The ''Onset'' Phase of CSD Initiation of CSD
Leao's observations were made in rabbits under barbiturate anaesthesia, and the stimulus to the cortex was bipolar electrical current delivered from an induction coil, but several other stimuli are also effective. Dialysis through an implanted microcatheter or superfusion of the exposed cortex with potassium chloride (KCl) at 130 mM or more is effective in the rat brain , as is local application of KCl with a wick. Neurosurgeons should also be aware that needling of the cortex is effective, and it seems inescapable that more complex surgical manipulations of similar, susceptible tissue are likely to be effective if, as seems clear from the recent findings in patients , CSD does indeed occur in the human brain. There is also experimental evidence that spreading depression occurs in the spinal cord . What determines susceptibility, by which is meant the frequency of occurrence of CSD (rather than vulnerability to damage from depolarisa-tions), is an important theme of this review. The factors which are currently believed to affect this are species differences, location in the brain, haemodynamic and metabolic conditions in the cortex, anaesthesia, and systemic metabolic variables (essentially - in the present state of knowledge - plasma glucose). All of these factors are best considered after we have first reviewed the basic electrophysiological, haemodynamic and metabolic properties of CSD.
For the purposes of a discussion focussed on brain injury, the CSD complex is best considered in its two phases, onset and recovery, since, as we shall see, it is probably deficiencies in the recovery process that underlie the differences between CSD and PIDs. When Leao measured the DC potential difference between a point on the cortex in the path of the propagating wave of depression and a remote reference point, he noted a transient negativity of some 10 to 15 mV. The observation has been repeated many times, and when sought, the DC potential transient is an invariable feature of both the CSD and PID patterns of depolarisation. The nature of the DC potential transient - presumably indicating a brief accumulation of negative charge in the cortex - is still unknown, although an increase in one or more anions, - lactate, amino-acids, or bicarbonate - has been suggested as a cause .
Mass Neuronal Activity: Grafstein - 1956
Studying areas of cerebral cortex isolated electrically by subpial transec-tion but with perfusion intact, Grafstein recorded unit activity with an extracellular microelectrode, and found a short phase of intense firing at the onset of the DC potential change, followed by prolonged silence  (Fig. 2). There is no suggestion that this transient neuronal activity conveys any physiological information, and as we shall see it is initiated by local changes in the extracellular environment. The observation suggests that excitatory or depolarising influences on neurones - not necessarily synaptic - contribute to the initiation of the CSD event as it reaches a new locus. It is of interest for current researchers studying models of stroke that
Grafstein was able to suspend and then restore resolution of the DC potential transient by first occluding and then releasing the middle cerebral artery (MCAO) in rabbits, showing that resolution of the DC potential is energy-dependent. Her proposal that potassium ion liberated by neuronal depolarisation caused subsequent depolarisation of adjacent neurons still forms the basis of current thinking on mechanisms of CSD propagation (see below).
Thus Grafstein's 1956 paper demonstrated or inferred three of the key features of CSD - mass neuronal depolarisation, mediation by potassium ion, at least in part, and the dependence of recovery on availability of perfusion and energy. The findings and inferences have remained substantially unchallenged, and form the foundation of our understanding of depolarisation events in the cerebral cortex; the paper is perhaps one of the key contributions to neuroscience in the past 50 years.
Changes in Extracellular Ion Concentrations [K+]e, [Na+]e, [Cl-]e, [Ca2+]e
A transient, marked increase in Ke from the normal 3 mM to 60 mM or more is a striking and regular feature of CSD, and lasts for approximately 30-40 seconds in total, often resolving with an undershoot below the baseline . There are accompanying decreases in [Nae] , in [Cl ]e and in [Ca2+]e .
Changes in Membrane Potential and Conductance During CSD
The first intracellular recordings from a neuron during passage of a wave of CSD were made by Collewijn and Van Harreveld, who concluded, after allowing for the simultaneous change in extracellular potential, that neuronal membrane potential (Vm) reached zero briefly . This pattern of depolarisation to zero volts is different from that of the action potential, where Vm may reach +20 mV, and could imply simultaneous opening of several or all membrane conductances in CSD; this might represent mechanical opening of membrane pores with no ion-specific conductance properties, but according to Somjen it is not necessary to postulate such special channels in order to explain the membrane potential changes in CSD .
Redistribution of Water: Tissue Impedance
An increase in electrical impedance of tissue is largely a measure of cell swelling, and Leao and Martins-Ferreira demonstrated an increase in impedance during CSD in 1953 . Measurements of extracellular space volume using indicators such as tetramethylammonium, together with morphological evidence, support this, and the basis most probably lies in the excess of the decrease in [Na+]e over the increase in [K+]e . This would imply a net movement of ions into cells, accompanied by osmotically obliged water, in turn raising impedance to current flow in the ECS. However, Somjen points out  that there is evidence that when impedance is measured some current flow is through rather than around cell membranes, perhaps more especially glia, and that there is also evidence of a marked drop in neuronal membrane resistance during SD [17, 18]. Whatever their precise nature, these changes in CSD appear closely related to the transient reduction in apparent diffusion coefficient (ADC) that can be detected in the rat  and cat  brains during CSD using magnetic resonance diffusion-weighted imaging (see also page 33: Section on Occlusive Stroke).
Early experimental studies of propagation of spreading depression were aided by the use, principally by Martins-Ferreira and Oliveira Castro , of the isolated chick retina, in which the presence and propagation of spreading depression is evident to the naked eye from a transient change in optical properties. They were able to establish a "ring" of retina in which the phenomenon could be constrained to propagate in circular fashion, at a rate measured at 3.7 mm/minute - similar to that originally described by Leao, and they found that alkaline conditions, or increased Ke or Cle, all accelerated propagation, whereas acidification or an increase in Mge slowed it.
In one of her 1956 papers , and noting the likelihood of neuronal depolarisation (as the basis for the brief phase of spontaneous spike discharges), Grafstein suggested that the resulting liberation of potassium ion into the ECS could occur in sufficient concentration to cause adjacent neurones to depolarise, thus causing - or at least supporting -propagation. The simultaneous reduction in ECS volume (see above) would increase the effective [K]e, thus facilitating depolarisation of neurones in the path of the wave.
The separate idea that potassium ion might diffuse slightly further in the ECS and cause depolarisation in non-contiguous neurones was explored in detail by Gardner-Medwin, who determined a rate for cortical extracellular diffusion of K+, and showed that this was slower than that of CSD propagation . A further argument against extracellular diffusion of Ke as the basis of propagation is that in CSD, no increase in Ke can be recorded in the cortex prior to the DC depolarisation (unlike PIDs, where a gradual, prior increase in Ke does occur [23, 24].
A second candidate agent explaining propagation is glutamate released into the extracellular space (ECS) by mass neuronal depolarisation, and in turn depolarising adjacent neurons. Van Harreveld induced CSD by application of compounds in brain extracts, one of which was glutamate , and he and Fifkova later demonstrated release of glutamate during CSD in the retina . However, glutamate dialysed into the cortical ECS does not elicit CSD, nor does inhibition of glutamate reuptake [8, 27].
Propagation of CSD via Glial and/or Neuronal Gap Junctions
The possible roles of intercellular coupling either of neurones or of astro-cytes in initiation and propagation of CSD have received much attention in the last few years. In the case of astrocytes, it is now abundantly clear that in cultures of astrocytes studied with intracellular calcium-sensitive dyes, waves of transient increase in intracellular calcium ion (Cai) can be initiated - by glutamate , nitric oxide (NO)  or mechanical stimulation  - and will then propagate across the culture at a rate very similar to that of CSD in the intact cortex . Nedergaard has shown that in mixed glia-neuronal cultures, such glial waves are associated with elevations in neuronal calcium concentrations . Transmission of calcium waves through glial cultures is believed to occur through glial gap junctions - specialised and specific membrane openings whose molecular structure is now well-characterised and which are usually readily permeable to ions and compounds of smaller molecular weight; examples are inositol trisphosphate (IP3) and potassium. IP3 is thought to mediate propagation of Cai waves through its role as a ligand for IP3-receptor-Ca-conductance complexes on the endoplasmic reticulum, and glial gap junctions are also thus a probable substrate for the mechanism of ''spatial buffering'' of increases in Ke, as envisaged by Somjen . Propagation is also mediated by an extracellular agent, ATP .
At least in the cell culture preparations in which glial communication has been studied, the capacity to propagate Cai waves seems exceptionally well supported, by a range of agents that include ATP , nitric oxide , and inositol trisphosphate (IP3), the latter via glial gap junctions . The demonstration that an intracellular calcium wave precedes the arrival of spreading depression  also lends support to the idea that CSD propagation is mediated primarily by glia. A further argument for the concept is based on the fact that halothane, which blocks glial gap junctions , also reduces the frequency of CSD in the gyrencephalic brain , and reduces MCAO infarct volume and PID frequency by an effect either on perfusion or on intrinsic PID susceptibility .
Other findings argue against this hypothesis. First, CSD is more readily elicited in areas of grey matter with relatively lower glia: neuron ratio, such as the CA1 layer of the hippocampus (in experimental studies) , and the occipital cortex in humans [40, 41] (if it is accepted that migraine with visual aura is a manifestation of CSD, as discussed below). Secondly, the use of specific agents toxic to glia such as fluorocitrate or fluoroacetate fails to prevent CSD [42, 43]. Third, CSD can occur in the absence of Ca; waves .
The Recovery Phase of CSD, and the Responses of Cerebral Metabolism and Blood Flow to CSD
Resolution of the cation transients might in theory be due either to restitution of normal, resting distributions by active transport, or in the case of the increased [K]e, to diffusion through the extracellular space (which would necessarily be slower than the observed resolution rate ), to spatial buffering by the astrocytes through gap junctions , or to passive elution through cerebral perfusion (probable only under conditions of energy failure ). Grafstein's experiment with MCAO described above is perhaps the earliest evidence for a role for energy-dependent active transport in the recovery phase, and evidence for the concept has steadily accumulated. Demonstration of the cation transients that are an integral feature of CSD makes it almost inevitable that restoration of resting cation distributions should necessitate a considerable increase in ATP utilisation.
Indeed CSD, and epileptic seizures, are perhaps the most extreme forms of activation challenge to reactivity of cerebral metabolism and blood flow (CBF).
Detailed studies by Rosenthal & Somjen and their colleagues of CSD in the normally perfused brain indicated transient oxidation of the mitochondrial respiratory chain . In the light of subsequent work demonstrating transient increases in perfusion  and in tissue pO2 during CSD , one simple interpretation of Rosenthal's work is that the redox potentials of the respiratory chain coenzymes are in equilibrium, and are determined by the balance between the rate of ATP hydrolysis and availability to mitochondria of molecular oxygen from cerebral perfusion.
Studies of normal, functional activation in the human brain using positron emission transverse tomography  indicated for the first time that the rate of glucose utilisation increased in greater proportion to oxygen utilisation, suggesting upregulation of glycolysis rather than of oxidative glucose utilisation. The finding of transient increases in brain lactate of some 30% in experimental studies of somatosensory activation  supported this interpretation, and suggested a degree of dependence on gly-colytic generation of ATP during activation. The very large cation shifts that occur in CSD make it highly likely that similar and greater - but still transient - changes in glycolysis would occur during repolarisation after CSD. However, an extracellular lactate transient need not necessarily mean a shift to anaerobic metabolism, and Back and colleagues showed that in the normally perfused brain CSD is accompanied by an increase in partial tissue pressure of oxygen ; this may be attributed to the hyperaemic response to CSD which is described below.
The model of the cerebral metabolic response to activation developed by Magistretti and colleagues  envisages that glycolytic activity is predominantly in the astrocytic compartment (where almost all glycogen in the brain is held [52, 53]), stimulated by an increase in extracellular glutamate during functional activation. It is further proposed that astrocytes deliver lactate to neurons, which, relying on lactate dehydrogenase activity in reverse, convert lactate to pyruvate. This pyruvate is then metabolised via the tricarboxylic acid cycle. Glucose transport across the blood brain barrier is highly efficient, to the extent that total unidirectional flux into the brain under non-activated conditions is approximately twice the rate of utilisation by glycolysis . This, allied with the hyperaemic response to CSD discussed below, endows the cortex with its capacity to meet the challenges of activation. It is not appropriate to pursue further this important topic in this context, and the reader is referred to work by Magis-
tretti and colleagues , to a review questioning some aspects of this ''compartmented glial glycolysis'' model , and to the review by Chen and Swanson of astrocytic function and changes in brain injury . Changes in glucose metabolism in focal ischaemia and during PIDs are described later in this review.
Fig. 3. Schematic diagram illustrating current concepts of the role of astrocytes in cerebral perfusion and metabolism (adapted with permission from Tsacopoulos and Magistretti : Copyright 1996 by the Society for Neuroscience). Cerebral capillaries are extensively invested by astrocyte end feet, and extraction of glucose from blood to brain (probably the astrocyte compartment) is highly efficient (arrow 2). During activation, and especially in cortical spreading depression, the glycolytic pathway in astrocytes is upregulated, and the different kinetics of glial and neuronal lactate dehydrogenases favour net movement of lactate from astrocytes to neurons; the (limited) brain glycogen pool is located in astrocytes. Under resting conditions, glycolysis in neurons may be sufficient to meet energy demands (arrow 1). Neurotransmitter glutamate released into the synaptic cleft is re-accumulated into astrocytes by high-affinity cotransport with Na+ ion, making use of the normal electrochemical gradient generated by Na+/K+ ATPase.
Several mechanisms regulate cerebral perfusion, with a prominent role proposed for astrocytes (arrow 3) . First, their high membrane conductance for K+ allows astrocytes to buffer the increased extracellular levels resulting from activation, with a direct vasodilator effect of K+ on the microcirculation via the astrocyte cytosol and end feet. Adenosine- and nitric oxide-based mechanisms also contribute. Recent work by Zonta et al. now supports an additional mechanism of astrocyte-mediated vaso-
dilation during activation 
Leao himself was the first to demonstrate hyperaemia in association with CSD; he observed a doubling in width of pial surface arterioles during CSD . If CSD induced in the prefrontal region of the rat is assumed to propagate anteroposteriorly in the cerebral hemisphere at a constant rate, serial coronal sectioning of the hemisphere after it has been frozen at a single time point will provide in the section sequence a time series of the response of the brain to the propagation wave. Using autoradiography for CBF, and reasoning in this way, Lauritzen et al. showed that CSD is closely followed by an intense (> 200%) but brief (2 minutes) transient hyperaemia . An extended phase of mild hypoperfusion (80-90% control) follows, lasting for some 60 minutes. This feature of CSD was later used by the same group to allow mapping with isotope scanning of a phase of hypoperfusion associated with migraine with aura that propagated forwards in the cerebral hemisphere at a rate in accordance with that of CSD - a finding that argues quite strongly for CSD as the basis of migraine with aura .
A careful histological study by Nedergaard & Hansen  found no evidence of classical ischaemic pathological changes in the cortex following CSD in the normally perfused cortex of rats. As will be described later, the situation is very different in focal ischaemia.
Expression after induction of CSD of some of the immediate early genes (IEG) that respond to stress has been studied extensively, principally in rats, mice and transgenic mice. The IEG responses to MCAO have also been studied. In many such MCAO studies, increases in gene expression extend to the whole hemisphere rather than remaining within the core and penumbral regions. It is generally believed that such widespread upregu-lation represents a response to a depolarisation event that started as a PID in the ischaemic territory but then propagated throughout the rest of the hemisphere as CSD. According to Sharp et al.  this applies to c-fos and jun-B. Cyclooxygenase-2 is also induced by CSD . In some cases, the association is relatively specific: for example, the degree of induction of the mRNAs encoding brain-derived neurotrophic factor and heat-shock protein-72 in response to CSD induced in the rat is dependent on the number of CSDs . It needs to be stated that in MCAO other gene expression patterns may relate more to cell damage than to CSD. Thus HSP70, a heat shock protein, behaves as a protein chaperone, increasing in the presence of denatured proteins , although expression in the infarct core may be limited by ATP depletion .
That cerebral ischaemia causes an increase in levels of interleukin-1b (IL-1b: an inflammatory cytokine) in the brain is well established [66-68]. CSD has a similar effect: Jander and colleagues recently showed that mRNA levels for IL-1b and tumour necrosis factor-a (TNF-a, also an inflammatory cytokine) are increased 24- and 60-fold respectively 4 hours after CSD induction with KCl . Expression of the IL-1b protein was largely confined to microglia in the superficial cortical layers. These authors suggest that ''cytokine expression following CSD forms part of a physiological stress response that contributes to the development of ischaemic tolerance in this and other preconditioning paradigms'' (see below). That IL-1b can promote CNS repair has also been shown . Another view of the effects of IL-1b comes from the work of Blamire et al., who examined the effects of microinjection of recombinant IL-1b into the striatum of 3-week-old rats, and found significant reduction in apparent diffusion coefficient (ADC) and increases in cerebral blood volume and blood brain barrier permeability . ADC reductions are usually attributed to a shift of water from extra- to intracellular compartments, but a reduction in water mobility in the intracellular compartment may occur ; both explanations are in keeping with an adverse effect of IL-1b.
Pre-Ischaemic Conditioning with CSD as Protection in Experimental Stroke
In experimental studies of stroke in rats, it is possible to confer a degree of protection from the effects of a period of ischaemia by prior induction of CSDs . Levels of mRNAs for FOS, BDNF, and tPA, are increased by preischaemic conditioning with CSD . TNF-a and IL1-b are believed to contribute to increased tolerance of ischaemia [75, 76], and an antagonist to nuclear factor k-B (NF k-B) blocked NF k-B activity and reduced the pre-conditioning effect .
It seems very likely that one or more of the currently identified expression cascades - or other(s) still to be detected, underlie the protective effect of preischaemic conditioning with CSD, and increasing understanding of the molecular response to CSD may in time allow us to identify which of the several genes upregulated by CSD is/are responsible for the protective effects of preconditioning, and so perhaps lead to novel therapy for cerebral ischaemia, or at least better protection of the brain when some degree of prospective risk exists.
Factors Determining Ease of Induction of CSD
It has long been clear that CSD is more readily induced - and its repetition maintained - in rats than in larger experimental animals , with primates seen as the most ''resistant'' group of species. However, it is certainly possible to induce CSD in the primate brain . A specific attempt to compare PID frequency in cats and squirrel monkeys after MCAO showed that PIDs do indeed occur spontaneously in a primate species, but failed to confirm a species difference in frequency of PIDs because of wide variability within both species . However, the results revealed a clear dependence of PID frequency on plasma glucose level: this is discussed below in the context of PIDs.
One of the most widely canvassed explanations for species differences starts with the observation that lissencephaly is characteristic of the CSD-prone species, whereas the more resistant brains are gyrencephalic. There are however also regional differences in susceptibility within the brain of a given species, with the hippocampus particularly liable to CSD, together with - in migraineurs-with-aura - the occipital cortex. A clue to the puzzle comes from consideration of the cytoarchitecture and glial: neuronal ratios of different brain regions and in the brains of different species. Thus neurons are particularly tightly packed, glia relatively sparse, and CSD frequent, in the CA1 layer of the hippocampus. Migraine with aura typically commences with a visual aura on or near the fixation point (although auras apparently arising from the somatosensory cortex also occur), and the glia: neuron ratio in the occipital cortex is lower than elsewhere in neocortex [40, 41].
Tower and Young compared the glia:neuron ratio with brain size in a group of mammals ranging from mice to whales and elephants, and, using a log: log plot, demonstrated a convincing hierarchy in which the glia:neuron ratio increases in proportion with brain size . Primates are distributed appropriately for their brain size within this hierarchy, rather than all of them possessing a high glia : neuron ratio independent of brain size, as might be predicted on ''evolutionary'' grounds. The issue is of interest in relation to the discussion above on mechanisms of CSD propagation, and the relationship of CSD propensity with Tower and Young's hierarchy is more in keeping with a homaostatic role for glia in the context of CSD than with one in which they propagate CSD.
Spreading depression has also been observed in experiments on the spinal cord [9, 81], and the possibility therefore arises that peril-esion depolarisations might contribute to the evolution of spinal cord damage - at least in grey matter - not only in trauma but also in vascular lesions.
The reduction of CSD frequency by agents known to block glial gap junctions such as halothane and propofol has been referred to above. The role of increased Ke in initiating CSD has been referred to above, and glutamate and other excitatory amino acid agonists can also effect this [25, 47, 82]; conversely, it is widely recognised that CSD and/or PID frequencies can also be reduced by the action of some excitatory amino acid antagonists, notably antagonists of the n-methyl-D-aspartate (NMDA) class of glutamate receptors [83-86]. Such mechanisms may operate in ischaemic and traumatic brain injury, although with no proven definitive therapeutic benefit in humans to date, and are considered below in the context of PIDs.
Evidence favouring CSD as the basis of the migraine aura has gradually accumulated since Leao and Morison first suggested this , and is reviewed in Migraine page 29. We can learn something of the mechanisms of CSD induction from descriptions from migraineurs (with aura) of the precipitating factors they implicate. Sometimes onset follows relaxation after a period of intense concentration or physical exercise. The onset is attributed to hunger by some migraineurs with aura, and we may speculate in the light of discussion below (page 25: Relationship of Cortical Glucose Availability with PID Frequency) that hypoglycaemia is responsible in these individuals. The various other precipitating factors do not at present appear relevant in this context.
Familial patterns of migraine incidence and inheritance are well recognised, and there is evidence that familial hemiplegic migraine is due to a calcium channelopathy . Migraine with (non-hemiplegic) aura is much more common, and appears sometimes to have a familial element. It seems likely that other gene/ion-channel abnormalities will emerge in due course. A patient with the appropriate genotype seems likely to be at increased risk of depolarisations occurring in association with stroke, subarachnoid haemorrhage or serious head injury.
The role of ischaemia, trauma, increased Ke and glucose availability to the cerebral cortex in stroke and head injury will be considered below in relation to PIDs. Disturbances of magnesium metabolism have also been invoked as an additional factor increasing migraine risk .
Was this article helpful?
Are you fed up with your frequent headache pain? 101 Simple Ways to Attack Your Headache BEFORE the Pain Starts Guaranteed No Pain, No Fear, Full Control Normal Life Again Headaches can stop you from doing all the things you love. Seeing friends, playing with the kids... even trying to watch your favorite television shows.