Bench-to-bedside review: Microvascular dysfunction in sepsis –hemodynamics, oxygen transport, and nitric oxide

The microcirculation is a complex and integrated system that supplies and distributes oxygen throughout the tissues. The red blood cell (RBC) facilitates convective oxygen transport via co-operative binding with hemoglobin. In the microcirculation oxygen diffuses from the RBC into neighboring tissues, where it is consumed by mitochondria. Evidence suggests that the RBC acts as deliverer of oxygen and 'sensor' of local oxygen gradients. Within vascular beds RBCs are distributed actively by arteriolar tone and passively by rheologic factors, including vessel geometry and RBC deformability. Microvascular oxygen transport is determined by microvascular geometry, hemodynamics, and RBC hemoglobin oxygen saturation. Sepsis causes abnormal microvascular oxygen transport as significant numbers of capillaries stop flowing and the microcirculation fails to compensate for decreased functional capillary density. The resulting maldistribution of RBC flow results in a mismatch of oxygen delivery with oxygen demand that affects both critical oxygen delivery and oxygen extraction ratio. Nitric oxide (NO) maintains microvascular homeostasis by regulating arteriolar tone, RBC deformability, leukocyte and platelet adhesion to endothelial cells, and blood volume. NO also regulates mitochondrial respiration. During sepsis, NO over-production mediates systemic hypotension and microvascular reactivity, and is seemingly protective of microvascular blood flow.

severity of the immediate arteriolar vasoconstriction response to endotoxin. Moreover, it recounted the subsequent phases of changing microvascular tone and cardiovascular collapse that occur with progression of sepsis.
With the realization that the systemic inflammatory response to bacterial infection (sepsis) or endotoxin (endotoxemia) caused plasma volume depletion came the introduction of fluid resuscitation [2]. With fluid resuscitation, both septic patients and animal models have been found, in general, to exhibit a hyperdynamic state [3][4][5]. This state is characterized by elevated cardiac output, increased oxygen delivery (DO 2 ), decreased systemic vascular resistance (with or without decreases in mean arterial blood pressure [MAP]), and increased tissue oxygen consumption (VO 2 ), but impaired oxygen extraction capacity [6,7] and lactic acidosis. The latter two observations have led to the concept that microvascular injury, impaired microvascular control, and maldistribution of microvascular blood flow induce a tissue oxygen debt [8], and the theory that inefficient matching of microvascular oxygen supply (qO 2 ) to oxygen demand (dO 2 ) impairs oxygen extraction [9,10]. This is manifested by a pathologic oxygen supply dependency, whereby VO 2 is dependent on DO 2 , and a decreased critical oxygen extraction ratio (O 2 ER) [11,12].
Global measures of hemodynamic and oxygen transport parameters, including cardiac output, arterial pressure, vascular resistance, blood gases, oxygen consumption, oxygen extraction and lactate, provide whole body information on the status of the cardiovascular system in the critically ill patient. Although serum lactate has been widely used as an indicator of tissue hypoperfusion, decreased tissue oxygenation, and anaerobic metabolism [13], recent evidence suggests that blood lactate concentration may also reflect altered pyruvate dehydrogenase [14,15] and Na + ,K + -ATPase [16] activity, and increased glycolysis rate [17]. Global oxygen transport parameters, however, fail to measure or assess the status of the microcirculation [18], which is vital to organ function because it is the microcirculation that delivers and distributes oxygen, nutrients, and inflammatory and coagulation factors throughout the tissue and removes metabolic waste products, heat, and carbon dioxide.
To obtain information on the functional state of the microcirculation, tissue oxygenation, bioenergetic, or redox status of an organ, a variety of techniques have been used both experimentally and clinically. Laser Doppler flowmetry provides a relative signal of red blood cell (RBC) flow [19,20] from an unknown tissue volume, whereas intravital videomicroscopy (IVVM) provides real-time images of microvascular geometry and blood flow, from which microvessel diameter [21][22][23][24], functional capillary density [19,[25][26][27], and intercapillary area [28,29] can be determined. Combined with video analysis tools and dual wavelength spectrophotometry, IVVM has allowed the quantification of capillary hemodynamics (RBC velocity, RBC lineal density, and RBC flux/supply rate) [19,30,31] and erythrocyte hemoglobin oxygen saturation (RBC SO 2 ) [30]. Analysis of microvascular blood velocity in the finger nail-bed [32] and, more recently, orthogonal polarization spectral imaging of the sublingual microcirculation [33,34] have brought microvascular imaging to the bedside, but with reduced spatial resolution.
The microcirculation as a functional system for distributing blood flow The cardiovascular system circulates blood throughout the body, but it is the microcirculation, in particular, that actively and passively regulates the distribution of RBCs and plasma throughout individual organs (Fig. 1). Blood flow into an organ is controlled upstream of the capillary networks or vascular beds by the vascular tone of the resistance vessels comprising the arteriolar network. These vessels are surrounded by smooth muscle that either constricts or relaxes in response to the balance between pressor and dilatory stimulation. Locally, vascular tone controls the diameter of the resistance vessels, the blood flow through them, and the pressure drop across the microvascular beds. Globally, vascular tone controls the systemic and pulmonary blood pressures.
Downstream of the arterioles, microvascular RBC flow is passively distributed throughout the capillary networks [51] and other vascular beds such as the liver sinusoids, according to local vessel resistance (diameter and length) and hemorheologic factors (blood viscosity and RBC deformability). RBCs are forced to deform and travel single file, often separated by plasma gaps, as they pass through vessels that are of smaller diameter then their own. This distinctive microvascular flow behavior maximizes the surface area available for gas exchange between the RBC and the local environment. Surprisingly, passive rheologic mechanisms appear to play a greater role than arteriolar heterogeneity in determining capillary heterogeneity and functional capillary density, especially at low flow states [51]. Passive rheologic mechanisms are also responsible for the Fahreaus effect (i.e. the drop in vessel hematocrit along the arteriolar tree to the capillary bed). In the skeletal muscle of septic rats, we have observed that stopped-flow capillaries have lower hematocrit, or lineal density (RBC/mm), than do neighboring flowing capillaries. Neither the implications nor the cause and effect relationship of this phenomenon is clearly understood.
In hyperdynamic sepsis progressive arteriolar constriction in the small intestine occurs at all levels of the microvascular arteriolar network, causing a decrease in blood flow to this organ [23] and thus contributing to impaired villus microcirculation [29,45]. In cremaster muscle [24,41] and the diaphragm [40], however, there is a differential arteriolar response in which arterioles with larger diameter vasoconstrict and those with smaller diameter vasodilate. Similarly, bacteremia causes a differential response in liver microvessels in which two-thirds of the inlet periportal sinusoids and portal venules and outlet distal centrilobular sinusoids and central venules dilate, whereas the remaining microvessels constrict [46]. In addition, Zhao and coworkers [22] reported that sepsis induces a differential sensitivity to acetylcholine in the arteriolar networks of the splanchnic circulation.
The endothelial cell is a critical component of the highly integrated microvascular system and plays an obligatory role as a signal transducer of shear stress [52,53] and vasoactive substances [54], including acetylcholine, catecholamines, prostaglandins, endothelin, bradykinin, thromboxane, adenosine, nitrosothiols, and ATP. Moreover, endothelial cells conduct and integrate local stimulatory signals throughout the microcirculation [55] via cell-cell communication. During endotoxemia endothelial dysfunction disrupts the microvascular communication system [56] and seemingly contributes to abnormal tissue perfusion [57].
The microcirculation is an integrated system designed to ensure that oxygen delivery meets or exceeds cellular oxygen demand throughout the tissue. Within the context of the cardiovascular system, oxygen transport can be considered to be a flow of oxygen from the lungs (high 'PO 2 ) to the tissues Available online http://ccforum.com/content/7/5/359   (low 'PO 2 ), involving both convective and diffusive transport mechanisms (Fig. 2).

Diffusive oxygen transport
Oxygen diffuses over relatively short distances down its partial pressure (PO 2 ) gradient. PO 2 gradients exist along the blood vessels and into the tissue (Fig. 2). In 1919 the physiologist August Krogh and mathematician Karl Erlang developed a mathematical model of oxygen transport based on simple cylindrical geometry and the assumption that each capillary supplied a unique tissue volume [58]. Today, we know that oxygen diffuses from arterioles [59] and capillaries in any direction based solely on the local PO 2 gradient; however, the Krogh model is still instructive, especially under conditions of diffusion limitation. Oxygen diffusion is limited by oxygen solubility (k), oxygen diffusivity (D), and the PO 2 gradient (dPO 2 /dr). The critical oxygen diffusion distance, which is the maximum distance that mitochondria can be away from an oxygen source without impaired function, is determined by these oxygen diffusion parameters and by capillary PO 2 and tissue oxygen consumption. Typical diffusion distances may range from 10 to hundreds of microns. Whether tissue is adequately oxygenated is then ultimately determined by local microvascular oxygen delivery, critical oxygen diffusion distance, and intercapillary distance.
The PO 2 gradient drives the net movement of oxygen from a region of high PO 2 to a region of low PO 2 ; as such, as the PO 2 gradient increases so to will the flux of oxygen, or the amount of O 2 that diffuses out of the vessel per unit surface area per unit time. Oxygen flux is described mathematically by Fick's first law of diffusion: The negative sign in the expression converts the negative slope of the gradient to a positive value. It is diffusion that facilitates the movement of oxygen in the lung alveolus to RBCs in capillaries surrounding the alveolar wall and from the RBCs in the microvasculature into the tissue.

Convective oxygen transport and the erythrocyte
Because oxygen has a low solubility in plasma, it is RBC flow specifically, and not 'blood' (plasma and RBCs) in general, that determines oxygen delivery; accordingly, the oxygen carrying capacity of the erythrocyte, facilitated by hemoglobin (Hb), is essential to the convective, or bulk, transport of oxygen over large distances by the blood. Within the RBC, as it circulates between the lungs and the tissue, oxygen binds co-operatively with Hb in a manner that alters its tetrameric conformation, switching it from a relaxed, high oxygen affinity structure to a tense, low oxygen affinity structure as Hb alternately acquires oxygen and releases it to the local environment. The physiologic significance of the Hb-oxygen interaction is reflected in the sigmoidal nature of the oxygen dissociation curve. The affinity of Hb for oxygen can be affected by temperature, pH, the Bohr effect, and NO [60,61], in which S-nitrosohemoglobin increases the affinity of Hb for oxygen [62].
In vitro, in vivo, and theoretical evidence [63][64][65][66] suggests that the RBC releases vasoactive ATP and nitrosothiols in response to increased PO 2 gradients and mechanical deformation [67]. In theory, the erythrocyte 'senses' the local PO 2 gradient through a conformational change in the Hb molecule and signals the microvasculature to vasodilate. Thus, the RBC is integrated into the microvascular system as both a deliverer and sensor of oxygen (Fig. 3).
During sepsis the mechanical properties of the RBC, including membrane deformability and shape recovery, are progressively altered such that the RBC becomes less deformable [26,68,69]. Condon and coworkers [70] reported that 'elderly' RBCs, comprising 20% of the circulating erythrocytes, were most susceptible to decreased deformability and that Hb content decreased in a large fraction of the RBCs during sepsis. The accumulation of rigid erythrocytes suggested to the authors that the erythrophagocytic capacity of mononuclear phagocytes had been overwhelmed during sepsis [70]. Although a cause and effect relationship has not been demonstrated, a change in blood rheology does appear to be a factor in the loss of functional capillary density [26] and peripheral shunting [68] during sepsis. What effect sepsis has on RBC oxygen sensing and signaling mechanisms is unknown.
In larger blood vessels, convective oxygen transport is calculated as the product of blood flow (Q [ml/s]), Hb concentration (g/dl), Hb SO 2 (%), and the oxygen-binding capacity of Hb (C [oxygen/Hb]).
Convective oxygen transport = Q × Hb × SO 2 × C In the microcirculation, capillary hemodynamics can be quantified as either RBC flux [31] or as an RBC supply rate (SR) [30]. RBC flux or SR (RBC/s) account for both RBC velocity (V [µm/s]) and capillary hematocrit or RBC lineal density (LD [RBC/mm]). Since RBC flux implies movement of RBC per unit area, we prefer the term RBC SR.
Similar to convective oxygen transport in larger vessels, oxygen flow in capillaries (qO 2 ) can be calculated from the RBC SR, the RBC SO 2 , and the oxygen carrying capacity of a single RBC (K = 0.0362 ml oxygen/RBC at 100% SO 2 ) [30].
Capillary oxygen transport parameters, capillary O 2 ER (O 2 ERc), and capillary oxygen flux (O 2 flux c ) can be calculated from capillary oxygen flow rates at the capillary entrance (en) and exit (ex) and the capillary surface area, as determined by local capillary dimensions length (L) and diameter (d), respectively.
Local differences in capillary hemodynamics give rise to microvascular flow heterogeneity and subsequently oxygen flow heterogeneity within the organ (Fig. 1). The relationships between microvascular geometry, capillary hemodynamics, functional capillary density, and oxygen transport are of particular importance in the pathophysiology of sepsis because the systemic inflammatory response induces remote microvascular derangements and dysfunction that may contribute to tissue injury and ultimately organ failure in septic patients.
The loss of functional capillary density in skeletal muscle occurs early during the progression of experimental sepsis and is associated with both a loss of RBC deformability and an over-production of NO [26]. Although Laser Doppler flowmetry has detected the presence of dysfunctional microcirculation on the basis of attenuated reactive hyperemia in skeletal muscle [19,20], it is unable to discriminate fundamental capillary stopped-flow or flow heterogeneity induced by sepsis. While the exact mechanism of microvascular stasis is still to be determined, it is clear that sepsis causes local regions of ischemia in the tissue by virtue of capillary stopped-flow.

Figure 3
The red blood cell (RBC) as a deliverer and 'sensor' of oxygen (O 2  swelling, reduced or altered driving pressures in the microcirculation, or the development of putative microthrombi [74]. Neither tissue edema nor leukocyte adhesion in postcapillary venules is the apparent cause, at least in skeletal muscle [25,50]. In liver, increased leukocyte rolling and adhesion is associated with decreased sinusoidal flow velocity [75], raising the possibility that leukocytes are affecting microvascular blood flow in this organ, although the precise relationship between adhesion and flow is unknown. Although it has been proposed that platelet-fibrin clots occlude microvessels, the role that platelets play in capillary stopped-flow is also unknown. Capillary stopped-flow appears to be independent of arterial pressure because both diaphragmatic capillaries and mucosal capillaries were more likely to shut down in septic relative to control animals with a similar degree of hypotension [31,40]. In our 6-hour acute and 24-hour chronic sepsis models [26,30] we also observed reversal of capillary flow in skeletal muscle, indicating that significant fluctuations in pressure gradients have occurred across the vascular bed.

Fluid resuscitation, tissue oxygen tension, and microvascular derangements
Sepsis induces decreases in liver, gut, and skeletal muscle tissue PO 2 [35,36,76,77]. Fluid resuscitation has been demonstrated to apparently rescue skeletal muscle microcirculation in terms of oxygen transport by improving PO 2 [35,36]. However, despite this apparent improvement in microvascular function tissue ATP remained depressed and lactate remained elevated [35] (Fig. 5), suggesting that tissue PO 2 was not a reliable indicator of bioenergetic status during sepsis, or conceivably of anaerobic metabolism. It has been argued that increased tissue PO 2 during sepsis indicates a decrease in the ability of tissue to consume oxygen [35,78] and that this may be due to mitochondrial failure or cytopathic hypoxia [79].
Subsequently, a more detailed evaluation of skeletal muscle oxygen transport at the individual capillary level in a fluid resuscitated, 24-hour normotensive rat model of sepsis  demonstrated that the microcirculation was not rescued by fluid resuscitation [30]. Using in vivo spectrophotometric imaging that simultaneously determined skeletal muscle microvascular geometry, capillary hemodynamics, and erythrocyte Hb SO 2 in normally perfused capillaries (i.e. capillaries with RBC velocity 20-325 µm/s), we found that O 2 ERc actually increased by a factor of three in septic rats, indicating an increase in oxygen flux out of the capillary into the surrounding tissue.
In the same muscle capillary beds we observed a threefold increase in the ratio of fast flow (i.e. RBC velocity > 325 µm/s) to normal flow capillaries, and a fourfold increase in capillary density of stopped-flow. Technical limitations pre-vented the determination of hemodynamic parameters in the fast flow vessels, but their appearance indicated that some local regions of tissue were clearly over-supplied with oxygen whereas other areas -those supplied by capillaries exhibiting increased O 2 ERc -were clearly under-supplied with oxygen. This was evidence of a maldistribution of RBC flow at the capillary level and a mismatching of local oxygen supply with local oxygen demand. These findings also indicated that the septic microcirculation had lost its ability to regulate capillary blood flow because it was unable to redistribute RBCs to regions of low PO 2 and increased oxygen demand. The presence of an increased proportion of fast flow vessels, adjacent to stopped-flow capillaries, suggested that oxygen might have been shunted through the capillary beds via these Available online http://ccforum.com/content/7/5/359

Figure 5
Relationships between global hemodynamics, tissue oxygen tension (PO 2 ), and skeletal muscle lactate and ATP. In a 6-hour rat model of sepsis (cecal ligation and perforation [CLP]), fluid resuscitation was found to prevent decreases in systemic oxygen delivery (DO 2 ) and tissue PO 2 in skeletal muscle (tPO 2 ). This apparent rescue of the muscle microcirculation was belied by elevated lactate and reduced ATP. *P < 0.05, versus control. FR, albumin fluid resuscitation; MAP, mean arterial pressure. By permission from Circ Shock 1988, 26:311-20 [35]. This putative shunting mechanism in skeletal muscle is suspected to be quite different from that proposed to exist in either the gut or the heart. In these organs it has been proposed that fundamental 'microvascular weak units' [38] exist that are bypassed during sepsis or low flow conditions. Although the exact microvascular geometry of such microvascular weak units has not been demonstrated, entire microvascular beds, rather than single capillaries, may be compromised. In the liver, Unger and coworkers [46] reported an intrahepatic redistribution of blood flow in which blood was channeled away from constricted vessels to vasodilated vessels. Taken together, these data demonstrate that sepsis creates microvascular heterogeneities or inefficiencies in the matching of local oxygen delivery to local oxygen demand in these organs, which may contribute to a local tissue oxygen debt and organ dysfunction.

Capillary stopped-flow and increased capillary oxygen extraction
Although increased local oxygen consumption [6,7] could account for increased capillary oxygen extraction in skeletal muscle in 24-hour septic rats [30], a positive correlation between capillary oxygen extraction and the degree of regional capillary stopped-flow (Fig. 6) indicates that, as stopped-flow increases, the remaining functionally normal capillaries offload greater amounts of oxygen to the surrounding tissue. There is no evidence of capillary recruitment. That vessels with high flow were unable to compensate for the loss of perfused capillaries is supported by the correlation between increased oxygen extraction and stopped-flow. The dysfunctional aspect of oxygen transport in the septic microcirculation is realized with the Krogh cylinder model (Fig. 4c). The effective tissue volume supplied by a capillary actually increases as the degree of capillary stopped-flow increases; hence, local capillary oxygen extraction necessarily increases to meet the increased oxygen demand imposed on it by the increased tissue volume surrounding it.
A loss of functional capillary density has been found to increase intercapillary distances in skeletal muscle and intestinal villi [19,31], but whether convective oxygen flow was sufficient and oxygen diffused over these increased distances is unknown. In the ileum of septic rats, oxygen transport is further compromised by increased flow motion with longer periods of capillary stopped-flow [29]. Although Hotchkiss and coworkers [39] did not detect cellular hypoxia using [18F]fluoromisonidazole in either the skeletal muscle, brain, liver, heart, or diaphragm of septic rats, it is possible that a mismatch in tissue qO 2 /dO 2 impaired oxygen extraction in the gut and heart [9,10].
It is important to realize that fluid resuscitation in the animal model did not prevent loss of functional capillary density, or restore microvascular regulation or the qO 2 /dO 2 mismatch in the tissue [30]. A similar impairment in microvascular perfusion was detected in septic patients, despite restoration of intravascular volume [33]. A question that remains unanswered is at what stage of sepsis does the microcirculation actually fail to ensure adequate oxygenation. The report that early goal-directed therapy in septic patients reduces 60 day mortality, and fewer patients succumbed to cardiovascular collapse (in-hospital mortality) [80], raises the possibility that microvascular dysfunction persists throughout the entire septic process. The severity of microcirculatory dysfunction may therefore ultimately determine the severity of organ dysfunction. In light of the above observations, it is important to consider local geometry, hemodynamics, and oxygen transport as well as the average tissue PO 2 when assessing the extent of tissue oxygenation or the functional state of the microcirculation during sepsis.

Nitric oxide properties: synthesis, diffusion, and transport
The gaseous molecule NO is a potent regulator of vascular tone, a cytotoxic agent, a neurotransmitter, an antioxidant (in that it reacts with superoxide anion to form peroxynitrite), and, seemingly, a modulator of overall microvascular integrity, function, and oxygen transport. This review focuses primarily on the latter role of NO.
NO is synthesized by the L-arginine pathway in a variety of cell types including macrophages, neutrophils, platelets,

Figure 6
Local capillary oxygen extraction ratio (O 2 ERc) versus regional capillary stopped-flow (%CDstop). In a 24-hour rat model of sepsis (cecal ligation and perforation [CLP]), capillary oxygen extraction from normally flowing capillaries (capillary velocity between 20 and 325 µm/s) was found to correlate with the degree of regional capillary stopped-flow. As capillary stopped-flow increased, the RBCs offloaded increased amounts of oxygen. (CLP group: y = 0.018x -0.18; r 2 = 0.64; P < 0.05.) By permission from Am J Physiol Heart Circ Physiol 2002, 282:H156-H164 [30]. Under normal conditions, a small amount of NO is produced in endothelial cells by the constitutively expressed NOS enzyme in response to receptor mediated and shear stress vasodilatory stimuli [81,82]. As a gas, NO diffuses down its concentration gradient, from endothelial cell to smooth muscle cell, where it opposes sympathetic or chemically mediated vasoconstriction of arterioles by relaxing smooth muscle in a guanylate cyclase mediated reaction. NO also diffuses into the microvascular lumen where it regulates RBC [83] and leukocyte deformability [44], leukocyte-endothelial adhesion in mesenteric and skeletal muscle postcapillary venules [84,85], and platelet adhesion and aggregation [86,87]. NO has also been reported to both maintain and, during endotoxemia, increase vascular permeability in the intestine, heart, liver, and kidney [88,89]. Taken together, NO is an important factor in maintaining the integrity of blood flow through the microcirculation by regulating resistance vessel diameter, blood rheology, interaction between cellular blood elements and the vascular wall, and blood volume.
NO has been reported to be transported from the lungs to the tissue via S-nitrosation of the Hb β-chain cysteine 93 thiol group (SNO-Hb) [65,90] and released in the microcirculation as a low molecular weight nitrosothiol (RSNO) with vasorelaxant properties. The mechanisms of formation and release of NO from RSNO species are not completely understood [91]. Feelisch and coworkers [92] have also found considerable species variation in the amounts and ratios of RSNOs, nitrosamines, and nitrosylhemes. Although providing evidence that intravascular transfer of NO occurs between SNO-Hb and nitrosated albumin, Gladwin and coworkers [93] concluded that in human circulation these nitrosated species had little effect on regulating vascular tone.
In human NO breathing experiments, in which exogenous NO counteracted the vasopressor effect of nonspecific NOS inhibition in the forearm, Cannon and coworkers [94] reported that NO reacted predominantly with the heme moiety of Hb forming either nitrosyl(heme)hemoglobin (Hb[FeII]NO) with deoxyhemoglobin or, consistent with a NO scavenging role by the RBC [95], methemoglobin (FeIII) and nitrate (NO 3 -) with oxyhemoglobin. The authors argued that remote vasodilation was mediated either by direct NO release from Hb(FeII)NO, indirectly via a SNO-Hb intermediate, or by nitrite (NO 2 -) bioconversion to NO [94]. In similar human studies, intravenous infusion and bolus injection of NO and RSNO in the brachial artery increased both artery diameter and forearm blood flow [96,97]. Here, the authors concluded that NO was transported as both free and nitrosative, RSNO, forms.
Caveats with respect to interpreting data from functional studies investigating NOS inhibition are that NO levels in tissue or plasma are often unknown, inhibition of NO overproduction is significant but incomplete, and the degree and specificity of NOS inhibition are unknown. Additionally, the reader should bear in mind that, during sepsis, NO functions in a 'sepsis milieu' of increased levels of reactive oxygen species and endogenous vasopressors, especially endothelin, which is upregulated [98] during sepsis and independent of NO inhibition [99]. Comparison of data can also be confounded by differences in animal models and the timing of the NOS inhibition (i.e. pretreatment versus delayed administration).

Nitric oxide over-production and time course during sepsis
During the progression of sepsis, proinflammatory cytokines tumor necrosis factor-α and interleukin-1, and lipopolysaccharide (LPS) stimulate the upregulation of iNOS [100] throughout the organs [101] and nNOS in brain and skeletal muscle [102]. Various LPS and cecal ligation and perforation (CLP) models of sepsis have shown rapid but transient increases in iNOS and nNOS mRNA expression followed by or commensurate with increased NOS activity and increased levels of NO oxidized metabolites, namely NO 2 and NO 3 -(NO x -), in tissue and plasma.
Following LPS treatment, iNOS mRNA in rat kidney appeared by 60 min, peaked at 2-4 hours along with plasma NO x -, and declined by 16 hours [103]; nNOS mRNA in rat brain increased by 2 hours whereas iNOS mRNA increased by 3 hours, with both NOS isoforms returning to baseline by 12 hours [104]; and iNOS mRNA in the small intestine was detected at 1 hour, peaked at 4 hours, and was faint at 24 hours [105]. In dogs treated with LPS, iNOS activity was increased in the liver by 4 hours and in the heart by 6 hours [106]. Infusion of Staphylococcus aureus cell wall components caused progressive increases in iNOS activities in thoracic aorta, lung and liver, and plasma NO x from 2 to 6 hours [107].
In rats subjected to CLP, Sheih and coworkers [108] found that tissue NO x levels increased from 5 to 10 hours in kidney, gut, heart, liver, and lungs commensurate with increased iNOS mRNA and increased plasma NO x -. Also in CLP rats, thoracic aorta and lung iNOS protein and activity increased by 6 hours and remained elevated for 48 hours, whereas cNOS protein and activity progressively decreased from 6 to 48 hours. Concomitantly, plasma NO x increased by 6 hours, peaked at 12 hours, and remained elevated for 48 hours [109]. The progressive loss of cNOS protein was consistent with downregulation of ecNOS protein and mRNA in bovine coronary venular endothelial cells 8 hours after exposure to LPS [110].
In septic patients no such NOS/NO x time course profiles have been determined; however, increased plasma NO x - [111] and increased skeletal muscle tissue NO x - [78] have been reported. In a longitudinal patient study, fluctuating plasma NO x levels were associated with recurring bacterial infection over a 28-day period [112]. Plasma NO x was also reported to be inversely related to systemic vascular resistance [113] and positively related to cardiac output [114] in septic patients.

Effect of nitric oxide over-production on microvascular reactivity
Under septic conditions iNOS and nNOS are upregulated in endothelial and muscle cells, respectively [102,115], leading to over-production of NO in the microvasculature and arteriolar dysfunction. In a rat CLP model of sepsis, Gocan and coworkers [102] reported that microvascular reactivity to acetlycholine, quantified as changes in arteriolar diameter and downstream capillary RBC velocity, was impaired in skeletal muscle by NO and restored by nNOS inhibition. Interestingly, acetylcholine increased the proportion of perfused microvessels in the sublingual microcirculation of septic patients [34]. Hollenberg and coworkers [42,43] found that NOS inhibition in CLP rats reversed arteriolar hyporesponsiveness to catecholamines and endothelin in cremaster muscle, as measured by changes in arteriolar contraction. In another study, iNOS deficient mice were resistant to vascular hypocontractility [116]. All of these findings suggest that NO over-production was a factor in refractory hypotension in sepsis.

The effect of nitric oxide synthase inhibition on hemodynamics and oxygen transport in sepsis
In 1990, Kilbourn and coworkers [117] reported that tumor necrosis factor induced hypotension in dogs could be reversed by nonspecific NOS inhibition. Subsequently, numerous clinical [118][119][120][121][122] and animal studies of sepsis [4,5,123] demonstrated that NOS inhibition prevented or reversed hypotension, even in septic patients unresponsive to conventional vasoconstrictor therapy [118,122]. In 48-hour sheep models of sepsis, both nonspecific and iNOS specific NOS inhibition reversed hyperdynamic sepsis and normalized cardiac indices, systemic vascular resistance, and oxygen extraction [124][125][126] by peripheral vasoconstriction. In a clinical study of septic shock patients, Broccard and coworkers [118] reported that nonspecific NOS inhibition increased MAP and decreased cardiac output, allowed gradual withdrawl of α-adrenergic support, decreased DO 2 and increased O 2 ER, but had no effect on VO 2 , lactate, or splanchnic oxygenation. Despite a favorable outcome on MAP in septic shock patients, a nonspecific NOS inhibition phase III clinical trial was terminated because of increased mortality arising from increased cardiovascular failure [127]. Interestingly, activated protein C, which is currently the only treatment to reduce patient mortality in severe sepsis [128], has been found to reduce increased lung iNOS mRNA and activity and prevent hypotension in a 3-hour LPS rat model of sepsis [129].
The effect of inhibiting NO over-production on microvascular geometry, hemodynamics, and oxygen transport is less well characterized. Nonspecific NOS inhibition normalized cardiac output but exacerbated vasoconstriction in the small intestine of rats infused with Escherichia coli [130]. In dogs treated with LPS for 3 hours, Walker and coworkers [131] reported that nonspecific NOS inhibition normalized gut and hind limb vascular resistance with negligible effect on oxygen extraction or oxygen uptake. In ex vivo rat endotoxemic hearts, nonspecific NOS inhibition resulted in decreased coronary blood flow and myocardial ischemia as measured by increased NADH fluorescence [132]. The same reduction in coronary blood flow in normal hearts had no effect, suggesting that hyperdynamic sepsis masked an underlying myocardial microvascular dysfunction. In liver, nonspecific NOS inhibition in 2-, 3-, and 8-hour models of sepsis decreased sinusoidal blood flow and increased both the number of nonperfused sinusoids and leukocyte adhesion in sinusoids and postsinusoidal venules [48,49,75]. Huang and coworkers [133] also reported that both nonspecific and partially specific NOS inhibition with aminoguanidine exacerbated 7 hour LPS induced decreases in blood velocity and hemoglobin oxygenation in liver sinusoids. Liver hypersensitivity to the pressor effect of endothelin may account, in part, for NOS inhibition exacerbating decreased liver hemodynamics during sepsis [98]. Although fluid resuscitation restored skeletal muscle tissue PO 2 in a 3.5-hour rat LPS model of sepsis, concurrent nonspecific NOS inhibition reduced tissue PO 2 [36].
Maintaining plasma NO x at baseline in a 6-hour rat CLP model of sepsis using the iNOS inhibitor aminoguanidine attenuated the loss of functional capillary density in skeletal muscle [26]. In the ileum, aminoguanidine was found to attenuate LPS induced decreased oxygen consumption [134]. However, the specific iNOS inhibitor 1400W had no effect on decreased intestinal perfused villi or oxygen extraction in a 24-hour pig LPS model of sepsis [135]. The effectiveness of NOS inhibition in this study, however, is uncertain because portal venous NO x levels remained elevated. In 6-hour endotoxemic pigs, the more selective iNOS inhibitor aminoethylisothiourea normalized hepatic artery blood flow and partially restored portal venous flow [136]. Although liver oxygen extraction remained above baseline, interestingly, liver oxygen consumption increased. In an 8-hour rat ischemia/reperfusion LPS model in which nonspecific and partially specific iNOS inhibition by aminoethyl-isothiourea reduced plasma NO x by 33% and 44%, respectively, liver microvascular blood flow, as assessed by Laser Doppler flowmetry, decreased dramatically in both cases [137].
Alternatively, evidence from NO donor studies suggest that increased NO production may preserve or protect microvascular blood flow. Sodium nitroprusside attenuated the loss of perfused liver sinusoids in a rat LPS model of sepsis [47], administration of nitroglycerin to fluid resuscitated septic patients increased sublingual microvascular flow index [33], and the NO donor 3-morpholinosydnonimine decreased critical oxygen delivery and increased critical oxygen extraction ratio in dogs treated with LPS [4], indicating that increased NO improved the matching of microvascular oxygen delivery with oxygen demand. NO over-production may also have caused an apparent rebound in liver perfusion at a later stage of sepsis [138]. Taken together with the deleterious effects of NOS inhibition, these results suggest that NO over-production may have a protective effect on microvascular blood flow by counteracting increased endogenous vasopressor activity during sepsis; clearly, however, more microcirulatory research needs to be conducted to assess the role of endogenous NO during sepsis.

Relationship between nitric oxide, mitochondrial dysfunction, and oxygen supply
Oxygen is primarily consumed at the inner membrane of the mitochondrion by the redox reaction, which simultaneously oxidizes cytochrome c and reduces oxygen to water. The reaction is catalyzed by the terminal enzyme of the electron transport chain, namely cytochrome c oxidase (or cytochrome a,a 3 ). Several in vitro studies have reported that NO inhibits cytochrome c oxidase [139][140][141][142]. Moreover, Torres and coworkers [140] found that the degree of inhibition is determined by the oxygen concentration (Fig. 7). Their data also indicate that inhibition is reversible and suggest that NO and oxygen compete for the same binding site on cytochrome c oxidase, although the precise nature of the inhibition is unknown. In vitro studies of rat aorta endothelial cells have shown that NO inhibits mitochondrial respiration in an oxygen dependant manner [143]. From the standpoint of the septic microcirculation, these findings raise the intriguing possibility that sepsis induced microvascular oxygen transport dysfunction and NO over-production contribute to both tissue hypoxia and mitochondrial inhibition.
Near infrared spectroscopy of baboon forearm muscle in an E. coli infusion and fluid resuscitated model of sepsis found progressive changes in the redox state of cytochrome a,a 3 that correlated with ultrastructural changes in the mitochondria despite minimal changes in global DO 2 , VO 2 , and oxygen extraction [144]. The authors concluded that abnormalities in muscle oxygen metabolism and mitochondrial function were the result of an early defect in oxygen supply followed by a progressive loss of cytochrome a,a 3 function. In human septic patients it was recently reported that increased NO levels in muscle (measured as NO x -) are associated with mitochondr-ial dysfunction, decreased ATP concentration, organ failure, and eventual outcome [78].

Conclusion
Since the mid-1950s it has been known that sepsis induces profound derangements in cardiovascular function. More recently, information acquired on the functional state of the microcirculation in intestine, liver, and skeletal muscle has shown that sepsis induces profound changes in microvascular geometry, hemodynamics, and oxygen transport. Increased microvascular stopped-flow results in a maldistribution of RBC flow within the microcirculation and a mismatching of local oxygen delivery with oxygen demand. Remaining functional capillaries compensate for decreased functional capillary density by offloading more oxygen to the surrounding tissue; nevertheless, increased oxygen flow heterogeneity seemingly impairs oxygen extraction by increasing critical oxygen delivery and decreasing the critical oxygen extraction ratio. Abnormal microvascular oxygen transport also indicates that regulatory mechanisms have become dysfunctional and suggests that local cellular environments, as such, have been dramatically altered. The loss of capillary blood flow may Available online http://ccforum.com/content/7/5/359

Figure 7
Dependence on oxygen concentration of the percentage of inhibition of cytochrome c oxidase by nitric oxide (NO). For equal concentrations of cytochrome c oxidase and NO (1.2-1.8 µm) oxygen was found to compete with NO when the enzyme enters turnover. Each data point represents a different experimental setup. The results suggest that, under physiologic conditions, in which oxygen concentration is low, nanomolar concentrations of NO can regulate mitochondrial respiration. By permission from Biochem J 1995, 312:169-173 [140].
potentiate the effects of proinflammatory mediators by increasing their residence time in the microcirculation and tissue.
The potent vasodilator NO plays an important, yet complex, role in microvascular homeostasis. During sepsis, over-production of NO has been associated with decreases in blood pressure, impaired microvascular reactivity, abnormal RBC deformability, decreased functional capillary density, and reduced oxygen consumption. Although inhibiting NO during sepsis increases blood pressure, it also reduces microvascular blood flow and exacerbates abnormal oxygen transport. Evidence that NO donors improve microvascular hemodynamics would seem to suggest that NO over-production protects microvascular flow and oxygen transport during sepsis, but clearly more microcirculatory research must be performed to assess the role of endogenous NO during sepsis.