The role of angiogenesis in dural arteriovenous fistulae: the story so far

Abstract

Intracranial dural arteriovenous fistulae are a commonly encountered pathology that can present with a variety of different clinical symptoms. Although there is a significant body of work relating to the natural history and treatment of dural arteriovenous fistulae the exact underlying pathogenesis remains elusive. Various different pathogenetic models have been put forward but there is now a growing body of evidence implicating angiogenesis and the involvement of angiogenetic factors. In this review we attempt to show how the various animal and human studies performed over the past two decades have contributed to the proposed hypothesis on the development of dural arteriovenous fistulae.

Introduction

Intracranial dural arteriovenous fistulae (dAVF) are a relatively frequently encountered vascular shunting lesion. They can present with a variety of different clinical symptoms ranging from pulsatile tinnitus to dementia and intracranial haemorrhage. It is also not an infrequent incidental finding in patients subjected to neuroradiological investigations for unrelated reasons.

The exact pathogenesis of these lesions is yet to be fully understood. A number of different models have been proposed; there is a growing body of evidence to suggest that angiogenesis plays a role in their formation. In this review we attempt to show how the various animal and human studies performed over the past two decades have contributed to the proposed hypothesis on the development of dAVF.

Sinus thrombosis, venous hypertension or both?

The close association between sinus thrombosis and dAVF led to the idea that venous sinus thrombosis itself could result in the formation of dAVF by means of raised venous pressure and the opening of pre-existing arteriovenous communications within the dura.^1–3 According to this hypothesis, the derangement to the venous outflow caused by persistent obstruction results in enlargement of the pre-existing vascular channels and eventually to incompetence of any valves that may be present, which may then lead to the development of dAVF even after the resolution of the thrombus.

In order to determine whether venous sinus thrombosis or local venous hypertension, which itself may be caused by thrombosis, was a prerequisite to the formation of dAVF, Terada et al.⁴ performed a series of animal experiments. Using three groups of rats different anatomical constructs were created. In group 1 (n=22) the left common carotid artery (CCA) and external jugular vein (EJV) were exposed and an anastomosis was created between these two vessels using an end-to-end anastomosis (CCA–EJV anastomosis) in a similar fashion compared to previous investigators.⁵ In group 2 (n = 13) the same CCA–EJV anastomosis was created, followed by occlusion of the contralateral posterior facial vein, which would cause an increase in the venous pressure compared to group 1. The rats in group 3 (n = 11) served as a control group without the formation of a CCA–EJV anastomosis and occlusion the EJV with or without occlusion of the contralateral posterior facial vein. Newly acquired AVFs were seen in three of the animals in group 1 (13.6%), three of the animals in group 2 (23.1%) and in none of the animals in group 3. In addition to the higher rate of arteriovenous fistulae (AVF) formation the authors also noted an altered venous drainage in group 2, as would be expected, in comparison to the other groups, and more prominent venous engorgement in the face and head. Interestingly the authors also noted that in two of the rats in the second group there was regression of the AVF after the CCA–EJV anastomosis was ligated. The authors concluded that venous hypertension alone, without additional thrombosis, can cause the development of dAVF and AVF in the soft tissue of the head and neck. The authors did, however, acknowledge that sinus thrombosis may be a predisposing factor. This study was the first to demonstrate that sinus thrombosis itself is not a prerequisite for the development of AVF, indicating that the observed association between sinus thrombosis and dAVF may be related to the venous hypertension that results from venous outflow obstruction.

The role of venous hypertension and angiogenesis

The experimental evidence published by Terada et al.⁴ supported the theory that venous hypertension may be a prerequisite for AVF formation; however, the nature of the relationship between venous hypertension and AVF remained elusive. Lawton and colleagues⁶ conducted a series of elegant animal experiments that pointed to angiogenesis playing a role in the formation of dAVF. Forty rats underwent a surgical procedure to induce venous hypertension. The investigators created a CCA–EJV anastomosis followed by venous outflow occlusion by means of sagittal sinus thrombosis in addition to bipolar coagulation and cauterisation of the left transverse sinus. In the non-hypertensive control cohort, superior sagittal sinus thrombosis and bipolar occlusion of the left transverse sinus were performed; however, a CCA–EJV fistula was not created and rather the right CCA and EJV were occluded using bipolar coagulation. Postoperatively, there was a significantly raised superior sagittal sinus pressure in the hypertensive cohort compared to their preoperative baseline (Δ 21.9 mmHg, P > 0.0001) as well as in comparison to the non-hypertensive group (P > 0.0001). There was no significant change in the pre and postoperative sinus pressures in the rats without venous hypertension, nor was there a significant difference in the mean arterial pressures in either group. None of the rats without venous hypertension developed either dAVF or facial AVM; however, 57% of the rats in the hypertensive cohort developed dAVF and 70% formed facial AVM. In addition to the angiographic findings, several of the rats from each group underwent dural sampling. Small pieces, approximately 2 × 2 mm, of parietal dura were removed at either 1, 2 or 3 weeks postoperatively. The samples were then immediately implanted into the cornea of the same rat and one week later an angiogenesis index was calculated.⁷ Briefly, the vascular density, scored on a four-point scale, was multiplied by the length of the new vessels, measured from the cornea–sclera junction to the leading edge of the front of the vessels. The angiogenic activity correlated positively with the sagittal sinus pressure (≥0.74) and the mean angiogenesis index for rats with dAVF was 2.3, which was significantly greater compared to the angiogenesis index of rats without dAVF (P = 0.04). The authors had thus shown for the first time that venous hypertension was related to angiogenic activity and the formation of dAVF. This group of authors went on to propose a sequence of events, beginning with venous outflow obstruction that causes venous hypertension, decreased cerebral perfusion, the release of hitherto unidentified angiogenic factors and the development of a dAVF. They also hypothesised that restoration of normal venous pressure may result in either a more benign clinical course or even dAVF regression, the latter having been reported both spontaneously⁸ and following venous flow reconstruction.⁹ Furthermore, if venous hypertension and not thrombosis was the true trigger for the disease, one can explain why only a relatively small subset of patients with sinus thrombosis develop dAVF.

Which angiogenic factors are involved and when?

Although the work of Lawton et al.⁶ had demonstrated that angiogenesis appears to play a role in the formation of dAVF, some questions remain unanswered. One of those was which angiogenic factors were involved. Shin et al.¹⁰ showed that there was an association between dAVF and the expression of vascular endothelial growth factor (VEGF). They induced venous hypertension in 40 rats, which were then divided into two groups. One group (n = 15) underwent immunohistological staining for VEGF one week postoperatively and the other group (n = 15) underwent angiography 90 days postoperatively. Ten rats underwent a sham procedure and served as controls. Neither VEGF expression nor dAVF were observed in the sham group. In the immunohistological study group, VEGF expression in the endothelium and the connective tissues of the dura mater was seen in five rats (33%) and in the neurons in 11 rats (73%) of the cerebral cortex and the basal ganglia were identified. In the angiography group, dAVF formed in six among 15 rats (40%). Subsequently, Shin et al.¹¹ sought to clarify the time course of the release of VEGF and its relationship to venous hypertension. Using male Sprague–Dawley rats (n = 45) the authors created three different anatomical constructs. In group 1, a CCA–EJV anastomosis without venous outflow obstruction was formed. In group 2, there was occlusion of the left draining vein of the transverse sinus and thrombosis of the sagittal sinus but no formation of a CCA–EJV anastomosis. In group 3, there was thrombosis of the sagittal sinus, bipolar occlusion of the left draining vein of the transverse sinus (as in group 2) and a CCA–EJV anastomosis (similar to group 1). Based on these differing anatomical constructs, the animals in group 3 were supposed to harbour the most severe degree of venous hypertension followed by the animals in group 1, with the least venous hypertension in group 2. The investigators, using a western blot technique, showed that the expression of VEGF correlated with the degree of venous hypertension, with the strongest intensity seen in group 3 (P = 0.002), a lower intensity seen in group 1 and the least expression in group 2. In addition, they also looked at the expression of VEGF over time. The expression of immunoreactive VEGF was restricted to the connective tissue and the endothelial layer of the dura matter, cerebral cortical tissue and neurons of the basal ganglia. As may be expected, they showed that the expression of VEGF peaked early on and was highest in the first week after the induction and then steadily dropped.

The expression of VEGF is stimulated by tissue hypoxia,^12,13 which may be caused by venous hypertension, as well as increased internal pressure within blood vessels.¹⁴ Although the investigators did not specifically measure tissue hypoxia, previous studies have demonstrated that their models did result in reduced cerebral perfusion pressure as a consequence of the induced venous hypertension.^15,16 The authors noted that the release of VEGF could be caused by either venous ischaemia or the vascular distension caused by an arteriovenous anastomosis, venous occlusion or a combination of both. Therefore, these investigators showed that VEGF was released in animal models of dAVF and this release occurred principally early on after the formation of the fistula and was related to the degree of venous hypertension. Subsequently, Zhu et al.¹⁷ showed that hypoxia inducible factor 1 (HIF-1) was released very rapidly after venous hypertension was induced. This factor is known to be an upstream regulator of VEGF and the peak of HIF-1 was seen to occur earlier than that of VEGF, 24 hours compared to 7 days. This group also demonstrated that cerebral ischaemia was not required for the release of these factors and that venous hypertension alone was sufficient. The authors monitored cerebral blood flow (CBF) using a laser Doppler needle probe. This confirmed that the CBF was unaltered following surgical intervention as compared to baseline. Thus, these two studies demonstrated that venous hypertension alone can cause the release of VEGF and its upstream regulators and that the degree of venous hypertension may regulate the expression of VEGF.

In humans, Klisch et al.¹⁸ studied the plasma levels of VEGF in 10 patients with dAVF both before and after endovascular treatment. Five of the patients were classified as Cognard I lesions and the remaining five patients were Cognard IIa/IIa+b/III. In their small prospective study the authors found that 80% of the patients had elevated plasma VEGF levels prior to treatment (above 2 standard deviations of published normal values) and interestingly the two patients with normal VEGF levels were both patients with Cognard I lesions. Furthermore, there was a statistically significant difference (P < 0.05) between the plasma VEGF levels between the Cognard I (113.86 pg/ml) and lumped Cognard II+III patients (487.67 pg/ml). There was a greater than 40% drop in the plasma VEGF levels following endovascular treatment which, however, was not statistically significant (P = 0.06). This may be because the study was underpowered. While this is a small study it highlights a possible increase in circulating VEGF levels in patients with dAVF, and this appears to be related to the venous drainage of the lesion. Type I lesions drain anterogradely and type II lesions all have some form of retrograde venous drainage, with subclassification dependent on the exact type of retrograde venous flow, with cortical venous reflux representing the extreme end of retrograde flow. Patients with type II fistulae have higher venous pressure than those with type I fistulae, which was demonstrated previously by Cognard et al.¹⁹ It is also worth noting that VEGF, or other angiogenic factors, may play a role in the progression of low-grade lesions to higher-grade lesions. It has previously been shown that the degree of hypoxia can alter the level of VEGF, with higher levels seen with more pronounced hypoxia.²⁰ Therefore, higher levels of VEGF, which may be related to individual venous architecture and venous collateralisation, may promote a greater degree of neovascularisation as has been seen with other conditions.²¹ Therefore, certain patients may develop higher flow lesions based on their individual VEGF response. It has recently been shown that arterial inflow is related to venous intimal hyperplasia, and so it is possible that higher flow lesions are more likely to occlude secondarily²² and become higher grade. Alternatively, in certain patients, dependent on the venous and arterial flow, there may be a stabilisation of the lesion or possibly even resolution.²³ Further work needs to be conducted on these hypotheses, but we believe that the growing literature on the putative role of angiogenesis in these lesions now supports this kind of work.

But does VEGF release actually cause the formation of dAVF?

The data presented so far have demonstrated the association between venous hypertension and VEGF release but not that this release can actually cause the formation of dAVF. However, in the last couple of years this final piece of the puzzle may have been answered. Li et al.²⁴ sought to address the question of whether VEGF was involved in the induction of dAVF or whether it was merely associated with venous hypertension caused by sinus thrombosis or any associated entity that can trigger dAVF formation. The authors proposed to answer this question by selectively activating or inhibiting the VEGF receptor signalling pathways using a VEGF recombinant adenovirus and a VEGF receptor inhibitor, respectively. They created similar anatomical constructs as previously described, but in one group they injected a VEGF recombinant adenovirus into the distal EJV prior to the formation of the CCA–EJV anastomosis. In another group the brain was lavaged with the VEGF inhibitor valatanib prior to and subsequent to the surgery to create the CCA–EJV anastomosis. Twelve weeks after the surgery angiography was performed and this showed that dAVF formation had occurred in 65% (13/20) of the rats transfected with the VEGF recombinant adenovirus, but that this did not reach statistical significance over the rats with the CCA–EJV anastomosis but without adenovirus infection (43%). However, only two of 21 rats lavaged with valatanib (10%) developed dAVF and the rate of induction in this group was significantly lower than all the other groups, aside from those animals that underwent a sham operation and served as the control group. For the first time it had been shown that the use of VEGF inhibition could inhibit the formation of dAVF, and the results suggest that VEGF released in response to venous hypertension and/or cerebral hypoxia plays a causative role in the pathogenesis of dAVF.

Conclusion

Although the role of angiogenetic factors has long been postulated to play a role in the development of dAVF it is only recently that animal studies suggest a causative role rather than an associative role. The translation of these data to humans will require further studies on patients harbouring dAVF.

Author contribution

PB: generation of hypothesis, manuscript preparation; LLY: manuscript preparation; TK: review and editing: HH: review and editing; MS: guarantor.

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.