Age-Dependent Lethality in Novel Transgenic Mouse Models of Central Nervous System Arteriovenous Malformations
Background and Purpose—The lack of an appropriate animal model has been a limitation in studying hemorrhage from arteriovenous malformations (AVMs) in the central nervous system.
Methods—Novel mouse central nervous system AVM models were generated by conditionally deleting the activin receptor-like kinase (Alk1; Acvrl1) gene with the SM22-Cre transgene. All mice developed AVMs in their brain and/or spinal cord, and >80% of them showed a paralysis or lethality phenotype due to internal hemorrhages during the first 10 to 15 weeks of life. The mice that survived this early lethal period, however, showed significantly reduced lethality rates even though they carried multiple AVMs.
Results—The age-dependent change in hemorrhage rates allowed us to identify molecular factors uniquely upregulated in the rupture-prone AVM lesions.
Conclusions—Upregulation of angiopoietin 2 and a few inflammatory genes were identified in the hemorrhage-prone lesions, which may be comparable with human pathology. These models will be an exceptional tool to study pathophysiology of AVM hemorrhage.
- activin receptor-like kinase 1
- arteriovenous malformation
- hereditary hemorrhagic telangiectasia
- intracranial hemorrhage
Hemorrhage from arteriovenous malformations (AVMs) in the central nervous system (CNS) may cause severe neurological deficit or death. Limited research in the development of therapeutic medications for AVMs has been due, in a large part, to an absence of an appropriate animal model. In humans, a pathological ALK1 mutation causes hereditary hemorrhagic telangiectasia (HHT), a disease characterized by AVM formation in the CNS and other organs. Although AVM mouse models have been generated by conventional and conditional Alk1 deletions, all models caused severe vascular malformations in diverse organs and inevitable lethal hemorrhages.1–4 Recently, a mouse brain AVM model was developed by a viral vector-mediated method using the Alk1 conditional deletion mice; however, this model does not cause hemorrhage or neurological deficit.5
We report novel CNS AVM mouse models exhibiting hemorrhage, paralysis, and partial lethality. All of these mice developed AVMs in their brains, and a majority of them died or were paralyzed due to internal hemorrhages before reaching 10 weeks of age. However, a subset of mice survived much longer despite carrying multiple AVMs. In addition, we identified variegated expression of angiopoietin 2 (Agpt2) and a few inflammation-related genes in the rupture-prone AVM walls.
Detailed Supplemental Methods are available in the online-only Supplement (http://stroke.ahajournals.org).
Animals and Blue Latex Injection
All mouse procedures carried out were reviewed and approved by the Georgia Health Sciences University Institutional Animal Care and Use Committee. Floxed-Alk1 conditional deletion mice Alk1(flox/flox),3,4 Alk1 deletion mice Alk1(+/−),2 and transgenic Tg(SM22-Cre) deleter mice (Jackson Laboratory, Bar Harbor, Maine)6 were intercrossed to generate SM22Cre-del models. For identification of Cre-recombined cells, R26R Cre-reporter mice (Jackson Laboratory)7 were crossed with the previously mentioned mouse lines. Littermate control mice were used for all experiments. Blue latex solution was systemically injected through the cardiac left ventricle, and tissues were imaged as previously described.3,4
Relative Quantitative Polymerase Chain Reaction
Total RNAs from the most AVM-prone segment of the cerebral hemisphere, roughly right striatum and basal ganglia regions, were isolated from 4 groups of animals (n=6 each): lethal period (LP) Flox-SM22Cre-del mice (4–8 weeks of age), LP littermate control mice, stable period (SP) SM22Cre-del mice (29–36 weeks of age), and SP littermate control mice. TaqMan Gene Expression Assays (Applied Biosystems) were used for the relative quantification of transcripts, and results are shown as fold difference compared with the SP control.
Localizations of brain AVMs were identified by MR angiography, and tissues were collected without any cardiac perfusion. Colorimetric immunohistochemical staining was performed to observe unpredictable AVM histology and localization of specific protein on a single section. The antibodies used in immunohistochemical staining are listed in the Online Supplement. Single and double immunohistochemical staining was performed with commercially available kits with and without hematoxylin counter staining, respectively. X-gal and immunohistochemical double staining on frozen tissue samples was performed as described.2
Proportions of genotypes in offspring were tested with the χ2 test. The log rank significant test was performed for Kaplan-Meier survival curves. The ΔΔCt values from quantitative polymerase chain reaction assays were analyzed with 1-way analysis of variance followed by a post hoc test.
Two closely related CNS AVM mouse models were generated by conditionally deleting the Alk1 gene using Tg(SM22-Cre) deleter mice. In the first Alk1(flox/flox);Tg(SM22-Cre) model, both copies of Alk1 were deleted by Cre recombinase expressed from the SM22-Cre transgene (Flox-SM22Cre-del mice hereafter). In the second Alk1(flox/−);Tg(SM22-Cre) model, 1 of the 2 copies of the Alk1 gene was constitutively deleted in all cells, mimicking human HHT and the second Alk1 copy was deleted by the SM22-Cre transgene (HHT+SM22Cre-del model hereafter). Of note, although the SM22-Cre transgene was initially reported to induce Cre recombination in smooth muscle cells and cardiac muscles in embryos,6 a wide-range and inconsistent Cre activation was observed in adults (Supplemental Figures I and II). Furthermore, an involvement of both Alk1-null and Alk1-intact cells in AVM walls was identified (Supplemental Figure II).
Both SM22Cre-del models caused partial lethality before 2 weeks of age (Supplemental Tables I and II). Intriguingly, the difference in the ratio of surviving pups at 2 weeks of age (Flox-SM22Cre-del pups: 47% versus HHT+SM22Cre-del: 15%, χ2: P<0.01) was the only notable difference between the 2 SM22Cre-del models, and their gross phenotypes after 2 weeks of age were virtually indistinguishable. Therefore, the following analyses were performed mainly using the Flox-SM22Cre-del mice unless otherwise specified.
Survival curves of the 2 SM22Cre-del models after 2 weeks of age were essentially identical (log rank significance: P=0.31), and a majority of mice suffered from spontaneous death, hindlimb paralysis, or whole body paralysis during the next 8 to 13 weeks (Figure 1A–B). There was no significant difference between female and male survival curves in either model (both P>0.3, data not shown).
CNS hemorrhages were identified in the paralyzed SM22Cre-del mice. Hindlimb-paralyzed mice showed spinal cord hemorrhage in all cases, with some of them showing AVM histology (Supplemental Figure III). Severe intracranial hemorrhage was found in many of the completely paralyzed animals (Figure 1C–D). Because the blue latex solution injected through the cardiac left ventricle cannot enter microcapillaries due to its particle size,4,5 only arterial vessels should have been filled with latex. However, cerebral veins of these animals were filled with blue latex, indicating presence of AVMs. Overall, these findings suggest that Alk1-deletion by the SM22-Cre transgene induced malformed vessels in the CNS, which caused internal hemorrhages and hence the early partial lethality.
Interestingly, the survival of the mice was considerably improved when the Flox-SM22Cre-del and HHT+SM22Cre-del mice surpassed 10 and 15 weeks of age, respectively (Figure 1). The mice that survived this early LP (≤10 weeks of age) continued to show partial lethality, however, at much reduced rates during the SP (≥16 weeks of age). To elucidate the cause of this change, we considered 2 alternative hypotheses. The first is that a small population of mice does not develop AVMs, and these non-AVM-bearing mice live much longer. The second is that most mice develop AVMs but the AVM walls stabilize once they reach SP age and become protected from lethal hemorrhages. The first hypothesis implies few or no AVMs in SP mice, whereas the second suggests presence of (multiple) AVMs in SP mice.
The cardiac blue latex injection was used to determine whether SP mice carry AVMs. All brain tissues in 15 SP mice (18–102 weeks of age) were found with various numbers, morphologies, and sizes of malformed vessels (Figure 2; Supplemental Figure IV). A number of malformed vessels showed direct connections between arteries and veins (AVMs), but there were also many tortuous vessels (TVs) that did not show an apparent connection to veins. In addition, many of the AVMs/TVs were accompanied by adjacent brown pigments, the hemosiderin deposits formed due to prior hemorrhages. Intriguingly, sometimes relatively large AVMs were free of neighboring hemosiderin (Figure 2B), whereas some of the smallest TVs were accompanied by hemosiderin (Figure 2D), indicating that any malformed vessels could cause hemorrhage. On a side note, we found malformed vessels in intestines of 4 SP animals (data not shown). In contrast, no AVMs were found in 22 littermate control mice except some TVs mostly in the cerebellum region. These findings were confirmed by scoring the severities of AVMs/TVs using a newly developed grading system (Supplemental Table III). The AVM/TV scores clearly showed that all SP SM22Cre-del mice carried a number of AVMs/TVs and most mice experienced hemorrhage (Supplemental Figure V).
Because only the vasculature in LP SM22Cre-del mice was prone to bleed compared with the SP SM22Cre-del and LP/SP control mice, the LP SM22Cre-del brain may have a unique molecular environment that enhances the chances of hemorrhage from their vascular walls. Transcript levels of 31 candidate genes that may contribute to such an environment were examined by quantitative polymerase chain reaction using the brain tissues from these 4 groups regardless of AVM presence (Supplemental Figure VI). Interestingly, Agpt2 was the only gene significantly upregulated in LP SM22Cre-del brains compared with the 3 other groups (all P<0.01). In addition, interleukin-1β and tumor necrosis factor α were significantly upregulated in LP SM22Cre-del brains compared with LP control brains (P<0.05).
Of significance, proteins of some of these factors were induced in the AVM walls in a variegated fashion. Localizations of AVMs in LP brain (5–6 weeks of age) were identified by MR angiography, and their histological sections were prepared (Supplemental Figure VII, Supplemental Movie I). The nuclei of AVM wall cells were occasionally stained with cell proliferation markers (Supplemental Figure VIII). Like in other Alk1-deletion models,3–5 smooth muscle cell coverage of AVMs was inconsistent and variegated (Figure 3; Supplemental Figure VII). Similarly, Agpt2 and interleukin-1β expression was found in a variegated pattern in AVM walls. In addition, variegated expression of cyclooxygenase 2 (COX2; PTGS2), a downstream molecule in inflammatory pathway, was also observed. Although smooth muscle cell coverage and the expression of these proteins were both variegated, Agpt2 and COX2 expression were actually often higher in the endothelial cell layer (Supplemental Figure IX). Interestingly, variegated focal recruitments of neutrophils and macrophages are reported in human AVM lesions.8 A future study will address if similar inflammatory cell infiltration is involved in mouse AVM lesions.
The novel SM22Cre-del CNS AVM mouse models showed distinctive phenotypes and allowed us to uncover some of the molecular differences in hemorrhage-prone and hemorrhage-protected AVM lesions. However, many questions on the natural history of these AVMs still remain unanswered: when AVMs start to form; which cell type and what stimulus is responsible for initial AVM formation (in addition to Alk1 deletion); if and when AVM growth stops; what is the role of Alk1-intact cells in AVM growth; and so on. Previous studies, including the “response-to-injury” hypothesis for AVM pathogenesis,9 may help us address some of these questions, and follow-up studies are currently conducted.
The significance of these novel AVM mouse models is apparent. Perhaps the most advantageous attribute is the transition from a hemorrhagic (LP) to stable and less hemorrhagic (SP) phenotype. This change led us to identify variegated expression of Agpt2 and inflammatory-related proteins in hemorrhage-prone AVM walls. Importantly, an induction of Agpt2 and inflammatory factors in AVM lesions were consistent with human AVM tissues, confirming the validity of these models.10,11 Although the limited availability of SP mice restricted the extent of this initial analysis, more comprehensive investigation is currently underway.
In addition, these mice will be an excellent model of human hemorrhagic CNS AVM. Because the LP mice are hemorrhage-prone, they can be used as a drug treatment model for hemorrhagic AVM. In addition, because human AVMs are stable and seldom bleed, the stable AVMs in SP mice may best replicate the human CNS AVM condition. We will be able to induce various pathophysiological insults on SP mice that cause vascular inflammation and/or injury, including hypertension, diabetes, and physical trauma, to test if these stimuli induce AVM hemorrhage.
Sources of Funding
This work was supported by the Georgia Health Sciences University Cardiovascular Discovery Institute Seed Award.
Alk1(flox/flox) and Alk1(+/−) mice were provided by Dr S. Paul Oh (University of Florida). We thank Dr Hua Su and Dr William L. Young (both University of California San Francisco) for critical comments on the article. We thank the Georgia Health Sciences University Georgia Research Pathology Services and Histology Core Laboratory for histology sample preparation services.
The online-only Data Supplement is available with this article at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/STROKEAHA.111.647024/-/DC1.
- Received December 2, 2011.
- Accepted December 12, 2011.
- © 2012 American Heart Association, Inc.
- Oh SP,
- Seki T,
- Goss KA,
- Imamura T,
- Yi Y,
- Donahoe PK,
- et al
- Seki T,
- Yun J,
- Oh SP
- Park SO,
- Lee YJ,
- Seki T,
- Hong KH,
- Fliess N,
- Jiang Z,
- et al
- Holtwick R,
- Gotthardt M,
- Skryabin B,
- Steinmetz M,
- Potthast R,
- Zetsche B,
- et al
- Hashimoto T,
- Lam T,
- Boudreau NJ,
- Bollen AW,
- Lawton MT,
- Young WL