The encephalopathy in sepsis is considered a diffuse cerebral dysfunction as a consequence of the systemic inflammatory response to an infection, with no direct central nervous system infection (Figure 1). The response to stress is physiologically triggered by an activating signal that is mediated by two pathways. The first one is the vagus nerve, which can detect visceral inflammation through its axonal cytokines receptors: inflammatory products produced in damaged tissues activate afferent signals that are relayed to the nucleus tractus solitarius in the brainstem. Subsequent activation of vagus efferent activity inhibits cytokine synthesis in damaged tissues through a cholinergic anti-inflammatory pathway (the inflammatory reflex) 8 . The vagus nerve is also connected to other autonomic nuclei, notably the paraventricular nucleus that controls adrenal axis and vasopressin secretion 9 . The second pathway involves the circumventricular organs (CVOs), which are located near neuroendocrine and neurovegetative nuclei. CVOs are deprived of a blood–brain barrier (BBB) and express components of innate and adaptive immune systems. Once visceral or systemic inflammation is detected by the first or the second pathway, the activating signal will spread to behavioral, neuroendocrine, and neurovegetative centers. Sepsis enhances the transcription of several pro- and anti-inflammatory cytokines and chemokines in the brain, including tumor necrosis factor alpha (TNFα), interleukin-1 beta (IL1β), transforming growth factor beta (TGF β), and monocyte chemoattractant protein 1 (MCP1) 10 . These mediators modulate the expression of α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptors (AMPARs) and N-methyl-D-aspartate receptors (NMDARs) on neurons, inducing brain dysfunction 11 . Recent studies have suggested the novel importances of IL1β and High Mobility Group Box 1 on the development of cognitive impairment in sepsis survivors 12 13 . These cytokines also modulate NMDARs, with functional consequences on cognition and behavior 14 .

Microglial activation may represent one of the earliest changes observed in sepsis-associated encephalopathy and prolonged microglial activation may negatively affect other brain cells 15 . Early microglial activation in sepsis was evidenced in mice models within 4 hours following LPS injection, as assessed by the increased proinflammatory cytokine IL1β level in microglia 16 . Using Positron Emission Tomography (PET) imaging in nonhuman primates, another study demonstrated microglia activation only 1h after LPS-induced systemic inflammation 17 . Moreover, experimental studies suggest that aging may increase the intensity of microglial activation and the production proinflammatory cytokines in the hippocampus, notably IL1β 18 19 . The IL1β-mediated inflammatory process in the hippocampus was confirmed in different models of systemic inflammation, including cecal ligation and puncture and peripheral surgery 20 21 . Numerous experimental studies suggest that these proinflammatory mediators released in the central nervous system at the onset of sepsis will in turn lead to neuronal loss within vulnerable areas of the brain, including the hippocampus 22 23 24 . Collectively, these findings represent a neuropathological basis for persistent cognitive impairment, hippocampal atrophy, and electroencephalographic changes observed in sepsis survivors 25 .

Sepsis induces activation of cerebral endothelial cells, which result in BBB dysfunction and release of various mediators into the brain. Experimental data indicate that at the early phase of sepsis, endothelial nitric oxide synthase-derived nitric oxide exhibits proinflammatory characteristics and contributes to the activation and dysfunction of cerebrovascular endothelial cells 26 . The activated endothelium relays the inflammatory response into the brain by releasing proinflammatory cytokines and NO that are able to interact with surrounding brain cells. The other consequences of endothelial activation may include microcirculatory dysfunction, which might compromise cerebral perfusion 27 . Other studies suggest that endotoxemia leads to inflammation in brain, with alteration in BBB, up-regulation of aquaporin 4 (AQP4) and associated edema, neutrophil infiltration, astrocyte activation, as well as apoptotic cell death, all of which appear to be mediated by TNF-alpha signaling through TNF-receptor 1 28 . Alterations of BBB also have been evidenced in patients with septic shock, with help of brain MRI 29 . BBB breakdown can be localized in the cortex around the Virchow-Robin spaces or have a more diffuse pattern in the whole white matter. It also can predominate in posterior lobes, being consistent with a posterior reversible encephalopathy syndrome 30 . Noninvasive assessment by MRI allowed the identification of new aspects of brain damage in experimental models of sepsis, including cytotoxic and vasogenic edema as well as neuronal damage. These findings highlight the potential applications of MRI techniques for the diagnostic and therapeutic studies in sepsis 31 . Finally, BBB alterations might facilitate the passage of potential neurotoxic factors from the peripheral circulation to the brain. For instance, plasma tryptophan levels are associated with delirium in critically ill patients 32 33 . More recently, increased kynurenine pathway activation, assessed by plasma kynurenine and kynurenine/tryptophan ratio, was found to be associated with fewer days without acute brain dysfunction 33 .

Endothelial activation alters vascular tone and induces both microcirculatory dysfunction and coagulopathy, which will in turn favor ischemic and/or hemorrhagic lesions 34 . Neuropathological studies performed in nonsurvivors of septic shock suggest that ischemia is consistently observed in brain areas susceptible to low cerebral flow and that hemorrhages can be found in approximately 10% of cases 34 35 . Furthermore, it has been recently shown that SAE is rather associated with disturbed autoregulation than with altered cerebral blood flow or tissue oxygenation 36 .

Deficits in cholinergic function have been postulated to cause delirium and cognitive decline 37 . Global hypocholinergia result from several mechanisms, including impaired acetylcholine synthesis and cholinergic synaptic dysfunction (impairment of presynaptic, synaptic, or postsynaptic functions of acetylcholine). Nicotinic receptors within the brain bind acetylcholine to modulate cognitive functioning, arousal learning, and memory. Use of anesthetic drugs that decrease acetylcholine release (e.g., isoflurane) may impact on cognitive function after surgery 38 . Anticholinergic medications and their metabolites also may induce delirium through competitive antagonism of postsynaptic muscarinic receptors that are more widely distributed throughout the brain. It was recently demonstrated that chronic cholinergic hypoactivity in the basal forebrain represents a major factor of acute brain dysfunction under systemic inflammation 39 .

Other significant neurotransmitter alterations have been described during experimental sepsis, involving brain beta-adrenergic, gamma-aminobutyric acid, and serotoninergic pathways 40 . These phenomena seem to predominate in cortex and in hippocampus, and may be mediated by NO, cytokines and prostaglandins 41 . Neurotransmitter balance also is altered by different circulating molecules, such as ammonium, tyrosine, tryptophan, and phenylalanine, whose plasma levels are increased secondary to liver dysfunction and muscle proteolysis 42 . Imbalance between dopaminergic and cholinergic neurotransmission is considered a major mechanism of delirium in critically ill patients. It also has been hypothesized that reduced cholinergic inhibition of microglia is involved in delirium 15 . However, administration of rivastigmine, a pharmacological agent that may restore cholinergic control of microglia, did not decrease duration of delirium and might have increased mortality of critically ill patients with delirium 43 . Data from studies performed in critically ill patients receiving prolonged mechanical ventilation also suggest that use of GABA-agonists, such as benzodiazepines, is associated with an increased risk of brain dysfunction 44 . Noradrenergic neurotransmission also might be particularly involved in SAE as dexmedetomidine, a selective agonist of alpha2-adrenoceptors expressed in the locus coeruleus, is associated with less brain dysfunction and better outcomes in septic patients compared with midazolam 45 46 . These findings were recently confirmed in a recent, multicenter trial, where dexmedetomidine compared with midazolam, improved patient’s ability to communicate 47 . Beneficial effects of dexmedetomidine notably include significant preconditioning and postconditioning effects against ischemic brain injury 48 49 . More recent experimental data suggested that neuroprotective effects of dexmedetomidine against glutamate-induced cell death were mediated by an increase in astrocyte expression of brain-derived neurotrophic factor through an extracellular signal-regulated kinase-dependent pathway 50 .

Experimental data suggest that oxidative damage, assessed by the thiobarbituric acid reactive species and the protein carbonyl assays, occurs early (after 6 hours) in the course of sepsis 51 . Moreover, the combined use of antioxidants (N-acetyl-cysteine and deferoxamine) attenuates oxidative damage in hippocampus 6 hours after sepsis induction 52 . Mitochondrial-mediated apoptosis has been evidenced in experimental sepsis and might be related to a decrease of intracellular anti-apoptotic (bcl-2) and an increase of pro-apoptotic (bax) factors 22 . In patients who had died from septic shock, neuronal and microglial apoptosis have been detected in neurovegetative and neuroendocrine nuclei as well as in amygdala 35 . Intensity of apoptosis correlated with expression of endothelial iNOS. Recent data suggested that Caspase-3 is involved in part in apoptosis in the dentate gyrus cell layer of the hippocampus in septic rats 53 . Additionally to nitric oxide, other pro-apoptotic factors have been incriminated, such as glutamate, TNFα, and hyperglycemia 54 .

A recent study performed in critically ill patients suggested that high baseline inflammatory biomarkers of systemic inflammation at admission, such as C-reactive protein and procalcitonin, predicted prolonged periods of acute brain dysfunction, irrespective of whether patients had sepsis or not 55 . In another study performed in 50 patients with systemic inflammatory response syndrome (SIRS), IL-8 was independently associated with delirium 56 . These findings suggest that sustained systemic inflammation may contribute to prolong or aggravate brain dysfunction. Stress hyperglycemia, which commonly develops in critically ill patients, also may aggravate sepsis-induced brain injury. In a recent neuropathological study performed in nonsurvivors of critical illness, hyperglycemia was shown to aggravate critical illness-induced neuropathological changes 57 . Patients with uncontrolled hyperglycemia showed increased microglial activation, an important reduction in the density and activation status of astrocytes, increased neuronal damage, and apoptosis in hippocampus and frontal cortex. Most of these abnormalities were virtually absent with normoglycemia. Interestingly, moderate hyperglycemia was shown to be associated with adverse cognitive outcomes in survivors of the acute respiratory distress syndrome 58 . Prolonged hypoxemia also may contribute to brain dysfunction and injury during sepsis. Recently, in a study performed in survivors of prolonged critical illness, a low partial pressure of arterial oxygen during ICU stay was associated with cognitive and psychiatric impairment at 12 months 59 .

Brain lesions in septic shock have been described in areas of the brain susceptible to ischemia (including Ammon’s horn, the frontal junctional cortex, the lenticular nuclei, the dentate nucleus, and the medullary olive), in the hypothalamic nuclei (supraoptic and paraventricular nuclei), the amygdala, the locus coeruleus, and the medullary autonomic nuclei (nucleus tractus solitarii, ambiguous, and parabrachial nuclei) 34 35 . Hippocampal lesions due to inflammatory but also ischemic, hypoxic, or dysglycemic insults 57 60 may explain long-term psychological and cognitive disorders observed in survivors of critical illness 25 61 . Interestingly, reduction of oxidative stress in hippocampus is associated with less cognitive dysfunction in septic rats 52 . There also are arguments for a global brainstem dysfunction during sepsis. First, a recent study suggested that abolition of cough reflex and oculocephalic responses in sedated critically ill patients are associated with death and delirium, respectively 62 . Second, the brainstem nuclei are liable to apoptosis 35 and use of dexmedetomidine, which have antiapoptotic properties 63 , is associated with less delirium in septic patients 64 . Finally, impaired sympathetic control of heart rate is frequent and associated with increased mortality in septic patients suggesting a central autonomic regulatory dysfunction 65 .

In case of focal neurological signs, a brain computed tomography should be performed to rule out ischemic or hemorrhagic brain injury 29 34 . Brain MRI should be considered systematically in case of persistent encephalopathy after control of sepsis and exclusion of major confounding factors. With use of diffusion-weighted imaging and gradient echo sequences, this tool has a higher sensitivity than CT for detecting acute CNS disorders, such as recent ischemic or hemorrhagic stroke. Several changes have been described in septic patients (Table 1). Moreover, additionally to the importance of etiological diagnosis, assessment of the nature, and extent of brain damage may also influence patient’s treatment.

In case of seizure(s), including palpebral myoclonus, an EEG will be required to rule out subtle status epilepticus. Several electrographic patterns have been described in septic patients (Table 2). Sepsis can be associated with electrographic seizures or periodic epileptiform discharges 73 . In a recent study, generalized periodic discharges were strongly associated with nonconvulsive seizures and nonconvulsive status epilepticus 74 . However, the impact of generalized periodic discharges on neurologic outcomes remains to be elucidated. Other EEG abnormalities include increased theta rhythms, triphasic waves and, less often but more pejorative, burst suppression patterns 75 .

Standard laboratory tests should be performed systematically to detect and correct common metabolic disturbances that can cause delirium or coma (such as hypoglycemia, hypercalcemia, hypo- or hypernatremia). In addition to sedatives and analgesics, many classes of drugs currently administered in critically ill patients can induce acute brain dysfunction, notably a number of antibiotics, steroids, and cardiac drugs (Table 3).

CNS central nervous system.

Benzodiazepines and opioid withdrawal syndromes may represent an important cause of delirium after discontinuation of sedation. Alcohol withdrawal syndrome often is evoked in a patient with a history of alcohol dependence who develops encephalopathy 78 . The predominance of psychomotor agitation and autonomic signs are suggestive of the diagnosis. The benefit of alcohol withdrawal prophylaxis is unproven. Early and aggressive titration of medication guided by symptoms is the only feature associated with improved outcomes. In malnourished or alcoholic patients, Wernicke’s encephalopathy must always be evoked and treated with intravenous thiamine, especially if there is evidence of ophthalmoplegia or ataxia 79 . Thiamine deficiency can be aggravated by infusion of glucose. Tobacco dependency is a risk factor for delirium in critically ill patients, which may be prevented by use of nicotine patch in chronic smokers 80 . Numerous iatrogenic and/or environmental factors also may aggravate brain dysfunction, such as use of physical restraints, excessive noise, or underexposure to light in the ICU 81 82 .

In a patient with unexplained neurologic symptoms (focal neurologic sign or encephalopathy) and bloodstream infection, infective endocarditis should be systematically ruled out as this condition often is associated with neurologic complications 83 . Of note, endocarditis has to be ruled out in the presence of cerebral microbleeds on MRI 84 . Finally, air embolism is an iatrogenic cause of sudden coma, agitation, seizure, or focal neurological signs, and for which hyperbaric oxygen is recommended. It must be emphasized that SAE often is a multifactorial condition. Finally, reappearance or persistence of encephalopathy may indicate that sepsis is not controlled.