Endoglin Deficiency Impairs Stroke Recovery
Background and Purpose—Endoglin deficiency causes hereditary hemorrhagic telangiectasia-1 and impairs myocardial repair. Pulmonary arteriovenous malformations in patients with hereditary hemorrhagic telangiectasia-1 are associated with a high incidence of paradoxical embolism in the cerebral circulation and ischemic brain injury. We hypothesized that endoglin deficiency impairs stroke recovery.
Methods—Eng heterozygous (Eng+/–) and wild-type mice underwent permanent distal middle cerebral artery occlusion (pMCAO). Pial collateral vessels were quantified before pMCAO. Infarct/atrophic volume, vascular density, and macrophages were quantified in various days after pMCAO, and behavioral function was assessed using corner and adhesive removal tests on days 3, 15, 30, and 60 after pMCAO. The association between ENG 207G>A polymorphism and brain arteriovenous malformation rupture and surgery outcome was analyzed using logistic regression analysis in 256 ruptured and 157 unruptured patients.
Results—After pMCAO, Eng+/– mice showed larger infarct/atrophic volumes at all time points (P<0.05) and showed worse behavior performance (P<0.05) at 15, 30, and 60 days when compared with wild-type mice. Eng+/– mice had fewer macrophages on day 3 (P=0.009) and more macrophages on day 60 (P=0.02) in the peri-infarct region. Although Eng+/– and wild-type mice had similar numbers of pial collateral vessels before pMCAO, Eng+/– mice had lower vascular density in the peri-infarct region (P=0.05) on day 60 after pMCAO. In humans, ENG 207A allele has been associated with worse outcomes after arteriovenous malformation rupture or surgery of patients with unruptured arteriovenous malformation.
Conclusions—Endoglin deficiency impairs brain injury recovery. Reduced angiogenesis, impaired macrophage homing, and delayed inflammation resolution could be the underlying mechanism.
Endoglin (CD105) is an auxiliary receptor for the transforming growth factor-β receptor complex. Upregulated endoglin expression has been reported during wound healing and tumor vascularization and in inflammatory tissues and developing embryos.1–3 Mutations in ENG cause type 1 hereditary hemorrhagic telangiectasia (HHT1), a disease characterized by arteriovenous malformation (AVM) in multiple organs and telangiectasia (small AVM) in the skin, mucous membranes.4 Furthermore, a common genetic polymorphism in ENG (207G>A) has been associated with increased risk of sporadic brain AVM.5 Although this polymorphism is synonymous (does not alter encoded amino acid), the A allele reduces the predicted binding score of SRp40, a splicing protein,6 which could influence endoglin protein production.
Patients with HHT1 have a higher prevalence of lung AVMs than other patients with HHT.7 One of the major consequences of lung AVM is the increased risk of paradoxical embolism in the systemic circulation through arteriovenous shunting, which can cause embolic ischemic brain injury. The effect of ENG mutation on recovery from stroke and brain surgery is not clear.
In this study using a mouse ischemic stroke model, we tested the hypothesis that ENG deficiency weakens brain injury repair through impaired angiogenesis and macrophage function. We further analyzed the association between a functional polymorphism in endoglin and outcomes after brain AVM rupture and surgical resection of unrupture AVM in patients with sporadic AVM.
Materials and Methods
The Materials and Methods are described in detail in the online-only Data Supplement.
Animal experimental procedures were approved by the Institution of Animal Care and Use Committees of the University of California, San Francisco, and Duke University and conformed to National Institutes of Health Guidelines for the use of animals in research. Mice were fed standard rodent food and water ad libitum and were housed (5 per cage) in sawdust-lined cages in an air-conditioned environment with 12-hour light/dark cycles. Adult Eng+/– mice8 and their wild-type (WT) littermates were used. Tail genomic DNA was used to identify mouse genotype through polymerase chain reaction using primers: R(R9b): 5′-tgagcctgacgggaaactg-3′, and F(F9a): 5′-accatcttgtcctgagtagcg-3′. Eng expression in the brain of Eng+/– mouse is reduced to ≈59% of the level in WT mouse.9
Eligible patients with AVM with modified Rankin Scale outcome data and DNA for genetic studies were identified from the University of California, San Francisco, Brain AVM Study Project, an ongoing prospective longitudinal cohort study. Patients (n=245) who presented with an intracerebral hemorrhage and had outcome data available after hemorrhage but before treatment were referred to as the natural history cohort. Patients (n=353) who underwent microsurgical excision of a brain AVM between 2000 and 2011 with outcome data available before and after treatment were referred to as the surgical cohort. All patients signed informed consent to participate, and the study was approved by the Committee for Human Research at University of California, San Francisco.
One-way ANOVA was performed followed by Bonferroni post hoc to analyze the difference of the means of infarct/atrophic volume, vessel density, and the number of CD68+ cells at the peri-infarct region among groups.
For analyses of the behavior test, Gaussian distribution was tested using d’Agostino and Pearson omnibus normality test. Equalities of variances were tested using the F test. For dual comparisons, t tests (Student, Mann–Whitney for non-Gaussian distribution, and Welch correction for unequal variances) were used when appropriate. To test the effect of genotype (WT versus Eng+/–) in the acute phase (day 3 after permanent distal middle cerebral artery occlusion [pMCAO]), we performed a Student t test. To test the effect of genotype in the postacute phase (days 7–60 after pMCAO), we used a mixed-effects linear regression model (taking into account the genotype and the number of days since the insult, after checking 2-way interactions involving the genotype) to predict the specified behavior outcome.
Sample sizes were 10 per group for the behavior test, 13 Eng+/– and 12 WT mice for infarct volume measurement on day 1, 7 per group for infarct/atrophic volume on days 3 and 60, 3-week-old and (6 WT and 6 Eng+/–) and 11-week-old (8 WT and 4 Eng+/–) mice for collateral quantification, and 6 per group for vessel density and CD68+ cell quantification. Data are presented as mean±SD. A 2-tailed P value <0.05 was considered statistically significant.
For human data, univariable and multivariable logistic regression analyses of outcomes were performed. Results are reported as odds ratios (ORs) and 95% confidence intervals (95% CIs). Before constructing multivariate models, we tested whether the effect of ENG 207G>A genotype on outcomes was modified by other predictors (ie, interactions) using likelihood-ratio testing. A significant interaction (P=0.05) was observed in the surgical cohort between ENG 207G>A genotype and hemorrhagic presentation, thus all surgical results are presented stratified as ruptured versus unruptured. Predictors in the multivariable models included ENG 207G>A genotype, age at presentation, sex, race/ethnicity (whites versus nonwhites), AVM size, exclusively deep venous drainage, and eloquent location.
Endoglin Deficiency Delayed Functional Recovery
Adhesive removal and corner tests were used. After the training phase, the baseline behavior results were similar in all groups. At the acute phase of stroke (day 3 of pMCAO), both WT mice and Eng+/– mice took longer to remove the adhesive from their right paws (WT, 25±15 versus 7±2 s; P=0.001 and Eng+/–, 38±7 versus 6±1 s; Figure 1A) and made more left turns (lesion site, WT, 75±8% versus 48±6%; P<0.001 and Eng+/–, 76±8% versus 47±3%; P<0.001; Figure 1B) when compared with their baseline performance. The adhesive removal time from the left paw was not affected (P=0.73; Figure IA in the online-only Data Supplement). There is no difference between Eng+/– and WT mice (adhesive removal, P=0.13 and corner tests, P=0.78; Figure 1A and 1B).
In the postacute phase (days 15–60 post-pMCAO), Eng+/– mice took longer to remove tapes from the right paw on day 15 (16±6 versus 33±12 s; P=0.002), day 30 (13±6 versus 29±17 s; P=0.02), and day 60 (11±4 versus 25±17 s; P=0.03; Figure 1C), and made more left turns than WT mice on day 15 (66±6% versus 79±11%; P=0.002), day 30 (60±5% versus 77±10%; P<0.001), and day 60 (60±9% versus 72±11%; P=0.01; Figure 1D) after pMCAO. Both genotype (OR, 15.5; 95% CI, 8.0–11.5) and the timing post-pMCAO (OR, 0.14; 95% CI, 0.05–0.24) influenced performance in removing adhesive from the right paw and in the corner test (genotype: OR, 14.1; 95% CI, 8.5–19.7; P<0.001 and timing: OR, 0.13; 95% CI, 0.05–0.20; P=0.001). Left paw adhesive removal was not affected (Figure IB in the online-only Data Supplement).
Eng+/– Mice Had More Severe Brain Injury
Eng+/– mice exhibited larger infarct volume than WT mice (19.7±6.5 versus 12.6±8.9 mm3; P=0.03) on day 1 (Figure 2A; Figure IIA in the online-only Data Supplement) and on day 3 (22±6% versus 16±6%; P=0.04; Figure 2B; Figure IIB in the online-only Data Supplement) and larger atrophic volume than WT mice on day 60 after pMCAO (21.27±5% versus 13.4±6%; P=0.03; Figure 2C; Figure IIC in the online-only Data Supplement). These data indicate that ischemic insult caused more severe brain damage in Eng+/– mice.
Endoglin Deficiency Impairs CD68+ Macrophage Recruitment or Clearance
CD68+ macrophages were quantified in the peri-infarct region right outside the infarct border on days 3 and 60 after pMCAO. Eng+/– mice had fewer microphages on day 3 (34±6% of total cells versus 40±4%; P=0.009) and more macrophages on day 60 (17±4% versus 13±3%; P=0.02) when compared with WT mice (Figure 3), indicating that endoglin deficiency delayed macrophage homing and clearance or promoted macrophage homing at the postacute stage of ischemic stroke.
Endoglin Deficiency Reduced Angiogenic Response to Ischemic Brain Injury
Among the factors determining infarct volume after stroke, collateral vessel anatomy (especially the collateral vessel number between the anterior cerebral artery and the MCA) has been proposed as a major factor determining infarct volume.10 We quantified the collateral vessel numbers before pMCAO and found no difference between the Eng+/– and WT mice at 3 (21±1.8 versus 21±0.8 vessels/mm2; P=0.56) and 11 (20±2.1 versus 21±2.9; P=0.6; Figure 4A and 4B) weeks. However, microvascular density in the peri-infarct area was lower in Eng+/– mice than in WT mice (417±69 versus 490±52 vessels/mm2; P=0.05; Figure 4C and 4D) on day 60 after pMCAO, indicating that endoglin deficiency reduced postischemic angiogenesis.
ENG 207G>A Allele Is Associated With Poor Outcome After Brain AVM Surgical Resection in Unruptured Patients
Characteristics of patients with good and bad outcomes, stratified by hemorrhage, are presented in Table I in the online-only Data Supplement. In the unruptured surgical group, patients with poor outcome had higher Spetzler–Martin scores (P=0.05). Those with poor outcome were also more likely to be carriers of the A allele of ENG 207G>A (P=0.01) and were at higher risk of poor outcome when compared with unruptured patients carrying only the G allele (univariate OR, 3.57; 95% CI, 1.22–10.43; P=0.02). A corresponding effect was not seen in surgical patients with ruptured AVM. Multivariable logistic regression analysis confirmed the association between the A allele and poor functional outcome, independent of other risk factors, among unruptured (OR, 3.65; 95% CI, 1.15–11.54; P=0.03) but not ruptured (OR, 0.67; 95% CI, 0.18–2.51; P=0.55; Table 1) patients. A sensitivity analysis restricted to white subjects (n=139) yielded similar results (data not shown).
ENG 207G>A Allele Is Associated With Increased Risk of Poor Neurological Status in the Natural Course of Brain AVM Rupture
Poor outcome patients were also more likely to carry the A allele (P=0.05) and had shorter latencies between hemorrhage and treatment (P<0.01). There were no notable differences in patient age, race/ethnicity, Spetzler–Martin distribution, AVM size, drainage pattern, or location (Table II in the online-only Data Supplement) for those included in the natural history cohort.
Logistic regression results (Table 2) show that the A allele of ENG 207G>A was associated with having a modified Rankin Scale >2 after AVM rupture after adjustment for baseline modified Rankin Scale and time between hemorrhage and modified Rankin Scale assessment (OR, 2.88; 95% CI, 1.10–7.75; P=0.03). Multivariable analysis showed this effect to be independent of other risk factors (OR, 3.21; 95% CI, 1.19–8.68; P=0.02).
In this study, we demonstrated that endoglin deficiency impairs brain ischemic injury repair. Eng+/– mice had more severe functional defects and larger atrophic volume than WT mice, which is associated with delayed macrophage homing and clearance or enhanced macrophage homing at the subacute stage. Although Eng+/– and WT mice had a similar number of collaterals between anterior cerebral artery and MCA at baseline, Eng+/– mice had lower vessel density at the peri-infarct region than WT mice 60 days after pMCAO, which suggests that endoglin deficiency leads to reduced response to angiogenic stimulation. Furthermore, we found that in humans, there was an association between functional polymorphism in endoglin and outcome of brain AVM rupture and surgery.
Endoglin is highly expressed in proliferating vascular endothelial cells11 and is elevated in the settings of inflammation and wound healing.12 After ischemia-reperfusion injury and myocardial infarction, endoglin is upregulated in the ischemic area and border zone.13 Microvascularity within the infarct zone was strikingly lower in Eng+/– mice than in WT mice. We also found that Eng+/– mice had lower vessel density in the peri-infarct region after pMCAO. Expression of vascular endothelial growth factor (VEGF) is reduced in cultured Eng+/– endothelial cells.14 Both VEGF receptor 1 and 2 are reduced in Eng+/– macrophages at baseline and after VEGF stimulation.15 Eng+/– macrophages and WT macrophages express similar level of Mmp9 at baseline. Mmp9 expression is increased in WT, but not in Eng+/–, macrophages after VEGF (50 ng/mL) treatment.15 In the brain, VEGF-induced upregulation of VEGF receptor 2 expression is greatly impaired in Eng+/– mice.16 Together, these data suggest that reduced angiogenesis in response to injury is one of the mechanisms for impaired tissue repair in Eng+/– mice after pMCAO.
Interestingly, Eng+/– mice had more severe brain injury than WT mice even on day 1 after pMCAO, which could not be explained by impairment of tissue repair. Hypoxia induces endothelial endoglin expression,17 and endoglin prevents apoptosis in hypoxia endothelial cells.18 Hence, vascular damage in Eng+/– mice could be more severe than in WT mice after ischemic insult. Endoglin haploinsufficiency has been associated with reduced nitric oxide production and increased endothelial nitric oxide synthase–derived superoxide production.19 Bioavailability of nitric oxide is lower in Eng+/– mice than in WT mice.20 Nitric oxide produced by endothelial cell induces vascular relaxation.21 Thus, reduced vessel relaxation and increased vascular damage, together with oxidative stress (more superoxide production), in Eng+/– mice enhance brain injury during the acute stage of ischemic stroke.
In addition to endothelial cells, several other cell types also express endoglin. For example, endoglin is present in monocytes and is upregulated during the monocyte to macrophage transition.22 Recruitment of human monocytes to the infarcted murine heart and subsequent vessel formation was severely impaired when HHT1 monocytes were used.13 Endoglin deficiency in endothelial cell reduced leukocyte adhesion and transmigration23 and impaired the endothelial-autonomous capacity to upregulate stromal cell–derived factor 1 (SDF-1) expression in response to hindlimb ischemic injury.24 In this study, we found that Eng+/– mice had fewer CD68+ cells in the peri-infarct area at 3 days and more CD68+ cells at 60 days after pMCAO. Taken together, endoglin deficiency seems to impair monocyte adhesion and migration. Additional studies are needed to determine whether increased macrophage in Eng+/– mice at the postacute stage is the result of delayed clearance or prolonged/enhanced macrophage homing at this stage.
The roles of postischemic inflammation are bidirectional.25 The postischemic inflammatory response contributes to secondary brain injury, and inflammation can be detrimental, because the influx of inflammatory cells amplifies brain cell death. However, during recovery inflammation may be construed as plastic forms of tissue remodeling.26 How the impaired macrophage homing and clearance impair brain injury recovery need to be explored in future studies.
ENG is a causative gene of HHT1. The prevalence of brain AVM in patients with HHT1 is 1000-fold higher than the prevalence in the general population (10/100 000).27 A common genetic polymorphism in ENG (207G>A) has also been associated with increased risk of sporadic brain AVM.5 Furthermore, patients with HHT1 have a high prevalence of lung AVMs,7 which increases the risk of embolic ischemic brain injury. An insertion/deletion polymorphism in ENG has also been associated with sporadic primary intracranial hemorrhage.28 Our study is the first to demonstrate that ENG polymorphisms are associated with recovery after brain AVM rupture or surgery in patients. Thus, endoglin may play a crucial role in all types of injury repair, including ischemic stroke, hemorrhagic stroke, or iatrogenic injury. Using an ischemic stroke mouse model, we showed that reduced angiogenesis, impaired macrophage homing, and delayed inflammation resolution could be the underlying mechanism. Additional studies are needed to determine whether the same mechanism applies to all brain injuries.
In this study, we have demonstrated that endoglin deficiency exacerbates ischemic brain injury and delays neurobehavioral recovery in mice, which is associated with reduced angiogenesis, impaired CD68+ cells homing, and delayed inflammation resolution. Endoglin may also play a role in surgical and hemorrhagic recovery in patients with human brain AVM. Understanding how endoglin is involved in different kinds of brain injury could provide new therapeutic opportunities to improve outcomes in patients.
We thank Voltaire Gungab, MA, for assistance with article preparation, and members of the University of California, San Francisco, Brain AVM Study Project (http://avm.ucsf.edu) for their support.
Sources of Funding
This study was supported by grants from the National Institutes of Health: R01NS027713 and P01NS044155 to Dr Su; R01NS034949 and P01NS044155 to Dr Kim, and R01HL097281 to Dr Marchuk. Additional support was provided by a grant to Dr Young from the Michael Ryan Zodda Foundation.
Dr Maze is a Board Member of the Foundation for Anesthesia Education and Research, Smarttots, and the Society for Anesthesia and Sleep Medicine and is supported by a grant from the National Institutes of Health.
The online-only Data Supplement is available with this article at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/STROKEAHA.114.005115/-/DC1.
- Received February 7, 2014.
- Revision received April 10, 2014.
- Accepted April 17, 2014.
- © 2014 American Heart Association, Inc.
- Dallas NA,
- Gray MJ,
- Xia L,
- Fan F,
- van Buren G 2nd.,
- Gaur P,
- et al
- Pawlikowska L,
- Tran MN,
- Achrol AS,
- Ha C,
- Burchard EG,
- Choudhry S,
- et al
- Cartegni L,
- Wang J,
- Zhu Z,
- Zhang MQ,
- Krainer AR
- Satomi J,
- Mount RJ,
- Toporsian M,
- Paterson AD,
- Wallace MC,
- Harrison RV,
- et al
- van Laake LW,
- van den Driesche S,
- Post S,
- Feijen A,
- Jansen MA,
- Driessens MH,
- et al
- Jerkic M,
- Rodríguez-Barbero A,
- Prieto M,
- Toporsian M,
- Pericacho M,
- Rivas-Elena JV,
- et al
- Choi EJ,
- Walker EJ,
- Degos V,
- Jun K,
- Kuo R,
- Pile-Spellman J,
- et al
- Zhu Y,
- Sun Y,
- Xie L,
- Jin K,
- Sheibani N,
- Greenberg DA
- Li C,
- Issa R,
- Kumar P,
- Hampson IN,
- Lopez-Novoa JM,
- Bernabeu C,
- et al
- Toporsian M,
- Gros R,
- Kabir MG,
- Vera S,
- Govindaraju K,
- Eidelman DH,
- et al
- Rossi E,
- Sanz-Rodriguez F,
- Eleno N,
- Düwell A,
- Blanco FJ,
- Langa C,
- et al
- Young K,
- Conley B,
- Romero D,
- Tweedie E,
- O’Neill C,
- Pinz I,
- et al