigate if the loss of BBB integrity is associated with ultrastructural changes, we performed electron microscopy. We imaged large calibre vessels and capillaries (Fig. 4e; Extended Data Fig. 4c) and examined cellular junctions, basement membrane and caveolae (defined as uncoated vesicular profiles of <100nm diameter). While tight junctions between endothelial cells remained intact (Fig. 4f), Caveola density along the large vessel aneurysms in pdgfrb^uq30bh mutants was significantly increased. This increase was almost exclusive to the abluminal side of the ECs (Fig. 4g). This abluminal accumulation of caveolae may be more associated with changes in the mechanical state of ECs (as caveolae are highly regulated by membrane tension^25,26) than changes in intracellular transport. Furthermore, we detected thickening of basement membrane (BM) (Fig. 4g), gaps in the BM and fluid accumulation outside the vessel wall at aneurysms in mutants. We found no obvious difference in the caveolae density of capillary ECs of pdgfrb^uq30bh mutants compared with siblings (Extended Data Fig. 4d). Thus, leakage hotspots at aneurysms are associated with a failure of the normal flattening of ECs, caveola induction and disruption of the BM, all factors that could contribute to or be observed concurrent with vessel wall rupture.

Brain vasculature is endowed with mural cells that control vascular development and stability^3,4,13,27. Studies in rodents have shown that pericytes regulate BBB integrity^5,6. This is currently thought to be caused by hotspots of increased EC transcytosis¹⁴, rather than mechanisms such as altered cell-cell adhesion (eg, Cldn5¹⁴), control of Mfsd2a^19,28, Netrin1- Unc5B signalling²⁹ or vitronectin regulation of integrin receptors³⁰. However, some studies in rodents have reported loss of pericytes without vessel leakage³¹. We show mural cells are involved in cerebrovascular patterning in larval, juvenile and young adult animals but these vessels maintain BBB function in the absence of mural cells. In our mutants, major arteries progressively harbour aneurysms. This aneurysm is likely associated with loss of vSMCs¹³. At later stages, adult mutants show leakage at hotspots distributed along aneurysms, which are focal haemorrhages. In some PDGFB/PDGFRβ mouse mutants, haemorrhages have been reported, but not in mild or hypomorphic mutants such as those used in BBB studies^3,4,12. In our model, BBB leakage appears to only be observed once haemorrhage is present. Of course, these observations could represent a species-specific divergence from the mammalian BBB. However, the zebrafish BBB appears highly conserved with that of mammals¹⁷ and we would argue that the importance of BBB function for vertebrate brain physiology throughout evolution makes divergence unlikely. We suggest that a deeper analysis of the nature of hotspot leakage in mural cell deficient mammalian models is warranted. Altogether, a deeper understanding of the fundamental biology of the BBB will pave the way for increasingly sophisticated efforts to target the BBB in disease in the future.

Acknowledgements

This project was supported in part by funding from Bright focus (A2018807S) and the Australian Research Council (ARC) (DP210102712) and The Brain Cancer Centre (founded by Carrie’s Beanies for Brain Cancer). B.M.H was supported by a National Health and Medical Research Council Senior Research Fellowship (1155221). O.F.B was supported by the Brain Cancer Centre Blood-Brain Barrier Program. R.G.P was supported by an ARC Laureate Fellowship (FL210100107). We thank the Centre for Advanced Histology and Microscopy (RRID:SCR_025432) at the Peter MacCallum Cancer Centre and Microscopy Australia Research Facility at the Centre for Microscopy and Microanalysis at The University of Queensland. We also thank Kelly Smith and Marcos Sande Melon for academic discussions and technical suggestions.

Additional information

Author contributions

O.B performed, analysed experiments and co-wrote manuscript. B.M.H, _onceptualized experiments, analysed data and co-wrote manuscript. A.U.G, S.D, W.W, M.C.R.G, Y.W.L, J.R, R.P, A.L and A.F performed and analysed experiments.

Supplementary materials

Zebrafish husbandry

Zebrafish work was conducted in compliance with animal ethics committees at the Peter MacCallum Cancer Centre, The University of Melbourne and The University of Queensland. Previously published transgenic lines were TgBAC(pdgfrb:egfp)^{ncv22 1},Tg(kdrl:Hsa.HRAS-mCherry)^{s843 2}, Tg(kdrl:EGFP)^{s843 3} and Tg(gata1:DsRed)^sd24.

Genome editing and genotyping

The pdgfrb^uq30bh mutant strain was generated using CRISPR/Cas9-mediated genome editing. The resulting allele carries a 39-bp deletion and a 5-bp insertion in exon 3 of pdgfrb (ENSDARG00000100897), causing a frameshift and premature stop codon.

The gRNA sequence and genotyping primers used were: gRNA binding site: 5′ GATGGTGACTAAGACGCGA 3′ Forward genotyping primer: 5′ CTTCCTTAGATCCTGACGTGTG 3′ Reverse genotyping primer: 5′ TATTGATGGGTTCGTCACCAG 3′

The pdgfrb F0 crispants used were generated as F0 Animals by CRISPR/Cas9-mediated genome editing using predesigned Alt-R CRISPR-Cas9 gRNAs (IDT). The gRNA sequence and genotyping primers used were: Dr.Cas9.PDGFRB.1.AA: 5′ GATGGTGACTAAGACGCGAG 3′

Forward genotyping primer: 5′ CTTCCTTAGATCCTGACGTGTG 3′ Reverse genotyping primer: 5′ TATTGATGGGTTCGTCACCAG 3′ Dr.Cas9.PDGFRB.1.AB: 5′ CTCGGTGCACACATAAACCC 3′

Forward genotyping primer: 5′ GACGAGAACATCCCAGACTTTC 3′ Reverse genotyping primer: 5′ GCGTGTAAACAAATCCTAACGG 3′

Tracer dye injections and imaging

For all injections at larva and juvenile stages, NHS Ester–Alexa Fluor 405 (Thermo Fisher: A30000) or 10 kDa Dextran–Cascade Blue (Thermo Fisher: D1976) or 70 kDa Dextran– Fluorescein (Thermo Fisher: D1822) were mixed and co-injected with 2000 kDa Dextran– Tetramethylrhodamine (Thermo Fisher: D7139). This allowed for the normalisation of dye signal relative to the larger molecular weight 2000 kDa tracer to ensure that the amount injected intravenously was appropriately controlled for in all imaging experiments. A final concentration of 5 mg/ml for each tracer was used in all injections.

7 dpf larvae were anesthetised with tricaine and mounted laterally with 0.5% low-melting point agarose (Bio-Rad, 1613112) in E3 embryo water on a 35-mm glass bottom dish (MatTek: P35G-1.5-20-C). 5nl of the tracer mix was injected into the posterior cardinal vein (PCV) and larvae were recovered from agarose using a glass pipette into E3 embryo water. 14 dpf larvae were anesthetised with tricaine, placed laterally on a 3% agarose injection mold. 10 nl tracer mix was injected into the PCV and larvae were recovered in E3 embryo water.

For live imaging, larvae were mounted ventrally with 0.5% low-melting point agarose in E3 embryo water on a 35-mm glass bottom dish. All larvae were carefully live imaged at 2 hours post injection, with a maximum variation in timing of 15 minutes due to the sequential imaging, using Nikon TiE with Yokogawa CSU-W1 spinning disk with Andor Sona sCMOS camera, 40X 1.15 NA water immersion objective for 7 dpf larvae and 20X 0.75 NA objective for 14 dpf (1 um z-step for each stage).

Injections at juvenile and adult stages were performed retro-orbitally using a glass capillary needle as previously described⁵. For juvenile stage (2-month post fertilisation), brains were extracted 2 hours post injection and whole brains were tissue-cleared using the CUBIC-based method previously described⁶. Briefly, brains were fixed in 2% paraformaldehyde (PFA) at 4°C overnight, washed in PBS, incubated in CUBIC-L solution at 37°C overnight, washed and incubated in PBS at 4°C overnight, then incubated in CUBIC-R solution at room temperature overnight. Brains were mounted ventrally with 2% agarose in CUBIC-R on a 35-mm glass bottom dish and imaged using Nikon TiE with Yokogawa CSU-W1 spinning disk with Andor Sona sCMOS camera, 4X 0.2 NA objective, 10X 0.45 NA objective, and 20X 0.75 NA objective. For adult stage (>2.5-months post fertilisation), Cascade Blue 10- or 70 kDa Dextran were injected as previously described⁵. Brains were extracted 2 hours post injection, fixed in 2% PFA overnight, washed in PBS. For vibratome sectioning, brains were embedded in 7% low- melting agarose in PBS in cryomolds (Tissue-Tek) and were sectioned coronally using a vibrating microtome (Leica VT1000 S) with 200 µm thickness setting. Selected sections were placed on a microscope slide and were incubated with RapiClear 1.52 solution (SunJin Lab, RC152201) for 3 hours at room temperature. Imaging was performed using Olympus FV3000 confocal microscope with 10X 0.4 NA objective with 5 µm step size and 40X 1.4 NA oil immersion objective with 1 µm step size. For whole brain imaging, brains were tissue cleared as described earlier and were imaged using Olympus FV3000 confocal microscope with 10X 0.4 NA objective with 5 µm step size. Videos highlighting the “hotspot” leakage sites were generated using Imaris 10.1 software (Bitplane) and Adobe Premier Pro 2024.

Live imaging of pericytes and blood vessels was conducted using Olympus FV-MPERS multiphoton microscope with 25X 1.05 NA water dipping objective with 3 µm step size. To ensure optical transparency, live imaging was conducted in F0 knockout animals for the gene slc45a2 (ENSDARG00000002593) as previously described⁷ or 0.003% PTU was applied to the embryo water at 1 dpf.

Transmission electron microscopy

Brains were extracted from euthanised 5-month-old zebrafish, fixed in 2.5% glutaraldehyde for 1 hour at room temperature and washed with PBS. For sectioning, the brains were embedded in 7% low-melting point agarose in PBS in cryomolds (Tissue-Tek) and were sectioned coronally using a vibrating microtome (Leica VT1000 S) with 200 µm thickness. Tissue sections were processed as described previously⁸.

Briefly, tissue sections were immersed consecutively in a series of aqueous solutions: 1.5% potassium ferricyanide and 2% osmium tetroxide, 1% thiocarbohydrazide, 2% osmium tetroxide, 2% uranyl acetate and 0.06% lead nitrate using a Pelco Biowave at 80W under vacuum for 3min each. Vibratome sections then underwent serial dehydration in increasing concentrations of ethanol, before serial infiltration with increasing concentrations of Procure 812 resin. Ultrathin sections were obtained using a Leica UC64 ultramicrotome and imaged on a Jeol JEM-1011 at 80kV.

Quantifications

Vessel length, diameter and branching points

Vessel length, diameter and branching points were quantified using Imaris 10.1 software (Bitplane). For larval stages, lumenized midbrain central artery (kdrl:Hsa.HRAS-mCherry) branching from the middle mesencephalic arteries was traced using the filament tool and the total filament length (µm) was plotted. The branching points within the traced blood vessels were quantified by manual scoring in 3D view. For adult stage, 100-µm thick maximum intensity projections (MIP) were generated from whole brain imaging, using the continuation of the left middle mesencephalic central artery as an anatomical anchor point as the consistent region measured. For diameter measurements, 10 capillary diameters were measured per sample. Vessel anatomy was identified using the study by Isogai et al⁹ as a reference.

Tracer intensity

Tracer intensity quantification was performed on genotype blinded image sets using Fiji¹⁰. For larval stage experiments, the tracer accumulation in brain parenchyma (70-or 10 kDa Dextran or 1 kDa NHS Ester) was calculated using average intensity from 4 midbrain parenchymal regions in 50-µm thick MIPs, which were generated starting from 50 µm below dorsal longitudinal vessel (DLV) to avoid detecting potential leakage from surface vessels. These measurements were normalized to the vascular tracer intensity of 2000 kDa Dextran, which was calculated using average intensity from 4 regions within the DLV lumen. As such, for each data point: intensity=parenchymal dextran (1, 10 or 70 kDa) intensity/luminal 2000 kDa Dextran intensity.

For juvenile stage experiments, 30-µm thick MIPs were generated starting from 60 µm below the most superficial vessel to potential leakage from the surface vessels. Brain parenchymal 70 kDa Dextran intensity was calculated using average intensity from 4 midbrain parenchymal regions and was normalized to vascular 2000 kDa Dextran intensity, which was calculated using average intensity from 4 regions within the lumen of superficial large calibre vessels. For adult stage experiments, 50-µm thick MIPs were generated and dextran intensity outside the vessels was measured at every 100 µm from medial to lateral direction within 1 mm. Branching point of left and right middle mesencephalic arteries was determined as the starting point.

TEM quantification

Imaged sections were quantified using Fiji¹⁰. Caveolae were defined as circular profiles of less than 100 nm in diameter and were scored as luminal or abluminal based on proximity to each surface membrane (within 500 µm of each surface or in a thin walled vessel the caveolae closest to each surface). Basement membrane thickness was scored in 6 different locations per vessel that were selected at random. N=3-8 vessels per group were analysed as indicated in the Figure legends.

Statistical analysis

Statistical analyses and graphical representations were conducted using GraphPad Prism 10 software. ns: P>0.05, *P≤0.05, **P≤0.01, ***P≤0.001, ****P≤0.0001. For Figure 3e, we applied the non-parametric two-sample Kolmogorov-Smirnov test using Python software to evaluate whether the distribution of tracer accumulation for each position differed from that of the control. Further details on statistics can be found in each figure legend.

Extended Data Fig. 1:

Extended Data Fig. 2:

Extended Data Fig. 3:

Extended Data Fig. 4: