Journal of Neurophysiology

Dr. Bernice Grafstein’s paper on the mechanism of spreading depression

Anthony J. Strong


This essay looks at the historical significance of one APS classic paper that is freely available online:

Grafstein B. Mechanism of spreading cortical depression. J Neurophysiol 19: 154–171, 1956 (

Albert Einstein is sometimes credited with the dictum that no important new idea will be accepted unless at first it is thought crazy. Certainly none of the reviewers thought this of Dr. Bernice Grafstein’s paper in 1956 (4), in which she established the fundamentals of the mechanism of cortical spreading depression (CSD), but it seems there were “sharp differences of opinion” about her interpretation of her data. Grafstein’s conclusion—that propagation of CSD depends on the liberation of potassium ions from depolarized neurons—has been challenged by a model centered instead on glutamate release, but in nearly 50 years it has never been refuted. In the same paper, Grafstein (Fig. 1) also demonstrated that recovery from spreading depression is an oxidative, energy-dependent process, and this conclusion is now central to concepts of the way in which acute brain injury evolves, both in experimental models and now increasingly in patients receiving neurocritical care.

FIG. 1.

Dr. Bernice Grafstein in 1956.


Working toward her doctorate under the supervision of Benedict Delisle Burns in the Department of Physiology of McGill University (Montreal, Quebec, Canada), an experience she continues to value highly, Grafstein used a cerveau isolé preparation in cats (midbrain transection under ether anesthesia). A slab of cortex on the exposed suprasylvian gyrus was then isolated surgically from underlying physiological input, but remained normally perfused (from the surface). A system was used that allowed recording of both the direct current (DC) potential from wick electrodes and also the unit activity from a 30-μm metal core micropipette within the cortex. The cortex was protected with mineral oil. Several types of experiments were carried out with this basic preparation and reported in the paper.

Spreading “Excitation”

First, the occurrence of the typical slow negative potential change of CSD was confirmed, in accordance with Leão’s original description of this (7), in a follow-up after his initial paper (8). Second, the effect of CSD on local responses in the slab to single electrical stimuli (1 ms, bipolar surface electrodes) was tested. The surface-positive response to such stimuli (mediated by synapses) declined quite early during the onset of the DC negativities, but was extremely slow to recover from the depression, indicating involvement of synapses or of interneuronal connections by the CSD.

The second—and possibly the fundamental—finding in this paper was that microelectrode recording showed for the first time that, at the initiation of the DC negativity, there was a brief burst of intense single unit activity lasting some 2–3 s, followed by prolonged silence (Fig. 2 of Ref. 4; reproduced here as Fig. 2). Similar findings were recorded at all depths of the cortex below layer 1. Importantly, Grafstein found that “if this [intense spontaneous] activity is interfered with by [ongoing or prior] stimulation of the cortex, the spread of depression may be arrested” (4). She concluded from this that “the phase of neuronal excitation is an essential link in the spreading depression mechanism” (Fig. 2).

FIG. 2.

A reproduction of Fig. 2 in Grafstein’s 1956 paper (4). The schema illustrates location of extracellular microelectrode recordings (panel 1) and surface wick electrode recordings (panel 2) from a perfused cortical slab in which cortical spreading depression was induced at the opposite end of the slab and allowed to propagate toward the recording electrodes. There is a brief period of intense neuronal firing (panel 1b) at the start of the slow negative potential (direct current) transient shown in panel 2.

Role of K+

A group of experiments within the paper provides good support for Grafstein’s hypothesis that release of K+ into the cortical extracellular space is the critical, self-propagating event in CSD. First, she showed that repetitive cortical stimulation (of a strength insufficient on its own to initiate a CSD wave), if started at least a minute before arrival of the wave at the point of stimulation, could block further propagation. The finding was compatible with the notion that a substance required for depolarization during propagation of the wave was being redistributed across some membranes by the prior subthreshold stimulation. Grafstein identified leakage of K+ from intra- to extracellular compartments as the candidate, although of course depletion of a neurotransmitter from the presynaptic compartment, or an intracellular shift of Na+, could equally have explained this particular observation. However, Grafstein’s next experiment noted that amplitude of the slow DC negativity accompanying CSD would gradually decline with consecutive inductions of CSDs. She was able to restore the negative amplitude by delivering additional K+ either to the cortical surface or through the circulation. No such effect was observed with other ions, even in hypertonic solution.

Further experiments in which the cortical slab was polarized provided additional evidence for the role of a positively charged particle such as K+. The cortical slab along which CSD would propagate was polarized tangentially, positive at the front and negative at the rear, or the reverse. Although in general propagation velocity tended to increase during the course of an experiment, when the polarization field was reversed between successively induced CSDs, there was clear evidence of acceleration when the cortical slab was polarized positively where the CSD wave was being initiated, clearly favoring the suggestion that the active agent is positively charged. Grafstein also argued persuasively that the alternative hypothesis of CSD propagation based on an electrical mechanism via dipoles, however they might be orientated, was unlikely.

Energy Dependence of the Recovery Phase of CSD

The effects of occlusion of the middle cerebral artery on different phases of the slow potential transient were examined. First, with the cortex stable and prior to any induction of CSD, surface-positive responses to pulsed cortical stimulation disappeared upon occlusion, without any change in DC potential. If occlusion occurred instead during the rising phase of the DC negative potential, peak amplitude was increased by some 50%. However, if occlusion started near the beginning of the recovery phase of the DC potential, recovery was arrested and then resumed when occlusion was removed after 35 s. A challenge alternative to proximal arterial occlusion, replacement of inspired oxygen with pure nitrogen, produced the same effect as occlusion: this indicated that the effect of occlusion was not simply due to failure of elution of a substance released by the depolarization.


Thus, in this single paper, Grafstein demonstrated, or argued convincingly and correctly, that 1) there is an initial phase of neuronal depolarization (actually excitation first, depression only afterwards); 2) K+ release into the extracellular space would depolarize adjacent neurons, thus causing propagation; and 3) recovery from this process depends on energy availability. To establish all of this correctly and in a single paper was a remarkable achievement. None of the large volume of work that has followed in the nearly 50 years since Grafstein’s paper has refuted her model. Specifically, Obrenovitch et al. (10) considered the alternative hypothesis that glutamate is the primary agent, but came to the conclusion that this is indeed K+. Moreover, in addition to establishing a mechanism for CSD, Grafstein’s paper (which appeared at a time when the importance of chemical transmitters in the cerebral cortex was barely recognized, and the nature of those transmitters unknown) set the stage for later introduction of the concept of excitotoxicity, which continues to influence current concepts of stroke and head injury pathophysiology.

Experimental studies of models of stroke and head injury have shown that depolarization events with the electrophysiological properties of CSD do not need to be induced in these models; they occur spontaneously with varying degrees of frequency and are almost certainly a major cause of neuronal loss (1, 2, 15). It is only in the past 10 years that conclusive proof has emerged that CSD-like events occur in the human brain, in migraine with aura (6) and in head injury (9, 14), and represent significant challenges to its metabolism and viability (11). CSD has been the subject of several recent reviews (3, 12, 13), and Grafstein returned to the field in 2000, reporting evidence which raises the possibility that gap junction-mediated communication between astrocytes and the overlying pia-arachnoid might provide one basis for the headache associated with migraine (5).


I thank Dr. Grafstein for helpful and interesting discussions and background information.


View Abstract