MicroRNA26 attenuates vascular smooth muscle maturation via endothelial BMP signalling

As small regulatory transcripts, microRNAs (miRs) act as genetic ‘fine tuners’ of posttranscriptional events, and as genetic switches to promote phenotypic switching. The miR miR26a targets the BMP signalling effector, smad1. We show that loss of miR26a leads to hemorrhage (a loss of vascular stability) in vivo, suggesting altered vascular differentiation. Reduction in miR26a levels increases smad1 mRNA and phospho-Smad1 (pSmad1) levels. We show that increasing BMP signalling by overexpression of smad1 also leads to hemorrhage. Normalization of Smad1 levels through double knockdown of miR26a and smad1 rescues hemorrhage, suggesting a direct relationship between miR26a, smad1 and vascular stability. Using an in vivo BMP genetic reporter and pSmad1 staining, we show that the effect of miR26a on smooth muscle differentiation is non-autonomous; BMP signalling is active in embryonic endothelial cells, but not in smooth muscle cells. Nonetheless, increased BMP signalling due to loss of miR26a results in an increase in acta2-expressing smooth muscle cell numbers and promotes a differentiated smooth muscle morphology. Similarly, forced expression of smad1 in endothelial cells leads to an increase in smooth muscle cell number and coverage. Furthermore, smooth muscle phenotypes caused by inhibition of the BMP pathway are rescued by loss of miR26a. Taken together, our data suggest that miR26a modulates BMP signalling in endothelial cells and indirectly promotes a differentiated smooth muscle phenotype. Our data highlights how crosstalk from BMP-responsive endothelium to smooth muscle is important for smooth muscle differentiation.

Introduction Vascular smooth muscle cells (vSMCs) provide structural integrity to the vessel wall. Guided control of signalling cascades, including Platelet derived growth factor (Pdgf), Notch, and Transforming Growth Factor-β/Bone morphogenic Protein (TGF-β/BMP) recruits and induces differentiation of perivascular mural cells (vSMCs and pericytes) to create a two-layered vessel wall with an internal endothelial cell lining and a muscle cell covering [1][2][3]. Once the vSMCs surround the vessel, they begin depositing extracellular matrix (ECM) proteins Laminin, Collagen IV and Fibulins to support the vessel wall [4]. vSMCs then take on a mature phenotype that stabilizes the underlying endothelial cells through induction of quiescence, expression of junctional and attachment proteins, and expression of contractile proteins to provide myogenic tone [2,[5][6][7].
vSMCs maintain phenotypic plasticity and can undergo a phenotypic switch from a quiescent contractile state to a proliferative synthetic state in response to cellular stimuli [4,8]. Contractile vSMCs are defined by an elongated and thin 'spindle-shaped' morphology and low rates of proliferation. The expression of key differentiation markers such as smooth muscle (α)-actin (Acta2), smooth muscle β-myosin heavy chain (Myh11), and transgelin (Sm22α) allows vSMCs to perform their contractile function and provide vascular tone. In contrast, the immature synthetic vSMCs have reduced expression of contractile genes, produce ECM proteins, are highly proliferative, and have a rhomboid or rounded morphology [9][10][11][12].
Numerous studies have demonstrated that BMP signaling through Smad1 modulate vSMC plasticity (reviewed by [13]). Defective BMP signalling can affect both endothelial and vSMC cells [14][15][16][17][18][19][20][21]. Aberrant vSMCs phenotype switching plays a critical role in the pathogenesis of vascular diseases such as hereditary hemorrhagic telangiectasia (HHT) and pulmonary arterial hypertension (PAH). In canonical Smad-mediated BMP signaling, Smad1 is phosphorylated by the serine-threonine kinase activity of a type 1 BMP receptor (ACVRL1 (ALK1)/ BMPR1A, BMPR1B) allowing it to associate and dimerize with the co-mediator Smad4 and translocate to the nucleus to control gene transcription. Murine homozygous null mutants for BMPR-1a (Activin like kinase 3, ALK3) or the type II receptor BMPR-2 (which is mutated in human patients with PAH) [22,23] and their ligand Bmp4 or downstream co-Smad4 are embryonic lethal, and present with vascular deformities attributable to a loss of Smad1 mediated signalling [24]. Mutations in ALK1 lead to HHT2, a disease characterized by arteriovenous malformations (AVMs) [25]. Deletion of Alk1 in mice leads to cranial hemorrhages, AVM-like fusion of micro-vessel plexi, dilation of large vessels and reduced coverage of vessels by vSMCs [26]. In zebrafish, disruption of Alk1 signalling results in pathological arterial enlargement and maladaptive responses to blood flow that generate AVMs. Potential vSMCs defects in this model have not been assessed [27].
Here, we investigate the role of microRNA26a (miR26a) in regulating vSMC dynamics using the zebrafish model of vessel stabilization. miR26a regulates proliferation, migration and differentiation of vSMCs and has been shown to target smad1, a key intracellular mediator of BMP signalling, in cultured vSMCs in vitro [42][43][44][45].
miR26a expression is altered during abdominal aortic aneurysm (AAA) and neointimal lesion formation [43,45]. However, the role of miR26a in vivo in an intact animal in the context of developing vSMCs are largely unknown. Using a combination of genetic gain and loss of function methods to understand the role of miR26a in vivo, we show that miR26a acts within a BMP responsive pathway to fine tune vSMC maturation via targeting smad1. Interestingly, we find that active BMP signalling and changes in Smad1 activation are observed within endothelium in vivo, and not in smooth muscle cells. Together the evidence suggests that miR26a plays a role in regulating blood vessel stabilization via a non-autonomous mechanism.
Results miR26a is expressed in developing blood vessels miR26a targets smad1 and thereby directly regulates BMP signalling ( [36,43] Fig 1A). To observe the spatial gene expression pattern of miR26a in developing embryos we used in situ hybridization. At 48 hpf, miR26a has a ubiquitous expression pattern ( Fig 1B and 1B'), however by 4 dpf expression becomes enriched in the ventral head of the embryo, with strong expression in the pharyngeal region, bulbus arteriosus and ventral aorta ( Fig 1C). miR26a is expressed in and around the blood vessel endothelium where it could potentially play a role modulating BMP signalling in blood vessels (compare to kdrl: GFP stain; Fig 1C', inset). In order to further analyze the cell specific expression of miR26a, we used fluorescent-activated cell sorting (FACS) to isolate EGFP +ve vSMCs and mCherry +ve endothelial cells from 4 dpf Tg (acta2:EGFP;kdrl:mCherry) embryos. In keeping with the in-situ hybridization data, RT-qPCR showed that miR26a is indeed expressed in both cell types, although it is not significantly enriched in endothelial cells (S1 Fig). FACS sorting efficiently separated vSMCs and endothelial cells; we find that acta2: EGFP +ve vSMCs cells have an average 37.4-fold enrichment in acta2 expression, and minimal expression of alk1 or smad1 when compared to kdrl: mCherry endothelial cells. However, smad1 is 14-fold enriched and acvrl1 is 3.5-fold enriched in mCherry +ve endothelial cells while there is nearly no expression of acta2 (S1 Fig). Thus, a miR26a target, smad1 is enriched in endothelial cells.

BMP signalling is active in the aorta endothelium
We next tested the relationship between mir26a expression and activated BMP signaling using an in vivo reporter of Smad1/5 activity. Tg(BRE:EGFP] transgenic fish encode EGFP driven by an upstream Bmp Response Element (BRE) that contains multiple short Smad-binding sites from the id1 promoter, a major transcriptional target of canonical Bmp/Smad1 signaling [46,47]. We crossed Tg(BRE:EGFP] to endothelial Tg(kdrl:mCherry) or vSMC Tg(acta2: mCherry) lines to observe BMP activation in endothelial and vSMCs, respectively (Fig 1D and  1G). We use the 4 dpf time point as vSMC cells first differentiate and begin to express the mature marker acta2 between 3 and 4 dpf [7,48]. Surprisingly, although miR26a has been implicated in controlling Smad1 regulated vSMC dynamics directly, we found that transgenic BRE:EGFP signals are restricted to the endothelium of the vessel wall and have co-localized expression with kdrl:mCherry ( Fig 1E, 1E', 1F and 1F'). The acta2:mCherry-positive vSMCs lie directly adjacent to BRE:EGFP-expressing cells, with no detectable expression of BRE:EGFP in vSMCs on the ventral aorta or in pharyngeal aortic arch arteries (Fig 1G, 1G" and 1H, both ventral and lateral projections are shown). Similarly, acta2:mCherry-positive cells are closely dpf miR26a is expressed in the pharyngeal arches, bulbous arteriosus and ventral aorta. C') Cross section of the head showing miR26a expression in blood vessels (purple; punctate stain) compared with endothelial stain (brown; kdrl:GFP transgenic). Inset is an enlargement of image in C'. D) Ventral view of the pharyngeal region of a 4 dpf double transgenic Tg(BRE:EGFP);Tg(kdrl:mCherry) embryo shows BRE:EGFP (green) expression within endothelial cells in aortic arches (red, white arrowheads in E'-E"') and ventral aorta (red, white arrowheads F'-F"'). G-H) Ventral and lateral views of a 4 dpf double transgenic Tg(BRE:EGFP); Tg(acta2:mCherry) zebrafish shows that acta2 positive cells are in direct contact with BMP-responsive endothelial cells but do not express BRE:EGFP. Scale bar represents 50μm. https://doi.org/10.1371/journal.pgen.1008163.g001 MicroRNA26 attenuates vascular smooth muscle maturation via endothelial BMP signalling associated with pSmad1-positive endothelial cells but do not show pSmad1 staining (S1D Fig). Together, our data suggest that in early development, miR26a and smad1 are expressed within endothelial cells where BMP signaling is also active, as visualized by two methods of detection.

Loss of miR26a leads to upregulation of smad1 mRNA
The highly conserved miR-26 family constitutes miR26a-1, miR26a-2, miR26a-3 and miR26b [44] as identified by their seed sequences and accessory sequence. In zebrafish and humans, miR26a-1, miR26a-2 and miR26a-3 have the same mature sequence, and only differ from the mature miR-26b sequence by two nucleotides [42,49]. To investigate the role of miR26a in vascular development in vivo, we knocked down miR26a using an antisense morpholino that targets the mature miRNA seed sequence of all three miR26a isoforms. A 6bp mismatch scrambled control morpholino was used as a control. 1 ng doses of morpholino were used, as suggested by current guidelines [50,51]. In parallel, we designed a second genetic knockdown approach using CRISPR interference (CRISPRi) [52] to target the pri-miR hairpin structure using the complementary sequence to the mature miRNA (Fig 2A). RT-qPCR shows a 26% (0.74±0.65) reduction in miR26a following miR26a MO knockdown and 34% (0.65±0.10) reduction of miR26a using CRISPRi (Fig 2B and 2C) confirming that both knockdown methods result in decreased miR26a expression. smad1 is a demonstrated target of miR26a in vitro [36,43]. In support of smad1 being a miR26a target in vivo, miR26a knockdown results in increased smad1 expression in 48 hpf and 4 dpf injected embryos as compared to controls by in situ hybridization ( Fig 2D) and resulted in an average 1.6-fold and 1.8-fold increase by RT-qPCR, respectively (Fig 2E and 2F).To further determine whether miR26a can regulate smad1 expression in vivo, we designed a sensor assay and fused the smad1 3 0 UTR to EGFP (EGFP: smad1pA). This was co-injected with an internal mCherry control into single-cell zebrafish embryos in the presence or absence of a miR26a morpholino. When fluorescence levels were examined at 24 hpf, injections of EGFP: smad1pA sensor mRNA alone resulted in EGFP expression; however, this fluorescence was enhanced by over 65% by co-injection of miR26a morpholino (S2A and S2B Fig).
At 4 dpf, upregulation of smad1 in miR26a morphants and CRISPRi knockdown embryos is more prominent in the ventral pharyngeal region, with staining in the ventral aorta, aortic arches and bulbous arteriosus ( Fig 2D, highlighted areas), similar to where miR26a is expressed most strongly (Fig 1B and 1C). In a complementary approach, we injected a miR26a mimic to overexpress miR26a and observed an increase in miR26a expression by RT-qPCR (S2C Fig

Loss of miR26a leads to increased phosphorylated Smad1 in endothelium
We next tested whether the increased expression of smad1 mRNA in miR26a knockdown embryos leads to enhanced Smad1 phosphorylation. Wildtype immunostaining showed pSmad1/5/9 is high in endothelium but not in vSMCs (S1D-S1D" Fig).

Increased levels of smad1 lead to vascular stability defects
Loss of miR26a leads to compromised vessel integrity at 48 hpf. miR26a morphants have an average 13±2% hemorrhage (Fig 4B and 4D) and CRISPRi embryos have an average 15±1% hemorrhage (Fig 4C and 4D) as compared to 2-3% rate of the controls. The phenotype is dose-dependent as higher doses of morpholino lead to an increase in hemorrhage to 40% and a 1.  MicroRNA26 attenuates vascular smooth muscle maturation via endothelial BMP signalling MicroRNA26 attenuates vascular smooth muscle maturation via endothelial BMP signalling not been previously connected to vascular stability defects, we next tested whether overexpressed smad1 could lead to hemorrhage. Injection of smad1 mRNA into single cell stage embryos resulted in significantly higher hemorrhage rate of 12±0.9% in injected embryos as compared to uninjected controls ( Fig 4E, 4F and 4I).
Further, as miR26a knockdown leads to increased smad1 levels, we predicted that reduction in smad1 would rescue hemorrhage in miR26a knockdown embryos. Double knockdown by co-injection of smad1 [53] and miR26a morpholinos reduced hemorrhage rates to below 5 ±0.8% (Fig 4H, top embryo and Fig 4I). Of note, smad1 MO alone did not result in hemorrhage; however it did result in a range of phenotypes associated with smad1 knockdown including dorsal-ventral axis defects and hydrocephalus as previously reported [53]. smad1 knockdown led to an average a 77±8.6% of embryos with a mild (V1) ventralized defect and 12 ±17% with a more severe (V2) phenotype (Fig 4G-4I). Both defects were reduced in double knockdown embryos (Fig 4G and 4H, bottom). Thus, reducing miR26a or increasing smad1 in vivo leads to leads to a loss of vascular stability and hemorrhage.

Loss of miR26a leads to increased numbers of acta2-positive vSMCs and upregulation of pathways involved in endothelial-vSMC crosstalk
To demonstrate functional consequences of increased endothelial BMP signalling on vSMCs, we next investigated vSMC investment on the ventral aorta and pharyngeal arch arteries of Tg (BRE:EGFP;acta2:mCherry) embryos in miR26a knockdown embryos. This assay allowed us to make three key observations. Firstly, BRE:EGFP signal intensity is enhanced in miR26a morphants (Fig 5A', 5B' and 5C), which correlates with the increased pSmad1 staining we observed in endothelial nuclei of knockdown embryos (Fig 3). Secondly, the number of acta2:mCherry positive cells along the ventral aorta and pharyngeal arch arteries (PAA) is increased in miR26a knockdown embryos (33.8±1.6 in controls vs 47.2 ±2.2 in miR26a knockdown, Fig  5A", 5B" and 5D). Thirdly, the increase in acta2 positive cell number is accompanied by a change in cell morphology in miR26a knockdown embryos (Fig 5E-5G). In control embryos, acta2 positive vSMCs have a rounded, punctate morphology, and 'sit' high on the vessel wall with an average height of 4.5±0.4 μm, above the underlying endothelium. In miR26a morphants, vSMCs are have a significantly reduced vSMC height of 3.4 ± 0.1μm, and appear flatter and more closely apposed to the endothelium when compared to control embryo vSMCs. These data suggest that loss of miR26a results in increased vSMC coverage along blood vessels and a shift to a differentiated morphology.
In parallel, we quantitated gene expression for vSMC differentiation genes. RT-qPCR using isolated embryonic head mRNA at 4 dpf showed a 1.7-fold increase in acta2 and 1.8-fold increase in myh11a mRNA in miR26a morphants ( Fig 5H). Further, using in situ hybridization, we found that miR26a morphants have increased expression of acta2 and myh11a in the pharyngeal region (S5 Fig), similar to the location of increased smad1 staining (Fig 2D). Conversely by 4 dpf, miR26a mimic injected embryos had reduction in acta2 and sm22 expression by in situ hybridization (S5 Fig). The Bmp/ Notch3/ Pdgf signalling axis is an important regulator of vSMC proliferation and subsequent differentiation. We found that the vSMC notch receptor notch3 has a 2.0-fold increase in miR26a morphants ( Fig 5H). Furthermore, the endothelial expressed ligand pdgfba and its mural cell receptor pdgfrβ, had a 3.6 and 2.0-fold increase, respectively, in miR26a morphants as compared to controls. This suggests that increased vSMC numbers could potentially arise from enhanced proliferation via activation of the Pdgfrβ pathway downstream of Smad1 activation. Increases in acta2, myh11 and notch3 may therefore reflect increased cell numbers in addition to increased vSMC differentiation.

Endothelial overexpression of smad1 promotes vSMC differentiation
To demonstrate that smad1 expression in endothelial cells promotes vSMC differentiation, we expressed smad1 under an endothelial promoter in a transposon vector (TolCG2:kdrl:smad1, hereafter smad1 ECOE ; Fig 6A). The vector and transposase or transposase alone control were injected into Tg(BRE:EGFP;acta2:mCherry) embryos and scored at 48 hpf and 4 dpf. At 48 hpf, 10% of smad1 ECOE embryos hemorrhage, similar to the increased hemorrhage observed in mirR26 knockdown and global smad1 mRNA overexpression (Fig 4). Higher doses of the vector result in significant cranial and pericardial edemas (S6A and S6B Fig). At 4 dpf, RT-qPCR of smad1 ECOE embryos shows a 1.9-fold increase in smad1 and 1.4-fold increase in the BMP responsive gene id1 expression as compared to control (Fig 6B). Similarly, BRE:EGFP fluorescence is also increased in smad1 ECOE embryos by 30% as compared to controls (Fig 6C", 6D" and 6F). Together, the data show that activation of Smad1 was significantly increased in smad1 of vSMC height (yellow axis) from the endothelium (white dashed line). Representative images of ventral aorta (from insets), Scr. Control (E) and miR26a morphants (F). G) Quantification of average vessel heights along the length of the VA (N = 3, miR26a MO n = 18, Scr. Control n = 13, Student's two-tailed t-test, ���� p< 0.0001 as compared to control, error bars = SEM). H) RT-qPCR quantification of vSMC differentiation genes in injected controls and miR26a morphants (n = 3). RT-qPCR data show the mean ± SEM, Student's two-tailed t-test � p < 0.05, n, number of biological replicates.
https://doi.org/10.1371/journal.pgen.1008163.g005 The total number of acta2:mCherry positive vSMCs along the ventral aorta in smad1 ECOE embryos was significantly increased from 15±0.6 in controls to an average 24±1.4 cells (Fig  6C', 6D' and 6E). The percent vSMC coverage of the ventral aorta is also increased by 20% with an average of 80 ± 2.05% in smad1 ECOE as opposed to 65.3±2.1% in controls (Fig 6C', 6D' and 6H). Our data suggests that upregulation of smad1 in endothelial cells is sufficient to increase vSMC number and coverage of the ventral aorta of 4 dpf embryos.

BMP inhibition rescues the effect of on vSMC differentiation after miR26a knockdown
Our results showed that miR26a knockdown leads to an increased number of acta2 positive vSMCs on the ventral aorta and upregulation of Smad1 activation in the endothelium. To further investigate the interplay between BMP signalling in endothelial cells and vSMC differentiation, we tested whether the increase in vSMC number and differentiation after loss of miR26a could be reversed by blocking endothelial BMP signalling. K02288 is a selective and potent small molecule inhibitor of BMP signalling that blocks Smad1 phosphorylation by type I receptor Activin like kinase 1 (Alk1) and Alk2 [54,55]. We show that alk1 expression is enriched in endothelial cells at this developmental stage, but not vSMCs (S1C Fig). We selected a time point for drug application when the endothelium of the major blood vessels is patterned [56], but when vSMC coverage of the ventral aorta and PAA is only starting [7]. Tg(acta2: EGFP;kdrl:mCherry) embryos were treated with 15μM K02288 from 52 hpf to 4 dpf. As expected, miR26a morphant and miR26a CRISPRi treated embryos have significantly more vSMCs than wildtype embryos (Fig 7B, 7B', 7C, 7C' and 7G). Wildtype embryos treated with K02288 show a 62% reduction in the total average number acta2:EGFP positive cells compared to vehicle control alone (Fig 7A', 7D' and 7G. 29±1 to 11±3). In miR26a knockdown embryos ( Fig 7B' and 7C'), the effects of K02288 were rescued; miR26a morphants had a non-significant 17% reduction in vSMC numbers (35±5 to 29±5) and miR26a CRISPRi embryos had a nonsignificant reduction from 39±1 to 34±3 (Fig 7E', 7F' and 7G). We also found that BMP inhibition not only affects vSMC number, but also reduces ventral aorta length by 55% in K02288 treated wildtype embryos, from an average 159±2.08 μm to 77±19.5 μm (Fig 7A, 7D and 7H). However, miR26a morphants treated with K02288 are rescued and have a ventral aorta length not significantly different than wildtype. miR26a CRISPRi embryos showed a smaller rescue and had a 20% decrease in length when treated (Fig 7B, 7C, 7E and 7F, 177±8.6 to 136±5.6). Of note, there was no statistical difference in ventral aorta length between miR26a knockdown and control embryos, which supports our finding that endothelial cell number is not affected by loss of miR26a.
We next tested whether pSmad1 levels are rescued in K02288 treated miR26a knockdown embryos as compared to controls (Fig 7I-7L). Using the endothelial nuclear marker fli1a: nEGFP, we confirmed that there was no significant difference in endothelial cell number between untreated control and miR26a morphants (Fig 7M). In control embryos, treatment with K02288 (Fig 7I, 7I', 7J and 7J') significantly reduced the number fli1a:nEGFP positive cells to 16± 0.2, which is 20% less than controls. Similarly, the proportion of pSmad1 positive/ nEGFP nuclei also decreased from 43±1% to 28±2% (Fig 7N). Although K02288 treated miR26a morphants have a slight reduction in the total number fli1a:nEGFP positive cells ( Fig  7M), there is no significant decrease in the proportion of pSmad1 positive/fli1a:nEGFP nuclei (Fig 7K, 7K', 7L, 7L' and 7N), and they remain similar to untreated controls. Taken together, our results further suggest that the endothelial cell is a critical site of Smad1-mediated BMP signalling and blocking its activation can significantly affect vSMC coverage. Loss of miR26a is able to rescue these defects to maintain both endothelial signalling and vSMC coverage.

Discussion
Compromised structural vascular integrity, vessel weakening and rupture (hemorrhage) can result from aberrant BMP signalling [57][58][59]. Hemorrhage ultimately results from weak endothelial junctions, however defects in mural cell coverage (attachment and ECM secretion) are implicated in the pathological progression of vascular diseases. We show that the endothelium of the ventral aorta in zebrafish has activated pSmad1 at 4 dpf, but that pSmad1 is not detectable in mural cells. At a stage when mature vSMC are normally present, embryos with loss of miR26a have upregulation of pSmad1, increased vSMC coverage and a change in vSMC morphology, with no observable changes in the number or morphology of the pSmad1-expressing endothelial cells. We show that inhibition of BMP signalling reduces both vSMC coverage and the length of the ventral aorta while dual miR26a knockdown and BMP receptor inhibition leads to a rescue such that animals maintain normal vSMC number, length of the ventral aorta, and vSMC coverage. We therefore suggest that miR26a modulates BMP signalling in endothelial cells to control vSMC differentiation via a paracrine mechanism potentially involving Notch and/or Pdgfrβ signalling. We propose that miR26a therefore functions in vivo to fine tune endothelial signals to the vSMCs (Fig 8).
Studies in cultured vSMC have suggested that miR26a controls Smad1-mediated BMP signalling within vSMCs to modulate their phenotype [28]. However, these studies do not address whether the levels of pathway activation in vitro are relevant to tissues in vivo. Additionally, data collected from in vitro culture systems do not address the role of cell to cell communication (autonomous and non-autonomous signalling) that is critical in vivo [1,9]. We therefore sought to use an in vivo model of vascular development with intact tissue and cellular contexts to assess how loss of miR26a and subsequent increases can affect vSMC coverage. We show multiple lines of evidence that suggest that endothelial pSmad1 levels correlate with increased vSMC coverage of blood vessels. Our use of BMP-reporter transgenic fish reveals that during normal development, and under physiological conditions, vSMCs directly contact BRE and pSmad1 positive endothelial cells but have undetectable BRE or pSmad1 signal themselves.
In parallel to loss of miR26a resulting in a subsequent increase in smad1 and vSMC coverage, we also demonstrate that endothelial specific overexpression of smad1 (smad1 ECOE ) results in increased vSMC coverage. Our data therefore inversely complement the murine knockout models of HHT that have noted reduced vSMC coverage when endothelial Smad1 signalling is reduced. Of note, endothelial specific knockdown of Alk1 or Smad4 leads to a reduction of αSMA/Acta2 coverage on larger arterial vessels. Interestingly, there is a contextdependant shift in vSMC coverage in these studies, as ectopic expression of vSMCs is seen on venous and capillary vessel beds [60]. This hypervascularization was presumed to be in response to increased flow from AVM affected vessels into finer retinal vessels. Similar shifts are seen when BMP9/10 blocking antibodies are used [61]. Our study did not address changes in vSMC coverage in venous beds, but it would be interesting to see if overexpression of smad1 leads to increased vSMC coverage across both arterial and venous vessel beds.
Our data suggest that the normal function of miR26a is to reduce Smad1 protein activation within the endothelium, and indirectly inhibit vSMC differentiation in early development. Treatment with K02288, a potent ALK1/2 inhibitor, significantly reduced both acta2-positive vSMC coverage and reduced the length of the ventral aorta. These effects could be rescued by loss of miR26a. Thus, we suggest that enhanced Smad1 activation in these embryos compensates for receptor inhibition. ALK1, ALK2 and ALK3 are expressed in both endothelial and vSMCs [17][18][19][20][21], however in zebrafish alk1 is highly expressed only in the endothelium at 36 hpf [47]. Violet beauregarde (vbg ft09e ) alk1 loss of function zebrafish mutants develop striking cranial vessel abnormalities by 48 hpf due to increased endothelial cell proliferation [19]. vbg ft09e are also unable to limit the diameter of arteries carrying increasing flow from the heart [27]. Based on our data involving indirect control by endothelial signalling, we would predict there is an additional defect in vSMC recruitment in alk1 mutants, although this remains to be tested.
Endothelial and mural cells signal through several paracrine pathways to stabilize vessels [62,63]. BMP signalling in endothelial cells activates an axis of BMP/ Notch3/ Pdgf signalling to promote the expression of contractile vSMCs genes such as Acta2 and Myh11a in in vitro co-culture systems [64]. Specifically, BMP9 signalling via endothelial cells induces NOTCH3 in vSMCs, which in turn induces expression of Pdgfrβ and maintains the proper response to Pdgf ligands [17,65]. There is evidence that mouse MiR26a is modulated by Pdgf-BB signalling [45]. For instance, neointimal hyperplasia results in elevated levels of Pdgfbb associated with upregulation of MiR26a and accumulation and proliferation of vSMC at sites of injury. Furthermore, treatment of primary mouse aortic vSMCs with MiR26a mimic drives cells to a synthetic vSMC state [45]. We found that notch3, pdgfrβ and contractile vSMC markers were significantly increased in miR26a knockdown embryos, suggesting that increases in endothelial Smad1 in zebrafish may be transmitted to vSMCs through a BMP/ Notch3/ Pdgf signalling axis. Pdgf ligands are primarily released by endothelial cells, and we observe an increase in pdgfba in miR26a morphants, providing a potential mechanism by which active BMP signaling in endothelium can recruit and induce vSMC differentiation via paracrine non-autonomous signalling pathways.
While we found increased differentiation of vSMCs at the later stage 4 dpf time point, at 48 hpf loss of miR26a results in hemorrhage. The 48 hpf to 4 dpf window is a common window for vascular instability phenotypes to emerge in zebrafish [3,[66][67][68]. BMP signalling is initiated in endothelium at this time point and perturbations can affect endothelial cell junction development [63]. We have previously shown mural cells present around vessels by 48 hpf, although they are mesenchymal and immature [3]. These cells express pdgfrβ but have no expression of mature vSMC markers [69], suggesting the 48 hpf time point represents a critical window for vascular mural cell attachment to endothelium and differentiation to a mature phenotype. It is paradoxical then that we see increased maturation of vSMCs at 4 dpf when mir26a is reduced. We suggest that the altered receptor and ligand expression in miR26a morphants may promote morphological change towards maturation, but may not regulate all aspects of maturation, leading to destabilization. For instance aberrant ECM deposition would not be visible in our assays and could lead to vascular instability at the earlier time points [63].
As critical modulators of vascular cell function and with roles in cell differentiation, contraction, migration, proliferation and apoptosis, miRs are attractive targets of therapeutic treatments aimed at modulating the vSMC phenotypic switch. Specific to TGF-β/BMP signalling, the miR-145/143 family has direct involvement in SMC differentiation by repressing the Klf4 to induce a contractile morphology and reduced rates of proliferation [40]. miR-21 controls vSMC differentiation through cross-talk with miR-143/-145 [35] and by mediating TGF-β/ BMP induction to promote miR-21 cleavage to its mature form and a more contractile phenotype (Fig 5). miR26a is unique in this group in that it represses smooth muscle differentiation, likely via a paracrine signalling from endothelial cells. As drug delivery to the endothelium is relatively straightforward, modulation of miR26a might be therapeutically useful for post-transcriptional control of key genes involved in vSMC phenotypic switching.

Ethics statement
All animal procedures were approved by the University of Calgary Animal Care Committee (AC17-0189). Anesthesia and euthanasia used MS-222 (Tricaine) at 10-40 mg/L.

Plasmid construction
For endothelial specific overexpression, smad1 was amplified from zebrafish cDNA using primers that incorporate attb1/b2 recombination sites (5'-GGGGACAAGTTTGTACAAAA AAGCAGGCTTCACCATGAATGTCACCTCACTCTTTTCC-3' and 3'-GGGGACCACT TTGTACAAGAAAGCTGGGTGCTAGGACACTGAAGAAATGGGGT-5' and inserted into pDONR221 to create a pME-smad1 vector. Three way Tol2 gateway cloning [75] was used to insert smad1 downstream of the kdrla promoter to achieve a TolCG2:kdrla:smad1 vector. One-cell stage zebrafish embryos were injected with a solution consisting of 5-20 ng/ μl kdrla-smad1 plasmid and 50 ng/ μl transposase mRNA. For the in vivo sensor test, smad1 3 0 UTR forward and reverse oligos (IDT) were designed incorporating BamHI and Bsrg1 sites using the prediction software TargetScan [76] for miR26 targets within the 3'UTR of zebrafish smad1 (underlined). (5'GCGTGTACACCGGATGACT AGAGGGTTAGGTTGTGTACTACTTGAAGGCAGTTTGTTAGGGTGGGGGTCATCGA ATCTGGCTGAAGAGTCCTCAGTTTTCAGCCCGTGAGAATCTGGAAGATACTTGAC AACTCTGTGGCCGGATCCATA-3' and 3'-TATGGATCCGGCCACAGAGTTGTCAAGT ATCTTCCAGATTCTCACGGGCTGAAAACTGAGGACTCTTCAGCCAGATTCGATGAC CCCCACCCTAACAAACTGCCTTCAAGTAGTACACAACCTAACCCTCTAGTCATCCG GTGTACACGC-5'). Oligos were digested and ligated into the p3E-polyA vector. This construct was then recombined into pDestTol2pA2 by Gateway cloning to achieve a CMV-SP6 promoter upstream of EGFP: smad1 3 0 UTR:EGFP or a control EGFP: p3E-polyA 3 0 UTR. Sensor mRNA and mCherry mRNA were in vitro transcribed from the pCS2 Gateway compatible vector (39) by using the mMessage Machine SP6 kit (Ambion). One-cell zebrafish embryos were injected with 150 pg sensor mRNA and 100 pg mCherry mRNA. When applicable, miR26a MO or miRNA mimic were added. Live embryos were imaged with an identical exposure time at 24 hpf (n = 10/group). The average pixel intensity for fluorescence was measured as described (17) Cell sorting, RNA isolation and RT-qPCR For FACS analysis~200 embryos were collected from 4 dpf Tg(acta2:EGFP;kdrl:mCherry) fish. Embryos were anesthetized with 0.4% Tricaine (Sigma) and heads dissected and pooled. Single cell dissociation was performed according to Rougeot et al. 2014. Briefly dissected embryo heads were washed once with calcium-free Ringers Solution and gently triturated 5-10 times before dissociation solution was added and incubated in a 28.5˚C water bath with shaking and periodic trituration for 45 min. The reaction was stopped, centrifuged and resuspended in Dulbecco's Phosphate-Buffered Saline (GE Healthcare Life Sciences, Logan, Utah, USA, centrifuged and resuspended in fresh resuspension solution. The single cell suspension was filtered with 75 μm, followed by 35 μm filters. Cells were then sorted with a BD FACSAria III (BD Bioscience, San Jose, USA) and collected.
The ΔΔCt method was used to calculate the normalized relative expression level of a target gene from triplicate measurements. Experiments were repeated independently at least three times, unless stated otherwise.

Small molecule inhibition
K02288 was used at a dose of 15μM (SML1307, Sigma). DSMO (D8418, Sigma) was used as a vehicle and control. Drug stocks were heated for 20 min at 65C and then diluted in E3 embryo medium. Drug or control was applied to the media from 52 hpf until 4 dpf. Embryos were grown at 28.5C in the dark until imaging, and drug changed once.

In situ hybridization and immunostaining
All embryos were fixed in 4% paraformaldehyde in PBS with 0.1% Tween-20 at 4˚C overnight, followed by 100% methanol at −20˚C. Digoxigenin (DIG)-labeled antisense RNA probes were used for in situ hybridization. Probes for smad1 (construct described by [53]) sm22a, acta2, myh11a were synthesized from PCR fragments previously described [7,48]. Probes were synthesized by using SP6 or T7 RNA polymerase (Roche, Basel, Switzerland). miR26a double-DIG-labeled LNA probe was obtained from Exiqon, (Copenhagen, Denmark. In situ hybridization was performed as described [32]using a Biolane HTI robot (Holle and Huttner AG, Tubingen, Germany). For microRNA in situ hybridization, a double-DIG-labeled Locked Nucleic Acid (LNA) probe (Exiqon) was used to detect the mature miR26a in whole-mount embryos as recommended by the manufacturer with the modification that hybridization was at 54˚C.

Imaging and data analysis
For imaging, embryos were immobilized in 0.004% Tricaine (Sigma) and mounted in 0.8% low melt agarose on glass bottom dishes (MatTek, Ashland, MA). Confocal images were collected on a Zeiss LSM 700 inverted microscope. Image stacks were processed in Zen Blue and are presented as maximal intensity projections and analyzed using FIJI/ImageJ [78] For cell counts images were converted to 16-bit using ImageJ and the threshold adjusted to allow counting of cells over a region of the VA from the anterior bulbous arteriosus to the most anterior PAA.
To measure intensity, total cell fluorescence (CTCF) was calculated using the formula: CTCF = Integrated Density-(Area of selected cell X Mean fluorescence of background readings). The area for measurement was gated by tracing the aorta from bulbous where the bulbous arteriosus merges with the ventral aorta to the distal tip of the ventral aorta or to the bifurcation point of the ventral aorta using the free form drawing tool, whichever was shorter [56].
For measurement of vSMC cell heights, measurements were made from the endothelial kdrla:EGFP expression to the highest point of the vSMC. 8 measurements were taken for each sample where possible. Ventral head measurements were taken from the ventral aorta and the aortic arch arteries. Measurements represent mean vessel diameter ± standard deviation in micrometers.

Statistical analysis
Distribution of data points are expressed as mean ± standard error of the mean (S.E.M.), or as relative proportion of 100% as mentioned in the appropriate legends. Depending on the number of the groups and independent factors, student's t-tests, one-way or two-way analyses of variance (ANOVA) with non-parametric tests were used as indicated in the figures. Two treatment groups were compared using Student's t-test, using Welch's correction. Three or more treatment groups were compared by one-or two-way ANOVA followed by post hoc analysis adjusted with a least significant-difference correction for multiple comparisons using GraphPad Prism version 7.00 (La Jolla California USA). Results were classed as significant as follows: � P < 0.05, �� P < 0.01, and ��� P < 0.001.