Introduction

Brain microvessels supply billions of neurons and other brain cells with the oxygen and nutrients they need. Pathologies affecting brain microvasculature progress slowly and silently over a lifetime but have devastating consequences from decreased perfusion and vascular destabilization. Cerebral small vessel disease (CSVD) encompasses progressive heterogeneous changes in brain microvessels; it is the most common cause of vascular dementia and a significant contributor to stroke and cognitive decline (Østergaard et al., 2016). 25% of all strokes are the result of CSVD, yet effective targeted treatments remain elusive (Østergaard et al., 2016). This is in part due to the inability to both detect and assess progressive damage in the brain.

While there are several genes implicated in familial CSVD (i.e., NOTCH3, HTRA1, FOXC1, COL4A1 and COL4A2), there is a lack of suitable in vivo models for studying disease development and progression. Research has predominantly focused on NOTCH3, but over the last decade, FOXF2 has emerged as a risk locus for CSVD (Chauhan et al., 2016; Duperron et al., 2023). SNPs in the intergenic region between FOXF2 and FOXQ1 decrease FOXF2 expression and significantly increase stroke risk due to the variant decreasing the efficiency of ETS1 binding to a novel FOXF2 enhancer (Ryu et al., 2022).

Foxf2 promotes mural cell differentiation and vascular stability in the zebrafish brain and is expressed highly in brain pericytes (Ahuja et al., 2024; Chauhan et al., 2016; Reyahi et al., 2015; Ryu et al., 2022). Brain pericytes interact closely with endothelial cells, contributing to extracellular matrix (ECM) deposition and blood-brain barrier (BBB) formation, in addition to providing vasoactivity and stability (Bahrami and Childs, 2020; Daneman et al., 2010; Dave et al., 2018; Stratman et al., 2009). In animal models, an absence of brain pericytes results in hemorrhages and accelerates vascular-mediated neurodegeneration (Bell et al., 2010; Wang et al., 2014). Foxf2 is clearly important for vascular stability across species, as loss of Foxf2 in mice and zebrafish leads to increased brain hemorrhage and alterations in brain pericyte numbers and differentiation (Chauhan et al., 2016; Reyahi et al., 2015; Ryu et al., 2022). Brain tissue from patients with aging-related dementias (i.e. post-stroke dementia, vascular dementia, Alzheimer’s disease) has reduced deep white matter pericytes and associated BBB disruption (Ding et al., 2020) suggesting that pericytes should be examined as mediators of CSVD progression in patients with FOXF2 deficiency.

We previously showed that complete loss of foxf2 in foxf2a^ca71; foxf2b^ca21 double- homozygous mutants in late embryogenesis leads to reduced brain pericyte numbers (Ryu et al., 2022). However, stroke susceptibility in humans is associated with reduced, but not absent, FOXF2 expression. Genome-wide association (GWA) indicates that carrying a minor allele of an SNP in a FOXF2 enhancer leads to reduced, but not absent, FOXF2 and is associated with stroke (Ryu et al., 2022). For this reason, we model CSVD using a zebrafish with reduced Foxf2 dosage using single homozygous foxf2a mutants. Zebrafish foxf2a and foxf2b genes are the result of genome duplication in zebrafish ∼430 million years ago, and have similar gene expression (Arnold et al., 2015; Chauhan et al., 2016). We have detected no difference in function between the two genes, and therefore foxf2a loss of function may be similar to human heterozygous loss of FOXF2 function, a state that is observed in the population in GnomAD (Chen et al., 2024).

Strikingly, while pericytes in embryonic foxf2 mutants are clearly affected, foxf2 mutants can survive until adulthood, albeit with a reduced lifespan. How pericytes change across the lifespan while CSVD progresses is unknown.Here we find that foxf2a mutants have significantly reduced brain pericyte numbers as embryos that do not recover over time. Pericytes in mutant embryos and larvae exhibit morphological abnormalities, including increased soma size, longer processes, and degeneration. We show that processes and soma in the adults are also abnormal, though their morphology differs over the lifespan. Although the initial pool of pericytes is smaller, mutants can regenerate pericytes after ablation. Our analysis suggests that foxf2 is required within pericytes to modulate numbers but also has a strong effect on morphology. We show that brain pericytes may contribute to the pathological progression of genetic CSVD, starting in embryonic development and continuing across the lifespan. Understanding the early developmental aspects of late-onset vascular conditions like CSVD will aid in the development of effective therapeutic strategies.

Results

Pericyte number is consistently lower in foxf2 mutant embryos and larvae

Embryonic phenotypes can lead to lifelong consequences. We have previously studied foxf2a;foxf2b double mutants only at a single embryonic stage at 3 days post fertilization (dpf). To understand how a pericyte and cerebrovascular phenotype evolves and/or resolves over development, we conducted serial imaging of individual brains of foxf2a mutants at embryonic stages (3 and 5 dpf), and at larval stages (7 and 10 dpf). Mutants were live imaged using endothelial (kdrl:mCherry) and pericyte (pdgfrβ:Gal4, UAS:GFP) transgenic lines.

Embryonically, brain pericytes have a thin-strand morphology and are closely associated with, and extend processes over, the endothelium in the midbrain and hindbrain of zebrafish (Fig. 1A-A’’). In wildtype embryos, the number of pericytes increases progressively from 3 through 10 dpf (Fig. 1B). However, foxf2a mutants show significantly fewer pericytes on brain vessels at 3 dpf (Fig. 1C; mean 21 in wildtype and 10 in mutants) and this pericyte deficiency persists through 5, 7, and 10 dpf (Fig. 1C-D). The reduction in pericyte numbers shows variable penetrance, with some foxf2a mutants having pericyte numbers only slightly reduced from wildtype, and others that are severely diminished. The same pattern of pericyte reduction is seen with foxf2a mutants from a homozygous or heterozygous incross, suggesting there is no maternal effect (Fig. 1E-F). Serial imaging of mutants with regional absence in earlier stages shows that defects in pericyte coverage persist into later stages, suggesting that the size of the initial pericyte population is a key determinant of later coverage (Fig. S1). Double foxf2a;foxf2b mutants have a fully penetrant phenotype with significantly fewer brain pericytes in mutants than wildtype at every stage (Fig. S2; mean 19 in the wildtype and 9 in mutants at 3 dpf). Incomplete penetrance in foxf2a single mutants could be due to genetic compensation from foxf2b. While foxf2b is not significantly upregulated in foxf2a mutants on average, individual embryos have highly variable foxf2b expression (Fig. S3).

Figure 1.

Since Foxf2 conditional knockout mice show reduced expression of the pericyte marker Pdgfrβ, we tested whether pdgfrβ expression is reduced in foxf2a mutants, as this might introduce inaccuracies in pericyte counting. We used two methods, quantitative hybridization chain reaction (HCR) in situ hybridization and integrated density of the pdgfrβ transgene expression in wildtypes and mutants carrying only a single copy of the transgene. We show that pdgfrβ mRNA nor transgene expression is not reduced in foxf2a mutants (Fig. S3). Further, we show that pdgfrβ has complete overlap in expression with other brain pericyte markers such as ndufa4l2 and foxf2b using HCR (Fig. 2A-D). Thus, pdgfrβ transgene or mRNA expression can reliably be used to count zebrafish brain pericytes in foxf2a mutants.

Figure 2.

Although we focus on mural cells, Foxf2 is expressed in adult mouse brain endothelium (Vanlandewijck et al., 2018). We assessed foxf2a and foxf2b mRNA expression patterns using HCR in situ hybridization. At 3 dpf, foxf2a is co-localized with the pericyte marker ndufa4l2a in brain pericytes and shows low level co-localized expression with the blood vessel marker kdrl in brain endothelium (Fig. 2A). The expression pattern of foxf2b is similar. It is expressed in pericytes (pdgfrβ) and only weakly in endothelial cells (kdrl) (Fig 2B). This is supported by the DanioCell atlas of single-cell sequencing of multiple embryonic stages that shows expression of both genes is strongest in mural cells, pericytes and vascular smooth muscle cells and low in endothelial cells (Sur et al., 2023) (Fig. S4). Thus, while foxf2a and foxf2b are principally expressed in pericytes they are lowly expressed in endothelium during brain development.

Pericyte loss or impairment leads to alterations in vascular patterning in the mouse retina (Eilken et al., 2017). To assess if the endothelial network is affected in foxf2a mutants, we employed a Python workflow using Vessel Metrics (McGarry et al., 2024) (Fig. 2E). We found no statistical difference in total network length between wildtype and mutants at 3 dpf (Fig. 2F) or hindbrain central artery diameter (Fig. 2G). However, pericyte density (number of pericytes divided by the total network length) is reduced by 40% in foxf2a mutants (Fig. 2H), reflecting the loss of pericytes with no change in vessel network length. Similarly, pericyte coverage of vessels (total process coverage from brain pericytes divided by the total network length) is reduced by 39% in mutants (Fig. 2I). Our data suggest that during early development, foxf2a depletion primarily affects pericytes.

Early defects in brain vessel development have lifelong consequences

foxf2 mutant animals can survive to adulthood, albeit with a reduced lifespan (∼1 year vs. >2 years). Are early pericyte deficiencies repaired, or is loss of pericytes unimportant to survival to adulthood? To understand how brain pericyte phenotypes evolve over the lifespan, we dissected adult wildtype and foxf2a mutant brains on pericyte and endothelial double transgenic backgrounds and imaged after iDISCO clearing. Gross measurements of standard length of the fish (snout to tail) show no significant differences between wildtype and mutants, except that female mutant brain length and width are significantly smaller at 11 months post fertilization (mpf) (Fig. S5). However, there is no significant difference in the proportional brain length/standard length ratio in foxf2a mutants vs. wildtype.

Projected 3D views of light sheet images of the whole brain show striking defects in pericyte density, coverage and vascular pattern in foxf2a mutants vs. wildtype adults (Movie 1-2; Fig. 3A-B). Pericyte distribution is irregular, and blood vessel density is visibly reduced in mutant brains (Fig. 3C-D). Using a machine learning workflow (Fig. S6), we segmented pericyte soma, or the vessel backbone for the entire adult brain to count the total pericyte number in comparison to the total endothelial network length. We find a significant reduction in pericyte numbers in 3 mpf foxf2a mutant brains, which have only 45% of the pericytes of a wildtype brain (average of 24567 pericytes/brain vs. wildtype 54833 pericytes/brain; n=3 of each genotype; Fig. 3E). However, at this stage, the vessel network length is not statistically different (Fig. 3F). The density of pericytes on vessels is significantly decreased from 0.01 pericyte/µm to 0.006 pericytes/µm in foxf2a mutants showing a clear deficit (Fig. 3G). In contrast, the mean vessel diameter across vessel segments is not significantly different at 3 mpf when all diameters are considered, nor when vessel diameters are grouped in 5 µm bins (n=522037 wildtype and 381544 foxf2a mutant diameters; Fig. 3H-I).

Figure 3.

Similarly, cleared dissected brains at 11 mpf were imaged and show similarly striking vascular defects (Fig. S7). Adult pericytes have a clear, oblong cell body with long, slender primary processes that extend from the cytoplasm, and secondary processes that wrap around the circumference of the blood vessel (Fig. 3J). These pericytes form a continuous network of processes to cover the blood vessels in the brain. In foxf2a 11 mpf brains, pericyte somas lose their oblong shape, and cell bodies cannot be easily distinguished from processes. Mutant cells also extend thickened linear processes, with no secondary processes encircling the vessels.

Mutant pericytes do not form an extensive network with other neighbouring pericytes. We note that there is variable phenotypic penetrance in foxf2a adults at 3 mpf that is reflective of the incomplete penetrance in foxf2a embryos (Fig. S7).

To view cellular morphology, we sectioned adult brains and immunolabeled pericytes and endothelium. In wildtype adult brains, we identified three subtypes of pericytes, ensheathing, mesh and thin-strand, previously characterized in murine models (Berthiaume et al., 2018a) (Fig. S8). In comparing brain sections from both wildtype and foxf2a mutants, on smaller vessels, mutant pericytes exhibit more linear processes with barely distinguishable soma, markedly differing from the characteristic appearance of wildtype adult pericytes (Fig. S9). Similarly, pdgfrβ-expressing cells on large-calibre vessels (mural cells, likely vascular Smooth Muscle cells (vSMCs)) show alterations in morphology and coverage (Fig. S9).

Although mutant embryos do not exhibit apparent abnormalities in their endothelium, large aneurysm-like structures are evident in the adult brain (Fig. S10). These structures also appear to have decreased kdrl expression. Thus, both whole-mount and sectioned tissue show that brain vascular mural cell number and morphology are severely impacted in adult foxf2a^-/- mutant brains, with increasing involvement of the endothelium, suggesting a worsening phenotype throughout life.

As foxf2a is expressed in vSMCs, we tested the effect of loss of foxf2a on vSMCs using confocal imaging of larvae and light sheet microscopy of cleared dissected 6 mpf adult wildtype and foxf2a mutant brains stained for transgenic endothelial (kdrl) and vSMC (acta2) markers. acta2-positive vSMCs are present on brain vessels around the Circle of Willis at embryonic (5 dpf) and larval (10 dpf) stages in both mutant and wildtype (Fig. 4A,C). Further, the number of vSMCs does not differ between mutant and wildtype (Fig. 4B,D). In adult brains, we find no significant difference in the total length of vSMCs in the brain, or in vSMC coverage (proportion of vessels covered by vSMCs) (n=3 of each genotype; Movie 3-4; Fig. 4E-H). We note that at 6 mpf, there is a significant decrease in vessel network length however (Fig. 4G).

Figure 4.

Morphological abnormalities emerge in the larval brain of mutants

Considering the altered pericyte morphology of adult mutants, we revisited developmental stages to identify when these abnormalities first arise. We analyzed morphology at 3 and 10 dpf using in vivo confocal imaging to understand defects in the cell body (soma) and processes (Fig. 5A). We find no significant difference in the soma area at 3 dpf, but at 10 dpf, there is a significant increase in mutant pericyte soma area compared to wildtype (Fig. 5B). In parallel, wildtype pericytes undergo a slight reduction in soma size from 3 to 10 dpf.

Figure 5.

The low number of brain pericytes in foxf2a mutants means that distinct pericyte processes can be distinguished and measured. However, in wildtypes, pericyte processes are not easily distinguished. For accurate process length measurements in wildtype we used multispectral labelling with a pericyte-specific Zebrabow transgenic line (pdgfrβ:Gal4, UAS:Zebrabow) (Pan et al., 2013; Whitesell et al., 2019). Cre mRNA was injected at the one- cell stage to activate random recombination, allowing us to visualize individual neighbouring pericytes (Fig. 5C). Wildtype processes have a mean length of 81.3 µm vs 102.9 in the mutant at 3 dpf and 96.3 µm vs 147.6 in the mutant at 10 dpf (Fig. 5D). Thus, the pericyte process lengths are significantly increased in foxf2a mutants at 3 and 10 dpf (Fig. 5D).

Through study of wildtype processes using multispectral labelling, we observed some differences in the behaviour of adjacent pericyte processes from that published in the adult mouse brain (Berthiaume et al., 2018b). While direct contact with no overlap between pericytes is the most common interaction in developing zebrafish, similar to mouse adult brain pericytes (Fig. 5F), we also see some overlap (Fig 5E). The average overlap between wildtype pericyte processes at 10 dpf was 5.9 µm (Fig. 5G). Together, our data show overlap in brain pericyte processes in wildtype animals.

To further explore mutant pericyte behaviour, we conducted serial imaging during larval development. Over time, some mutant pericyte processes form disconnected bead-like blebs with cell bodies that disappear over time (Fig. 6A). This pericyte phenotype can occur on vessels with full patency and is not associated with endothelial regression. The bead-like remnants are highly prevalent during larval development in foxf2a^-/- mutants but not present in wildtype (Fig. 6B). To visualize process degeneration in real-time, we time-lapse imaged from 4- 5 dpf (Fig. 6C). The pericyte can undergo a process reminiscent of cell death with soma blebbing and process degeneration phenotypically resembling neural dendrite degeneration or pruning (Fig. 6D).

Figure 6.

In summary, larval foxf2a mutant pericytes show reduced numbers, increased soma size, and elongated processes with evidence of process degeneration. In addition, we make the novel observation of overlapping pericyte processes during zebrafish development.

Foxf2a mutants do not have an impaired capacity to repopulate brain pericytes

Zebrafish have regenerative capacity in various tissues (Becker et al., 1997; Lepilina et al., 2006; Otteson and Hitchcock, 2003), yet foxf2a mutant embryos that are pericyte-deficient maintain strong brain pericyte defects through aging which suggests that either they may not be able to replenish absent/damaged pericytes, or that the smaller size of the initial pool in embryogenesis limits repair such that numbers never can catch up to wildtype. To differentiate these hypotheses, we tested whether foxf2a mutants lack the capacity to regenerate pericytes. We reduced pericyte numbers in a foxf2a heterozygous incross using a cell ablation strategy.

Zebrafish expressing pdgfrβ:Gal4, UAS:NTR-mCherry and flk:GFP transgenes were treated with 5 mM metronidazole (MTZ) at 50 hpf for 1 hour which ablates most pericytes. MTZ is a prodrug substrate that elicits cell death in nitroreductase (NTR)-expressing cells due to its cytotoxic derivatives. We then imaged and counted brain pericytes at 3 dpf (a day after treatment) and 10 dpf (recovery).

At 3 dpf, in the vehicle (DMSO) control group there is the expected significant difference between wildtype and mutant pericyte numbers (Fig. 7A-B). When treated with metronidazole, both mutants and wildtypes both have a similar, severe reduction in pericytes at 3 dpf post- ablation, and are not significantly different from each other (Fig. 7B). By 10 dpf, pericytes are partially repopulated in both wildtype and mutant ablated groups (Fig. 7C). Surprisingly, given the initial pericyte number defects in foxf2a mutant embryos, we find no significant difference between any groups (Fig. 7D; mean 70 in wildtype DMSO treated, 53 in mutant DMSO treated, 43 in wildtype MTZ treated, 33 in mutant MTZ treated). This shows that foxf2a mutants retain their ability to repopulate pericytes following ablation. This data is important to distinguish the timing and mechanism of pericyte reduction in foxf2a mutants. While foxf2a mutant pericytes can regenerate following induced catastrophic loss, it cannot compensate for the initially smaller pool of pericytes during embryogenesis. This data suggests that foxf2a plays a critical role in the establishment of a properly sized initial pericyte pool during early development. Regeneration mechanisms, although intact, cannot fully compensate for this initial reduction in pericytes.

Figure 7.

Discussion

Our reduced dosage model of Foxf2 demonstrates disease processes at the cellular level in intact animals, giving insight into pathological changes that occur during CSVD that have not been observed in other models or humans. Our data supports a developmental origin for this type of CSVD, which then progresses and evolves across the lifespan (Fig. 8).

Figure 8.

Dosage sensitivity of Foxf2

The logic for using a reduced dosage of Foxf2 in our studies is to better match the effect of common population variants leading to CSVD. We previously modelled the effect of a high-risk CSVD and stroke-associated SNV on the expression of FOXF2 in human cells, showing that it can reduce expression by ∼50% (Ryu et al., 2022). Since zebrafish have two FOXF2 orthologs (foxf2a and foxf2b), foxf2a homozygotes have equivalent FOXF2 dosage as human heterozygotes, assuming that foxf2a and foxf2b not only exhibit similar expression but also share similar functions. Foxf2 is a dosage sensitive gene as zebrafish foxf2a hets show a significant reduction in brain pericytes. Similarly, mouse Foxf2 is also dosage sensitive (Reyahi et al., 2015). We find that foxf2a mutants are variably penetrant, while foxf2a;foxf2b double mutants are fully penetrant. Genetic compensation is common with gene duplication (Kok et al., 2015; Rouf et al., 2023) and would explain variability in disease phenotypes in different individuals with the same mutation.

Impaired embryonic pericyte coverage in Foxf2 deficiency

We show that numbers of brain pericytes are reduced at multiple developmental and adult stages when foxf2a and/or foxf2a;foxf2b are lost. The deficiency is significant at every stage, including 3 dpf, the earliest time point that pericytes are robustly observed in development. Deficient numbers could be due to a reduction in the pericyte precursor population (i.e. nkx3.1 positive cells (Ahuja et al., 2024)), or to an inability of foxf2a-deficient cells to differentiate into pdgfrβ-positive pericytes. Our data does not allow us to distinguish these, although pdgfrβ expression in foxf2a mutants is transcriptionally unchanged, suggesting that pericyte number in early development is the primary phenotype. Additional evidence for pericytes being the primary cells affected is that there is no difference in total vessel network length or hindbrain CtA diameter at 3 dpf in foxf2a heterozygotes or mutants when pericyte numbers are reduced. A reduction in pericytes is expected to have early consequences, as pericytes provide signals to the endothelium for quiescence and arterial-venous identity (Mäe et al., 2021), as well as allowing contractility of cerebral blood vessels in early development (Bahrami and Childs, 2020). Fewer pericytes distributed on a normal-sized endothelial network length results in reduced vessel coverage. It is intriguing therefore, that pericyte process length is increased at both 3 and 10 dpf, potentially to compensate for low pericyte density. A similar elongated pericyte phenotype and reduced coverage is seen in mice with constitutive Pdgfβ^ret knockout (Mäe et al., 2021). The convergent phenotypes after manipulation of two genes (foxf2a in fish and Pdgfβ ^ret in mice) that reduce pericyte number suggest that the intrinsic pericyte response to depletion of pericyte density is to elongate, perhaps to attempt to cover ‘naked’ vessels. We note that previous mouse Foxf2 knockouts using a conditional Wnt1-Cre driver saw contrasting results to ours, with increased pericyte numbers, although a similar loss of vascular stability is seen in both models (Reyahi et al., 2015). The reason for the difference in the direction in pericyte number may be experimental due to the different knockout technique (conditional in mice which only removes neural crest-derived pericytes vs. full knockout which removes mesodermal, neural crest-derived and endothelial foxf2a in fish) or the difference could reflect a species- specific difference.

Insights on fundamental pericyte process properties

The reasons behind the hypergrowth of pericyte processes in foxf2a^-/- mutant brains is unclear. Berthiaume et al (2018a) propose that repulsive interactions among pericytes in the adult brain establish boundaries between adjacent pericytes, preventing their processes from overlapping. In this case, one might hypothesize that foxf2a mutant pericytes, which are less dense along vessels, would lack feedback from adjacent pericytes, leading to uncontrolled growth.

In mice, adult brain pericytes form a non-overlapping network along capillaries, although their processes occasionally approach each other without overlapping (Berthiaume et al., 2018b; Hill et al., 2015). Overlap between pericyte processes have not been well studied the developing brain, but pdgfrb^+ve cells in developing zebrafish fin also overlap (Leonard et al., 2022).

Strikingly, using multispectral (Zebrabow) labelling, we find processes overlap in the developing zebrafish brain. Overlaps occur on capillary segments and at branch points. Previous studies in mice were conducted in the adult brain, and it is possible that overlapping processes in the developing brain are transient, zebrafish-specific, or were potentially overlooked in the mouse experiments.

We have also observed forked extensions at the tips of some pericyte projections at 10 dpf in wildtype embryos, where pericytes are not in contact. Whether these unique structures are involved in attractive/repulsive signals between pericytes or help determine the direction and extent of process growth is unclear, but understanding the fundamental mechanisms governing pericyte-pericyte process interactions would yield valuable insights into development and disease, as well as into how pericyte depletion results in abnormally long processes.

Morphological abnormalities in pericyte soma in foxf2a mutants

A second morphological abnormality of foxf2a deficient pericytes is that their soma size is increased at 10 dpf, which may also be a compensatory change. Pathological changes in soma size have not been observed in brain pericytes. However, increased neuronal soma size is observed in Amyotrophic Lateral Sclerosis or Lhermitte-Duclos disease (Dukkipati et al., 2018; Kwon et al., 2001) and is linked to mTor signaling (Kwon et al., 2003). In neuronal populations, moderate soma swelling can be an adaptation for survival, while rapid, drastic swelling indicates imminent death (Rousseau et al., 1999). It is therefore not surprising that we observe degeneration of pre-existing pericytes in foxf2a mutant but not wildtype animals. This degeneration phenotypically resembles neuronal dendrite remodelling and pruning during development (Fukui et al., 2012; Greenwood et al., 2007). Further work to test whether pericytes share similar mechanisms of degeneration in response to stress or cellular damage may provide further insight into phenotypic progression.

Changes in foxf2a mutants across the lifespan

For the first time, we have examined how vascular phenotypes change in the adult brain using iDISCO clearing. We find that adult soma take on very unusual, ‘stiff’ shapes and become almost indistinguishable from processes. The adult processes are shorter and more linear than the embryonic processes. Hypertrophic embryonic pericytes either do not survive to adulthood or undergo morphological changes over time. Given the limited knowledge in this area, the underlying factors driving this shift in pericyte morphology are unclear. Actin is present near the plasma membrane of cells, where it provides structural and mechanical support, enabling motility and determining cell structure. Given the abnormal shapes of adult pericytes, alterations in actin may contribute to their irregular morphology.

In the adult foxf2a mutant brain, abnormal pdgfrβ-expressing cells are observed on large-calibre vessels and likely represent vSMCs. We observed large-calibre aneurysm-like vessels, suggesting that vSMCs may lose functionality over time, thereby increasing the vulnerability of underlying vessels to dilation and weakening. Loss of acta2 and, subsequently, vSMCs, has been associated with conditions such as thoracic aortic aneurysms and dissections (Guo et al., 2009). However, we found no significant change in either vSMC cell number in embryonic development or in network length of vSMCs in the adult brain suggesting that the primary phenotype in foxf2a mutants is in pericyte cells, and that any changes to vSMCs are likely secondary.

Similarly, brain vessel diameter and network length are not significantly altered in embryonic foxf2 mutants, but in 6-month-old adults, the vessel network length is significantly decreased. Our data suggests that foxf2 deficiency contributes to cumulative vascular defects. Adult endothelial defects may be secondary to pericyte defects, although foxf2a is expressed in both pericytes and endothelial cells, and both cell types may be affected autonomously.

The initial pericyte pool is critical for lifelong vascular health

To understand the mechanism by which foxf2a influences pericyte numbers, we needed to distinguish between its role in early development and later roles. Our lab recently demonstrated that pericyte precursors express nkx3.1 and foxf2a before pericyte differentiation and prior to expression of canonical markers such as pdgfrβ (Ahuja et al., 2024). Here we show that the pericyte pool in foxf2a mutants is reduced at the earliest embryonic stage that it can be measured. Zebrafish are remarkably regenerative and able to repair cardiac, retinal, spinal cord and pancreatic damage among other tissues (Poss and Tanaka, 2024). We therefore anticipated that foxf2a mutants might be able to regenerate new pericytes to replace lost pericytes. This was tested using genetic ablation of pericytes. Ablation is very efficient in mutants and wildtype. Surprisingly, 7 days after ablation, foxf2a mutants show recovery of pericyte numbers. Thus, when placed under extreme stress, embryonic foxf2a mutants can regenerate pericytes; yet, under baseline conditions, foxf2a mutant pericytes do not replenish. This experiment is important to distinguish an underlying mechanism of the foxf2a pericyte phenotype. Our data suggests that the earliest and most important difference in foxf2a mutant pericytes is their low initial number. Even though some repair is possible, it is not enough to compensate. Reduced pericyte density is associated with elongated processes, enlarged soma, and degeneration in mutants, which are likely signs of cellular stress. Over a lifetime, the deficiency leads to progressive damage to the vascular network. Determining that there is a critical embryonic window for developing a robust pericyte precursor pool helps us focus on early interventions to mitigate these deficits before damage over the lifespan worsens. The necessity of Foxf2 during development for the establishment of proper brain vasculature was also seen in a mouse Foxf2 knockout (Reyahi et al., 2015). Further research into the early role of foxf2a in establishing initial pericyte specification/differentiation, and the impact of early vascular defects on disease progression, will be crucial for developing strategies to prevent and treat cerebrovascular conditions.

Materials and Methods

Zebrafish Husbandry and Strains

All procedures were conducted in compliance with the Canadian Council on Animal Care, and ethics approval was granted by the University of Calgary Animal Care Committee (AC21-0169). Embryos were maintained in E3 medium (5mM NaCl, 0.17mM KCl, 0.33mM CaCl2, 0.33mM MgSO4, pH = 7.2) (Westerfield, 2007) at 28°C. Larvae up to 10 dpf were maintained in an incubator with a light cycle (14 hours light, 10 hours darkness), with daily water changes and feeding. As zebrafish sex is not determined until 28 dpf, sex is not considered in embryo and larval experiments. For adult experiments, results were compared by sex where sufficient n’s were available. Results from the progeny of het in-crosses were blinded in that they were quantified before genotyping. All reagents and strains are listed in Table 1 (Key Resources Table). All experiments included wildtype as a comparison group, and developmental stages, n’s, and genotypes are noted for each experiment (Table 2).

Table 1:

Table 2.

Genotyping

Adult fish were anesthetized in 0.4% Tricaine (Sigma) and placed on a cutting surface. A small portion of the tip of the fin was clipped using a razor blade, and the fish were returned to the system to recover. For developmental DNA isolation, whole embryos, or larvae (up to 10 dpf) were anesthetized in 0.4% Tricaine prior to sampling.

Genomic DNA (gDNA) was extracted using the HotSHOT DNA isolation protocol, adapted from Kosuta et al., 2018. To isolate gDNA, tissue or embryo was placed in 50 μL of Base Solution and incubated at 95°C for 30 minutes. 5 μL of Neutralization Solution was added to neutralize the reaction. Samples were spun down in a mini centrifuge, and DNA was sampled from the top portion of the solution to avoid undigested samples.

Zebrafish were genotyped for target genes or the presence of transgenes using the KAPA2G Fast HotStart Genotyping Mix as per the manufacturer’s instructions. foxf2 mutants were genotyped using foxf2a primers (Table 1). A custom TaqMan probe for foxf2b^ca22 (ANAACEC) was obtained from Applied Biosystems. Samples were genotyped using the QuantStudio 6 Flex Real-Time PCR System (Applied Biosystems) with the wildtype allele reported in FAM and the mutant allele reported in VIC.

In Situ Hybridization

Custom probes for foxf2a (PRD069), nduf4al2 (RTD146), kdrl (PRI089), pdgfrβ (PRA654), and foxf2b (PRF462) were obtained from Molecular Instruments (Los Angeles, CA) for Hybridization Chain Reaction (HCR) in situ hybridization. In-situ staining was performed according to the manufacturer’s instructions. Samples were permeabilized with proteinase K (1 mg/mL stock; Invitrogen, 4333793) at various times and concentrations, depending on their age.

Integrated Intensity

pdgfrβ and foxf2b intensity was obtained using ImageJ from in-situ stained embryos at 3dpf. Mean fluorescent intensity was measured using the freehand selection tool to circle each pericyte. The background mean intensity was subtracted from the pericyte mean intensity to standardize each measurement, and the intensities of each fish were averaged. The same protocol was used to measure the intensity of the pdgfrβ transgene in 3dpf live embryos.

Total RNA extraction and cDNA synthesis

Total RNA was extracted using the Research RNA Clean & Concentrator-5 Kit (Zymo Research, R1015) with some modifications. 50 dechorionated embryos were collected at 48 hpf, and homogenized in TRIZOL (Ambion, 15596026) reagent with a syringe needle. Briefly, samples were centrifuged at 16,000 rpm for 2 minutes to remove pigment. After separating the solution from the precipitate, chloroform was added, and the mixture was centrifuged at 12,000 rpm for 15 minutes at 4°C. The aqueous layer containing nucleic acids was removed, and an equal volume of 95-100% ethanol was added, then mixed well. The rest of the protocol followed the manufacturer’s instructions using the Zymo-spin IC column. After eluting RNA with RNase/DNase-free water, samples were quantified using a Nanodrop to evaluate RNA quality. For cDNA synthesis, qSCRIPT cDNA Supermix (QuantaBio, 95048-100) was used according to the manufacturer’s instructions and stored at -20°C.

RT-qPCR

To assess relative gene expression, real-time quantitative PCR (RT-qPCR) was carried out using custom-designed primers. PowerUP SYBR Green Master Mix (Applied Biosystems, A25742) was utilized on a QuantStudio 6 Flex Real-Time PCR System (Applied Biosystems). Cycle conditions followed the protocol of a PowerUP SYBR Green standard reaction: 50°C for 2 minutes, 95°C for 2 minutes, and 40 cycles of 95°C for 15 seconds and 60°C for 60 seconds. To calculate the fold change in gene expression, mutants were compared to wildtype using ΔΔCT calculations (Livak and Schmittgen, 2001).

Brain Dissection

Zebrafish were first euthanized in Tricaine and mounted on a Sylgard gel plate with dissection pins to stabilize them. Scissors were used to sever the brain stem entirely at the base of the head, posterior to the hindbrain. The eyes were removed with forceps/micro-scissors, and the optic stalks were cut. An incision was then made at the posterior portion of the skull plate anteriorly. Forceps were used to pull back the skull plate and orbital bones. Any nerve connections were severed, and the brains were removed and immediately fixed in ice-cold 4% PFA overnight at 4°C.

Brain Tissue Clearing

Whole zebrafish brains were cleared by either CUBIC (Susaki et al., 2015), or a modified iDISCO+ (Renier et al., 2016) protocol. For CUBIC, samples were incubated in 1:1 ratio of Reagent 1 (25% Quadrol (Sigma, 122262), 25% Urea, and 15% TritonX-100):H2O overnight at 37°C. Next, brains were incubated in 100% Reagent 1 at 37°C until the tissue was transparent. Finally, brains were rinsed in PBS, mounted in 2% low melting point agarose and re-placed in 100% Reagent 1 at 37°C until the tissue was transparent.

For iDISCO, the samples were first dehydrated in methanol:H₂O dilution series for 30 minutes each, then chilled at 4°C. Next, samples were incubated overnight in a 1:3 ratio of dichloromethane (DCM; Sigma, 270997):methanol at room temperature. The following day, samples were washed in methanol and then chilled at 4°C before bleaching in fresh 5% H₂O₂ in methanol overnight at 4°C. Then, samples were rehydrated in methanol:H2O dilution series for 30 minutes each, then washed in PTx.2 (0.2% TritonX-100 in PBS) twice over two hours at room temperature. Next, samples were incubated in permeabilization solution (0.3 M glycine, 20% DMSO in 400 mL PTx.2) at 37°C for up to two days, after which the samples were rinsed in PBS twice over one hour. Samples were then washed three times over two hours in 0.5 mM SDS/PBS at 37°C for three days before being incubated in primary antibody (Table 1) in 0.5 mM SDS/PBS at 37°C for two more days. The primary antibodies were refreshed in PTx.2 and left for another 4 days before overnight washing in PTwH (10 μg/ml heparin and 0.2% Tween-20 in PBS) at 37°C. Samples were left in secondary antibodies (Table 1) in PTwH/3% NSS at 37°C for 3 days, refreshed and left for another 3 days before washing overnight in PTwH. After immunolabeling, brains were mounted in 2% low melting agarose (LMA) and dehydrated in methanol:H₂O dilution series for 30 minutes each and left overnight at room temperature. Next, brains were incubated in a 1:3 ratio of DCM: methanol at room temperature for three hours before washing in 100% DCM for 15 minutes twice. Finally, brains were incubated in Ethyl cinnamate (Sigma, NSC6773) for two hours before replacing the solution and incubating overnight at room temperature.

Tissue Sectioning

For vibratome sections, brains were fixed in 4% PFA, washed and mounted in 4% LMA (Invitrogen, 16520-050) before sectioning with a vibratome (Leica, VT1000S) to obtain a series of transverse sections at 50 µm.

Cre mRNA Injections

Wildtype fish on a Zebrabow transgenic background (Pan et al., 2013) were injected with 1pg of Cre mRNA at the one-cell stage. Embryos were kept in E3 at 28°C until imaging.

Immunofluorescence

Sections for immunofluorescence were briefly washed in PBS, followed by 2% TritonX-100 in PBS for permeabilization. Sections were moved from the permeabilization solution into blocking buffer (5% Goat serum, 3% BSA, 0.2% TritonX-100 in PBS) for 1 hour before incubation in primary antibody. Sections were then washed and left in secondary antibodies for 3 hours at room temperature. After incubation, sections were washed, mounted and cover slipped with Fluoromount-G™ Mounting Medium, with DAPI or without counterstain.

Drug Treatments

Embryos were dechorionated prior to treatment, and all drug treatments included DMSO at an equivalent concentration to the drug solution as a control (Sigma, D8418). Treatments were performed in a 24-well plate with approximately 15 embryos per well. For pericyte ablation experiments, 5 mM Metronidazole (MTZ; Sigma, M3761) was applied at 50 hpf for 1 hour with a DMSO control.

Microscopy

Imaging fluorescent transgene expression in live embryos, antibody staining, and fluorescent staining were completed using an inverted laser scanning confocal microscope (LSM900; Zeiss) with a 10X (0.25 NA), 20X (0.8 NA), 40X water (1.1 NA) or 60X (1.4 NA) oil objectives. Laser wavelengths included blue (405 nm), green (488 nm), red (561 nm), and far red (640 nm).

Embryos were maintained in Phenylthiourea (PTU) from 24 hpf onwards to prevent pigment development, anesthetized in 0.4% tricaine and mounted in 0.8% LMA dissolved in E3, on a clear imaging dish. In some cases, live samples were retrieved from the agarose following imaging for further imaging at later time points, genotyping, or other data collection. Images were processed using Zen Blue and Fiji (Schindelin et al., 2012) software.

Imaging of whole adult brains was completed using the Light-sheet microscope (Ultramicroscope II equipped with a SuperPlan module; Miltenyi Biotec) with a 4X (0.35 NA) objective at 1.0 zoom. Laser wavelengths included green (488 nm) and red (561 nm). Images were processed using Zen Blue, Fiji and Imaris software.

Vessel network quantitation

Confocal images from embryos were analyzed using the Python-based software tool Vessel Metrics (McGarry et al., 2024). The total vessel network length was measured starting from below the middle cerebral vein and dorsal longitudinal vein until the emergence of the basal communicating artery. Vessels included in measurements are as follows: middle mesencephalic central arteries, posterior mesencephalic central arteries, the primordial hindbrain channels and the posterior region of the basilar artery. The forebrain vessels (i.e. anterior cerebral veins) were not included. Blood vessel diameter was restricted to the horizontal hindbrain central arteries, which were comparable between all images.

Adult vessel network length was measured using Ilastik (Berg et al., 2019) and Imaris (version 10.2.0; Bitplane AG, Oxford Instruments) software. Vessels were annotated in Ilastik to obtain a probability map. The probability map was imported into Imaris as a new channel for each image. The surface tool with machine learning was then used to annotate the vessels a second time, using the Ilastik map as the training guide. A mask was created from this surface. The filament tool was then used to create a 3D network of the vessel, using the surface mask as a training guide. The total network length and the diameter of each vessel segment were exported from the filament statistics tab.

Smooth muscle cell coverage was measured using the same pipeline; however, the raw channel was used for the surface tool annotation without using Ilastik. Total network length was exported from the filament statistics tab.

Pericyte quantitation

Embryonic pericytes were manually counted using the Fiji counting or tracing tool on flattened Z-stacks from confocal images of the whole zebrafish head. Pericyte counting and analysis were restricted to the mid and hindbrain regions for all metrics. For pericyte process lengths, individual processes extending from a single soma were measured and summed to determine the total process length per pericyte (µm). In instances where two processes appeared to merge or cover the same blood vessel, half of the total length was added to each pericyte. For Zebrabow images, only pericytes with processes distinguishable from neighbouring pericytes were measured. For soma size, individual cell bodies were traced, and the area was determined by the software in µm².

Adult pericytes were counted using the spot tool in Imaris. Using the raw channel, the spot tool was trained to identify pericyte cell bodies. The total spot number, equivalent to the number of pericyte cell bodies, was exported from the spot statistics tab. A mask of the spot tool was created to better visualize pericytes.

Statistics

All statistical analyses were performed using GraphPad Prism 10, with significance determined by p < 0.05. Statistical tests conducted are included in figure captions. If no significance is indicated, it is not significant. The D’Agostino-Pearson test was used to assess the normality of data sets, and in cases where the data did not pass the normality test, non-parametric statistics were used. Experimental N’s are reported in Table 2. Results are expressed on graphs as mean ± standard deviation (SD). Only significant p-values are indicated.

Supplementary figures

Supplemental Figure 1.

Supplemental Figure 2.

Supplemental Figure 3.

Supplemental Figure 4:

Supplemental Figure 5.

Supplemental Figure 6.

Supplemental Figure 7.

Supplemental Figure 8.

Supplemental Figure 9.

Supplemental Figure 10.

Acknowledgements

This work was funded by the Canadian Institutes of Health Research PJT-183631. MFG received a Canada Graduate Scholarship Master’s from Canadian Institutes of Health Research, the Alberta Graduate Excellence Scholarship (AGES) for Master’s Research from the Province of Alberta, and a Biochemistry and Molecular Biology Department scholarship from the University of Calgary. EH received the Alvin Libin Graduate Scholarship in Cardiovascular Research, the Alberta Children’s Hospital Research Institute (ACHRI) Graduate Scholarship for Master’s Research, the ACHRI Graduate Scholarship for Doctoral Research, and the Alberta Graduate Excellence Scholarship (AGES) for Doctoral Research from the Province of Alberta. We acknowledge the Alberta Children’s Hospital Research Institute and Hotchkiss Brain Institute Imaging facilities for microscopes and technical support.

Additional information

Data availability

Table 2 contains the numerical data for all experiments.

Funding

Canadian Institutes of Health Research (CIHR) (PJT-183631)

• Sarah J Childs

Additional files

Movie 1. Rotating view of a cleared wildtype brain at 3 mpf with pdgfrβ (blue) and kdrl (red).

Movie 2. Rotating view of a cleared foxf2a mutant brain at 3 mpf with pdgfrβ (blue) and kdrl (red).

Movie 3. Rotating view of a cleared wildtype brain at 6 mpf with acta2 (blue) and kdrl (red).

Movie 4. Rotating view of a cleared foxf2a mutant brain at 6 mpf with acta2 (blue) and kdrl (red).