Mouse models of cerebral arteriovenous malformation

Characteristics of human brain AVM and mouse models of the disease

Arteriovenous (AV) malformation (AVM) is a vascular anomaly capable of both hemorrhagic and ischemic insults, leading to seizures, headaches, stroke, and even death.¹ BAVM prevalence is estimated at 0.05%,² often occurring in young people between 20 and 40 years of age.³ BAVMs account for 50% of hemorrhagic stroke in children⁴ and 1–2% of all strokes in the population.⁵ Brain AVMs (BAVMs) can cause life-threatening intracerebral hemorrhage (ICH) (Figure 1⁶). 50% of patients are first diagnosed upon ICH,¹ with 1% and 5% annual hemorrhage rate for previously unruptured and ruptured AVMs, respectively.^{7, 8} Following BAVM rupture, reported mortality rates range from to 15–29%,⁷ and long-term morbidity rates range from 16–56%.^{1, 9} Thus, BAVM is defined by vascular features and accompanying neurological deficits.¹

Figure 1. Features of human brain AVM.

(A) An AVM is visualized on the lateral temporal surface of a human brain. (B) Left ICA angiography (lateral view) reveals a left lateral temporal AVM with a large feeding artery and draining vein. (C) Cartoon of this subtype (lateral view), indicating feeding arteries and draining veins. Reprinted from Lawton.⁶ Copyright 2014, Thieme Medical Publishers.

AVM is characterized by high-flow AV connections that shunt blood directly from arteries to veins, displacing intervening capillaries with a nidus of enlarged and tortuous vessels. BAVM clinical characteristics include: (1) AV shunting, the presence of direct connections between arteries and veins, displacing intervening capillaries; (2) abnormally high blood flow through the feeding artery, AV shunt, and draining vein; (3) the presence of a focal nidus consisting of enlarged, tangled vessels; (4) ICH and ischemia and/or increased endothelial permeability; and (5) neurological deficits, including seizures, headache, unsteadiness, and stroke. Therefore, mouse models relevant to translational BAVM research should exhibit these anatomical, functional, and symptomatic features of the human disease.

Two different approaches have led to progress in genetically engineered mouse models of BAVM (Table 1). “Bedside-to-bench” – heritable risk alleles in human patients have been mutated in mouse counterparts; conversely, “Bench-to-bedside” – genes identified in embryonic AV specification have also been mutated in mice to provide new insight into the human disease. Here, we provide an overview of mouse models of BAVM developed by both approaches. (Table 1; please see http://stroke.ahajournals.org, Table I).^10–32

Table 1.

Genetic mouse models exhibit features of brain AVM (BAVM)

Genetic manipulation	Description of manipulation	Incidence of BAVM Penetrance of BAVM	Dilated vessels	*AV shunts	*High flow	*Nidus	*Hemo- rrhage	*Neuro. dysfunc.	Lethality	Non- brain AVM	Ref.
Eng+/− (CD1 background)	Embryonic germline heterozygous deletion		□	□	□	□	□	□		□	10
Eng+/− (129/Ola background)	Embryonic germline heterozygous deletion	14%	□	□	□	□	□	□	14% by 1yr	✓	10
Eng+/− (129/Ola background)	Embryonic germline heterozygous deletion	25%	□	□	□	□	□	□	25% by 2yrs	✓	11
Eng+/−(B6 or 129 background)	Embryonic germline heterozygous deletion	25–40wks 30%	✓	✓	✓	✓	□	□		□	12
Alk1+/−	Embryonic germline heterozygous deletion	40%	✓	□		□	□	✓		✓	13
Eng+/− + human recombinant VEGF	Embryonic germline heterozygous del. + adult angiogenic stimulus	2–4wks post-VEGF 89%	✓	□		✓	□	□		□	14
Eng+/− + AAV-VEGF	Embryonic germline heterozygous del. + adult angiogenic stimulus	by 6wks post-AAV	✓							□	15
Alk1+/− + AAV-VEGF	Embryonic germline heterozygous del. + adult angiogenic stimulus	by 6wks post-AAV	✓							□	15
L1-Cre; Alk1fx/fx	Endothelial cell deletion during embryonic development	100%	□	✓		□	□	□	embryonic day 18.5	✓	16
L1-Cre; Alk1fx/fx	Endothelial cell deletion during embryonic development	by 5d 100%	✓	✓	✓	✓	✓	□	5d	✓	17
Ad-Cre + Alk1fx/fx + AAV-VEGF	Local deletion in adults + angiogenic stimulus	by 8wks post-AAV/Cre	✓	✓			▫				18
Ad-Cre + Alk1Δ/fx + AAV-VEGF	Local deletion in adults/germline null + angiogenic stimulus	by 8wks post-AAV/Cre	✓	✓			▫				19
Ad-Cre + Engfx/fx + AAV-VEGF	Local deletion in adults + angiogenic stimulus	by 8wks post-AAV/Cre	✓				▫				20
Pdgfb(PAC)-CreERT2; Alk1fx/fx + AAV-VEGF	Endothelial cell deletion in adults + angiogenic stimulus	10d post-AAV/TAM	✓	✓			✓		6–13d post-TAM	✓	21
Cdh5(PAC)-CreERT2; Engfx/fx	Endothelial cell deletion at birth									✓	22
Cdh5(PAC)-CreERT2; Engfx/fx	Endothelial cell deletion in adults									✓	22
R26-CreERT2; Alk1fx/fx	Global deletion in adults	100%	□	□		□	□	□	9–21d post-TAM	✓	17
R26-CreERT2; Engfx/fx + AAV-VEGF	Global deletion in adults + angiogenic stimulus	8wks post-AAV/TAM	✓	✓		✓	✓			✓	23
R26-CreERT2; Engfx/fx	Global deletion in adults						▫	□	2mos post-TAM	✓	23
R26-CreERT2; Engfx/fx	Global deletion in adults						▫	□	8–10d post-TAM	✓	24
SM22α-Cre; Alk1fx/fx	Smooth muscle cell deletion in adults	10–15wks	✓				✓	✓	2–82wks		25
SM22α-Cre; Alk1fx/−	Smooth muscle cell deletion in adults/germline null	10–15wks	✓				✓	✓	2–58wks		25
SM22α-Cre; Engfx/fx	Smooth muscle cell deletion in adults	5wks 90%	✓	✓			✓		50% by 6wks	✓	23
Myh11-CreERT2; Engfx/fx	Smooth muscle cell deletion in adults						▫			✓	24
Scl-CreERT; Engfx/fx	Endothelial cell deletion in adults						▫		0% 1mo post-TAM	✓	24
Scl-CreERT; Alk1fx/fx	Endothelial cell deletion in adults						▫			✓	24
NG2-CreERTM; Alk1fx/fx + AAV-VEGF	Pericyte deletion in adults + angiogenic stimulus						▫				21
Tie2-tTA; TRE-Notch1*	Constitutive activation in endothelial cells at 21d	100%	□							✓	26
Tie2-tTA; TRE-Notch4*	Constitutive activation in endothelial cells at 21d	100%	□							✓	26
Tie2-tTA; TRE-Notch1*	Constitutive activation in endothelial cells at birth	14d 100%	✓	✓			✓	✓			27
Tie2-tTA; TRE-Notch4*	Constitutive activation in endothelial cells at birth	18d 100%	✓	✓	✓	✓	✓	✓	36d		27, 28
Tie2-tTA;TRE-Notch4* OFF	Constitutive activation in endothelial cells OFF at 12d	BAVM reg. from 24hrs 100% reg.	reg.	rec.	nor.	reg.	rec.	rec.	prevented		29
Cdh5(PAC)-CreERT2; Rbpjfx/fx	Endothelial cell deletion at birth	14d 100%	✓	✓		✓	✓	✓	21d	✓	30
Cdh5(PAC)-CreERT2; Alk1fx/fx	Endothelial cell deletion at birth			□						✓	31
Cdh5(PAC)-CreERT2; Alk1fx/fx	Endothelial cell deletion in adults			□						✓	31
Mgp−/−	Embryonic germline homozygous deletion	by 4wks 100%	✓	✓			✓			✓	32

^*Denotes clinically defined features of brain AVM

Abbreviations: Neuro.dysfunc. (Neurological dysfunction); Ref. (Reference); reg. (regress); nor. (normalize); rec. (recover)

Bedside to bench: Human mutations inspire mouse models of HHT-mediated AVM

Although most BAVMs are sporadic with no known genetic lesions, about 5% are associated with autosomal dominant disorders (please see http://stroke.ahajournals.org,Table II). Hereditary Hemorrhagic Telangectasia (HHT) is the most prevalent of these, and is characterized by AVMs in multiple organs, including the brain.³³ HHT is mainly caused by mutations in Endoglin (ENG) (HHT1), encoding a TGFβ binding protein,³³ and activin receptor-like kinase 1 (ACVRL1) (ALK1) (HHT2), encoding a cell-surface receptor for TGFβ ligands.³⁴ Both genes are expressed primarily by endothelial cells (ECs), but how deficiencies in either ENG or ALK1 lead to AVM pathology remains unclear. Additionally, mutations to MADH4, which encodes for Smad4, an effector of TGFβ signaling, cause a combined juvenile polyposis syndrome and HHT.³⁵ HHT can also result from mutations in BMP9³⁶ and two unidentified genes on chromosome 5 (HHT3)³⁷ and on chromosome 7 (HHT4).³⁸ Furthermore, PTPN14, which encodes for a non-receptor tyrosine phosphatase, shows genetic association with pulmonary AVMs in HHT.³⁹ These studies of familial HHT have revealed that multiple, heritable genetic lesions can lead to HHT-related AVMs.

Mutations in RASA1 and PTEN have been linked to AVM in humans. RASA1, which encodes for p120 Ras GTPase-activating protein (a negative regulator of Ras/MAPK pathway), is mutated in CM-AVM (capillary malformation-arteriovenous malformation).⁴⁰ CM-AVM is an autosomal dominant disorder that is characterized by cutaneous capillary malformations and AVMs, including BAVMs.⁴⁰ PTEN encodes a tumor suppressor in the phosphoinositide 3-kinase (PI3K) pathway. Mutations in PTEN cause Bannayan-Riley-Ruvalcaba and Cowden syndromes and result in AVMs as part of their clinical phenotype.⁴¹ Identification of these causal mutations holds promise for future discovery of molecular pathways attributable to AVMs.

Experimental mouse models were engineered with targeted mutations in the Eng (HHT1) and Alk1 (HHT2) genes. Eng or Alk1 knockouts exhibit embryonic vascular defects, including dilated and fused artery-vein pairs and die in utero.^{42, 43} Eng^+/− or Alk1^+/− heterozygous mice are viable and develop characteristics of HHT during adulthood;^10–13 however, features of BAVM, including AV shunts, niduses of dilated vessels, and rounded, misaligned EC nuclei, occur in 30% of Eng^+/− mice aged 25–40 weeks, similar to BAVM incidence in HHT1 patients.¹² Thus, loss of one allele of Eng or Alk1 is sufficient to induce BAVM in adult mice, but with incomplete penetrance.

The incomplete penetrance and focal BAVM development in Eng^+/− and Alk1^+/− mice led to the hypothesis that these genetic perturbations require a “second hit” – a corroborating process or genetic lesion – in AVM formation. Data from human BAVM patients support the second hit hypothesis: 1) BAVM typically presents in adolescence or adulthood, even though patients harbor germline mutations;⁴⁴ 2) a high level of angiogenic signaling near human AVM suggests that AVM may be triggered by angiogenesis;⁴⁵ 3) somatic loss of heterozygosity (LOH) has been observed in RASA1 mediated AVMs.⁴⁶ The finding that a genetic perturbation leads to BAVM in immature/remodeling but not mature/quiescent mouse brains provides the first experimental evidence that angiogenic remodeling may be a permissive factor for AVM formation.^{26, 28} Both classes of “second hit” candidates have been explored, resulting in more robust and tractable models of BAVM formation.

Local delivery of vascular endothelial growth factor (VEGF) results in local vascular dysplasia in Eng^+/− or Alk1^+/− mice. Recombinant human VEGF injection into Eng^+/− brains leads to microvascular abnormalities, including enlarged, tortuous, and clustered vessels, with 89% penetrance and 2–4 week latency.¹⁴ Similarly, focal adenoviral VEGF delivery into the cerebral cortex of Eng^+/− and Alk1^+/− adult mice results in abnormally enlarged capillaries and increased capillary density, with six-week latency.¹⁵ Notably, vascular defects are more profound in Eng^+/− mice than in the Alk1^+/− mice.¹⁵ Together, these studies support the possibility that VEGF-induced angiogenic stimulus can be a second hit for vascular dysplasia in Eng^+/− and Alk1^+/− mice.

The hypothesis that a somatic LOH increases AVM formation has been experimentally tested using genetic tools for tissue-specific, temporal gene deletion. Deletion of both alleles of Alk1 from embryos in a subset of Alk1 expressing cells results in late gestational or postnatal lethality with AVMs in the brain (Figure 2A^{17, 18, 20}), lung and intestine.^{16, 17} However, tamoxifen-dependent deletion of Alk1 from adult mice using R26-CreER^T2 results in lung and intestinal AVMs but is insufficient to induce BAVMs.¹⁷ Together, these studies suggest that deletion of both Alk1 alleles is sufficient to induce BAVM during development, but not adulthood.

Figure 2. HHT mutations lead to features of brain AVM in mice.

(A) Entangled, tortuous AV shunts in latex dye-perfused L1-Cre; Alk1^fx/fx P3 mouse brains. In Alk1^fx/fx control brains, latex dye labeled major arteries (left panel). In mutant brains, dye is found in veins and arteries (right panel). Reprinted from Park et al.¹⁷ Copyright 2009, American Society for Clinical Investigation. (B) In adult Alk1^fx/fx mice, viral Cre and VEGF induces large, tangled vessels near the injection site 8 weeks after virus delivery (left panel). Alk1 deletion by Ad-Cre, without AAV-VEGF, does not affect local vasculature (right panel). Reprinted from Walker et al.¹⁸ Copyright 2011, John Wiley and Sons. (C) In adult R26-CreER^T2; Eng^fx/fx mice, global Eng deletion and focal delivery of AAV-VEGF induces tangled vessels and increased vessel dysplasia near the injection site (right panel, white arrow) 8 weeks post-treatment, as shown by latex dye perfusion. Eng deletion without AAV-VEGF does not affect local vasculature (left panel). Reprinted from Choi et al.²⁰ Copyright 2014, Public Library of Science. (Scale bars: B, 100 µm; C, 1 mm)

Combination of local angiogenic stimulus and Alk1 or Eng deletion promotes BAVM formation in adult mice (refer to Table 1). Deletion of Alk1 or Eng, coupled with VEGF administration, results in signs of AVM, including: enlarged and dysplastic vessels (Figure 2B–C);^18–21 AV shunting;^{18, 20} irregular vessel aggregates;^{18, 20} and microhemorrhage.^{20, 23} These studies show that loss of either Alk1 or Eng alleles, in conjunction with angiogenic stimulation, may lead to AVM formation.

Endothelial deletion of Alk1 or Eng, in combination with angiogenic stimulus, results in features of non-brain AVM in mice. Endothelial Alk1 deletion in adult mice leads to gastrointestinal AVM and hemorrhage 6–14 days after induction of gene deletion.^{23, 24} Deletion of Eng from postnatal endothelium leads to AV shunting and increased EC proliferation in the developing retina.²² However, angiogenic matrigel implantation or wounding is required to induce vascular defects in endothelial-Eng deficient skin.^{22, 24} Thus, loss of Alk1 or Eng from postnatal endothelium can result in non-brain AVM under certain circumstances.

Recent work has raised the possibility that HHT mutations in perivascular cells may also contribute to AVM. Mice with Alk1 deficiency in smooth muscle cells (SMC) (SM22α-Cre; Alk1^fx/fx or SM22α-Cre; Alk1^fx/−) exhibit characteristics of BAVM by 10–15 weeks of age.²⁵ Both models, SM22α-Cre; Alk1^fx/fx and SM22α-Cre; Alk1^fx/−, develop tortuous vessels, large areas of hemorrhage, and hindlimb or whole body paralysis. Similarly, 90% of mice with SMC deletion of Eng during adulthood also develop characteristics of BAVM.²⁰ Enlarged, tortuous vessels assembling into focal tangles are observed, as well as direct AV shunting. These studies raise the possibility that loss of Eng or Alk1 from SMCs lead to AVM.

However, these data remain controversial – SM22α-Cre mediated recombination has been observed in some brain ECs,²⁵ potentially confounding the conclusion that gene deletion is confined to the SMC compartment. Whether perivascular HHT mutations drive AVM formation remains questionable: 1) Alk1 or Eng deficiency in adult ECs, but not SMCs, induces AVM in a skin wound model;²⁴ 2) deletion of Alk1 from ECs, but not pericytes, along with focal delivery of VEGF leads to BAVM formation in adult mice.²³ Together, these studies suggest that endothelial, but not perivascular, Alk1 or Eng deficiency can result in AVM in combination of VEGF stimulation. However, altered perivascular cell coverage is associated with BAVM in mice. Following VEGF-induced angiogenesis in Alk1-deficient brains, fewer pericytes, decreased PDGFR-β expression, and fewer vessels expressing αSMA are observed, suggesting reduced smooth muscle coverage.²³ The perivascular defects observed in animal models are similar to human AVM; however, the contribution of these defects to AVM progression remains unclear.

Bench to bedside: Mutations in the Notch pathway lead to hallmarks of AVM in mice

Investigations into the functions of genes regulating AV specification (Figure 3A²⁹) have led to potential roles in AVM formation. Perturbations to signaling pathways that disrupt normal AV specification often lead to vascular abnormalities that resemble AV shunting in mice (please see http://stroke.ahajournals.org, Table III). Differential AV expression patterns of Notch receptors persist in adult endothelium, suggesting that they are important in maintaining AV specification in adult. Carlson et al. first showed that upregulation of Notch signaling in postnatal endothelium elicits AVM formation.²⁶ Endothelial expression of a constitutively active Notch4 allele (Notch4*) in adult mice results in features of AVM in liver, skin, and uterus, but not in brain. Arterial marker expression is increased, suggesting arterialization of vessels in Notch4* adult mice. This seminal study opens the possibility that Notch, crucial in AV specification, may be important in AVM pathogenesis.

Figure 3. Notch signaling regulates endothelial AV specification.

(A) Signaling pathways regulating AV specification. Arterial EC identity is prompted by the pro-angiogenic factor VEGF. VEGF activates Notch signaling and leads to expression of the arterial marker Efnb2. Sox and Fox transcription factors contribute to Notch activation and arterial identity. Venous EC identity requires suppression of Notch signaling by the transcription factor COUP-TFII. Inactivation of Notch permits expression of the venous marker Ephb4. (B) In whole-mount brain, Ephb4^tau-lacZ is expressed by venous (closed arrowheads), but not arterial (open arrowheads), vessels. (C) Following endothelial Notch4* activation, Ephb4^tau-lacZ is downregulated in veins and AV shunts. (D) Four days post reversal of Notch4* activation, Ephb4^tau-lacZ expression is restored in veins and regressing AV shunts. (Scale bars, 100 µm) Reprinted from Murphy et al.²⁹ Copyright 2012, American Association for the Advancement of Science.

Endothelial expression of Notch4* in immature mice leads to hallmarks of BAVM.²⁸ Vascular lesions were completely penetrant when Notch4* was turned on from birth, causing lethality by P36 Lesions exhibited the hallmarks of human BAVM, including enlarged, tortuous vessels, AV shunting, increased flow in the feeding carotid arteries, nidus formation, hemorrhage and ataxia (Figure 4A^27–29). Endothelial Notch4* increases arterial marker expression (Efnb2, Connexin40, Jagged1, Dll4) and decreases venous marker expression (Ephb4) (Figure 3B–C), suggesting arterialization of the brain endothelium by Notch4*.^{28, 29} Thus, unlike in adult brains, Notch4* is able to induce AVMs in immature brains, suggesting immature brain vasculature is susceptible to Notch4*-induced AVM formation. Endothelial expression of constitutive Notch1 (Notch1*) in immature brains also leads to features of BAVM, indicating that increased activity of either Notch receptor is sufficient to cause BAVM.²⁷

Figure 4. Endothelial expression of constitutively active Notch leads to hallmarks of brain AVM.

(A) Endothelial expression of Notch4* induces enlarged, tangled blood vessels in cerebellum and midbrain as shown by vascular casting at P27 (right panel, arrowheads). Control brains exhibit normal vasculature (left panel). Reprinted from Murphy et al.²⁸ Copyright 2008, National Academy of Sciences, USA. (B) Notch4* initiates AV shunts through enlargement of capillary-like vessels, as shown by in vivo two photon imaging. Arrowheads indicate an AV shunt developing from a capillary-diameter AV connection between P14 and P19. Reprinted from Murphy et al.²⁷ Copyright 2014, National Academy of Sciences, USA. (C) Repression of Notch4* decreases AV shunt diameter and decreases blood flow velocity, as shown by in vivo two-photon imaging. Reprinted from Murphy et al.²⁹ Copyright 2012, American Association for the Advancement of Science. (Scale bars: B, C, 50 µm)

In vivo time-lapse imaging of BAVM formation in Notch4* mice shows that AV shunts arise from microvessels with capillary-like diameter and blood flow profiles, without a significant increase in EC proliferation (Figure 4B).²⁷ Clinical observations suggest that increased flow through low-resistance AV shunts encourages their growth, while “stealing” blood flow from adjacent higher-resistance vessels.⁴⁷ In this model, Notch4* permits “steal” and perpetuates a positive feedback loop, leading to selective growth of higher velocity at the expense of lower velocity AV connections.²⁷ Thus, Notch4* (and presumably Notch1*) promotes the initiation and progression of BAVM in mice (Figure 4A).

Conversely, blocking Notch signaling, via deletion of Rbpj, in postnatal endothelium also leads to features of BAVM in mice.³⁰ Endothelial deletion of Rbpj at birth results in tortuous vessels, AV shunting, vessel aggregates, hemorrhage and signs of neurological deficits by P14 in the brain. AV shunts show decreased Efnb2 and increased Ephb4 expression, suggesting acquisition of venous identity. Data from the gain- and loss-of-function Notch models are consistent with the model that tight regulation of Notch signaling is essential to prevent BAVM in mice.

Crosstalk between HHT and Notch signaling pathways

Gene expression changes in HHT mutant mice suggest a link between HHT and Notch signaling pathways in AVM formation. Loss of Alk1 function results in abnormal arteriovenous marker expression, both in embryonic and postnatal mice.^{18, 31, 42} Alk1 signaling also synergizes with activated Notch in the endothelium to induce expression of Notch target genes.⁴⁸ These data suggest that Alk1 may affect the expression of Notch downstream genes.

Alk1/Notch crosstalk also functions in BAVM development. Deficiency of the extracellular bone morphogenetic protein (Bmp) antagonist matrix Gla protein (Mgp) leads to BAVM formation.³² Alk1 is a receptor for Bmps, and thus an increase in available Bmp results in increased Alk1 signaling. Mgp^−/− mice develop features of BAVM by four weeks of age, with enlarged cerebrovascular vessels, AV shunting, and hemorrhage (Figure 5³²). Analysis of AV marker expression shows increased Efnb2 and decreased Ephb4 expression in Mgp^−/− brains. Notch ligands Jagged 1 and 2 are upregulated in Mgp^−/− brains, and Mgp^−/− BAVMs, and heterozygous deletion of Jagged 1 and 2 in Mgp^−/− mice suppresses BAVM formation. These findings extend to cultured brain ECs, where Bmp9 is sufficient to activate Notch signaling and induce arterial marker expression.³² Together, these data suggest cooperation of Alk1 and Notch pathways in BAVM pathogenesis.

Figure 5. Deficiency of BMP antagonist MGP leads to hallmarks of brain AVM.

MicroCT imaging shows enlarged vessels and AV shunts in Mgp^−/− but not Mgp^+/− or Mgp^+/+ mice. Colors represent vessel radii; asterisks represent AV connections. Reprinted from Yao et al.³² Copyright 2013, National Academy of Sciences, USA. (Scale bars: 1 mm)

Attempts in therapeutic treatment of AVMs

Mechanisms that underlie BAVM pathogenesis remain unclear, limiting the rational design of specific molecular interventions. To date, there are no specific or approved medical therapies to treat AVMs or to prevent AVM hemorrhage. Current treatment strategies include medical management⁴⁹ or invasive procedures such as surgical resection, stereotactic radiosurgery, or endovascular embolization.^{50, 51} Treatment strategies aimed at inhibiting angiogenesis and maintaining vascular integrity have led to novel therapeutic approaches for the treatment of vascular malformations, including BAVMs (please see http://stroke.ahajournals.org, Table IV and Supplemental text). Further therapeutic development depends on an improved understanding of mechanisms underlying AVM pathogenesis, such as those uncovered using mouse models of BAVMs.

Reversal of AVM by normalization of the causal Notch lesion in mice

Correction of a molecular lesion allows the regression of existing BAVMs and leads to the restoration of AV specification in an animal model. In the Notch4* model, symptoms of AVM are eliminated upon suppression of the Notch4* transgene,²⁸ along with rapid regression of the AV shunts, restoration of blood flow to distal arteries, and perfusion of the brain parenchyma (Figure 4C).²⁹ Additionally, normal AV specification is restored in concert with the regression of existing Notch4* AVMs – overexpression of arterial markers (Efnb2, Dll4, Jag1, Cx40) is decreased and venous marker expression (Ephb4) is restored (Figure 3D). The normalization of these high-flow AV shunts by a single genetic correction has conceptually changed our view on AVM treatment. The discovery that suppression of a causal gene can lead to AVM regression in mice, without hemorrhage or thrombosis, may change the way we think about AVM pathogenesis and treatment.

Concluding remarks

Mouse models of human BAVM disease provide a useful platform for elucidating the mechanisms of AVM pathogenesis and for exploring treatment options. Moving forward, the identification of additional genetic perturbations associated with BAVM, through continued investigation of both mouse and human genetic studies, will open new opportunities for the rational design and development of better treatment options for this disease.