Prof. Daniel De Backer, who was essential for the high expertise in the

Acknowledgements

First of all, I would like to thank Prof. Jean-Louis Vincent, who inspired me with his unbelievable strength and passion in the daily work and research. He supported since my first arrival in the Erasme and was present during all the steps of my scientific activity. He invited me to start a PhD project few years ago and significantly contributed to all my results. The best “Chef” and mentor I may wish to anyone.

Then, I’d like to thank all the people who helped me during this long and hard work;

Prof. Daniel De Backer, who was essential for the high expertise in the

microcirculatory field and who gave me all the necessary suggestions and support to develop this project and make it published; Dr. Fuhong Su, who showed me how difficult can be the experimental modelling and who guided all my experiments. I would also like to thank Drs. Xinrong He, Ali Abdelhadii, Charalampos Pierrakos, Katia Donadello, Laura Penaccini, Alessandro Devigili and all those who spent some time with me in the laboratory, discussing about research, medicine and the … meanings of life.

A special thank goes to my friend Dr. Sabino Scolletta, who was more than just a colleague during his two years in Erasme and who supported me during the whole work. Also, I’d like to thank Prof. Cathy De Deyne and Prof. Olivier Dewitte, because they instructed me on how to perform neurosurgery in animals and how to analyse brain metabolism with very advanced techniques.

I would like to thank Mr. Hassane Njimi for his time (and skills!!!) in the statistical analysis of the data; Dr. Karen Pickett, who had the hard task to correct my

“Mediterranean English”; Mrs. Marie-Rose André and all the ICU staff (Veronique, Colette, Alain, Aurelie, Ina, Dominique) for the precious help and presence.

I’d like to underline also the support I received from all my ICU colleagues and

in the laboratory and supported me when I finally came back to clinics, first as a Clinical Fellow, then as Senior ICU staff physician.

Also, this work could not have been realized without the help of the Fondation Erasme, which provided economical and technical assistance during three years for my research, and the European Society of Intensive Care Medicine, which gave me the Basic Science Research Award in 2009 and contributed with this grant to complete my project.

Last but not the least, I would like to thank my family in Italy (my parents, Walter and Virginia, and my brother, Massimiliano), who I had to leave more than 10 years ago for this “new life” in the cold Belgium and who supported me with their love during all this time.

Finally, I have to thank the persons who made my life better and who accepted the

long time I dedicated to research – and then, far from home – just to see my project

achieved. Thanks Vittoria, thanks Virginia and Massimo – without you, this would

have never been done. You are all my life.

Jury Members Prof. Jean-Paul SCULIER – President

Department of Intensive Care and Oncologic Emergencies Institut Jules Bordet – Université Libre de Bruxelles Bruxelles (Belgium)

Prof. Kathleen MCENTEE – Secretary Laboratory of Physiology, Faculty of Medicine Université Libre de Bruxelles

Bruxelles (Belgium)

Prof. Anne-Pascale MEERT

Department of Intensive Care and Oncologic Emergencies Institut Jules Bordet – Université Libre de Bruxelles Bruxelles (Belgium)

Prof. Hazim KADHIM

Department of Pathology – Neuropathology

Centre Hospitalier Universitaire Brugmann – Université Libre de Bruxelles Bruxelles (Belgium)

Prof. Jean-François PAYEN

Department of Anesthesiology and Critical Care Hôpital Albert Michallon

Grenoble (France) Prof. Bernard VIGUÉ

Department of Anesthesiology and Intensive Care Hôpital Kremlin-Bicetre – Université Paris-Sud Le Kremlin-Bicêtre (France)

Prof. Jean-Louis VINCENT - Advisor Department of Intensive Care

Hôpital Erasme – Université Libre de Bruxelles Bruxelles (Belgium)

Prof. Daniel DE BACKER – Co-Advisor Department of Intensive Care

Hôpital Erasme – Université Libre de Bruxelles

Bruxelles (Belgium)

Summary (in English and French) 11

INTRODUCTION 17

Brain dysfunction during sepsis 19

Pathogenesis of SAE: a short overview 22

Alterations in brain perfusion during sepsis 27

Cerebral Blood Flow: normal ranges, physiology and regulation How to monitor cerebral perfusion and autoregulation in septic patients Brain Perfusion in Experimental Sepsis

Brain Perfusion in Human Sepsis

Microcirculation 43

What is microcirculation?

How is microcirculation regulated?

How to monitor microcirculation?

Mechanisms of microcirculatory alterations during sepsis Microvascular abnormalities during sepsis

HYPOTHESIS 63

RESULTS 67 1. Cerebral autoregulation is influenced by carbon dioxide levels in patients with

septic shock.

2. Cerebral microcirculation is impaired during sepsis: an experimental study.

3. Sepsis is associated with altered cerebral microcirculation and tissue hypoxia in experimental peritonitis.

4. Effects of reversal of hypotension on cerebral microcirculation and metabolism in experimental sepsis.

DISCUSSION AND PERSPECTIVES 151

CONCLUSIONS 161

REFERENCES 165

List of Abbreviations

ALI: Acute Lung Injury ANOVA: Analysis Of Variance

APACHE II: Acute Physiology and Chronic Health Evaluation ARDS: Acute Respiratory Distress Syndrome

BBB: Blood-Brain Barrier

BOLD: Blood Oxygenation Level Dependent Contrast CA: Cerebral Autoregulation

CAI: Cerebral Autoregulation Index CBF: Cerebral Blood Flow

CI: Cardiac Index

CMRO

: Cerebral Metabolism Rate Of Oxygen CNS: Central Nervous System

COR: Cerebrovascular-CO

Reactivity CPP: Cerebral Perfusion Pressure CRP: C-Reactive Protein

CT: Computed Tomography

CVOs: Circumventricular Organs CVP: Central Venous Pressure CVR: Cerebrovascular Resistance DO

: Oxygen Delivery

EEG: Electroencephalography

eNOS: Endothelial Nitric Oxide Synthase

FiO

: Inspired Fraction Of Oxygen GCS: Glasgow Coma Score

HbT: Total Tissue Hemoglobin HES: Hydroxyethyl Starch

ICAM-1: Inter-Cellular Adhesion Molecule 1 ICU: Intensive Care Unit

ICP: Intracranial Pressure IFN-: Interferon Gamma

iNOS: Inducible Nitric Oxide Synthase L-NAME: N-Nitro-L-Arginine Methylester L-NMMA: N-Monomethyl-L-Arginine LPR: Lactate/Pyruvate Ratio

LVSWI: Left Ventricular Stroke Work Index MAP: Mean Arterial Pressure

MCA: Middle Cerebral Artery MD: Microdialysis

MFI: Mean Flow Index MOF: Multiple Organ Failure MRI: Magnetic Resonance Imaging NIRS: Near-Infrared Spectroscopy

NO: Nitric Oxide

NPC: Number of Perfused Capillaries NSE: Neuron-Specific Enolase

OPS: Orthogonal Polarization Spectral

PaCO : Arterial Carbon Dioxide Tension

PaO

: Arterial Oxygen Tension

PAOP: Pulmonary Artery Occlusion Pressure PbO

: Brain Oxygen Tension

PET: Positron Emission Tomography PSPV: Proportion Of Small Perfused Vessels PVD: Perfused Vessels Density

RL: Ringer’s Lactate

ROC: Receiver Operating Characteristic ROS: Reactive Oxygen Species

SAE: Sepsis-Associated Encephalopathy SAH: Subarachnoid Hemorrhage

SDF: Sidestream Dark Field

SOFA: Sequential Organ Failure Assessment

SPECT: Single Photon Emission Computed Tomography SSEP: Somato-Sensory Evoked Potential

StO

: Tissue Oxygen Saturation

SvO

: Mixed Venous Oxygen Saturation of Hemoglobin SVR: Systemic Vascular Resistances

SVRI: Systemic Vascular Resistances Index TBI: Traumatic Brain Injury

TCD: Trans-Cranial Doppler

TNF-: Tumor Necrosis Factor Alpha

TPC: Thoraco-Pulmonary Compliance

VMCA: Mean Velocity in the Middle Cerebral Artery

Summary

Brain dysfunction is a frequent complication of sepsis and is usually defined as “sepsis- associated encephalopathy” (SAE). Its pathophysiology is complex and related to a number of processes and pathways, while the exact mechanisms producing neurological impairment in septic patients have not been completely elucidated. Alterations in cerebral blood flow (CBF) have been suggested as a key component for the development of SAE. Reduction of CBF may be caused by cerebral vasoconstriction, induced either by inflammation or hypocapnia. More importantly, the natural mechanisms that protect the brain from reduced/inadequate CBF can be impaired in septic patients, especially in those with shock, and this further contributes to cerebral ischemia if blood pressure drops below a critical threshold. Hypercapnia is associated with a narrower autoregulatory plateau, which may potentially results in large CBF variations when mean arterial pressure (MAP) varies within usual targets. However, few data are available on the role of PaCO

on cerebral autoregulation (CA). Finally, as SAE occurs also in patients without hemodynamic instability, alterations in brain tissue perfusion could occur independently from hypotension; thus, alterations in cerebral microcirculation, which largely regulates regional flow and blood-cellular nutrients exchanges, could contribute to SAE. In septic animals, these microcirculatory abnormalities could be implicated in the development of electrophysiological abnormalities observed during sepsis and contribute to neurological alterations. However, these findings were limited by several factors, including the technique used to assess the microcirculation, the short time of observation and the limited amount of fluid resuscitation used in those models.

In the first part of this work, I evaluated CA and the potential influence of PaCO

on

CA in patients with septic shock. In 21 mechanically ventilated patients, I observed that 14 of

patients with PaCO

< 35 mmHg, 7/9 (77%) with PaCO

between 35 and 42 mmHg, and 3/3 (100%) with PaCO

> 42 mmHg had impaired CA. The Receiver Operating Characteristic (ROC) analysis showed that a PaCO

threshold of 38 mmHg had a sensitivity of 50% and a specificity of 100% for the prediction of impaired CA, with an area under the ROC curve of 0.76 (95% confidence interval: 0.52–0.91).

In the second part of this work, I hypothesized that altered cerebral microcirculation may occur in the early phase of sepsis and contribute to brain hypoxia. In a clinically relevant model of ovine fecal peritonitis, I showed that there was a progressive deterioration of cerebral microvascular flow in septic animals (n=10) when compared to sham animals (n=5), starting already after 6 hours from sepsis induction and becoming significant at 12 hours thereafter. Moreover, changes in the cerebral microcirculation were not related to changes in MAP, cardiac output or blood lactate levels, suggesting that these alterations in the brain may occur even when global perfusion pressure is maintained, i.e. in non-hypotensive conditions.

In a second study, including 10 septic and 5 sham animals, I found that cortical microvascular

alterations were associated with decreased cerebral oxygenation. Furthermore, cerebral

metabolic disturbances compatible with tissue hypoxia (i.e., increased brain lactate/pyruvate

ratio, LPR) occurred mostly during shock, suggesting that hypotension is a critical factor in

the development of anaerobic metabolism in the septic brain. Nevertheless, I showed in a

third study (n=8) that the reversal of hypotension using vasopressor agents, although

increased cerebral oxygenation and slightly reduced LPR, did not significantly influence the

alterations of cerebral microcirculation and was associated with an increase in glutamate and

glycerol, suggesting ongoing excitotoxicity and cellular damage. These alterations in cerebral

microcirculation, oxygenation and metabolism may then contribute to the pathogenesis of

SAE.

Résumé

La dysfonction cérébrale est une complication fréquente du sepsis et elle est généralement

identifiée comme « encéphalopathie associée au sepsis » (sepsis-associated encephalopathy,

SAE). La physiopathologie de la SAE est complexe et liée à des nombreux processus et voies

de signalisation, même si les mécanismes qui induisent cette dysfonction cérébrale chez les

patients en sepsis n’ont pas été clairement élucidés. Des anomalies du débit sanguin cérébral

(cerebral blood flow, CBF) ont été proposées comme une des déterminants pour le

développement de l’SAE. La réduction du CBF pourrait être induite par une vasoconstriction

cérébrale, élicitée pas l’inflammation ou par l’hypocapnie. De plus, les mécanismes qui

naturellement règlent le CBF pour qu’il soit ni diminué ni inadéquat aux besoins cellulaires

peuvent être altérés pendant le sepsis, particulièrement en cas de choc septique, et ceci

pourrait davantage contribuer au développement de zones d’hypoperfusion cérébrale si la

pression artérielle diminue au-dessous d’un seuil critique. Un autre point important est que

l’hypercapnie est associée à une diminution du plateau d’autorégulation du CBF, ce qui

pourrait potentiellement causer des larges variations du CBF endéans des valeurs de pression

artérielle considérés comme normaux en pratique clinique; malheureusement, très peu de

données sont disponibles sur le rôle de la PaCO

sur l’autorégulation cérébrale (cerebral

autoregulation, CA). Enfin, vu que l’SAE survient aussi chez des patients qui n’ont pas

d’instabilité hémodynamique, des anomalies de la perfusion cérébrale régionale pourraient se

produire en absence de toute hypotension artérielle ; en effet, des altérations de la

microcirculation cérébrale, qui règle le débit sanguin au niveau des tissues et l’échange

d’oxygène et nutriments entre la circulation sanguine et le cellules, peuvent aussi contribuer

au développement de la SAE. Dans des modelés expérimentaux de sepsis, les altérations

utilisée pour évaluer la microcirculation, le temps d’observation très court et la quantité limitée de fluides administrés au cours de la réanimation liquidienne dans ces modelés.

Dans la première partie de ce travail, j’ai décrit les anomalies de la CA et l’impact de la PaCO

sur la CA chez des patients en choc septique. En étudiant 21 patients en ventilation mécanique, j’ai pu observer que 14 d’entre eux avaient une CA altérée, y compris 7/14 avec une PaCO

< 40 mmHg et 7/7 avec une PaCO

≥ 40 mmHg (p = 0.046). De plus, 4/9 (44%) avec PaCO

< 35 mmHg, 7/9 (77%) avec PaCO

between 35 and 42 mmHg, and 3/3 (100%) avec PaCO

> 42 mmHg avaient une CA altérée. L’analyse selon la « Receiver Operating Characteristic » (ROC) montrait une sensibilité de 50% et une spécificité de 100% pour prédire une CA altérée, avec un seuil de PaCO2 de 38 mmHg (l’aire sous la courbe de l’analyse ROC était à 0.76 [95% ICs: 0.52–0.91]).

Dans la deuxième partie de ce travail, j’ai émis l’hypothèse que des anomalies de la

microcirculation cérébrale peuvent survenir dans la phase précoce du sepsis et contribuer au

développement d’une hypoxie tissulaire. Dans un modelé de péritonite fécale induite chez le

mouton, très proche de la situation clinique, j’ai pu montrer que il existe une détérioration

progressive de la microcirculation cérébrale chez les animaux septiques (n-=10) quand

comparés aux animaux témoins (n=5) qui commence déjà a la sixième heure de l’induction du

sepsis and devient significative après 12 heures. De plus, les changement de la

microcirculation cérébrale n’étaient pas corrélés à ceux de la pression artérielle, du débit

cardiaque ou des taux de lactate, ce qui suggère que ces anomalies peuvent se produire aussi

en conditions de stabilité hémodynamique. Dans une deuxième étude comprenant 10 animaux

septique et 5 témoins, j’ai observé que les anomalies microcirculatoires étaient associées à

une diminution de l’oxygénation cérébrale tissulaire. Toutefois, les anomalies du métabolisme

cérébral compatible avec une hypoxie tissulaire (des valeurs élevées du rapport

lactate/pyruvate, LPR) se développaient que dans la phase du choc septique, indiquant que

l’hypotension artérielle est le facteur déterminant pour ces anomalies métaboliques au cours

du sepsis. Cependant, dans une troisième étude sur 8 animaux en sepsis, j’ai démontré que la

correction de l’hypotension par administration de vasopresseurs, même si elle était associée à

une augmentation de l’oxygénation cérébral et une diminution du LPR, n’améliorait pas de

façon significative la microcirculation cérébrale et s’accompagnait par une augmentation des

taux de glutamate et glycérol, ce qui plaidait plutôt pour un excitoxicité persistante et une

progression des lésions cellulaires. Toutes ces anomalies de microcirculation, oxygénation et

métabolisme cérébral pourraient contribuer à la pathogenèse de l’SAE.

INTRODUCTION

Brain dysfunction during sepsis

Septic shock and related multi-organ failure (MOF) remain a major cause of morbidity and mortality in intensive care units (ICUs) worldwide [1]. The infectious stimuli, associated with a widespread reaction characterized by the release of numerous circulating pro- inflammatory molecules, can potentially impair the function of several organs [2]. Brain dysfunction occurs early during sepsis and is commonly characterized by the development of an altered mental state; however, cerebral abnormalities have been also described in late course of sepsis, often accompanied by MOF, hypotension and other systemic events [3, 4].

Hippocrates first reported the association between infections and cerebral dysfunction more than 2500 years ago [5], and sir William Osler also described, later on, the occurrence of

“delirium” in patients with ongoing sepsis [6]. Nevertheless, concomitant hepatic or renal failure, electrolyte and metabolic disturbances, altered glucose homeostasis, hypotension, hypoxemia, hypothermia or neurological side-effects of different pharmacological agents may concomitantly occur in septic patients, complicating the differentiation between sepsis-related brain dysfunction and encephalopathy from other causes [7].

Sepsis-associated encephalopathy (SAE) can be defined as a diffuse or multifocal brain

dysfunction associated with an infectious illness (a) without clinical and laboratory evidence

of intracranial infection and/or (b) without conditions unrelated to the infectious process that

would significantly alter brain function [8]. The diagnosis should therefore exclude structural

brain lesions or primary pathologies of the central nervous system (i.e. meningitis or stroke),

other neurological diseases (i.e. epilepsy), a toxic-metabolic cause, fat embolism syndrome

and anoxic brain injury [9]. Moreover, acute inflammatory encephalopathies, such as acute

disseminated encephalomyelitis or acute hemorrhagic leucoencephalopathy, should be also

The occurrence of SAE is variable but is one of the most common forms of encephalopathy encountered in critically ill patients [11-16]. In a large cohort of mechanically ventilated patients, mostly septic or with acute pneumonia, altered mental state was observed in 66% of them at some point during the ICU stay [11]. In a prospective study of 1758 patients admitted to a medical ICU, 217 of them experienced a neurological complication;

encephalopathy was present in one third of patients, and SAE was the most frequent etiology [12]. In a series on 69 non-sedated patients with bacteraemia, 70% of them had clinical signs of brain dysfunction and half of them showed severe encephalopathy [9]. In two large cohort studies of septic patients, acute mental state abnormalities were described in more than 20%

of them [13]. The heterogeneity of SAE rate among these studies may probably be due to the different definitions of sepsis and SAE used. The most common manifestation of SAE is an alteration of mental state, ranging, from mild disorientation or lethargy to stupor and coma [17]. Most patients have fluctuant changes in mental status, especially in the early course of the infectious process where SAE may often be the first manifestation of sepsis [18-22].

There are no diagnostic tests with high specificity for SAE, so that the clinical,

electrophysiological (EEG; Somato-Sensory Evoked Potentials, SSEPs), biochemical

(neuron-specific enolase, NSE; S-100B protein) or imaging (Magnetic Resonance Imaging,

MRI) criteria may be useful. Young et al. demonstrated that EEG is particularly sensitive in

the diagnosis of SAE, with mild diffuse slow waves even in patients without clinical

abnormalities [15]. Also, EEG could be useful in the differential diagnosis of coma in

critically ill septic patients, especially to detect unexpected clinically silent non-convulsive

status epilepticus, which occur in up to 20% of comatose septic patients [23, 24]. In one study

using SSEPs, an increase of peak latencies in cortical and sub-cortical component of dorsal

column-medial lemniscus pathway (84% and 34% of all cases) was recorded. The impairment

of these pathways was associated with severity of illness and was correlated with the degree

of brain dysfunction [25]. Finally, serum levels of NSE and S-100B protein have been correlated with poor outcome in septic shock patients [26]. Magnetic resonance imaging showed variable degrees of vasogenic edema, related to blood-brain barrier (BBB) breakdown, or ischemic lesions surrounding the Virchow-Robin spaces in septic brain [27].

Septic encephalopathy is not only a confusion state but represents a form of severe organ dysfunction and may contribute to poor outcome in septic patients [3, 13, 28]. The greater the severity of SAE, as assessed by the Glasgow Coma Score (GCS), the higher the risk of death, with a mortality rate of 16% when GCS was 15, 20% when GCS was 13-14, 50% when GCS was 9-12, up to 63% when GCS was below 9) [3]. In another study, septic patients with altered mental status had a higher mortality than those without it [13]. In a large cohort of mechanically ventilated patients, the majority of them admitted for sepsis, Ely et al. showed that ICU-related delirium was independently associated with a longer hospital stay, a worse cognitive recovery and a higher mortality [11]. It appears that SAE may also actively contribute to the development of long-term cognitive dysfunction after critical illness [29].

Unfortunately, there is no specific treatment of SAE and outcome depends on the appropriate

treatment of the underlying disease. Sadly enough, the use of activated protein C appeared to

be beneficial in SAE, but is no longer available [30]. Other drugs, including dopexamine,

inhibitors of the inducible nitric oxide synthase (iNOS), corticosteroids, magnesium and

antioxidants, have shown promising results in experimental sepsis, however no clinical data

on the efficacy of these interventions are currently available [31-34]. Given the absence of

specific therapeutic interventions, treatment of SAE is mainly supportive, aimed to limit

SAE-related secondary cerebral damage.

Pathogenesis of SAE: a short overview

SAE is still poorly understood. The pathophysiology is likely to be multifactorial (Figure 1) [35]. The concept of altered brain function related to the presence in the blood, and possibly in the brain, of micro-organisms and/or their toxins is considered obsolete as altered CNS function is observed also in patients without bacterial bloodstream invasion [36].

Disseminated cerebral micro-abscesses have been suggested as a cause of SAE [37], although they have not been reported in recent post-mortem studies [38]. The specific role of specific bacterial products like endotoxin is unlikely, as the incidence of SAE is similar in Gram- positive and Gram-negative bacteremia, as well as fungemia and even sepsis with unidentified pathogens [13]. Encephalopathy occurs also in non-infectious conditions, such as pancreatitis or trauma, suggesting that it would be mostly related to the systemic inflammatory response [39, 40].

The role of inflammation on brain dysfunction has been widely investigated in several in

vitro or experimental models of sepsis; although some of the reported findings have also been

confirmed in the human setting, the pathogenesis of brain dysfunction mostly relies on

laboratory data. Circulating pro-inflammatory mediators can promote vascular permeability of

the cerebral endothelium, which separates the blood from the brain parenchyma (so called

BBB), thus permitting to several substances to exert toxic effects on neuronal cells. Tumor

necrosis factor-alpha (TNF-) selectively altered BBB permeability in brain microvessel

endothelial cells without disrupting tight junctions [41]. Endothelial cells treated with

interferon-gamma (IFN-) also exhibited significant morphological changes with increased

permeability [42]. Inflammatory mediators have also direct effects on brain functions; Tureen

et al. showed that the subarachnoid injection of TNF- altered cerebral metabolism and

increased cerebrospinal fluid lactate, by activating the production of NO [43]. Also, TNF-

astrocytosis through activation of its receptor [44]. The intra-cerebral administration of interleukin-1 and IFN- in animals induced the same slow-waves EEG patterns than those observed in patients with SAE [45, 46].

More importantly, sepsis-induced systemic inflammation can directly affect brain homeostasis by triggering the two important pathways that are implicated in the response to stress and in the immune system modulation: the circumventricular organs (CVOs), which lack a BBB and have a direct communication with circulating mediators of sepsis and the vagus nerve, which is triggered by visceral inflammation [7]. Once systemic and/or visceral inflammation are detected by these pathways, an abnormal activation of brain signaling spread to behavioral, neuroendocrine and neurovegetative structures, affecting directly microglial and neuronal cells functions and modulating neurosecretion and neurotrasmission [47, 48]. Inflammation rapidly alters the function of hypothalamus, which regulates the endogenous production of corticosteroids and can modulate the inflammatory response [49]

and may significantly impair the reticular activating system functions, which control consciousness and attention [50]. The cholinergic and serotoninergic releases are altered, while an increase in the -aminobutyric acid receptor density is observed in the forebrain of septic rats [51, 52]. Brain tissular norepinephrine and epinephrine concentrations are decreased in the forebrain and brain stem of septic animals together with the down-regulation -adrenergic system [53]. Also, a significant and early increase of plasma aromatic aminoacids during abdominal murine sepsis was found to be associated with the development of encephalopathy [54]. The activation of CVO also promotes the entry of leucocytes into the CNS, with further increase of local inflammation [55].

Several other mechanisms may be involved in the development of SAE. Sepsis produces

peroxidation in the cerebral vessels and the surrounding parenchyma, which eventually cause structural membrane damage and promote inflammation [56]. Cerebral oxidative stress can also be triggered by the decrease of heat-shock factors or by hyperglycemia [57, 58]. Bacterial endotoxin and inflammatory cytokines increase cerebral tissue levels of glutamate via the up- regulation of iNOS; glutamate can trigger brain cells damage by allowing high intracellular levels of calcium and the activation of several enzymes that alter cell structures, such as components of the cytoskeleton, membrane, and DNA [59]. Within the brain parenchyma, the activation of astrocytes via the toll-like receptors by the mediators of inflammation [60] can further increase the vulnerability of neurons to glutamate and free radical-mediated injuries [56, 61, 62]. Moreover, sepsis also alters the synthesis of ascorbate, which may provide antioxidant and protective effects on neurons in response to glutamate and ROS release [63].

The complement system, which normally contributes to eliminate bacteria, may be

excessively activated during sepsis and have deleterious effect through the activation of glial cells, secretion of pro-inflammatory cytokines and generation of other toxic products on brain function [64]. Edema formation could be further promoted by the activation of aquaporin channels, which regulate the presence of water in the brain tissue [65]. Alterations in the glucose uptake and metabolism or regional deregulation of intracellular calcium homeostasis have also been suggested to contribute to the pathogenesis of SAE [50, 66]. All these

pathways cause mitochondrial dysfunction by inhibiting the mitochondrial electron transport chain and uncoupling oxidative phosphorylation, which ultimately leads to bioenergetic failure and apoptosis of brain cells [38, 58].

The complex signaling network that is implicated in these phenomena include the

production of NO, through the activation of the iNOS [67, 68], the release of pro-

inflammatory cytokines and their receptors from neurons, astrocytes and microglial cells [69,

70], as well as the production of prostaglandins, which are key mediators in the brain

response to inflammatory stimuli [71]. Finally, a number of other mediators are involved in

the cerebral brain response to systemic inflammation, such as chemokines, macrophage

migrating inhibitory factor, angiotensin, endothelin, platelet activating factor, superoxide

radicals and carbon monoxide [72].

Figure 1: Schematic representation of mechanisms involved in the pathogenesis of sepsis- associated encephalopathy. Pro-inflammatory cytokines are released during sepsis. They can either activate vagal fibers or enter the brain causing neurologic dysfunction (see text for details).

5-HT, serotonin (5-hydroxytryptamine); Ach, acetylcholine; BBB, blood-brain barrier; CNS, central nervous system; CO, carbon monoxide; CVOs = circumventricular organs; ECs, endothelial cells;

GABA, gamma-aminobutyric acid; GLT, glutamate; HPA, hypothalamic-pituitary-adrenocortical;

HSF, heat-shock factors; IL, interleukin; LPS, lipopolysaccharide; NF-kβ, nuclear factor kappa B; NO,

nitric oxide; NTS, nucleus tractus solitarius; PAF, platelet activating factor; PG, prostaglandin; PVN,

paraventricular nuclei; RAS, reticular activation system; ROS, reactive oxygen species; TNF-α, tumor

necrosis factor-alpha. Adapted from [73].

Alterations in brain perfusion during sepsis

A decrease in brain perfusion is a major determinant of SAE [74]. Alterations of systemic blood flow, associated with tissue hypo-perfusion and poor oxygen distribution, are a key feature of sepsis [2]; indeed, several studies have evaluated the link between cerebral perfusion abnormalities and brain dysfunction in sepsis. As brain perfusion is primarily dependent on mean arterial pressure (MAP), which is generally reduced in severe sepsis and septic shock, it would be important to monitor the adequacy of cerebral blood flow (CBF) and oxygenation during sepsis as well as to evaluate the integrity of flow regulation, in order to adapt blood pressure levels and avoid the development of secondary brain ischemic events in such patients [38].

Cerebral blood flow: normal ranges, physiology and regulation.

In the adult, CBF is typically 750 mL/min (or 50-55 mL/100 g of brain tissue/min) and

represents roughly 15% of the cardiac output. It ranges from 20 mL/100g/min in the white

matter to more than 70 mL/100g/min in the grey matter and is tightly regulated to meet

cerebral metabolic demands [75]. CBF regulation is extremely complex and is determined by

a number of factors, such as viscosity of blood, the diameter of cerebral blood vessels and the

net pressure of the blood flow into the brain, known as cerebral perfusion pressure (CPP),

which is calculated as the difference between MAP and intracranial pressure (ICP) or the

central venous pressure (CVP), whichever is the higher [76]. Cerebral vessels are able to

maintain CBF constant through a wide range of MAP (from 50 to 150 mmHg) by altering

their diameters in a process called “cerebral autoregulation” (CA); in normal conditions, a

raise of MAP induce vasoconstriction whereas a reduction of MAP produces vasodilatation

and potentially damage brain tissue. On the other hand, low CBF (<18-20 mL/100g/min) induces brain ischemia and tissue death when CBF falls below 8-10 mL/100g/min. Systemic administration of vasopressors would only minimally affect CBF, provided the BBB is not altered [78]. Pressure autoregulation is thought to be controlled by the baroceptor reflexes and both the upper and the lower limits of CA can be affected by many factors, including

sympathetic nerve activity, arterial carbon dioxide tension (PaCO

) and pharmacologic agents [79-82].

The control of CBF is dependent not only on MAP but also on four other major determinants, including chemical, metabolic, neural and myogenic factors (Figure 3) [77].

Although this division may be somewhat artificial and these control mechanisms probably operate in concert, it is useful to consider each separately. CBF is extremely sensitive to changes in arterial PaCO

(chemical), displaying marked increases during moderate hypercapnia and reduction during hypocapnia. Changes in CBF in response to changes in PaCO

are referred to as cerebral CO

-reactivity (COR).

The magnitude of cerebral circulatory response is approximately 5% change in CBF for each 1 mmHg change in PaCO

, within a range of 25 to 60 mmHg [83]. In contrast, acidosis and alkalosis, for blood pH values ranging from 6.7 to 7.6, have little effect upon the CBF [84]. Changes in PaCO

are detected by carotid artery chemoreceptors and this regulatory mechanism can be altered by a lesion of the tegmental reticular formation [77]. The effect of PaCO

seems to be related to changes in local brain perivascular pH rather than direct carbon dioxide action on smooth vascular cells, but could also be mediated by K

-dependent

channels or adenosine, while vasodilatation induced by hypercapnia can be abolished by prostaglandin synthesis inhibition [85-89]. Astrocytes contribute in maintaining the

homeostasis of the extracellular compartment in CNS and may profoundly influence central

CO chemoreactivity and respiratory control [90]. Importantly, cerebrovascular dilatation

induced by hypercapnia is markedly attenuated by moderate hypotension, which contributes to exhaust the capacity of cerebral vessels to further dilate [83]. On the opposite, with marked hypercapnia, the capacity to maintain a constant CBF during hypotension is lost, and CBF will decline as CPP declines [91]. Finally, CBF is less sensitive to changes in arterial oxygen tension (PaO

) over the normal physiological ranges. If PaO

falls below 50 mmHg, CBF increases markedly (at 30 mmHg CBF is almost twice as high as at 100 mmHg PaO

).

Conflicting results have been published on the effects of hyperoxia on cerebral perfusion;

moderate hyperoxia induced a mild decrease in CBF in healthy volunteers [92], while ventilation with 100% of oxygen has only minimal effects upon CBF in patients with traumatic brain injury [93].

Local CBF is also tightly coupled to neuronal activity (metabolic) so that CBF may adjust to the level of energy generation in the brain (“neurometabolic coupling”). The regional cerebral metabolic activity is represented by the metabolism of oxygen (CMRO

) and glucose [94]. The precise mechanisms responsible for this coupling remain elusive and alterations in the concentrations of local metabolites or the generation of several vasoactive chemical substances (i.e., H and K ions, adenosine, vasoactive intestinal peptide-VIP or products of arachidonic acid) have been proposed as pivotal mediators [95, 96]. Astrocytic processes extensively unsheathe cerebral arterioles and, through the link between neuronal synapses and the cerebral vasculature, are in a strategic position to convey cellular signals to the blood vessels [97]. The brain metabolism can also significantly affect the magnitude of CBF

modification to changes of PaCO

; indeed, factors that reduce neurons activity (e.g., sedation reduces oxygen and glucose consumption) reduce the cerebral circulation response to

hypercapnia [98], while mild hypercapnia itself may cause a suppression of cerebral

is that part of cerebral metabolism would use lactate generated by the metabolism of glucose in the astrocytes, through the activation of glycolytic pathways [100]; lactate itself has some vasodilating effect on cerebral arterioles and could contribute to modulated brain perfusion in response to cerebral cells demand [101, 102].

The third factor implicated in CBF regulation is the perivascular innervation (neuro- vascular reactivity) (Figure 3), which depends not only on the extrinsic nerve supply from the cranial ganglia to the cerebral vasculature, but also on the intra-cerebral neurons linked to these vessels. Most of the extrinsic neuron fibers are sympathetic and can be summarized in three different types, each with distinct origins and neurotransmitters. The first consists of sympathetic neurons arising principally from the superior cervical ganglion (producing norepinephrine and neuropeptide-Y, thus vasoconstriction); the second consists of

parasympathetic neurons in the spheno-palatine and otic ganglia (producing acetylcholine and VIP); the third consists of sensory fibers originating in the trigeminal ganglion (producing substance P and calcitonin gene-related peptide, thus vasodilation) [103]. Nevertheless, attempts to manipulate CBF by either cervical sympathectomy or symphathetic stimulation resulted in variables CBF changes in experimental models [104, 105]. Also, the

parasympathetic nerves may have some effect only in pain-mediated cerebral vasodilatation while trigeminal fibers can substantially affect CBF only in particular conditions, such as hypertension and seizures [103]. Thus, CBF appears to be primarily regulated by local metabolism with only minor modulation by extrinsic nerves. Among local neuromodulators, dopaminergic axons innervate the intra-parenchymal microvessels and dopamine may directly regulate cortical blood flow [106].

Bayliss et al. were the first to present the myogenic basis of CA, showing that

modifications of the arteriolar smooth muscle tone are elicited by changes in transmural

pressure [107]. Accordingly, an increase in transmural pressure would stretch the smooth

muscle of the vascular wall, stimulating the reflex contraction of radial fibers, which eventually result in vasoconstriction [108]. A growing body of evidence suggests that endothelium-dependent pathways are the primary mediators of myogenic control of CA [109].

The endothelium appears to act as the transducer of transmural forces that would stimulate the release of vasodilating substances, such as NO and L-arginine [110, 111]. The role of NO has been confirmed by the vasodilatation of large cerebral arteries and pial arterioles in response to the application of acetylcholine in vivo [112]; this process was dependent on the NOS activity and could be blocked by NOS antagonist, such as N-monomethyl-L-arginine (L- NMMA) [113].

Autoregulation can be totally lost during some acute cerebrovascular disease or preserved

but shifted (normally towards the right and higher values, for example in chronic

hypertension) [114-117]. In such abnormal conditions, the upper and lower ranges of

MAP/CPP or PaCO

, among which CBF remain constant, can be much narrower than in

normal conditions. Such ranges may be different among patients and change over time within

the same subject, so that optimal MAP targets to maintain adequate brain perfusion may be

difficult to predict. In sepsis, several mechanisms are implicated in altered brain perfusion and

CBF regulation in septic patients.

Figure 2: Schematic representation of cerebral blood flow (CBF) variations associated with

changes in mean arterial pressure (MAP, red line), arterial carbon dioxide tension (PaCO

,

green dotted line) and arterial oxygen tension (PaO

, bleu dotted line). Adapted from [73].

Figure 3: Schematic representation of multiple mechanisms of cerebrovascular control. The control of cerebral blood flow is dependent on five major determinants: mean arterial

pressure, and chemical, metabolic, neural and myogenic factors (see text for details). Adapted from [73].

L-arg, L-arginine; NO, nitric oxide; PaCO

, arterial carbon dioxide tension; VIP, vasoactive

intestinal peptide.

How to monitor cerebral perfusion and autoregulation in septic patients

Cerebral perfusion has been initially evaluated using the Kety-Schmidt technique, which applies the Fick principle (arterial and bulb jugular venous content at different time-points of a tracer is proportional to the global blood flow) to calculate CBF; different tracers, such as N

O, xenon or argon have been used [118]. Using the same Fick principle, the CMRO

and the cerebral metabolic rate for glucose could be also calculated [119]. Similarly, the indicator- dilution technique can also estimate CBF, through the injection of a dye solution (i.e.

indocyanine green) and arterial and bulb jugular dye concentrations, which are used to build dilution curves and calculate the mean transit time for the indicator [120]. These methods are rather invasive and time-consuming and have also a low temporal resolution.

Neuroimaging can be used to measure global and regional CBF. The CT-perfusion is nowadays routinely used in clinical practice and consists of the sequential scanning of selected brain areas during the injection of a bolus of contrast medium, when it passes

through the cerebral vasculature [121]. Additional imaging techniques include functional MRI and positron emission tomography (PET) scan that can both be used to measure global and regional CBF [122, 123]. Other techniques include single photon emission computed tomography (SPECT), which uses a gamma-emitting tracer (the 99-Technetium), and the Xenon-enhanced CT scan, which evaluates the brain distribution of Xenon, after the

administration through the ventilator of a mixture of air and 28% Xenon [124]. Neuroimaging techniques can reliably measure CBF but cannot be used continuously at the bedside and patients cannot be exposed to high doses of contrast media and/or irradiation; likewise it would be unsafe to transport them to the Radiology department. In this setting, trans-cranial Doppler (TCD), although it does not directly measures the CBF, is a suitable non-invasive tool to assess cerebro-vascular reactivity to blood pressure and PaCO

and to assess CA [125].

Other less studied non-invasive techniques include near infrared spectroscopy (NIRS) [126].

There is no general consensus on which is the best method to monitor CA. Testing CA requires to apply a hemodynamic stimulation, such as an increase of MAP through the administration of vasoactive agents, manipulating the ventilator to induce PaCO

changes, the modification of venous return (i.e., through thigh cuff release, application of negative body pressure, tilting test) or the compression of the carotid artery [127]. Thereafter, dynamic changes of CBF are recorded to quantify the reactivity of autoregulatory forces. This approach is limited because it only allows assessment of CA to precise time-points, i.e. during the different manipulations. Another option to assess CA, without the potentially harmful effects of hemodynamic manipulations, is to analyze the continuous dynamic trends of MAP and CBF over time. In this setting, the most commonly used tool to estimate CA is TCD, although some data suggest that cerebral NIRS could also be a valuable method [126]. Brain autoregulation could be continuously assessed by calculating the moving correlation coefficient between MAP and middle cerebral artery velocities (VMCA) (so called, Mx index) [128], or between MAP and cerebral oxygenation estimated by NIRS (Tox index) [126]. Briefly, values of MAP and VMCA are calculated every 10 seconds by bedside softwares and Mx/Tox indexes are obtained as the moving linear correlation coefficient over the last 30 consecutive values. A positive correlation coefficient indicates a close linear relationship between pressure (MAP) and flow (VMCA), thereby suggesting pressure dependency of CBF and impaired CA, while a coefficient close to zero or negative (<0.3) indicates intact CA.

Different stimuli have also been used to test the cerebro-vascular reactivity to CO

, such

as altering PaCO

by changing respiratory rate or using a breathing hold test. Recently, the

intravenous injection of acetazolamide, the reversible inhibitor of the enzyme carbonic

vasodilatation of cerebral arterioles and allows testing cerebro-vascular reactivity to CO

also in septic patients [129].

Brain Perfusion in Experimental Sepsis

Several experimental papers have investigated cerebral perfusion during sepsis. In one study on dogs, CBF showed a 30% decrease within 15 min after endotoxin administration, while the arterial blood pressure was still not markedly changed [130]. In a second canine study, CBF decreased immediately after the administration of endotoxin and consistently remained below control values [131]. Cerebrovascular resistances (CVR) initially decreased, then progressively increased to levels significantly higher than normal and were associated with the lowest CBF levels in the later stages of shock. Nevertheless, other studies also demonstrated that CBF was unchanged during sepsis induced in rats and sheep, while others reported an increase in CBF [132-135]. These seemingly controversial findings may be related to different models (endotoxin vs. bacteria) and species used.

Cerebrovascular reactivity to pressure changes was well maintained in several experimental model of sepsis [139, 136, 137], but in others reactivity to PaCO

changes was altered [138, 139]. In one study, Hinkelbein et al. [140] found no significant difference in the global CBF between non-septic and septic animals, despite the presence of significant

hypocapnia in the sepsis group. Although one possible explanation to these findings is that the cerebrovascular tone becomes unresponsive to carbon dioxide stimuli, authors also suggested that, during the hyperdynamic phase of sepsis, brain hyperemia might develop, which

counteracts hypocapnia-mediated reduction of CBF. The role of PaCO

is also essential in

maintaining normal CA during sepsis; in one study, septic normocapnic animals showed

higher increase in CMRO

than hypocapnic animals, suggesting loss of CA and uncoupling

between CBF and cerebral metabolism during sepsis at normal or high PaCO levels [130].

Local inflammation is another important factor influencing brain autoregulation in experimental models. In one study, CBF was increased with preserved autoregulation in rats with pneumococcal sepsis, even if there was a right shift of the lower threshold of MAP at which CBF was kept constant [141]. Importantly, if these animals had also a direct brain injury, represented by concomitant bacterial meningitis, CA was completely impaired, suggesting that pneumococcal bacteremia itself can trigger only cerebral vasodilatation but does not affect CA in the absence of direct brain inflammation. The importance of

inflammation on brain autoregulation was also underlined in a paper by Rosengarten et al., in which cerebral hyperemia induced by transient carotid compression in septic rats was

significantly impaired in those animals receiving high dose endotoxin and having lower MAP [142].

Although some controversy still exists, some studies suggest that sepsis can

significantly impair brain perfusion and CBF regulation. Whether modulating brain perfusion with hemodynamic augmentation and the use of vasopressors may improve brain perfusion after sepsis is still unclear. In one study on a model of sepsis induced by continuous infusion of Pseudomonas aeruginosa [136], CBF remained stable during the hypotension phase even if a redistribution of cardiac output in other organs than the brain was observed. Interestingly, when norepinephrine was used to restore normal MAP, cerebral perfusion was unaffected, and the same was observed after the administration of L-NMMA, which inhibits the NO production from iNOS. As Meyer and Lingnau [132, 133] observed a significant decrease in CBF after NOS inhibition using N-nitro-L-Arginine-methylester (L-NAME), it is still

possible that more selective iNOS inhibitors like L-NMMA, as compared to L-NAME, would

leave the constitutive NOS untouched and thus prevent excessive vasoconstriction.

in parallel with decreased EEG activity and was associated with reduced glucose uptake, measured by PET-scan, and increased inflammation [143]. These data suggested that an early drop of CBF could be related to regional changes in neuronal activity and energy demand and may be independent from MAP changes in septic animals. Although different areas of the brain showed significant differences in cerebral metabolic changes during sepsis (i.e. an increase of 27 to 33 % in the septal nucleus and raphe nucleus and a decrease of 14 to 27 % in the auditory cortex, lateral geniculate, superior colliculus, hippocampus, parietal cortex, and locus coeruleus) [50], the influence of cerebral metabolic changes on brain perfusion during a septic process may add a new key of interpretation.

Brain Perfusion in Human Sepsis

The concept of reduced cerebral perfusion as a major determinant in SAE development was supported by a retrospective analysis of patients developing sepsis after surgery, which showed that hypotension was the only predictor of delirium in [144]. Also, in an autopsy study analyzing the brain of patients who died from sepsis, multiple ischemic lesions could be identified in different areas of the brain and were attributed to hypotensive events, which may occur in the presence of preexisting cerebrovascular disease as well as in case of impairment of CBF autoregulation [38].

Several studies on healthy volunteers or septic patients have been conducted to understand the changes of CBF during a severe infectious process, as well as brain autoregulatory

capacity and metabolism (Table 1). In experimental sepsis using endotoxin injection in

healthy volunteers, Moller et al. [145] reported a reduction of CBF immediately after sepsis

induction, which was attributed to hypocapnia occurring because of general symptoms of

malaise. Cerebral oxidative metabolism was unchanged and no detectable cerebral flux of

cytokines was observed, despite high systemic concentrations of these molecules. In another

experiment on healthy volunteers, when sepsis was induced using an intravenous bolus of Escherichia coli endotoxin, CBF and CMRO

measured few hours later were preserved, despite a drop in systemic vascular resistance [146]. In septic patients, Bowton et al. [74]

showed reduced CBF measured by Xenon clearance technique; low brain perfusion occurred

independently from MAP values. In another small cohort (n=6) of septic patients with MOF,

CBF was within the normal range but significantly lower than awake controls; CMRO

was

also significantly reduced and these alterations were associated with a significant slowing of

EEG [130]. A recent study in mechanically ventilated septic patients with delirium [120],

Thees et al. also measured CBF by thermo-dilution technique and reported to be within

normal ranges (64 ± 29 mL/100g/min). Also, cerebral oxygenation was within the normal

limits in all patients. In 20 patients with septic shock, Straver et al. [147] showed a close

relationship between cerebral and systemic hemodynamics. This study showed that mean

VMCA measured by TCD significantly increased if the systemic vascular resistances (SVR)

decreased, suggesting that concomitant cerebral and systemic vasodilatation induced an

increased CBF. Also TCD abnormalities were strongly related to disease severity and

outcome.

Table 1. Clinical studies on cerebral hemodynamics in sepsis.

Pts Timing Age APACHE II Severity CBF CBF measure CA COR CMRO

Reference

10 < 24 hrs 60 23 S,SS,Sh - TCD Normal Normal - [148]

10 > 48 hrs 50 31 SS Normal Dilution - Normal Normal [120]

16 > 48 hrs 75 23 S,SS,Sh - TCD Altered * - Normal [125]

8 NA 60 33 SS Increased NIRS - Reduced - [149]

9 < 72 hrs 52 NA S Reduced Xenon-CT - Normal Reduced [74]

6 Day 2-10 48 NA SS Reduced KST - - Reduced [156]

12 > 24 hrs 68 18 SS,Sh - TCD - Variable - [150]

15 12-48 hrs 54 22 Sh Reduced C-Doppler Altered [152]

20 NA 58 21 Sh Increased TCD - - - [147]

21 < 72 hrs 65 20 Sh - TCD Altered - - [157]

20 < 72 hrs 65 37 Sh - TCD - Reduced - [151]

23 NA 68 22 SS, Sh - TCD/NIRS Impaired ** - - [126]

14 NA NA NA S - TCD - Reduced - [129]

30 <24 hrs 64 32 SS - TCD Impaired - - [153]

10 - 30 - HV Normal KST - - Unchanged [145]

8 - 25 - HV Reduced KST - Normal Unchanged [146]

10 - 28 - HV - TCD Enhanced *** - - [154]

10 - 32 - HV - TCD Enhanced *** - - [155]

* Only in patients with septic shock; ** In case of lower MAP and higher values of PaCO

; *** Associated with reduced PaCO

Pts = patients; APACHE = Acute Physiology And Chronic Health Evaluation; CBF = cerebral blood flow; CA = cerebral autoregulation; COR = carbon dioxide reactivity; CMRO

= cerebral oxygen consumption; hrs = hours; S = sepsis; SS = severe sepsis; Sh = septic shock; TCD =

transcranial Doppler; NIRS = near infrared spectroscopy; NA = not available; CT = computed tomography; KST= Keyt-Schmidt technique; NA

= not available; C-Doppler = Carotid Doppler; HV = healthy volunteers.

In mechanically ventilated septic patients [148], authors showed that there was only a moderate reduction of cerebrovascular reactivity to PaCO

when compared to values in awake controls, but consistent with values obtained during sedation and anaesthesia. Bowton et al.

[74] also reported a normal response to CO

changes in septic patients. More recently, in ongoing sepsis for more than 48 hours, Thees et al. showed CO

reactivity was to be intact in 10 septic patients treated by mechanical ventilation [120]. Hypocapnia was associated with a reduction of CBF without affecting CMRO

, which was already within low ranges at baseline.

None of the characteristics of patient population, including Acute Physiology and Chronic Health Evaluation (APACHE) II score, temperature, MAP, CI had any significant association with CO

reactivity. In contrast with these findings, Terborg et al. [149] observed, in a small cohort of brain injured patients developing severe sepsis and septic shock, that

cerebrovascular reactivity to PaCO

was severely impaired, independently of changes in MAP. Nevertheless, the pre-existing neurological illness of these patients may have affected the results. In another study by Bowie et al., on 12 patients, 3 of them had normal

cerebrovascular reactivity, 7 had impaired vasomotor tone and two had a greater response to CO

changes [150]. Recently, Szatmári et al. found septic patients with SAE had altered cerebro-vascular reactivity as assessed by an acetazolamide injection test [129]. Finally, the cerebrovascular reactivity to PaCO

in patients with septic shock was lower than in control patients, with a significantly lower reactivity in patients receiving dexmedetomidine than propofol [151].

Concerning CA, in a first study on 10 patients with sepsis and altered mental status, Matta et al. [148] showed that CA was intact within the first 24 hours after ICU admission, when phenylephrine infusion was used to increase MAP within the normal pleateau of

autoregulation. In contrast, Smith and colleagues [152] reported loss of CA in 15 patients with

correlated with changes in cardiac index (CI). In a more recent study on 16 patients with

various degrees of sepsis severity, Pfister and colleagues [125] found CA was altered in

patients with sepsis-associated delirium but not in patients without neurological symptoms,

despite similar systemic hemodynamics and baseline MAP. These alterations were associated

with higher levels of inflammatory biomarkers (CRP and IL-6), of brain injury biomarkers

(s100B) and poorer outcome, but not to the severity of disease, assessed by the APACHE II

score, and to catecholamine requirements. Finally, Schramm et al. [153] showed that CA was

impaired in the majority of patients with severe sepsis and these alterations persisted in more

than 40% of them up to the fourth day after sepsis diagnosis.

Microcirculation

Interestingly, as SAE occurs also in patients without hemodynamic instability, one may argue that alterations in oxygen supply to the cerebral tissue during life-threatening infections could occur independently from low cardiac output and hypotension. Thus, it would be the microcirculation, which largely regulates tissue perfusion and blood-cellular nutrients exchanges, the key component in the development of tissue hypoxia in septic patients.

What is microcirculation?

Microcirculation deals with the circulation of blood from the heart to arterioles

(terminal arteries), capillaries and venules; pre-capillary sphincters control the blood flow

throughout the microcirculation, while fluid and gasses exchanges take between the blood and

the tissues take place at the capillary bed (Figure 4) [158]. Arterioles, which are richly

innervated by sympathetic adrenergic fibers and highly responsive to sympathetic

vasoconstriction via both α

and α

receptors, are surrounded by smooth muscle cells, and its

diameter range between 20 and 100 µm. Arterioles carry the blood to the capillaries, which

are not innervated, have no smooth muscle, and have a diameter around 5-10 µm. Capillaries

have three structural classifications: a) continuous (found in muscle, skin, lung and central

nervous system), with a continuous basement membrane and tight intercellular junctions (i.e.,

the lowest permeability); b) fenestrated (found in exocrine glands, renal glomeruli, intestinal

mucosa), with basal fenestrations resulting in relatively high permeability; c) discontinuous

(found in liver, spleen, bone marrow), with large intercellular gaps resulting in extremely high

permeability [159]. Blood flows out of the capillaries into the venules, which have little

smooth muscle and are 20-200 µm. In addition to these blood vessels, the microcirculation

Microcirculation has then three major sectors: pre-capillary (resistive), capillary (swap) and post-capillary (capacitive). In the pre-capillary sector, arterioles and pre-capillary sphincters participate. Their function is to regulate blood flow and tissue perfusion before blood enters the capillaries and venules by the contraction and relaxation of the smooth muscle found on their walls. As a consequence, microcirculation blood flow remains constant despite of changes in systemic blood pressure. This mechanism is present in all tissues and organs of the human body. The second sector is the capillary sector, which is represented by the capillaries, where substance and gas exchange between blood and interstitial fluid takes place. Finally, the post-capillary sector is represented by the post-capillary venules, which also allows free movement of some substances [160]. Microcirculation also serves to regulate temperature, to activate the coagulation cascade and it is involved in the response to inflammation as well as in the angiogenesis process [158]. Most vessels of the microcirculation are lined by flattened cells, the endothelium, which acts as a continuous, selective, semi-permeable separation between the vessel lumen and the surrounding tissue, controlling the passage of water, ions and small molecules and the transit of white blood cells into and out of the bloodstream. Through the presence of several intercellular tight junctions and desmosomes, it maintains its integrity and limits the penetration of circulating pathogens into peripheral tissues [161].

How is microcirculation regulated?

Tissue perfusion is determined by vascular density (i.e. the diffusive component) and by flow

(i.e. the convective component) [162]. Capillary density increases in response to chronic

hypoxia or during training [163, 164]. In exercise, the maximal oxygen consumption is

proportional to muscle capillary density. However, this adaptive process may take several

weeks to occur; in more acute situations, there is a small reserve for capillary recruitment,

mostly because a few capillaries are shut down at baseline. Compared with baseline, the heterogeneity of the microcirculation increased by close to 10% during hypoxia or hemorrhage [165]. How do capillary flow and density adapt in normal conditions? In healthy conditions, the microcirculation is responsible for fine-tuning of perfusion to meet local oxygen requirements. This is achieved by recruiting and de-recruiting capillaries, shutting down or limiting flow in capillaries that are perfusing areas with low oxygen requirements and increasing flow in areas with high oxygen requirements. This process implies local control of flow, which needs to be driven by backward communication. Indeed, release of vasoactive or hormonal substances can only lead to downstream adaptation, but it is upstream adaptation that is required. Two mechanisms may help with this local communication:

perivascular sympathetic nerves [164], which mostly influence the control of arteriolar tone,

and backward communication along the endothelium, mediated by endothelial cells

themselves. In addition, red blood cells may act as intravascular sensors [166]. The decrease

in oxygen saturation that occurs as a result of oxygen offloading causes the local release of

nitric oxide, leading to capillary dilation at the site where it is needed. What drives blood flow

in the capillaries? According to Poiseuille's law, flow in a capillary is proportional to the

driving pressure (ΔP) and to the fourth power of the capillary radius (r), and inversely

proportional to capillary length (L) and blood viscosity (η) (Figure 4). Because capillary

length and viscosity cannot be actively manipulated, capillary flow can only be adapted by

local dilation and increased driving pressure. Because capillaries are situated downstream of

resistive arterioles, an increase in driving pressure can only be obtained by vasodilation of

resistive arterioles. Hence, in normal conditions, the organism is continuously fine-tuning

microvascular density and flow by subtle dilation/constriction of selected arterioles and

capillaries [167]. Also, arterioles respond to metabolic stimuli that are generated in the

vasodilation. Importantly, it should be remembered that capillary hematocrit is less than systemic hematocrit, due to the necessary presence of a plasma layer at the endothelial surface. Accordingly, hematocrit is proportional to capillary radius, so vasodilation will markedly increase local oxygen delivery as the result of a combined increase in flow and in oxygen content. Finally, it should be noted that adaptation of capillary perfusion at the organ level does not depend on systemic arterial pressure and cardiac output but, of course, will result in increased cardiac output if venous return increases as a result of a major increase in capillary flow, as during exercise or feeding.

The dynamic displacement of materials between the interstitial fluid and the blood is named capillary exchange. These substances pass through capillaries through three different systems or mechanisms: diffusion, bulk flow and transcytosis or vesicular transport [168].

The liquid and solid exchanges that take place in the microvasculature particularly involve capillaries and post-capillary venules. Capillary walls allow the free flow of almost every substance in plasma, except proteins, because of their high molecular weight. Diffusion is the first and most important mechanism that allows the flow of small molecules across capillaries.

The process depends on the difference of gradients between the interstitium and blood, with

molecules moving to low concentrated spaces from high concentrated ones. Glucose, amino

acids, O

and other molecules exit capillaries by diffusion to reach the organism’s tissues,

while CO

and other wastes leave tissues by the same but reverse process [169]. Diffusion

through the capillary walls depends on the permeability of the wall and on the Starling

equation, which describes the roles of hydrostatic and osmotic forces in the movement of

fluid across capillary endothelium. When the flow of substances goes from the bloodstream or

the capillary to the interstitial space or interstitium, the process is called filtration; this kind of

movement is favored by blood hydrostatic pressure and interstitial fluid osmotic pressure. If

the substances move from the interstitial fluid to the blood in capillaries, the process is called