Preclinical ischemic stroke is at the crossroads in search of reliable and robust simulation models as past experiences with their translation from the laboratory to the standard of clinical care have often been disappointing. The efficacy of neuroprotective agents is still elusive, and the use of thrombolytics alone is limited to the narrow time window of presentation from the onset of the deficit. Hence, the focus has shifted to interventional revascularization to salvage the parenchyma at the risk of infarction. As the burden of disease morbidity and mortality is so enormous, neurointerventionalists have adopted a more aggressive approach to mechanical revascularization with the limited approved tools available—the Penumbra and the MERCI retrieval system, and the recently incorporated stent retrievers. In fact, the interventional space is among the fastest growing fields in stroke research today. Assessing treatment efficacy in these scenarios is infinitely complex as the heterogeneity of the cerebrovasculature, physical and mechanical nature of the occlusive embolus and the time of presentation are all confounders in assessing treatment outcomes. As no single thromboembolic model is apt to address all of these questions, an integrated methodology with a combination of both in vitro and in vivo assessment needs to be adopted. This involves clinically relevant thromboembolic analogs in device evaluation in vascular replicas, thromboembolic stroke induction in large animal gyrencephalic ischemic stroke models for thrombolytic, imaging and neuroprotection research and a native cerebrovascular target for evaluation of the safety and efficacy of mechanical thrombectomy devices.
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
Ischemic stroke accounts for 87% of all strokes that may result either from an embolism of material originating elsewhere in the vasculature that impedes flow to the brain or from in situ thrombosis of a diseased cerebral vessel.1 A wide spectrum of reliable acute ischemic stroke (AIS) animal models exist and have been very well characterized for elucidating stroke pathophysiology. These models further serve to study and replicate the effects of reperfusion after restoration of blood flow through the occluded vessel and mechanisms of hemorrhagic transformation.
Despite recent progress in imaging, pathophysiological characterization, neuroprotection, thrombolytic and mechanical recanalization strategies, AIS remains challenging for the neurointerventionalist. Realistic preclinical AIS models have the potential to bridge the chasm and lead to an early translation of these advances to stroke therapy. The advent of endovascular mechanical retrieval and intra-arterial thrombolysis perhaps offers a paradigm change, allowing real time visualization and assessment of the response to therapy.
The central challenge for preclinical modeling of AIS is the assessment of safety and efficacy of mechanical thrombectomy in the absence of stroke. Currently, efforts at preclinical modeling designed to address this question are limited to extracranial vascular occlusion modeling with the efficacy endpoint that is angiographic evidence of revascularization and local vascular histopathology for safety. The composition of and environment around the systemic vasculature is different than that of the cerebrovasculature. Numerous relevant questions remain, such as which patients should receive intra-arterial treatment, how effective is a given device in the salvage of the penumbra, what factors participate in the success of a given thrombectomy device in reducing morbidity and mortality, does clot composition or stroke etiology favor one technology over another for mechanical revascularization? Furthermore, there is a cautionary tale of decades of flawed preclinical stroke modeling leading to failed clinical trials in neuroprotection. We believe that it is a challenge to the neurointerventional surgery community to identify and execute thoughtful research into preclinical modeling for intra-arterial intervention. Indeed, we would argue that it is a moral imperative.
Cerebral infarction is a heterogenous clinical entity with a variety of factors such as etiology, localization, severity of ischemia and coexisting systemic diseases determining the outcome. These factors make clinical stroke characterization in humans challenging. Many of these variables can be eliminated by employing an appropriate AIS ischemic model, enabling neuroscientists to focus on fundamental questions.
As with animal models of any disease state, it is essential to employ a model which is most appropriate for the question being addressed. In many instances, the hypothesis is too complex for any one model system to adequately provide definitive confirmation. Therefore, it is becoming more common to formulate a research plan that employs multiple models designed to study potential therapies. The study of animal models for ischemic stroke can be divided into a study of the mechanism employed to induce infarction and the species used for stroke induction.
Mechanism of ischemic injury
Focal AIS may originate from an embolism from an upstream source, decreased cerebral perfusion (ie, atherosclerotic narrowing or vasculitis) or a thrombotic occlusion. Various animal stroke models have been devised to simulate this pathophysiology.
Mechanical MCA occlusion
Focal AIS models have been characterized by decreasing cerebral blood flow (CBF) to a distinct arterial territory. This may be realized by mechanical occlusion of the middle cerebral artery (MCA) territory by advancing an occlusive material endovascularly, such as silicone coated suture material, or by directly exposing the vessel surgically and externally occluding it. The cerebral vessel occlusion produced is highly reproducible, and is often reversible allowing the study of tissue reperfusion after ischemic injury. These mechanical occlusion models however do not accurately replicate the hemodynamic features of thrombolytic reperfusion. They also do not allow for thrombolytic and mechanical thrombectomy device evaluation. Also, when craniectomy is employed in the modeling approach, decompression may have a profound beneficial effect on survival and functional outcome.
Two methods of AIS induction through thrombotic vessel occlusion in the animal models exist. One method is the direct induction of thrombus formation by in situ thrombin infusion to the MCA origin which has been well described in rats, mice and rabbits.2–4
In another well described method, photothrombosis, stroke induction proceeds by the focal illumination of the target cerebral vessel through an intact skull, following the systemic intravenous injection of a photosensitive dye Rose Bengal. Thrombosis is probably induced by the rapid reaction of the reactive oxygen species and subsequent local endothelial damage with local platelet adherence and aggregation producing microvascular occlusion.5 With a low level of invasiveness, small infarcts can be created in functionally distinct cortical areas in a reproducible fashion. This model remains important in studying the neuropathophysiology relating to cortical infarction. However, this model has significant limitations and does not completely simulate human thrombotic stroke. The rapidity and severity of the ischemic insult involving the microvasculature and macrovasculature results in a sharply defined infarct core and a rapid development of the breakdown of the blood–brain barrier. There is complete absence of penumbral tissue due to the inability to rechannel the blood flow through the collateral circulation which is different than what is seen in human subjects. Also, this model cannot be used to adequately study the effects of reperfusion and thrombolysis. MRI of the photothrombotic stroke shows early increases in T2 signal in combination with decreased apparent diffusion of water which indicates simultaneous vasogenic (extracellular) and cytotoxic (intracellular) edema—a pattern not representative of AIS.6
Microspheres, macrospheres and autologous/allogenic blood clots have been used to create embolic occlusion of the cerebral vessels.
Although technically more challenging, thromboembolic models that use prepared blood clots to create a focal cerebrovascular occlusion are very relevant in studying thrombolysis, neuroprotection and mechanical reperfusion strategies. They also allow for an evaluation of combined thrombolytic and mechanical revascularization strategies. Also, large animal thromboembolic models can be used for the evaluation of catheter directed thrombolysis and mechanical revascularization. However, due to the lack of control over the final destination of the injected thromboemboli, the occlusion produced is often variable both in size and location. Due to these inconsistencies, these models pose a challenge in AIS research. Further in this review, we will focus on the thromboembolic AIS models which are amenable to thrombolytic, neuroprotection and/or endovascular strategies.
Animal AIS models
Rigorous guidelines laid down by the Stroke Therapy Academic Industry Roundtable (STAIR) criteria for a robust and detailed preclinical evaluation of AIS therapies aim to provide an increased success rate for pharmacological and interventional therapies preceding clinical trials.7 These recommendations suggest that experiments should demonstrate therapeutic efficacy when performed across different species, and their pharmacokinetic and toxicology properties assessed. In the aftermath of the failure of NXY-059 to sustain and replicate the neuroprotection benefit in the SAINT clinical trials, a more critical review of the preclinical study design is mandated with the use of biologically relevant functional outcome measures and multimodal imaging to follow in vivo stroke progression.8
Lissencephalic stroke models
Rodents possess an agyric or a pachygyric (lissencephalic) brain.
Rat and mice stroke models
Rodent stroke models are by far the most prevalent for ischemic stroke modeling. Among their many advantages are a remarkable genetic homogeneity and the ability to induce reproducible infarcts. Their smaller size facilitates analysis of the physiology and brain histopathology, and there is general ethical acceptance for these stroke models. The rodent models offer a higher versatility and a lower cost which makes designing sufficiently powered scientific studies much more feasible.9 However, their small vascular size precludes their use in evaluating mechanical endovascular devices. A lissencephalic brain has a higher grey to white matter ratio which does not represent the cortical ratios in humans. An exciting area in preclinical stroke research involves testing of novel therapeutic strategies through gene manipulation in transgenic mice. Most rat stroke models can be easily adapted to mice, however their survival post-surgical procedures is lower than in rats and they have a tendency to become hypothermic post-stroke resulting in better histological and behavioral scores.10
The most widely employed model today to study thrombolytic and neuroprotection strategies is the thromboembolic clot injection, which was first described in dogs and later applied to the rat.11 Earlier adaptations of the model confirmed that the size (length and diameter) and the biological characteristics of the blood clot (fibrin rich) are critical for the relevance and reproducibility of the model. Busch et al developed a single autologous clot injection method which produced reliable occlusion of the proximal MCA and consistent reduction of CBF.12 Dynamic thrombus formation by withdrawing blood in a thrombin laced catheter positioned proximal to the MCA origin and subsequent embolization has also been described.13 Use of multimodal MRI imaging is extremely useful in the visualization of the anatomic, diffusion, perfusion and functional data temporally for an assessment of the response to lytic therapy.14 Despite widespread adaptation, no treatment regimen that has shown reduced cerebral infarction in the rat model has translated to the clinic.15
Rabbit stroke model
Reversible occlusion techniques using a silicon rubber cylinder embedded in a nylon suture to study reperfusion and surgical ligation or clipping of the MCA via the retro-orbital approach have previously been described. However, the rabbit cerebral clot embolism model in particular has distinct advantages over other animal models as their endogenous thrombolytic profile closely matches that of humans outside of the non-human primates (NHP). This has proved particularly useful in establishing the efficacy of thrombolytic therapies such as tissue plasminogen activator.16
Two popular rabbit embolic stroke models described are the rabbit small clot embolism model (RSCEM) or the rabbit large clot embolism model (RLCEM).17 The key advantage of both of these is the ability to induce cerebral ischemia in an awake, intact, normothermic New Zealand white rabbit. The emboli may be seeded with radiolabeled particles to quantify the amount of emboli that make it to the brain and their distribution. After anesthetizing the rabbits, one of the carotid bifurcations is exposed and the external carotid artery is ligated. A catheter is then introduced and secured in the common carotid artery. After closing the incision, leaving the distal end of the catheter extracorporeal, the animals are allowed to recover for a minimum of 3 h. Thereafter, the solution containing the emboli is injected and the animals are observed at various time points following the induction of cerebral ischemia. The outcome of these experiments is measured by simplified neurological assessments. Importantly, the rabbit embolic stroke model was used in the first report of the safety and efficacy of tissue plasminogen activator as an AIS treatment modality. The RSCEM causes a heterogenous stroke and is useful for making neurophysiological assessments. The RLCEM by contrast, produces a more homogenous stroke. Clot placement is confirmed by CBF measurements, neurological evidence of cerebral ischemia in awake animals or through measurement of radioactivity in the brain from the radiolabeled clot.
Another means of stroke induction is accomplished through injection of thermally coagulated autologous blood through a microcatheter under fluoroscopic guidance.18 The clot placement at the bifurcation of the common carotid artery can be confirmed angiographically. This allows real time assessment and localization of the occlusion. However, the diameter of the rabbit cerebrovasculature limits its use in the evaluation of thrombectomy devices (figure 1A, B).
Recently, a selective MCA occlusion model has been described in rabbits.3 Briefly, a 1.2 French microcatheter is advanced selectively into the rabbit internal carotid artery (ICA) beyond the posterior communicating artery under fluoroscopic guidance for slow thrombin infusion (10 NIH U/100 μl). There is minimal filling of the anterior cerebral artery (ACA) due to contralateral ACA flow. MCA occlusion is confirmed angiographically. Subsequent plasmin injection demonstrated rapid lysis of the clot. This model has potential for evaluating intra-arterial thrombolytic recanalization; however, slow delivery of the coagulant mixture may result in the occlusion of the distal MCA branches. In comparison, human thromboembolic proximal MCA occlusion allows for the preservation of collateral flow distal to the occlusion.
These lissencephalic models present with several limitations in AIS research. They are not anatomically representative of the human brain. Also, their cerebrovasculature is unfavorable for endovascular access and the evaluation of neurothrombectomy devices. Due to their small brain size they require high resolution small animal imaging MR, positron emission tomography and CT scanners, as the spatial resolution of the clinical systems is often not adequate. The STAIR recommendations recognize the translational limitations of the rodent AIS models and suggest a sequential hierarchy of drug and device testing from small animal models to large animal stroke models prior to initiating enrollment for clinical studies.7
Gyrencephalic stroke models
The time window for tissue salvage in cerebral white matter ischemia is larger when compared with grey matter stroke. In addition, the infarction thresholds for the grey and white matter are different.19 Hence it is suggested that modeling cerebral ischemia in an appropriate gyrencephalic stroke model is a more relevant endeavor. Ischemic stroke has been modeled in various large animal models, such as pigs, sheep, cats, dogs and NHP.
Swine and sheep models
Although these animals have a considerable advantage of size and cortical complexity, endovascular access of the anterior cerebrovasculature is impeded by the rete mirabile, a complex plexiform arterial network which re-coalesces to form the ICA. This endogenous vascular filter (arteriolar diameter <1 mm) precludes the induction of anterior thromboembolic strokes, and the intracranial use of thrombectomy devices is impossible in these animals (figure 1C, D).20
Canine AIS model
The canine thromboembolic AIS model described by Hill et al has historically been the most relevant, using autologous blood clots to occlude the canine MCA.11 Recent studies have returned to the original canine embolic stroke model, using autologous blood clots injected into the ICA via an angiographic catheter to create an occlusion of the MCA.21 The occlusion can be reversed by the intra-arterial administration of recombinant tissue plasminogen activator. These models have been used to drive nascent imaging advances in the c-arm CT measurement of cerebral blood volume, which when fully realized has the potential to provide an opportunity to triage, treat and track the dynamic progression of the changes in tissue viability temporally and in the response to therapy.22 23
One of the advantages of the canine model is its compatibility with non-specialized, clinical imaging equipment, unlike smaller animal models. Additionally, vessel size and the degree of vasospasm encountered in dogs are more similar to humans than rodent and porcine models and allows for a more targeted arterial embolization. The caliber of the intracranial vasculature is sufficient to permit the investigation of new endovascular devices, an area that in recent years has been gaining traction and of specific interest to readers of this journal. Vessel occlusion with autologous clots also makes this model appropriate for studying thrombolytic agents.
Having described what would appear to be at first blush a remarkable simulation of AIS, excitement must be tempered. Like previously described techniques, it has its own set of limitations. The tortuosity of the canine ICA is prohibitive for testing of endovascular devices designed for retrieval and obliteration of the thrombus. For this reason, autologous clots may also be delivered into the posterior cerebral circulation, occluding either the vertebral or basilar artery. The catheter is tracked through the anastomosis between the vertebral and anterior spinal arteries, and then through the vertebral–basilar junction to inject the clot into the basilar artery. This technique was used to evaluate the safety and efficacy of intra-arterially administered reteplase and intravenous alteplase.24 Levy et al investigated the potential use of intracranial stents in recanalizing large clot embolisms that had occluded the distal vertebral artery.25
The dog model of focal cerebral ischemia is challenging due to an extensive collateral circulation via the maxillo-carotid and meningocerebral anastomoses that makes it difficult to obtain reproducible infarction. Also, the canine brain contains less white matter than the human brain, which affects the metabolic demands of the brain tissue and threshold for cell death. This difference could make extrapolating canine data to human subjects difficult. Similar to humans, vascular occlusions that are anatomically identical in different animals can produce very different infarcts (figure 2). The heterogeneity of the infarct volume due to variability of pial collaterals, although highly representative of human stroke, introduces the requirement of large sample sizes to ascertain treatment effects.
Non-Human Primate model
NHP are the penultimate species for preclinical stroke simulation since they have the closest phylogenetic homology to humans. Extrapolation of data from pharmacological dosing regimens, toxicology and drug efficacy studies in the NHP may be better applied to clinical trials of therapy. The NHP have distinct advantages over other methods as their vascular anatomy is analogous to humans. Their gray matter to white matter ratios, neurophysiologic response to treatment and their complex behavior and recovery patterns are closer to humans.7
In early NHP thromboembolic models, several autologous blood clots (n=5, diameter 1 mm, length 5–7 mm) were injected into the ICA. This caused a heterogeneous pattern of intracranial arterial occlusion in the ACA, MCA and the posterior cerebellar artery territories of one and both hemispheres with variable degrees of alterations in EEG and sensory evoked potentials.26 In addition, this model was not very reliable and often had early spontaneous recanalization and inconsistent occlusions. However, an ideal embolic model to study interventional approaches to large vessel recanalization should entail a blood clot that lodges in the proximal segment of the MCA without spontaneous fragmentation. Smaller clots may embolize distally into the arterial tree whereas large clots may lodge too proximal to cause an MCA occlusion. Therefore, the size and characteristics (ie, flexibility) of the clot are important determinants for model success. Recently, a more selective intra-arterial approach with a single autologous clot injection (0.9 ml blood mixed with 0.1 ml of 20 units of thrombin for 15 min) via a 2.3 French catheter into the supraclinoid segment of the ICA of Rhesus monkeys (Macaca mulatta) or cynomolgus monkeys (Macaca fascicularis) has been described.27 Thrombolytic therapy with reteplase with or without coadministration of abciximab tended to improve recanalization compared with untreated controls. However, despite this selective approach, a second clot injection was required in 19% of the cases (one animal was excluded because no persistent occlusion was achieved after the second clot deposition), and highly variable degrees of recanalization and infarction (eg, no or small infarcts in 50% and large infarcts in 50% of control animals) were observed precluding detection of significant differences.27 A more predictable and reproducible clot model was developed by Kito and colleagues.28 This involved the injection of a single clot through the ICA of the cynomolgus monkey. The animals were monitored by observing a consistent post-occlusion reduction of CBF (less than 25% of pre-occlusion values) as measured by positron emission tomography in the temporal cortex and caudate putamen, histological damage in the ipsilateral MCA territory and stable neurological deficits. Visual inspection demonstrated no spontaneous clot lysis or recanalization within 24 h. Based on positron emission tomography studies in this model, the same group suggested that the threshold for ischemic infarction was approximately 12–15 ml/100 g/min.29 Furthermore, peri-infarct tissue (ie, the ischemic penumbra) may be present for several hours, indicating a relatively longer therapeutic window compared with rodents. Importantly, intra-arterial thrombolytic therapy with urokinase initiated within 1 h after embolism achieved partial to complete recanalization, significantly reduced final infarct volume and improved the neurological deficit score in this model.30 In conclusion, the single clot model probably best approximates human embolic stroke, which is caused by a single embolus in the MCA territory in a significant proportion of cases.31 Therefore, this model is highly promising for studying the pathogenesis of ischemic stroke and thrombolytic therapy.
Unlike cats and dogs, which rely heavily on the extracranial maxillo-carotid anastomosis and vertebrobasilar system, the angiogenic branching patterns of the ICA in NHP bear remarkable similarity to humans. They possess a complete circle of Willis, and an M1 occlusion results in basal ganglia and white matter ischemia with variable cortical infarction. Despite these similarities, a few distinct differences exist. These include the singular ACA and an occasional anomalous origin of the lenticulostriate arteries from the ICA in baboons. The angioarchitecture as expected permits exploration and selective catheterization with standard endovascular devices.32 However, the ICA and MCA luminal diameters are considerably smaller than humans, and are comparable with those of canines.33 Also, the financial undertaking involved in terms of personnel expertise, housing, enrichment and ethical considerations should not be underestimated.
Parameters for evaluation in an ischemic stroke model: a complete simulation
AIS is a complex pathophysiologic entity and no single model can simulate all the variables experienced in human AIS. The advantages and limitations of each model need to be kept in perspective. The experimental design must ascertain the relevant clinical questions which need to be addressed. Also, it must be established whether it is more appropriate to limit the variability or to validate the robustness of the treatment, despite the heterogeneity. The lack of translation of the therapeutic success achieved through AIS models may lie in the inappropriate interpretation of the data derived from the clinical designs rather than a limitation of the model itself.34 This is not a new phenomenon as various successful cancer therapies, myocardial regeneration or angiogenic preclinical therapies have failed to extrapolate their benefit onto large scale clinical trials.
Fortunately, the simulation of thromboembolic stroke in preclinical models can be segregated into various themes, and individual utilization of each model for optimization of the parameters involved may be employed. This approach entails a judicious and rational use of a combination of both in vitro and in vivo models for achieving predefined objectives.
The in vivo stroke induction model which best mimics the pathophysiologic mechanism of stroke is the thromboembolic model. Ideally, this should be a large animal gyrencephalic model which is best for advanced imaging and studying stroke pathophysiological mechanisms, and thrombolytic and neuroprotection strategies. These parameters are best met by the canine and NHP models.
Representative thromboembolic analogs
A crucial component, often overlooked in AIS modeling, is the characterization of the mechanobiology of thromboemboli35 and a close study of the interactions between the vascular wall and the thrombus. Simulation of these experiences with in vivo and in vitro research models will assist in the critical evaluation of the thrombectomy devices. Parameters analyzed for thrombectomy devices which may be influenced by these factors are the number of retrieval attempts for revascularization, amount of the thromboemboli removed and the clot fragmentation resulting in a distal embolic shower.
Assessment of device efficacy and safety
Vascular replicas representative of the human angioarchitecture are necessary for bench top and in vitro thromboembolectomy device design optimization and for improvisation of their efficacy.36 It is also essential to do in vivo device testing in preclinical animal models. The porcine extracranial circulation (figure 1 C,D) and the femoral artery are the most commonly utilized vascular territories for embolectomy device testing. The vessel diameters resemble the human cerebral arteries (2.5–3.0 mm), a similar distance from the catheter sheath (puncture site) and the target vessel and passage through the aortic arch and carotid arteries by the endovascular devices.37 They allow for the testing of the efficacy of these devices to retain and retrieve the thrombus while preventing the risk of clot fragmentation and distal embolization in a hemodynamically analogous environment. They also allow for an in vivo assessment of vessel vasospasm, stretching, dissection, perforation, thromboembolization and ex vivo histological analysis. However, human cerebrovasculature in comparison is more tortuous and less prone to vasospasm.20
An ideal vascular territory to overcome these limitations is the distal canine vertebral artery and the basilar artery, which provide a unique opportunity for testing the efficacy and safety of devices in the native intracranial vasculature (figure 3). The canine vascular response is closer to humans than the porcine vasculature with respect to vasospasm, vessel recoil, neointimal proliferation and thrombus formation. Also, the canine vertebrobasilar system offers sufficient arterial diameters to test devices navigated through the endovascular route.38
The confluence of powerful imaging tools for triage and better devices for treating stroke has the potential to mean great things for patients confronted with this devastating emergency. We believe that an opportunity for a real role of interventional treatment in AIS to reduce patient morbidity and mortality has never been better. The potential impact could be the toppling of AIS from the number one cause of disability in the USA. Many questions remain regarding all aspects of the treatment, yet clinical trials are rapidly moving forward to attain FDA clearance of various devices. The time is now that we as a community engage in deliberate, collaborative and scientifically valid preclinical modeling to optimize the treatment approach and patient selection.
Competing interests None.
Provenance and peer review Not commissioned; externally peer reviewed.