Despite the apparent restoration of blood flow through arteries supplying the penumbra, there may be protracted and heterogeneous penumbral cell injury, for unclear reasons. One possibility is that variable areas of brain receive penumbral levels of flow and the distribution of these areas in the brain is determined by variable degrees of collateralization from other blood vessels. Such collaterals generally connect in the pia between end-branches of the larger arteries or via the microvascular network. In contrast to the well-known sources of collateral flow among large arteries, an extensive network of capillary collateralization has been described in the mouse [29]. Whether such a rich source of microvascular collateral flow exists in humans remains to be shown.

Variability in blood-flow may influence the extent of vascular injury and inflammation during re-perfusion [30]. Throughout the time the microvessels suffer low flow, residual blood elements begin to coalesce (Fig. 1.2). Restoration of flow, if it comes too late, may exacerbate this process. Recent in vivo imaging studies of the cerebral microcirculation after transient MCAo demonstrate that re-perfusion is associated with accumulation of leukocytes along cerebral venule walls [31]. Such leukocyte accumulation can be blunted by hemodilution (albumin therapy) and with the improvement of venular re-perfusion, the final infarct volume is reduced [32]. Re-perfusion injury also worsens in a time-dependent manner to include platelet as well as leukocyte adhesion to vessel walls. This likely reflects a time-dependent increased expression of adhesive molecules which mediate platelet adhesion, such as P-selectin on cell surfaces of endothelial cells and platelets [33]. Other early markers of endothelial activation include E-selectin, ICAM-1, VCAM, and the urokinase plasminogen activator receptor (uPAR) [34, 35]. P-selectin in turn unleashes a cascade of inflammatory and thrombotic mechanisms [36]. Such data suggests vascular injury and subsequent microvascular occlusions may form a mechanism that underlies penumbral heterogeneity.

Another reason for reperfusion failure may be the accumulation of thrombus in the distal microcirculation, proposed originally as the no-reflow phenomenon [37, 38]. Several groups have observed clot in the microcirculation distal to larger artery occlusion, although it can be challenging to eliminate the possibility of post-mortem clot formation. Elegant studies in the baboon striatum document a clear relationship between endothelial cell “activation,” by which the authors mean up-regulation of inflammatory markers that promote or signify thrombosis, and neuronal cell death [39]. Such data suggest that anti-thrombotic therapy should help stroke patients, but clinical trials of anti-thrombotic therapy, including abciximab, heparin, and heparinoid, all failed [40, 41]. It is not clear if these trials failed due to weak design (overly generous time window for enrollment, heterogeneous patient selection) or because distal microcirculatory thrombosis contributes relatively little in the patho-anatomic mechanism of penumbral flow failure. As a result of the clinical trial failures, alternative mechanisms must be considered.

Another reason for no-reflow into the penumbra after recanalization is downstream vasoconstriction. There are several mechanisms for downstream vasoconstriction. First, proximal large artery occlusion, such as tMCAo, releases a number of substances from the endothelium and from the in situ thrombus. Many of these substances act distally to cause smooth muscle cell contraction. It is not yet clear whether this phenomenon contributes meaningfully to outcome, but in vivo measurements do confirm distal changes in blood flow after arterial infusion of some of these substances [42, 43]. Second, it is now clear that astrocytes play a role in mediating regional rCBF, perhaps by directly stimulating arteriolar vasoconstriction and dilation [44, 45]. Data suggests that astrocytes communicate via gap junctions, and the calcium transients may mediate the propagation of signaling events downstream [46]. Third, direct muscular contraction could propagate down the arterial tree, from the site of proximal injury to the distal branch arterioles. While such propagating waves of vascular wall spasm have been documented in ex vivo preparations, a direct link to rCBF has not been proven.

A final and speculative reason for heterogeneous patterns of reflow is the variable distribution of effector pathways in the cortex and striatum. For example, there is a well-known pattern of selective vulnerability to global ischemia [47]. In areas of greater vulnerability, there is a much greater concentration of excitatory synapses compared to inhibitory synapses [48, 49]. Although largely unexplored, no doubt there are significant differences in the distribution of important effector systems such as arteriolar receptors for vasoconstrictive substances.

The penumbra concept grew out of research directed at mapping the topography of cerebral blood flow. Traditional approaches to mapping rCBF (regional cerebral blood flow) include tracer studies (usually ¹⁴C-antipyrine), or hydrogen clearance, both of which are technically demanding. In fact, if one examines the data from the original studies that motivated the penumbra model, there are data points that fall outside the blood flow boundaries [1, 2]—it is possible that some electrodes were recording from islands of lower blood flow, while other electrodes were recording normal flow in the presumed penumbral shell. There is no way now to confirm this novel interpretation of that data, but studies will be needed to determine whether rCBF reduction is heterogeneous after MCAo.

Recent application of optical methods to the problem of in vivo blood flow has provided new insights into the effects of ischemia on microcirculation [50–53]. Via a cranial window, confocal laser microscopy has been used to visualize blood flow in individual cerebral vessels after infusion of fluorescently tagged blood cells in vessels up to 200 μm deep to the cortical surface. Moving erythrocytes are mapped, and from the time and distance moved, actual blood velocity is calculated directly. After 4-vessel occlusion, capillary flow was shown to be zero, which correlated with EEG flattening; flow returned after clamp release [54]. The application of a two-photon laser scanning microscopy (TPLSM) method by Kleinfeld and colleagues now allows visualization of flow in vessels up to 400 μm deep to the cortical surface and quantitation of flow in arterioles and veins, as well as capillaries [50, 51]. This methodology is compatible with experimental ischemia and has provided new insights into flow rearrangement following cortical microthrombosis [29]. The combination of classical immunohistochemical methods with in vivo imaging afford a unique opportunity to explore the mechanisms underlying penumbral injury. Both techniques will provide complementary evidence of blood flow changes and cellular injury in ischemic zones.

Until very recently, our concept of the penumbra simply included 2 zones, one surrounding the other, as shown in Fig. 1.1a. There are a number of problems with the classical penumbra concept. Despite the interesting data obtained by some groups, the vast majority of patients with focal ischemia do not exhibit such consistent patterns. Many more patients who respond to thrombolysis with a gratifying recovery do NOT show a pattern on PET or MRI consistent with a presumed core and a larger penumbra. It is increasingly clear that the penumbra concept should be expanded to include more complicated flow patterns [55]. For example, in Fig. 1.1b we illustrate a possible scenario in which a central core is surrounded by patches of penumbral flow, some of which include additional zones of very low, or core, flow. It is quite likely that variable areas of brain receive penumbral levels of flow and that the distribution of these areas in the brain is determined by variable degrees of collateralization from other blood vessels. Such collaterals generally connect in the pia between end-branches of the larger arteries. There is some limited pathological data to support this concept of “islands” of penumbral flow—the heterogeneous penumbra—near areas of complete infarction [55–57].

Support for the heterogeneous penumbra comes from a variety of sources. In reviewing some seminal papers, there is obvious heterogeneity in the published diagrams. For example, a diffusion MRI study from Stanford included a photograph of diffusion weighted images taken serially from one subject [16]. In Fig. 7 from that study, the image taken 1.5 h ischemia onset shows clear islands of abnormal signal in the striatum and overlying cortex. The islands begin to coalesce in the 2.5 h image, and are nearly confluent by 4 h. Similarly, PET images usually show a heterogeneous pattern of hypometabolism in the first hour or two following ischemia onset (see Fig. 1.3) [58].

No matter what the spatial pattern of blood flow changes, it is clear that cells in penumbral regions do not survive indefinitely. The exact duration of survival is unknown, and likely it is different under differing circumstances. Cells in zones receiving 10–20 cc/100 g/min likely survive for hours, but the number of hours is not known. Many factors may reduce the time such marginally perfused cells might survive. For example, it is known that hyperthermia of 1 or 2 °C will accelerate cell death [59]. Elevations of serum glucose also accelerate cell death. Such marginal levels of blood flow do not support neuronal survival indefinitely, but could be sufficient to deliver neuroprotectants into the ischemic zone. Therefore, pending recanalization and restoration of blood flow, a number of therapies could be directed at idling neurons in the penumbra. Such neuroprotectants might salvage brain by preserving neurons until blood flow is restored. The possible mechanisms of neuroprotection include interrupting ischemic excitotoxicity, blocking apoptosis, and blocking the inflammatory response that follows ischemia.

The elucidation of the excitotoxic cell death mechanism advanced stroke neurology considerably. An understanding of the biochemical events outlined below spawned a plethora of therapeutic trials. As of this writing (2013) no agent has yet proven successful in humans, despite considerable excitement from laboratory studies. Future successful neuroprotection depends upon further delineation of the steps in the ischemic cascade, and identifying candidate points for intervention.

As outlined above, there are a variety of interacting events proceeding simultaneously during ischemia and it now seems unlikely that a single agent will prove insufficient to salvage most of the ischemic brain in most patients [152]. On the other hand, multiple agents targeted at different receptors or events in the ischemic cascade may cooperatively improve outcome. Attempts have been made to design a combination of neuroprotective agents, with some success, but it is difficult to predict the doses of the two agents to use. Furthermore, the combination may manifest benefit as increased potency compared to the single agents or it may lengthen the treatment delay time window, or both (see Chap. 3). For example, when both rt-PA and a granulocyte inhibitor were given 30 min after ischemia onset, each was effective, as was the combination [123, 153]. After a treatment delay of 90 min t-PA and the combination were effective, but there was no synergistic benefit. After a delay of 180 min, t-PA was not effective when used alone, nor when combined with anti-ICAM given 5 or 175 min after ischemia. In a follow-up study, the same group studied longer time intervals [153]. In this study, t-PA given 2 h after embolization was not effective, but the combination of anti-ICAM 15 min after and t-PA 2 h after embolization was quite effective. Again, the dose of t-PA alone was not increased to see whether a maximal dose of the single agent was equipotent to the combination. However, this is highly unlikely because the dose of t-PA used, 3 mg/kg, is known to thrombolyse the majority of the injected emboli. Therefore, this study suggests that the use of the two agents conferred a benefit that could not be obtained from higher doses of either agent alone, i.e., synergism.

Hypothermia is the most potent neuroprotectant every studied (see Chap. 3). Mild hypothermia shows significant benefit after focal or global cerebral ischemia in animals and humans by reducing metabolic demand, limiting the inflammatory response to ischemia, reducing excitotoxin release and free radical generation, and limiting blood-brain barrier breakdown [154, 155]. Hypothermia treatment in ischemic rat animal models has been reviewed extensively in the literature [156] and improved neurologic outcomes have been demonstrated [157]. In models of global ischemia, the effect of hypothermia vastly depends on ischemia duration and time between ischemia and onset of hypothermia.

The optimal duration and depth of hypothermia remain controversial [158]. Data are accumulating to support the beneficial effect and safety of mild (34–35 °C) hypothermia in comparison to moderate/severe hypothermia that is often accompanied by increased systemic adverse effects [159–161]. The duration of hypothermia has been studied over a variety of maintenance periods (1–72 h), and substantial protection has been shown even at the shortest duration of 1 h [162]. The benefit of hypothermia in reducing infarct size and neuronal death (both cortical and striatal) has been shown to persist up to 6 months post-stroke [163]. This sustained benefit establishes that hypothermia may actually stop the ischemic cascade rather than simply delay inevitable damage.

In the treatment of stroke with hypothermia, rapid induction, precise temperature control, and ease of administration are critical requirements for the therapy. Cooling can be initiated through surface (skin) cooling or endovascular methods, such as catheter-based systems or intravenous infusions of chilled fluids [164]. Surface cooling using previous technologies may induce hypothermia slowly, however the latest generation of surface cooling devices has not been tried in awake stroke patients [165]. The average rate of cooling with surface methods ranged from 0.3 °C to 1.7 °C/h and can be significantly slower in patients with a high body mass index [166, 167]. Control around the target temperature was previously difficult with surface cooling. In the COOL AID study, target temperature overshoot was observed in all patients by as much as 5 °C [168]. More recent versions of surface cooling units may approach the cooling power of endovascular methods with greater control.

Other strategies for hypothermia induction include infusion of cold intravenous solutions, body cavity cooling with cold lavage, and extracorporeal cooling devices. These devices/methods have limitations. Body cavity cooling is either minimally effective or results in patient discomfort and complications. Gastric lavage with cold fluids can cause cramping and diarrhea. Bladder lavage is minimally effective. Use of cold intravenous solutions, combined with cooling water and air convection blankets, provides sufficient levels of hypothermia induction, and recent data suggest this approach can serve as a prelude to more definitive forms of cooling [169, 170]. Therefore, current protocols allow for hypothermia induction with 4 °C saline, followed by placement of an endovascular cooling catheter. Hypothermia with fluid loading alone is difficult to maintain for long periods of time as the patient is at risk for volume overload.

Endovascular cooling devices have been shown to reliably induce hypothermia, maintain target temperature, and provide stable re-warming without rebound temperature elevation [158, 171]. The ICTuS-L trial confirmed that hypothermia could be induced safely in patients receiving thrombolytic therapy with rt-PA [172]. The ICTuS-L trial also confirmed that very rapid and precisely controlled hypothermia can be induced in awake stroke patients with endovascular devices [171, 172]. The induction of hypothermia in stroke patients depends on the power of the cooling device, but the patient’s body mass index and age critically determine the speed of cooling [173]. While hypothermia may pose added risk, for example an increased risk of pneumonia, but further studies are need to confirm this [174].