ls
an
, Gl
orm
to the understanding of ischemic injury mechanisms but suffers from its
the
o fa
in i
of
utco
patients with cerebrovascular disease poses the fundamental
situation may arise at three stages of experimental research:
the selection of the appropriate animal model for the induction
2. Models of global ischemia
to squeeze the large mediastinal vessels against the inner wall
of the thorax.
The duration of brain ischemia is longer than that of cardiac
arrest because it takes some time until the heart is able to re-
store brain recirculation. Return of cardiac function, in turn,
* Tel.: þ49 221 4726 210; fax: þ49 221 4726 349.
E-mail address: hossmann@nf.mpg.de
Available online at www.sciencedirect.com
Neuropharmacology 55 (20
question of clinical relevance (Khaja and Grotta, 2007; Sacco
et al., 2007). Successful translation of experimental findings
can only be expected if the pathophysiology of the model rep-
licates the essential criteria of the clinical situation (Perel
et al., 2007). If not, injury and treatment concepts are gener-
ated which are valid for the model but not for the disease,
which explains why so many clinical trials have failed in the
past. Discrepancies between experimental data and the clinical
The clinically most dramatic manifestation of global brain
ischemia is cardiac arrest (Fig. 1). Experimentally, cardiac ar-
rest can be induced by intracardiac injection of potassium
chloride or other cardioplegic agents (Kofler et al., 2004), by
ventricular fibrillation (Bo¨ttiger et al., 1997), exsanguination
(Behringer et al., 2000), drowning (Makarenko, 1972) or as-
phyxia (Liachenko et al., 1998). A technically simple but
rather traumatic intervention is the Korpatchev method (Kor-
patchev et al., 1982), where a hook is introduced into the chest
interrupted. The use of these models for the study of injury
mechanisms and the development of treatment strategies for
2.1. Cardiac arrest
or incompletely, focally or globally, permanently or transiently
dations are given to improve the translational power of brain ischemia-related experimental research.
� 2007 Elsevier Ltd. All rights reserved.
Keywords: Brain ischemia; Focal ischemia; Global ischemia; Brain imaging; Multiparametric imaging; Review
1. Introduction
Brain ischemia comes in many forms, depending on the
kind and severity of circulatory impairment. To cope with
this variability many experimental models have been estab-
lished in which blood flow is acutely or slowly, completely
of ischemia, the selection of the appropriate method for the as-
sessment of ischemic injury, and the correct interpretation of
outcome data. This review summarizes the most widely used
approaches under this particular aspect.
Cerebral ischemia: Mode
Konstantin-Alex
Max Planck Institute for Neurological Research
Received 25 October 2007; received in revised f
Abstract
Experimental research into brain ischemia contributes substantially
limited relevance for clinical treatment startegies. One of the reasons is
replicate the pathophysiology of naturally occurring brain ischemia. T
most widely used experimental models and analytical methods of bra
iological particularities of the various ischemia models, the application
penumbra, benign oligemia and normal tissue, as well as to the final o
0028-3908/$ - see front matter � 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.neuropharm.2007.12.004
use of experimental models and methods that do not or only partially
cilitate the understanding and interpretation of experimental data, the
schemia are reviewed. Particular emphasis is given to the pathophys-
imaging methods for the reliable differentiation between infarct core,
me of experimental interventions. Based on this analysis, recommen-
, methods and outcomes
der Hossmann*
eueler Strasse 50, D-50931 Cologne, Germany
6 December 2007; accepted 10 December 2007
08) 257e270
www.elsevier.com/locate/neuropharm
erent
per-
imental data are translated to the clinical setting.
ischemic brain. This has been done by occluding, in addition
to the main arterial supply, the internal mammary arteries,
the pterygopalatine and other neck arteries, or by retrograde
rma
2.2. Selective arrest of cerebral circulation
To avoid cardiac interference on the outcome of global
brain ischemia, most presently used animal models are based
on the selective interruption of cerebral blood flow. Obviously,
this greatly improves the chances of post-ischemic recovery
induced brain damage therefore varies greatly under diff
experimental conditions which have to be known when ex
depends on the mode of cardiac arrest (cardioplegia is less
harmful than ventricular fibrillation or asphyxia) and on the
cardiac resuscitation procedure (extrathoracal compression is
less efficient than direct cardiac massage). Cardiac arrest-
Fig. 1. Global brain ischemia of 30 min duration induced in cat by cardiac ar-
rest. Recording of the electrocorticogram (EcoG), cerebral blood flow (CBF),
electrocardiogram (ECG) and systemic arterial pressure (SAP). Cardiac arrest
was induced by ventricular fibrillation and 30 min later reversed by direct car-
diac massage. Note abrupt cessation of brain perfusion during cardiac massage
followed by slowly progressing recirculation after return of spontaneous car-
diac function (data from Hossmann and Hossmann, 1973).
258 K.-A. Hossmann / Neuropha
but it should be recalled that under natural conditions global
brain ischemia in the absence of cardiac failure is a rare event.
The use of such models for the study of the pathophysiology
and therapy of global ischemia may, therefore, lead to results
which are of limited relevance for cerebrocirculatory arrest
arising under clinical conditions, and must be interpreted
accordingly.
2.2.1. Complete ischemia
In the rat the most widely established method of complete
or near-complete global brain ischemia is the coagulation of
both vertebral arteries, followed one day later by the transient
occlusion of both common carotid arteries (4-vessel occlusion,
Fig. 2) (Pulsinelli and Brierley, 1979). In gerbils near complete
forebrain ischemia can be produced by ligation of both com-
mon carotid arteries without additional occlusion of the verte-
bral arteries because in this species the circle of Willis is
incomplete, and there is no connection between the basilary
and the internal carotid artery (Levine and Payan, 1966). Other
experimental models for the selective interruption of cerebral
blood flow are the compression of the blood vessels in the
neck by strangulation or inflation of a pneumatic cuff (Kabat
and Dennis, 1938), the increase of intracranial pressure above
blood pressure by infusing fluids under high pressure into the
cisterna magna (Siesjo¨ and Zwetnow, 1970), bilateral common
carotid and subclavian artery occlusion (Hua et al., 2006) and
the intra-thoracic occlusion of the innominate and left subcla-
vian arteries which blocks the blood supply to both carotid and
vertebral arteries (Hossmann and Zimmermann, 1974).
In all models of selective cerebrocirculatory arrest care has
to be taken to avoid a residual collateral supply of blood to the
Fig. 2. Global brain ischemia of 30 min duration induced in rat by bilateral oc-
clusion of common carotid and vertebral arteries. Recording of electroenceph-
alogram (EEG), cerebral blood flow (CBF) and systemic arterial blood
pressure (SAP). Note biphasic return of blood flow after reversal of vascular
occlusion. The sharp rise of systemic arterial pressure at the onset of ischemia
is a brainstem-mediated response (Cushing’s response; data from Kloiber
et al., 1993).
cology 55 (2008) 257e270
drainage of the occluded vessels. Collateral flow can also be
reduced by lowering the blood pressure to hypotensive levels,
either by bleeding or the pharmacological application of vaso-
dilating agents.
For certain purposes, isolated heads or brains connected to
extracorporeal circulation are used for the induction of global
ischemia (Krieglstein et al., 1972; Niu et al., 2002). These
preparations are suited both, for the production of complete
or graded incomplete ischemia because the blood flow can
be varied over a wide range by simple adjustment of the
pump speed of the extracorporeal circulation system (Hinzen
et al., 1972).
Although the common pathogenic factor of all these models
is global interruption of cerebral blood flow, considerable
pathophysiological differences exist. When arterial blood sup-
ply is interrupted without simultaneous blockade of venous
outflow, most of the blood escapes from the brain during is-
chemia, leading to anemic ischemia. The additional interrup-
tion of venous outflow, e.g. during strangulation ischemia,
causes hyperemic ischemia with massive congestion of the
cerebral vasculature. This difference is of importance for the
recirculation after ischemia because the increased viscosity
of stagnant blood increases the risk of no-reflow and requires
a higher blood pressure for reperfusion than after anemic is-
chemia (Fischer and Ames, 1972).
Another pathophysiologically important factor is the vascular
and extracellular fluid volume which determines the severity of
ischemic cell swelling. In anemic ischemia this volume is smaller
than in hyperemic ischemia or in ischemia induced by cisternal
infusion of artificial CSF. Ischemic brain swelling is, therefore,
least pronounced during anemic ischemia which should be
cir-
alue
slightly below blood pressure. Other approaches are bilateral
organotypic cultures (Vornow et al., 1994) or brain tissue sli-
ces (Whittingham et al., 1984) are incubated in deoxygenated,
glucose-free medium in order to mimic the interruption of the
supply of oxygen and nutrients to brain parenchyma. Brain sli-
ces have also been prepared after a period of circulatory arrest
to study post-ischemic resuscitation ex vivo (Charpak and
Audinat, 1998). All these models share several major disad-
vantages. First, the preparation of the slice is associated not
only with severe tissue trauma but also with a period of circu-
259K.-A. Hossmann / Neuropharmacology 55 (2008) 257e270
common carotid artery stenosis (Shibata et al., 2004), the for-
mation of an arterio-venous fistula (Hai et al., 2002) or the
combination of servo-controlled common carotid artery con-
striction with bilateral subclavian artery occlusion (Boehme
et al., 1988). A characteristic pathophysiological feature of in-
complete global ischemia is the accentuation of the flow re-
duction in the peripheral regions of the supplying territories
of the cerebral arteries. Since these regions are located in
the borderzones between these territories, the resulting brain
injury is referred to as ‘‘border line’’ or ‘‘border zone’’ lesions.
A complicating factor of incomplete ischemia is the persis-
tence of a residual blood supply of glucose and water to the
brain which supports ongoing anaerobic glycolysis and en-
hances cell swelling. Both factors aggravate ischemic injury:
ongoing glycolysis by promoting, among others, acidosis (Si-
esjo¨, 1988), DNA fragmentation (Li et al., 2001) and gluta-
mate release (Li et al., 2000), and cell swelling by its
adverse effect on the microcirculation which contributes to
the development of a no-reflow phenomenon (Kempski and
Behmanesh, 1997). This explains why severe incomplete is-
chemia causes more pronounced brain damage than the
same duration of complete cerebrocirculatory arrest (Re-
hncrona et al., 1979).
2.3. In vitro ischemia and ex vivo resuscitation
To facilitate high throughput investigations of molecular in-
jury pathways of brain ischemia, in vitro models have found
increasing application over the past years. In these models
primary neuronal cultures (Goldberg and Choi, 1993),
rial blood pressure below the autoregulatory range of brain
culation or by increasing intracranial pressure to a v
remembered when edema-alleviating treatments are tested.
2.2.2. Incomplete brain ischemia
Incomplete global ischemia is produced by lowering arte-
Fig. 3. Autoradiographic measurements of cerebral blood flow in experimental mod
MCA: middle cerebral artery (courtesy of G. Mies).
latory arrest before the slice is brought into the incubation me-
dium. The control situation of such preparations represents
a post-traumatic, post-ischemic recovery state which may be
basically different from the intact brain. This is particularly
disturbing in studies of the hippocampus where brain trauma
or a few minutes of ischemia are known to cause long lasting
disturbances of protein synthesis and delayed neuronal death
(Widmann et al., 1991). Furthermore, in vitro preparations re-
quire incubation media which differ substantially from the
normal extracellular environment and which, in contrast to
the in vivo situation, provide an unlimited supply of extracel-
lular solutes, notably sodium and calcium. Finally, as these
preparations are mostly derived from pups and not from adult
brains, the results obtained have little in common with the in
vivo situation of adult brain ischemia and should be inter-
preted with caution.
3. Models of focal ischemia
The most important clinical equivalent of experimental
focal brain ischemia is ischemic stroke. According to the Fra-
mingham data 87% of strokes are caused by atherothrombotic
and cardioembolic occlusions, 14% by haemorrhages and 3%
by other or undefined reasons (Wolf et al., 1992). A total of
65% of strokes that result from vascular occlusion present le-
sions in the territory of the middle cerebral artery, 2% in the
anterior and 9% in the posterior cerebral artery territories,
15% in brainstem and cerebellum, and the rest in watershed
or multiple regions (Bogousslavsky et al., 1988). According
to these and many other epidemiological studies, focal ische-
mia in the territory of the middle cerebral artery is the domi-
nating e although not the sole e cause of clinical stroke.
In experimental stroke research, this situation is reflected
by the preferential use of middle cerebral artery occlusion
models (Fig. 3), but for technical, economical and ethical rea-
sons further models have been developed which will also be
briefly discussed.
els of focal brain ischemia in gerbil, rat and cat. CCA: common carotid artery;
tized, freely moving animal by gently pulling the thread
The disadvantage of the model is the variability of the is-
preserved, the penumbra which is characterized by a loss of
function without structural damage, and the ischemic core,
in which both functional and structural integrity are severed.
The precise localization of core, penumbra and benign olige-
mia depend on the animal strain and the method of vascular
occlusion, and may vary considerably under different experi-
mental conditions.
3.2.1. Transorbital middle cerebral artery occlusion
This model was introduced in the seventies for the produc-
tion of stroke in monkeys (Hudgins and Garcia, 1970), and
later modified for use in cats, dogs, rabbits and even rats.
The procedure is technically demanding and requires micro-
surgical skills. The advantage of this approach is the possibil-
ity to expose the middle cerebral artery at its origin from the
internal carotid artery without retracting parts of the brain.
Vascular occlusion consistently affects basal ganglia and can
be performed without the risk of brain trauma. It is therefore
the preferred stroke model for polygraphic physiological re-
cordings (Fig. 4). However, removal of the eyeball is invasive
and may evoke functional disturbances which should not be
ignored. Surgery may also cause generalized vasospasm which
may interfere with the collateral circulation and, hence, induce
rmacology 55 (2008) 257e270
chemic impact, depending on the caliber of the anterior com-
municating artery. It is, therefore, mandatory to measure blood
flow or to monitor neurological impairment in each individual
animal. Another shortcoming is the gerbil’s high excitability
which bears the risk of ischemia-induced seizures (Herrmann
et al., 2004). On the other hand, the inherent variability of the
vascular anatomy offers the opportunity to test the outcome of
ischemia over a wide range of flow values. By using multi-
modal imaging techniques (see below), the gerbil is therefore
an excellent model to study the threshold relationship between
blood flow and functional or biochemical disturbances (Pa-
schen et al., 1983).
3.1.2. Common carotid artery occlusion in combination
with anoxia (Levine’s model of anoxic-ischemic
encephalopathy)
In the rat the combination of extracranial common carotid
artery occlusion with repeated exposure to respiratory hypoxia
produces focal brain injury on the side of vascular occlusion
despite a functionally intact circle of Willis (Levine, 1960).
This model is also technically easy but the induction of hyp-
oxia results in systemic alterations which may modulate the is-
chemic impact in a complex way. The model is, therefore, not
recommended for stroke research. However, as it can be per-
formed in very small subjects, it is widely used for the induc-
tion of anoxic-ischemic injury in neonatal animals (Vannucci
et al., 1999).
3.2. Middle cerebral artery occlusion
Occlusion of the middle cerebral artery at its origin
interrupts blood flow to the vascular territory of this artery,
including the basal ganglia which are supplied by the lentic-
ulo-striate arteries. These MCA branches are end-arteries
which in contrast to the cortical branches do not form collat-
erals which the adjacent vascular territories (Zu¨lch, 1981).
As a consequence, the basal ganglia consistently suffer severe
reduction of blood flow whereas the cerebral cortex exhibits
a gradient of blood flow with decrease from the peripheral to-
wards the central parts of the vascular territory. According to
the threshold concept of ischemia, the declining flow rates
(Suzuki et al., 1983).
3.1. Common carotid artery occlusion
3.1.1. The gerbil model of common carotid artery occlusion
In gerbils and several mouse strains focal brain ischemia
can be produced by unilateral occlusion of the common ca-
rotid artery because the circle of Willis lacks a posterior com-
municating artery (Kelly et al., 2001; Levine and Payan,
1966). The operation is easy to perform and does not require
special surgical skills. By placing a suture around this artery
and passing the thread under the skin to the back of the animal,
vascular occlusion can even be performed in the unanaesthe-
260 K.-A. Hossmann / Neuropha
can be associated, in this order, with an area of benign olige-
mia in which functional and structural integrity of the brain are
variations in infarct size (Hossmann and Schuier, 1980). The
procedure therefore requires extensive training before repro-
ducible results can be expected.
3.2.2. Transcranial occlusion of the middle cerebral artery
Post- or retro-orbital transcranial approaches for middle
cerebral artery occlusion are mainly used in rats and mice
because in these species the main stem of the artery appears
Fig. 4. Effect of transorbital middle cerebral artery (MCA) occlusion in cat on
the hemodynamic and electrophysiological functions of cerebral cortex. Elec-
trocorticogram (ECoG), cortical steady potential (DC), pial arterial pressure
(PAP) and blood flow are recorded in the parietal cortex. Anoxic terminal de-
polarization is reflected by the negative shift of the cortical DC potential (data
from Shima et al., 1983).
on the cortical surface rather close to its origin from the inter-
nal carotid artery (Tamura et al., 1981). However, to produce
ischemic injury in the basal ganglia, care has to be taken to po-
sition the occlusion proximal to the lenticulo-striate branches.
The further distally the artery is occluded, the smaller and
more variable infarcts are. In rats, therefore, the zygomatic
bone must be removed to access the proximal segment of
the artery. In chron