Part III
Ischaemic Stroke Pathophysiology
How occluded flow becomes infarcted brain — the molecular cascade from energy failure through glutamate excitotoxicity, calcium overload, oxidative stress, and inflammation, and the conceptual heart of acute stroke therapy: the ischaemic penumbra.
1. Cerebral Blood Flow Physiology
The adult brain is ~2% of body mass but consumes ~20% of resting cardiac output, ~20% of oxygen, and ~25% of glucose. It is the body’s most metabolically extravagant tissue and the most intolerant of supply interruption. Mean global cerebral blood flow (CBF) in healthy adults is ~50 mL/100 g/min, partitioned into ~80 mL/100 g/min in grey matter and ~20 mL/100 g/min in white matter. The cerebral metabolic rate of oxygen (CMRO2) is ~3.5 mL O2/100 g/min.
CBF obeys a Poiseuille-like relation between perfusion pressure and cerebrovascular resistance:
where CPP is cerebral perfusion pressure (mean arterial pressure minus intracranial pressure) and CVR is cerebrovascular resistance, set primarily by smooth-muscle tone of pre-capillary arterioles. Three regulatory systems keep CBF appropriate:
Pressure autoregulation
Bayliss myogenic response keeps CBF roughly constant over MAP ~50–150 mmHg. Curve is right-shifted in chronic hypertension, narrowing tolerance for pressure drops.
CO2 reactivity
CBF rises ~2–4% per mmHg PaCO2; arteriolar pH-dependent. The basis for hyperventilation-induced vasoconstriction in raised ICP — effective but transient.
Neurovascular coupling
Local activity dilates feeding arterioles via astrocytic Ca²⁺ → arachidonic-acid metabolites (EETs, 20-HETE), NO, and K⁺ siphoning. Basis of fMRI BOLD signal.
Oxygen delivery is \(\mathrm{DO_2} = \mathrm{CBF} \cdot \mathrm{CaO_2}\), and at rest the brain extracts only ~30–40% of the oxygen that arrives (oxygen extraction fraction, OEF). This headroom is the brain’s first line of defence: when CBF falls, OEF can rise toward unity (“misery perfusion,” Baron 1981) before metabolism is forced to fail.
2. The CBF Thresholds
Heiss and Astrup’s classical baboon experiments (1981) established that cerebral function and cerebral structure fail at different CBF thresholds. That gap — tissue that has stopped working but not yet died — is the biological substrate of all acute stroke therapy.
| CBF (mL/100 g/min) | Tissue state | Marker / consequence |
|---|---|---|
| ~50 | Normal | Resting CMRO2; full function. |
| ~22–35 | Oligaemia | OEF rises; protein synthesis falls; clinically silent. |
| ~18–22 | Functional threshold | EEG flattens; evoked potentials lost; neurological deficit appears. |
| ~12–18 | Penumbra | Synaptic transmission off, ion homeostasis preserved — viable but at risk. |
| <10–12 | Core / infarction | Membrane failure; anoxic depolarisation within minutes; cell death. |
Critically, the thresholds are time-dependent: tissue at 15 mL/100 g/min survives perhaps an hour; the same tissue at 8 mL/100 g/min infarcts in minutes. Jones et al. (1981) captured this as a CBF×duration hyperbola — the “Jones curve” — whose practical meaning is that even modest improvements in residual flow (collaterals, BP support) can dramatically extend the window for salvage.
Oligaemia (~22–35)
Mild flow reduction. Rarely infarcts unless flow drops further or duration is prolonged (e.g. watershed under chronic stenosis).
Penumbra (~12–22)
Electrically silent, metabolically alive. This is what reperfusion saves. Imaged as “perfusion–diffusion mismatch” on MRI.
Core (<10)
Membrane failure already occurred. Reperfusion here cannot salvage tissue and raises haemorrhagic-transformation risk.
3. Energy Failure & Ion-Pump Collapse
Once CBF drops below ~10 mL/100 g/min, oxygen delivery cannot match CMRO2. Mitochondrial oxidative phosphorylation slows; the cell shifts to anaerobic glycolysis, generating lactate (~15–20 mM in the core; pH falls to ~6.2–6.5) and yielding only 2 ATP per glucose instead of ~30. Within ~2 minutes of complete ischaemia, ATP collapses to <10% of baseline.
The cell’s energy state is summarised by Atkinson’s adenylate energy charge:
EC is normally ~0.9. In severe ischaemia it falls below ~0.5, at which point ATP-dependent enzymes (kinases, ligases, ATPases) cease to function. The first casualty — and the central event of the ischaemic cascade — is the plasma-membrane Na⁺/K⁺-ATPase.
Anoxic depolarisation
When the Na⁺/K⁺-ATPase stops:
- K⁺ leaks out down its gradient; extracellular [K⁺] rises from ~3 mM to ~50–80 mM within ~2 min.
- Na⁺ and Cl⁻ rush in; water follows osmotically → cytotoxic oedema (the substrate for early DWI hyperintensity).
- Membrane potential collapses from ~−70 mV to ~−20 mV.
- Voltage-gated Ca²⁺ channels (L, N, P/Q types) open; presynaptic terminals release their neurotransmitter cargo — including a flood of glutamate.
- Na⁺/Ca²⁺ exchangers (NCX) reverse, importing yet more Ca²⁺.
This anoxic depolarisation propagates as cortical spreading depolarisations (SDs) that recur for hours, each one consuming residual ATP and progressively recruiting penumbral tissue into the core (Hartings et al. 2017).
For the underlying biophysics of the resting potential, the Na⁺/K⁺-ATPase, and the action potential the ischaemic cascade dismantles, see Cell Physiology Part V.
4. Glutamate Excitotoxicity
The term excitotoxicity was coined by John Olney (1969) to describe the paradox that an ordinary excitatory neurotransmitter, when present in excess, kills the neurons it normally signals. Dennis Choi’s seminal cortical-culture work (1985–1992) established glutamate excitotoxicity as the central mechanism of ischaemic neuronal death, and Stuart Lipton’s laboratory mapped much of its molecular architecture (Lipton & Rosenberg, NEJM 1994).
Resting extracellular [glutamate] in brain interstitium is held at ~1–3 µM by astrocytic uptake transporters (EAAT1/GLAST and EAAT2/GLT-1, encoded by SLC1A3 and SLC1A2; EAAT2 alone clears ~90% of synaptic glutamate). Within minutes of ischaemia, extracellular glutamate rises 10–100× (microdialysis: peaks of 30–500 µM in the core).
Why glutamate accumulates
- Vesicular release from depolarised pre-synaptic terminals (Ca²⁺-dependent SNARE-mediated exocytosis).
- Reverse transport by EAATs — the transporters are electrogenic (3 Na⁺ + 1 H⁺ in / 1 K⁺ + 1 glutamate out per cycle); when Na⁺ and K⁺ gradients collapse, they run backwards, dumping cytosolic glutamate into the extracellular space.
- Cell lysis in the core releases the entire cytosolic pool (~10 mM intracellular).
- Failed astrocytic clearance — EAAT2 shut-down in the depolarised, energy-starved astrocyte.
Three glutamate receptor families convert this extracellular flood into intracellular catastrophe:
NMDA receptor
GluN1 + GluN2A/B tetramer. Voltage-gated by Mg²⁺ block (relieved by depolarisation), highly Ca²⁺-permeable (PCa/PNa ~10). The principal Ca²⁺ entry route in excitotoxicity. GluN2B-containing extrasynaptic NMDARs are particularly toxic (Hardingham & Bading 2010).
AMPA receptor
GluA1–4 tetramer. Fast Na⁺ flux drives further depolarisation that unblocks NMDARs. GluA2-lacking AMPARs (induced by ischaemia) become Ca²⁺-permeable — the “GluA2 hypothesis” of delayed CA1 death (Pellegrini-Giampietro et al. 1992).
Kainate & mGluR
GluK1–5 ionotropic kainate receptors and metabotropic mGluR1–8 modulate the cascade. mGluR1/5 amplify Ca²⁺ release from IP3-sensitive ER stores; group II/III mGluRs (mGluR2/3, 4/6/7/8) are partially neuroprotective.
NMDA receptor (GluN1/GluN2 tetramer) — PDB 5ZMC
Structure of the heterotetrameric NMDA glutamate receptor — the principal calcium-permeable ionotropic glutamate receptor of the forebrain. Over-activation by extracellular glutamate flood is the central event of ischaemic excitotoxicity, driving the lethal Ca²⁺ surge dissected in the next section. NMDA antagonism was the most heavily pursued neuroprotection target of the 1990s–2000s; despite spectacular success in animal models, every clinical trial (selfotel, aptiganel/CNS-1102, gavestinel, magnesium in IMAGES, NA-1 in ESCAPE-NA1, dexanabinol, ifenprodil) failed to translate, due to a combination of narrow therapeutic windows, dose-limiting psychotomimetic effects, and the discovery that synaptic NMDARs are pro-survival while only the extrasynaptic pool is toxic — a pharmacological needle the field never quite threaded.
See Neuroscience for the broader physiology of glutamatergic signalling, NMDA-dependent plasticity, and the synaptic/extrasynaptic NMDAR dichotomy that determined why decades of neuroprotection trials failed.
5. The Ca²⁺-Dependent Cascade
Resting cytosolic free [Ca²⁺] is exquisitely tight: ~100 nM, against an extracellular ~1.2 mM and ER lumen ~0.5 mM. Within minutes of NMDAR over-activation, cytosolic [Ca²⁺] rises to 1–10 µM — a 10–100-fold elevation that detonates a wide network of Ca²⁺-activated enzymes, each individually capable of killing the neuron.
| Ca²⁺-activated effector | Cellular damage |
|---|---|
| Calpain-1 / -2 (cytosolic cysteine proteases) | Cleaves spectrin, MAP2, fodrin, tau; degrades Na⁺/Ca²⁺ exchanger 3 (NCX3) — locking Ca²⁺ in. |
| Phospholipase A2 (cPLA2, sPLA2) | Liberates arachidonic acid & lyso-phospholipids; substrate for COX/LOX → prostaglandins, leukotrienes, ROS. |
| Phospholipase C (PLC) | IP3 release amplifies Ca²⁺ from ER stores (CICR via RyR/IP3R). |
| Neuronal NOS (nNOS, NOS1) | Tethered to GluN2B via PSD-95; produces NO that combines with O2·− → peroxynitrite (ONOO⁻). |
| Endonucleases | DNA fragmentation (apoptotic signature). |
| CaMKII (autophosphorylating) | Persistent activation drives mis-targeted phosphorylation. |
| Mitochondrial Ca²⁺ uniporter (MCU) | Massive matrix Ca²⁺ uptake → mitochondrial permeability transition pore (mPTP) opening. |
The mitochondrial pivot
Mitochondria buffer cytosolic Ca²⁺ until matrix loading exceeds threshold; then the mitochondrial permeability transition pore opens. The inner membrane becomes non-selectively permeable to anything <1.5 kDa; the proton gradient collapses (ΔΨm → 0); ATP synthase reverses, consuming the last cytosolic ATP; matrix swelling ruptures the outer membrane, releasing cytochrome c, AIF, Smac/DIABLO, and Endo G into the cytosol — the apoptotic kiss-of-death signal. This is the molecular point of no return for the cell.
6. Reactive Oxygen & Nitrogen Species
Ischaemia, and especially reperfusion, generate a torrent of free radicals. Because the brain is ~10% lipid by weight, rich in polyunsaturated fatty acids (DHA, AA), and contains high concentrations of redox-active iron, it is uniquely susceptible to oxidative injury (Chan 2001; Lo et al., Nat Rev Neurosci 2003).
Sources of ROS
- Mitochondrial leak — complexes I & III, accelerated when Ca²⁺-loaded and the ETC is reduced.
- NADPH oxidase (NOX2/NOX4) on microglia, neutrophils, vasculature → O2·−.
- Xanthine oxidase (converted from XDH on Ca²⁺-activated proteolysis): hypoxanthine + O2 → uric acid + O2·−.
- nNOS / iNOS → NO, then peroxynitrite.
- COX-2, 12/15-LOX (prostaglandin/leukotriene biosynthesis).
- Free Fe²⁺ released from ferritin → Fenton reaction → ·OH (hydroxyl radical, virtually undetoxifiable).
Damage mechanisms
- Lipid peroxidation of PUFAs → 4-HNE, MDA, isoprostanes; membrane integrity loss.
- Protein oxidation / nitration — tyrosine nitration (3-NT) inactivates SOD, MnSOD, prostacyclin synthase, parkin.
- DNA strand breaks activate PARP-1; PARP hyperactivation depletes NAD⁺ & ATP — “parthanatos” (Yu et al. 2002).
- Ferroptosis — iron-dependent lipid-peroxidation cell death; GPX4-dependent.
- BBB injury via MMP-9 activation by ONOO⁻.
Endogenous antioxidants — SOD1/SOD2/SOD3, catalase, glutathione peroxidases, peroxiredoxins, and the Nrf2/Keap1 transcriptional axis — are overwhelmed within minutes. Glutathione (GSH), normally ~2–5 mM in neurons and astrocytes, is depleted; its synthesis is itself ATP-dependent.
7. Neuroinflammation
By hours–days post-occlusion, the ischaemic cascade has shifted from cell-autonomous toxicity to a sterile inflammatory response that recruits resident microglia, peripheral leukocytes, and the cerebral vasculature itself (Iadecola & Anrather, Nat Med 2011). Damaged neurons release DAMPs (HMGB1, ATP, S100B, peroxiredoxins, nucleic acids) that ligate pattern-recognition receptors (TLR2/4, RAGE, NLRP3) on microglia.
- Microglial activation — resident microglia transition from ramified surveillance to amoeboid phagocytic state within ~30 min; classical “M1-like” cells release IL-1β, IL-6, TNF-α, NO, and ROS, while “M2-like” cells release IL-10, TGF-β, BDNF and clear debris. The balance evolves over days.
- NLRP3 inflammasome — assembled in response to extracellular ATP (P2X7), K⁺ efflux, and ROS; activates caspase-1, which cleaves pro-IL-1β and gasdermin D → pyroptotic pores.
- Endothelial activation — TNF/IL-1 induce E-selectin, P-selectin, ICAM-1, VCAM-1; neutrophils tether, roll, firm-adhere, and transmigrate within hours.
- Neutrophil infiltration peaks at ~1–3 days; releases MMP-9, NE, cathepsin G, NETs — driving secondary BBB injury.
- Monocyte/macrophage entry peaks at days 3–7; lymphocytes (T cells, including a damaging IL-17-producing γδ-T-cell population) accrue subsequently.
- BBB breakdown — MMP-2/9 degrade tight-junction proteins (claudin-5, occludin, ZO-1) and basement-membrane laminin/collagen IV; oedema worsens; haemorrhagic transformation risk rises.
Cytokine signatures vary by cell-of-origin: TNF-α and IL-1β from microglia, IL-6 from astrocytes & endothelium, IFN-γ from infiltrating T cells. Transcriptomic time-courses (Cunningham 2019) show a biphasic profile — an acute destructive phase (0–3 days) and a delayed reparative phase (5–21 days) involving angiogenesis, neurogenesis, gliogenesis, and remodelling. Therapeutic modulation must respect this temporal duality — blanket immunosuppression in the repair phase is harmful.
8. Apoptosis vs Necrosis
Ischaemic cell death is not a single mechanism but a continuum, and the same neuron can show features of several. The dominant mode is set by ATP availability, severity of injury, and time:
Necrosis (the core)
ATP-independent; rapid; minutes to hours. Membrane rupture; release of intracellular contents; secondary inflammation. Predominates where CBF < 10 mL/100 g/min.
- Massive Ca²⁺ overload → calpain → membrane / cytoskeleton failure.
- mPTP opening, ATP collapse, mitochondrial swelling.
- Necroptosis: RIPK1/RIPK3 → MLKL pore-formation (Degterev 2005).
Apoptosis (the penumbra)
ATP-dependent; programmed; hours to days. Predominates where residual ATP >~25% permits caspase activation. Cells with intermediate injury can be rescued for many hours.
- Intrinsic: Bax/Bak → cyt c → Apaf-1 → caspase-9 → -3.
- Extrinsic: TNFR/Fas → caspase-8 → -3.
- Caspase-independent: AIF, Endo G translocate to nucleus.
Other death modalities recognised in stroke biology: parthanatos (PARP-1 hyperactivation → AIF release; Yu 2002), ferroptosis (iron-dependent lipid peroxidation, GPX4-regulated; Dixon 2012), pyroptosis (inflammasome → gasdermin D pores; caspase-1/-11), and autophagic cell death.
Spatial and temporal stratification matters: rapid necrosis in the core occurs in minutes; apoptotic and inflammatory secondary death extend across days–weeks into adjacent tissue. The latter creates the therapeutic window beyond reperfusion — a still largely unrealised target for neuroprotection, now reawakening with better trial design and tissue-clock imaging.
9. The Penumbra Concept
The single most consequential idea in modern stroke medicine is the ischaemic penumbra — tissue that is functionally silent but structurally intact, and whose fate hangs on whether perfusion is restored in time. Lindsay Symon, Niels Astrup, and Bo Siesjö framed the concept in baboon middle-cerebral-artery occlusion experiments (Astrup, Siesjö & Symon, Stroke 1981): around the lethally hypoperfused core sat a mantle of tissue with CBF below the functional threshold (~18–22 mL/100 g/min) but above the membrane-failure threshold (~10–12 mL/100 g/min). Function was lost; structure was not. They borrowed the term penumbra from astronomy — the partial shadow at the edge of an eclipse — because clinically the patient is dark, but biologically the tissue still has light.
Why the penumbra is the whole point
If the core were the entire infarct, acute therapy would be useless — the tissue is already dead before the patient reaches the hospital. The reason thrombolysis (1995) and thrombectomy (2015) work is that, in a typical large-vessel occlusion, the core occupies only a fraction of the territory at risk (often ~10–30% in the early hours) while the penumbra occupies the rest. Reperfusion converts viable penumbra into surviving brain. Failed reperfusion lets the penumbra collapse into core over hours.
Hossmann (1994) and Heiss (2003) summarised it best: the penumbra is tissue whose fate is determined not by the occlusion but by what we do about it.
Imaging operationalises the penumbra in two complementary ways — both detailed in Part VI: Imaging:
- PET (the gold standard, but impractical) — 15O-water CBF, 15O2 CMRO2, and OEF maps; the penumbra is tissue with elevated OEF >~0.7 (“misery perfusion”).
- CT perfusion / MR PWI — Tmax > 6 s defines hypoperfused tissue; CBF < 30% of contralateral defines core; mismatch volume = penumbra.
- DWI / PWI mismatch — DWI marks tissue where cytotoxic oedema (failed Na⁺/K⁺-ATPase) restricts water diffusion → core. PWI marks hypoperfusion. The PWI−DWI mismatch is the penumbra in shorthand. (See DEFUSE, EPITHET, EXTEND.)
The DAWN (NEJM 2018) and DEFUSE-3 (NEJM 2018) trials revolutionised stroke care by proving that imaging-defined favourable tissue physiology — not the wall clock — should select late-window patients for thrombectomy. Patients up to 24 h from onset, with a small core and large mismatch, gained absolute benefits in functional independence of ~36% in DAWN, ~25% in DEFUSE-3: among the largest treatment effects in modern medicine.
Tissue clock vs wall clock
Two patients with the “same” 6-hour MCA occlusion can have radically different penumbral physiology. Why?
- Collateral circulation — pial leptomeningeal anastomoses from neighbouring territories (ACA → MCA, PCA → MCA) sustain residual flow into the penumbra. Excellent collaterals (ASPECTS > 8, robust pial CTA) buy hours; poor collaterals collapse the penumbra in <1 h.
- Pre-morbid vascular reserve — chronic stenosis can recruit collateral pathways already; conversely, stiff hypertensive vessels lose autoregulatory range.
- Systemic factors — BP, glucose, temperature, haematocrit all modulate residual perfusion and metabolic demand.
- Genetic and individual variability — Circle-of-Willis configuration (only ~50% of people have the textbook ring) and individual EAAT2 / NMDAR / antioxidant capacities.
The practical consequence: every patient has their own time window. The wall clock recommends an upper limit; the tissue clock determines the actual one. This is the philosophical bridge from Part III pathophysiology to Part VI imaging-based selection and Part VII reperfusion therapy.
Tissue plasminogen activator (tPA, two-chain) — PDB 1A5H
Crystal structure of human tissue plasminogen activator — the molecule that, as recombinant alteplase (and now tenecteplase), drives clinical thrombolysis and gave the penumbra its first practical defender. tPA cleaves plasminogen → plasmin at the fibrin surface, dissolving the occluding clot. The 1995 NINDS rt-PA trial proved that within 3 h (later extended to 4.5 h by ECASS-3, 2008), pharmacological reperfusion turns penumbra into surviving brain — the first treatment to make stroke a treatable disease. Note the kringle (K1, K2) and serine-protease domains visible in the structure; fibrin binding via the K2 finger gives the enzyme its clot specificity.
10. Reperfusion Injury & Haemorrhagic Transformation
Reperfusion is the goal of acute therapy and is itself a source of injury — the great therapeutic paradox of stroke. Eng Lo’s “neurovascular unit” framework (Lo et al., Nat Rev Neurosci 2003; Stroke 2013) emphasises that reopening an artery dumps oxygen onto reduced mitochondria, returns leukocytes and complement to ischaemic endothelium, and spikes hydrostatic pressure across a compromised BBB.
Mechanisms of reperfusion injury
- Oxidative burst as O2 meets reduced ETC complexes — peak ROS production at minutes to hours post-recanalisation.
- Calcium re-loading — restored ATP energises pumps that pump Ca²⁺ into mitochondria → mPTP opening.
- Leukocyte / platelet recruitment — “no-reflow” phenomenon: capillary plugging by neutrophils & platelets, plus pericyte constriction (Hall et al., Nature 2014).
- Complement — C3a, C5a deposition activates microglia and lyses cells.
- BBB injury — biphasic opening at ~3 h (MMP-2) and 24–48 h (MMP-9); vasogenic oedema.
Haemorrhagic transformation (HT)
- ~10–40% of ischaemic strokes show some HT on MRI; ~2–7% develop symptomatic ICH after IV-tPA (NINDS, ECASS).
- HI-1 / HI-2: petechial; usually clinically silent.
- PH-1: parenchymal haematoma <30% of infarct, no mass effect.
- PH-2: >30% of infarct, mass effect — strongly associated with neurological deterioration and death.
- Risk factors: large core (low ASPECTS, large DWI), late time-to-reperfusion, hyperglycaemia, hypertension, tPA itself (proteolytic effects on BBB beyond fibrinolysis).
The molecular driver of post-reperfusion BBB rupture is matrix metalloproteinase-9 (MMP-9): activated by ONOO⁻ and pro-inflammatory cytokines, MMP-9 cleaves the basement-membrane collagen IV and laminin and degrades tight-junction claudin-5 / occludin. Its plasma level at 24 h correlates with HT risk; selective inhibitors are an active translational target.
The future of stroke therapy — combinations of reperfusion plus targeted cytoprotection (NA-1/nerinetide, uric acid, edaravone, normobaric oxygen, hypothermia) — rests on a precise molecular understanding of this cascade. Every step in this chapter is, in principle, a drug target; every failed neuroprotection trial since 1990 (and there are over 1,000 of them) is a lesson in how poorly preclinical models translate. Better stratification (imaging-defined penumbra, biomarker-defined inflammatory phenotype) is bringing the field back into motion.
Where to Next
Part IV (next)
Haemorrhagic Stroke
A different physiology: ICH, SAH, ICP, Cushing reflex, vasospasm.
Part VI
Imaging the Penumbra
DWI/PWI mismatch, CTP, ASPECTS, late-window selection.
Part VII
Acute Management
Reperfusion biology in clinical practice: tPA, TNK, thrombectomy.
Part II (prev)
Cerebrovascular Anatomy
The vascular substrate the cascade unfolds in.
Cross-course
Cell Physiology Part V
Resting potential, Na⁺/K⁺-ATPase, action potential — the machinery the cascade dismantles.
Cross-course
Neuroscience
Glutamate, NMDA receptor biology, synaptic vs extrasynaptic signalling.