Introduction
Hereditary hemorrhagic telangiectasia (HHT) is an autosomal dominant vascular disorder that affects 1 in 5000 people worldwide [
1,
2]. HHT patients commonly exhibit: spontaneous, recurring nosebleeds; small lesions on mucous membranes called telangiectasias; and/or larger visceral lesions known as arteriovenous malformations (AVMs) [
3,
4]. AVMs, which are direct connections between arteries and veins, are most commonly found in major organs such as the brain, liver or lungs. These lesions present a serious health risk and can lead to decreased quality of life and/or early death due to hemorrhaging, stroke and aneurysms [
3,
5‐
8].
Approximately 85% of HHT cases are linked to heterozygous loss-of-function mutations in the transforming growth factor beta (TGFβ) cell surface receptors activin receptor-like kinase 1 (
ALK1, HHT2) or endoglin (
ENG, HHT1) [
9,
10]. A small subset of HHT patients (~ 4%) exhibit haploinsufficiency of Mothers against decapentaplegic homolog 4 (
SMAD4, JP/HHT) and commonly present with juvenile polyposis syndrome (JP) [
11,
12]. SMAD4 is a transcription factor found in nearly all cell types [
13,
14], where it serves as the central conduit through which canonical TGFβ signaling proceeds, including ALK1 and ENG signaling [
15]. However, despite the key role of SMAD4 in the TGFβ pathway, the mechanisms by which it contributes to HHT pathogenesis remain unknown. In fact, virtually all HHT animal studies have focused on the
Alk1 and
Eng receptor interface of the TGFβ signaling pathway, whereby endothelial loss of
Alk1, or
Eng or blockade of the TGFβ pathway via
Bmp9/10 ligand-blocking antibodies results in HHT-associated phenotypes [
16‐
26]. What little we know about the in vivo role of SMAD4 in the vasculature comes from embryonic studies. These studies revealed that SMAD4 plays a critical role in blood vessel remodeling and maturation [
27], integrity of the blood-brain barrier endothelium [
28] and regulating coronary artery size [
29]. Conversely, nothing is known about SMAD4 function in the postnatal vasculature as homozygous loss of
Smad4 is embryonic lethal [
27]. Therefore, due to limited information on how SMAD4 contributes to the developing endothelium, it is unclear how SMAD4 defects lead to HHT phenotypes, such as AVM formation.
In order to better understand SMAD4’s contribution to HHT pathogenesis, we created an inducible, endothelial cell (EC)-specific Smad4 knockout mouse model (referred to as Smad4-iECKO). We find that induced deletion of Smad4 leads to various vascular defects including the formation of AVMs. In addition, we show that SMAD4 influences EC proliferation, EC size, mural cell coverage and artery–vein gene expression. Utilizing this new Smad4-iECKO model, we found that deletion of Smad4 leads to decreased levels of vascular endothelial growth factor receptor 2 (VEGFR2) expression. Furthermore, concurrent loss of endothelial Smad4 and Vegfr2 in vivo leads to an increased AVM severity. This work provides a new model for the HHT field and presents evidence that the TGFβ and VEGF pathways may be linked in AVM pathogenesis.
Discussion
Our studies are the first to report a Smad4 animal model of HHT (Smad4-iECKO). We showed that endothelial loss of Smad4 recapitulates vascular phenotypes seen in other HHT mouse models, particularly AVM formation. To better understand Smad4’s role in HHT pathogenesis, we performed a comprehensive characterization of Smad4-iECKO mice. Our results demonstrated that increased EC proliferation and size, alterations in mural cell coverage and disruption in AV gene expression are associated with Smad4-deficient blood vessels. We also provided evidence that loss of SMAD4 causes decreased VEGFR2 expression, and that loss of a single allele of Vegfr2 in the Smad4 null background leads to an increased severity of AVMs.
Considering
Smad4’s centralized role in TGFβ signaling, we aimed to test the universality of our
Smad4-iECKO model in relationship to HHT phenotypes. Consistent with previous reports, loss of TGFβ signaling through
Smad4 leads to vascular defects similar to those found in
Eng and
Alk1 mouse retinal models (Fig.
1; summarized comparisons of HHT models in Table
1). For example, blood vessel enlargement in
Alk1 and
Eng mutant mice has been linked to increases in EC proliferation [
17,
20,
34,
45], which we also observed in our
Smad4 mutants (Fig.
2). Moreover, loss of
Smad4 led to an increase in EC cell size both in vivo and in vitro, which was unreported in
Alk1 and
Eng models (Fig.
3). These findings are supported by a recent study showing that loss of
Smad4 caused an increase in the size and rates of proliferation of ECs in the coronary artery and under in vitro flow conditions [
29]. Furthermore, work in zebrafish has shown that in response to increases in flow,
Eng-deficient blood vessels enlarge [
25]. Interestingly, our in vitro data suggest that, in
Smad4 mutants, cell size changes may also occur in the absence of flow. Taken together, it appears that blood vessel enlargement in HHT models is affected not only by increases in EC proliferation but also by an increase in EC size itself. How these alterations may lead to AVM formation is unclear, although it is notable that defects in the NOTCH pathway (both gain-of-function and loss-of-function) cause AVM formation [
40,
41,
46,
47] via an initial increase in size of ECs [
48]. Whether AVMs in HHT patients form in a similar manner remains an open question, as evidence in zebrafish suggests that HHT-associated AVMs are not directly caused by alterations in NOTCH signaling [
49].
Table 1
Comparison of HHT mouse models
Associated with
| HHT2 | HHT1 | HHT/JP |
Percentage of mutants with AVMs
| | | 82% |
Angiogenic delay
| | | Reduced |
Vessel size
|
Artery size | | | Enlarged |
Vein size | | | Enlarged |
Proliferation in
|
Artery | | | Increased |
Vein | | Increased [ 17], not changed [ 34] | Increased |
Capillary | | Increased [ 17], not changed [ 34] | Increased |
Cell size
| | | |
Smooth muscle coverage
| | | Increased |
Pericytes
| Decreased (only in capillaries) [ 20] | | Decreased |
Artery identity
|
Dll4
| | No change [ 34]; not expressed in AVM | No change |
Ephrinb2
| Downregulated?? | No change [ 17], downregulated [ 34] | No change |
Hey1
| | | Downregulated |
Jagged1
| | No change [ 17]; increased [ 34] | No change |
Notch1
| | | Downregulated |
Venous identity
| | | |
A
pj
| No change [ 17]; upregulated [ 24] | No change (Expressed in AVM) [ 34] | No change; expressed in AVM |
Eph
b
4
| | No change (Expressed in AVM) [ 34] | Reduced expressed in AVM |
VEGFR2 levels
| | Altered VEGFA-induced kinetics [ 34] | Downregulated |
Respiratory distress
| | | 168–192 h post-Tx Inj |
Nonetheless, our work demonstrated that expression of NOTCH signaling components, which are associated with arterial identity, as well as genes connected to venous and tip cell identity, are disrupted in the absence of
Smad4 (Fig.
4). We also revealed that these changes can occur in arteries, veins and/or capillaries; however, it is important to note that the AVMs themselves expressed all genes examined regardless of whether the marker was up- or downregulated in other vessel types. When comparing these results to those obtained in
Alk1 and
Eng mouse models, we noted variations in AV gene expression between all three mutant backgrounds [
17,
20,
24,
34,
38]. These differences could be due to tissue-specific effects related to the source tissues examined and/or the vascular expression patterns of
Alk1,
Eng and
Smad4. For instance, some studies examined gene expression in isolated lung ECs [
24], while others utilized brain and/or retinal ECs [
20,
34]. Additionally, it is possible that expression levels in various vessel types play a role, as
Alk1 is highly expressed in arterial ECs [
50], while
Eng is only moderately expressed in arteries [
51].
Eng also is expressed highly in capillaries and weakly in veins [
52]. In comparison,
Smad4 is present in virtually all tissues [
13,
14]. However, despite these differences, it is clear that overall disruptions in AV gene expression are consistent between all three mouse models of HHT. Further examination is needed to address whether alterations in AV identity are a primary cause or secondary effect of AVM formation. To this point, our work does not address whether the observed phenotypes and molecular changes are a cause or an effect of AVM formation, as experiments were performed after AVMs developed. This cause/effect relationship has not been explored in
Alk1 and
Eng models of HHT either. Therefore, future studies addressing this issue will be important for identifying the underlying molecular defects that drive AVM pathogenesis versus those that are secondary effects of AVM formation.
It is also important to note that tamoxifen-inducible murine models of HHT have several limitations. HHT phenotypes arise in patients due to mutations (most commonly missense mutations) that lead to haploinsufficiency [
10]. In contrast, mouse models of HHT often utilize null genetic backgrounds because loss of one allele of
Alk1,
Eng or
Smad4 does not result in consistent presence of AVMs in predictable locations [
21,
26,
53‐
55]. Furthermore, HHT patients harbor germline mutations, which manifest from gestation and remain throughout adulthood. However, in mice, complete loss of
Alk1,
Eng or
Smad4 during gestation results in embryonic lethality making it impossible to study their postnatal impact on HHT [
16,
27,
33,
56,
57]. For this reason, the mouse retina has become an effective model to study AVM formation; the retinal vasculature forms directly after birth allowing researchers to assess developmental angiogenesis, similar to vessel growth that would be seen in a developing human. Although these models do not perfectly mimic the genetic background of HHT patients, retinal AVMs form at consistent rates and locations providing a reliable model to investigate the mechanisms of AVM formation.
In our
Smad4-iECKO retinas we noted delayed angiogenic outgrowth similar to
Eng mutants [
17,
34], while
Alk1 mutant retinas did not exhibit reduction in vascular outgrowth [
20]. Interestingly, our
Smad4-iECKO mice exhibit a significant reduction in
Eng transcript levels but show no changes in
Alk1 mRNA levels (Fig.
5A). This could account for the observed similarities in reduced vascular outgrowth between
Smad4 and
Eng, but not
Alk1 mice. However, this result also illustrates the complex association between the TGFβ pathway and HHT, as
Eng expression levels are reduced in the
Alk1 mouse models of HHT [
20,
24], yet show no changes in outgrowth. To our knowledge, it is unknown what happens to levels of
Alk1 expression in the
Eng HHT model, or whether
Smad4 levels are affected in either
Alk1 or
Eng mouse models. Moving forward, it will be important to understand the association between
Alk1,
Eng and
Smad4 in HHT because even though it is expected that all three cooperate in a linear manner in the TGFβ pathway, differences in phenotypes (Table
1) suggest this might not be the case.
The overall objective of our work was to develop a
Smad4 model of HHT that could be used to identify the TGFβ targets that drive AVM formation, as almost nothing is known about these downstream effectors. To this end, we explored a possible link with the vascular endothelial growth factor (VEGF) signaling pathway that has been previously suggested in other HHT models [
21,
24,
26,
34,
42,
53]. For instance, homozygous-induced deletion of
Alk1 or
Eng in adult mice requires the presence of exogenous VEGF before AVMs will form in the brain, suggesting that activation of the VEGF pathway is needed for AVM formation [
21,
26]. To this end, VEGF neutralizing antibodies have been shown to prevent wound-induced skin AVMs from developing in
Alk1-deficient mice [
53]. Furthermore, in the absence of
Alk1 and
Eng, several studies have reported increased
Vegfr2 expression and altered VEGFR2 kinetics in vitro [
24,
34,
42]. In contrast, our data showed that loss of
Smad4 led to a reliable and significant decrease in
Vegfr2 expression both in vitro and in vivo (Fig.
6). This is consistent with a previous study on human patients with cerebral brain AVMs where there was a marked decrease in
Vegfr2 expression [
58]. Contrary to other HHT studies, the reduction of
Vegfr2 in
Smad4-iECKO mice could potentially be attributed to the downregulation in
Nrp1, a VEGFR2 co-receptor. Studies have shown that decreased
Nrp1 levels correlate with reduced
Vegfr2 expression [
59,
60]. Although other HHT studies did not find reduced
Vegfr2 levels, homozygous deletion of both
Smad4 and
Vegfr2 produced similar results to those obtained in double
Alk1- and
Vegfr2-deficient retinas [
24]. In each study, deletion of both alleles of
Vegfr2 in the
Alk1 or
Smad4 null backgrounds resulted in inhibition of retinal vascular development, suggesting that appreciable loss of
Vegfr2 in the absence of either
Alk1 or
Smad4 overrides HHT-like phenotypes because the vasculature is severely underdeveloped (Fig.
7). We did note that AVMs still formed in both experiments at fewer and similar rates in
Alk1 and
Smad4 mutants, respectively. However, in further studies we demonstrated that loss of a single
Vegfr2 allele in the
Smad4 mutant background led to an enhancement of vascular phenotypes associated with
Smad4-iECKO retinas; the vascular front exhibited a consistent increase in density and AVMs showed a substantial enlargement. Alternatively, increased AVM size could be attributed to altered blood flow rates, hemodynamics forces and/or rates of oxygen diffusion caused by the overall stunted growth of the mutant blood vessels, rather than due to the loss of VEGFR2 directly. Future studies will be needed to understand how these processes are altered in TGFβ mutant backgrounds and how those contributions may affect severity of AVMs.
This SMAD4-VEGFR2 association is somewhat contrary to the clinical use of bevacizumab (also known as Avastin), which is a humanized anti-VEGF monoclonal antibody that sequesters VEGF to prevent it from binding both VEGFR1 and VEGFR2 subsequently hindering angiogenesis [
61,
62]. Bevacizumab is currently used as a palliative therapy for HHT where it alleviates symptoms such as chronic nosebleeds but is not considered a long-term therapy [
63]. Studies on the use of bevacizumab have been performed in mature vascular networks, namely that of adult humans and mice [
64,
65]. Little information is known about the effects of bevacizumab in children or developing/remodeling vascular networks. Our work suggests that the connection between SMAD4 and VEGFR2 is different during developmental angiogenesis, when AVMs are thought to form, as compared to mature, established vascular networks. Therefore, further research on the effects of bevacizumab in developing vascular networks is needed, as our results indicate that bevacizumab may enhance developmental HHT phenotypes.