Progressive decrease in constrictor reactivity of the non-absorbing intestine during chronic sepsis.

Chronic sepsis leads to an impaired intestinal microcirculation, which might reflect altered microvascular control. We hypothesized that intestinal microvascular sensitivity to norepinephrine (NE) is decreased during chronic sepsis. Chronic sepsis was induced by a polymicrobial inoculation of implanted subcutaneous sponges in rats. Septic rats were studied either 24 or 72 h after a single inoculation (1-hit) of bacteria. Other rats received a second inoculation (2-hit) of bacteria 48 h later and were studied at 24 h after the second inoculation. NE (0.01-1.0 microM) responses in the non-absorbing terminal ileal arterioles (inflow A1, proximal-p and distal-d premucosal A3) were measured by video microscopy. NE threshold sensitivity (pD(T20) = -log of 20% response dose) was analyzed. pD(T20) was significantly decreased in A1, pA3, and dA3 of 1-hit 24-h septic rats (P < 0.05), and was further decreased in all vessels of 2-hit 72-h septic rats (P < 0.05). In contrast, the pDT(T20) of all three vessels significantly returned toward normal values after 72 h in rats that had only 1 bacteria inoculation. We conclude that an initial bacterial challenge decreases vasoconstrictor reactivity of the intestinal microcirculation and that subsequent repeated bacterial challenge exacerbates this defect in vasoconstrictor control in the non-absorbing intestine.

PAF increases vascular permeability without increasing pulmonary arterial pressure in the rat.

In vivo pulmonary arterial catheterization was used to determine the mechanism by which platelet-activating factor (PAF) produces pulmonary edema in rats. PAF induces pulmonary edema by increasing pulmonary microvascular permeability (PMP) without changing the pulmonary pressure gradient. Rats were cannulated for measurement of pulmonary arterial pressure (Ppa) and mean arterial pressure. PMP was determined by using either in vivo fluorescent videomicroscopy or the ex vivo Evans blue dye technique. WEB 2086 was administered intravenously (IV) to antagonize specific PAF effects. Three experiments were performed: 1) IV PAF, 2) topical PAF, and 3) Escherichia coli bacteremia. IV PAF induced systemic hypotension with a decrease in Ppa. PMP increased after IV PAF in a dose-related manner. Topical PAF increased PMP but decreased Ppa only at high doses. Both PMP (88 +/- 5%) and Ppa (50 +/- 3%) increased during E. coli bacteremia. PAF-receptor blockade prevents changes in Ppa and PMP after both topical PAF and E. coli bacteremia. PAF, which has been shown to mediate pulmonary edema in prior studies, appears to act in the lung by primarily increasing microvascular permeability. The presence of PAF might be prerequisite for pulmonary vascular constriction during gram-negative bacteremia.

Hypercoagulability in patients with obstructive sleep apnea.

BACKGROUND: Obstructive sleep apnea (OSA) has been linked to cardiovascular complications such as stroke and myocardial infarction. Previous studies demonstrate that OSA patients show elevated fibrinogen levels and increased platelet aggregation that are reversed with 1 night of nasal continuous positive airway pressure treatment (NCPAP). Questioning overall coagulability in OSA, we examined whole blood coagulability in 11 chronically NCPAP treated OSA subjects, 22 previously untreated OSA subjects, and in 16 of these after 1 night of NCPAP treatment. PATIENTS AND METHODS: During full polysomnography, subjects from each group had blood drawn prior to bedtime (21:00 h) and upon waking in the morning (07:00 h). RESULTS: Untreated OSA patients had faster P.M. clotting times than chronically treated OSA patients (3.33+/-0.31 versus 6.12+/- 0.66 min, P<0.05 by ANOVA). A.M. values showed similar results (4.31+/- 0.34 min versus 7.08+/-0.52 min, P<0.05 by ANOVA) for the respective groups. One overnight treatment with nasal CPAP did not produce a significant change in A.M. whole blood coagulability (4.35 +/-0.43 to 5.31+/-0.53 min; n=16; P=0.1) in 16 treated subjects. CONCLUSIONS: These data indicate a relationship between obstructive sleep apnea and blood hypercoagulability status that appears to be reversed by chronic NCPAP treatment. These data suggest that NCPAP might protect against the development of cardiovascular complications in OSA patients.

Sustained infection induces 2 distinct microvascular mechanisms in the splanchnic circulation.

BACKGROUND: Altered intestinal blood flow during systemic inflammation leads to organ dysfunction. Mucosal ischemia occurs during sepsis despite an increase in portal blood flow. We hypothesized that separate mechanisms are active in the large resistance and small mucosal microvessels to account for this dichotomy. METHODS: Chronic infection was induced in rats by bacterial inoculation (Escherichia coli and Bacteroides fragilis) of an implanted subcutaneous sponge. Separate groups were studied at 24 and 72 hours after a single inoculation of bacterium or 24 hours after a second inoculation (ie, 72 hours of sepsis). Time-matched controls were used for each group. Intravital microscopy of the terminal ileum was used to assess endothelial-dependent vasodilation to acetylcholine (10(-9) to 10(-5) mol/L) in resistance (A(1)) and premucosal (A(3)) arterioles. Threshold sensitivity (-log of 20% response dose) was calculated from dose response curves for each animal. RESULTS: Vasodilator sensitivity to acetylcholine in A(1) arterioles was significantly decreased at 24 hours, and these changes persisted up to 72 hours after a single bacterial inoculation. There was no change in the dilator sensitivity of A(3) arterioles after a single inoculation. When there was a challenge with a second bacterial inoculation, there was a reversal of the A(1) dilator response and an increase in A(3) sensitivity. CONCLUSIONS: An initial septic event results in a decrease in dilator reactivity in the resistance A1 arterioles that persists for at least 72 hours. A sustained septic challenge results in increased dilator reactivity in both A(1) and A(3) vessels. This enhanced sensitivity during sepsis suggests that more than 1 therapeutic approach to preservation of intestinal blood flow will be necessary.

Lazaroid and pentoxifylline suppress sepsis-induced increases in renal vascular resistance via altered arachidonic acid metabolism.

Early sepsis leads to renal hypoperfusion, despite a hyperdynamic systemic circulation. It is thought that failure of local control of the renal microcirculation leads to hypoperfusion and organ dysfunction. Of the many mediators implicated in the pathogenesis of microvascular vasoconstriction, arachidonic acid metabolites are thought to be important. Vasoconstriction may be due to excess production of vasoconstrictors or loss of vasodilators. Using the isolated perfused kidney model, we describe a sepsis-induced rise in renal vascular resistance and increased production of key arachidonic acid metabolites, both vasoconstrictors and vasodilators, suggesting excessive production of vasoconstrictors as a cause for microcirculatory hypoperfusion. There is evidence of increased enzymatic production of arachidonic acid metabolites as well as nonenzymatic, free radical, catalyzed conversion of arachidonic acid. Pentoxifylline (a phosphodiesterase inhibitor) and U74389G (an antioxidant) both have a protective effect on the renal microcirculation during sepsis. Both drugs appear to alter the renal microvascular response to sepsis by altering renal arachidonic acid metabolism. This study demonstrates that sepsis leads to increased renal vascular resistance. This response is in part mediated by metabolites produced by metabolism of arachidonic acid within the kidney. The ability of drugs to modulate arachidonic acid metabolism and so alter the renal response to sepsis suggests a possible role for these agents in protecting the renal microcirculation during sepsis.

Regulation of intestinal blood flow.

The gastrointestinal system anatomically is positioned to perform two distinct functions: to digest and absorb ingested nutrients and to sustain barrier function to prevent transepithelial migration of bacteria and antigens. Alterations in these basic functions contribute to a variety of clinical scenarios. These primary functions intrinsically require splanchnic blood flow at both the macrovascular and microvascular levels of perfusion. Therefore, a greater understanding of the mechanisms that regulate intestinal vascular perfusion in the normal state and during pathophysiological conditions would be beneficial. The purpose of this review is to summarize the current understanding regarding the regulatory mechanisms of intestinal blood flow in fasted and fed conditions and during pathological stress.

Platelet-activating factor and bacteremia-induced pulmonary hypertension.

BACKGROUND: Acute lung injury is a common complication of gram-negative sepsis. Pulmonary hypertension and increased lung vascular permeability are central features of lung injury following experimental bacteremia. Platelet-activating factor is a prominent proinflammatory mediator during bacterial sepsis. Our previous studies have demonstrated that exogenous administration of platelet-activating factor (PAF) induces pulmonary edema without causing pulmonary hypertension. Interestingly, inhibition of PAF activity during Escherichia coli bacteremia prevents the development of both pulmonary hypertension and pulmonary edema. These data suggest that PAF contributes to pulmonary hypertension during sepsis, but that this is unlikely to be a direct vascular effect of PAF. The goal of the present study was to investigate the mechanism by which acute E. coli bacteremia induces pulmonary injury and to define the role that PAF plays in this injury. We hypothesized that the effects of PAF on pulmonary hypertension during bacteremia are due to the effects of PAF on other vascular mediators. Several studies suggest that PAF induces the expression of endothelin-1 (ET), a potent peptide vasoconstrictor. Further, our previous studies have implicated ET as a central mediator of systemic vasoconstriction during bacteremia. We therefore sought to assess whether ET is modulated by PAF. E. coli has also been demonstrated to increase endothelial production of nitric oxide (NO), which contributes to maintenance of basal vascular tone in the pulmonary circulation. We hypothesized that PAF might increase pulmonary vascular resistance during bacteremia by activating neutrophils, increasing expression of ET, and decreasing the tonic release of NO. Furthermore, we hypothesized that hypoxic vasoconstriction did not contribute to pulmonary vasoconstriction during the first 120 min of E. coli bacteremia. METHODS: Pulmonary artery pressure (PAP), blood pressure (BP), heart rate (HR), and arterial blood gases (ABG) were measured in anesthetized spontaneously breathing adult male Sprague-Dawley rats. E. coli (10(9) CFU/100 g body wt) was injected at t = 0, and hemodynamic data were obtained at 10-min intervals and ABG data at 30-min intervals for a total of 120 min. Sham animals were treated equally but received normal saline in place of E. coli. In treatment groups, a 2.5 mg/kg dose of WEB 2086, a PAF receptor antagonist, was administered intravenously 15 min prior to the onset of sepsis or sham sepsis. The groups were (1) intravenous E. coli (n = 5); (2) intravenous WEB 2086 pretreatment + intravenous E. coli (n = 5); (3) intravenous WEB 2086 alone (n = 5); and (4) intravenous normal saline (n = 6). Nitric oxide metabolites (NOx) and ET concentrations were assayed from arterial serum samples obtained at the end of the protocol. Lung tissue was harvested for measurement of myeloperoxidase (MPO) activity and pulmonary histology. RESULTS: E. coli bacteremia increased HR, PAP, and respiratory rate early during sepsis (within 20 min), while hypoxemia, hypotension, and hemoconcentration were not manifest until the second hour. Pretreatment with WEB 2086 completely abrogated all of these changes. E. coli bacteremia increased the activity of serum ET, lung MPO, and neutrophil sequestration in the lung parenchyma via a PAF-dependent mechanism. However, the mechanism of increased production of NO appears to be PAF independent. CONCLUSIONS: These data support the hypothesis that E. coli bacteremia rapidly induces pulmonary hypertension stimulated by PAF and mediated at least in part by endothelin-1 and neutrophil activation and sequestration in the lung. Microvascular injury with leak is also mediated by PAF during E. coli bacteremia, but the time course of resultant hypoxemia and hemoconcentration is slower than that of pulmonary hypertension. The contribution of hypoxic vasoconstriction in exacerbating pulmonary hypertension in gram-negative sepsis is probably a late

Glucose and glutamine gavage increase portal vein nitric oxide metabolite levels via adenosine A2b activation.

INTRODUCTION: Postprandial intestinal hyperemia is a complex vascular response during nutrient absorption. Many mediators have been studied including enteric reflexes, GI hormones, and absorption-stimulated metabolic mediators such as pH and adenosine. We have shown that nitric oxide (NO) mediates premucosal arteriolar dilation during glucose absorption and that glucose-induced portal vein NO metabolite production requires adenosine A2b receptor activation. We hypothesize that Na+-linked absorption of l-glutamine or l-glycine might also stimulate NO release in the enteroportal circulation via adenosine A2b receptors. METHODS: Male Sprague-Dawley rats (190-220 g) were anesthetized with urethane/alpha-chloralose and cannulated for hemodynamic monitoring and blood sampling. A right paramedian abdominal incision was made for access to both the stomach (gavage) and the portal vein (blood sampling). Animals received intragastric nutrient gavage (saline, d-glucose, l-glutamine, racemic glycine, or oleic acid) with and without adenosine A2b receptor blockade. NO metabolites (NOx) were measured by a fluorescent modified-Greiss assay at baseline and 30 min after nutrient gavage. RESULTS: Glucose and glutamine gavage increased portal NOx levels compared to baseline, while glycine and oleic acid gavage did not. Adenosine A2b antagonism returned NOx levels to baseline in both glucose and glutamine gavage animals, but did not alter portal NOx levels in glycine- or oleic acid-treated animals. CONCLUSIONS: These data suggest that nutrient-induced adenosine is involved in a signaling process from the intestinal epithelium to nitric oxide-producing cells elsewhere in the vasculature. Adenosine A2b receptors are required for NO production during Na+-linked glucose or glutamine absorption.

Complement activation mediates intestinal injury after resuscitation from hemorrhagic shock.

BACKGROUND: Endothelial cell injury after hemorrhage and resuscitation (HEM/RES) might contribute to intestinal hypoperfusion and mucosal ischemia. Our recent work suggests that the injury might be the result of complement activation. We hypothesized that HEM/RES causes complement-mediated endothelial cell dysfunction in the small intestine. METHODS: Male Sprague-Dawley rats (195-230 g) were anesthetized and HEM to 50% of baseline mean arterial pressure for 60 minutes. Just before RES, animals received either soluble complement receptor-1 (sCR1, 15 mg/kg) to inhibit complement activation or saline vehicle. Resuscitation was with shed blood and an equal volume of saline. Two hours after RES, the small bowel was harvested to evaluate intestinal nitric oxide synthase activity (NOS), neutrophil influx, histology, and oxidant injury. RESULTS: HEM/RES induced tissue injury, increased neutrophil influx, and reduced NOS activity by 50% (vs. SHAM), all of which were completely prevented by sCR1 administration. There were no observed differences in oxidant injury between the groups. CONCLUSION: Histologic tissue injury, increased neutrophil influx, and impaired NOS activity after HEM/RES were all prevented by complement inhibition. Direct oxidant injury did not seem to be a major contributor to these alterations. Complement inhibition after HEM might ameliorate reperfusion injury in the small intestine by protecting the endothelial cell, reducing neutrophil influx and preserving NOS function.

Small intestinal production of nitric oxide is decreased following resuscitated hemorrhage.

BACKGROUND: Small intestine microvascular vasoconstriction and hypoperfusion develop after resuscitation (RES) from hemorrhage (HEM), despite restoration of central hemodynamics. The responsible mechanisms are unclear. We hypothesized that the microvascular impairment following HEM/RES was due to decreased intestinal microvascular nitric oxide (NO) production. METHODS: Male Sprague-Dawley rats (195-230 g) were utilized and three experimental groups were studied: (1) SHAM (cannulated but no HEM), (2) HEM only, and (3) HEM/RES. HEM was to 50% of baseline mean arterial pressure for 60 min, and RES was with shed blood and an equivalent volume of saline. Ex vivo isolated intestinal perfusion and a fluorometric modification of the Greiss reaction were used to quantify production of NO metabolites (NOx). Perfusate von Willebrand factor (vWF) was used as an indirect marker of endothelial cell activation or injury. To assess the degree of NO scavenging by oxygen-derived free radicals, immunohistochemistry was used to detect nitrotyrosine formation in the intestine. RESULTS: Intestinal NOx decreased following HEM/RES (SHAM 1.35 +/- 0.2 mM vs HEM/RES 0.60 +/- 0.1 mM, P < 0.05), but not with HEM alone (1.09 +/- 0.3 mM). There were no differences in serum NOx levels between the three groups. Release of vWF was increased during the HEM period (SHAM 0.18 +/- 0.1 g/dl vs HEM 1.66 +/- 0.6 g/dl, P < 0.05). There was no detectable nitrotyrosine formation in any group. CONCLUSIONS: Intestinal NO metabolites decrease following HEM/RES. Elevated vWF levels during HEM and the lack of detectable nitrotyrosine suggest that this is due to decreased endothelial cell production of NO. HEM/RES-induced endothelial cell dysfunction may contribute to persistent small intestine post-RES hypoperfusion and vasoconstriction.

Glucose-induced intestinal hyperemia is mediated by nitric oxide.

UNLABELLED: Glucose-induced absorptive hyperemia of the intestine has been well demonstrated through microsphere blood flow experiments. We have previously demonstrated that glucose, when applied topically to rat ileal epithelium, restores microvascular vessel diameters and blood flow following Escherichia coli bacteremia or hemorrhage/resuscitation. However, the mechanisms of this hyperemia are not completely understood. We hypothesize that nitric oxide is a mediator of the microvascular response to glucose exposure on the rat intestinal epithelium. METHODS: Male Sprague-Dawley rats, 200-225 g, were monitored for hemodynamic stability with mean arterial blood pressure and heart rate. A 2-cm segment of the terminal ileum with intact neurovascular supply was exposed for intravital videomicroscopy. Intestinal arteriolar diameters (A1D, inflow; and A3D, premucosal arterioles) and microvascular blood flow (A1Q) were measured following topical application of isoosmotic glucose or saline, with or without l-NAME (LN, 100 mM), a competitive inhibitor of nitric oxide synthase. Statistical analysis was performed by ANOVA followed by Tukey-Kramer honestly significant difference test. RESULTS: All data are expressed as mean percentage changes from baseline +/- standard error of the mean. Hemodynamic variables did not change during the experimental procedure and there were no significant differences among group baselines. Addition of isotonic glucose to the bath solution caused a significant increase in A3D that persisted throughout the experiment (at 30 min, 19.2 +/- 4.2 vs -3.9 +/- 4.5, P < 0.05). This vasodilation was blocked by topical administration of LN (3.1 +/- 2.9, P < 0.05). A1D remained at baseline levels (saline and glucose) or constricted (LN) in all groups. Topical LN also attenuated A1Q in both the saline and glucose groups. CONCLUSIONS: These data demonstrate that glucose-induced intestinal hyperemia is primarily characterized by premucosal A3 arteriole dilation in this model and that nitric oxide is a mediator of glucose-induced intestinal hyperemia. These findings suggest that either (1) glucose directly causes endothelial nitric oxide production or (2) epithelial cells transduce a vasodilatory signal through vascular endothelial-derived nitric oxide during postprandial intestinal hyperemia.

Heme oxygenase-dependent carbon monoxide production is a hepatic adaptive response to sepsis.

The hemodynamic effects of sepsis have been attributed in part to increased nitric oxide (NO) production and activation of guanylate cyclase, resulting in increased cGMP and relaxation of vascular smooth muscle. Heme oxygenase-1 (HO-1), a heat shock protein, has been shown to increase intracellular cGMP levels by formation of carbon monoxide (CO). We hypothesized that HO may be an important mediator of the hepatic response to infection. Male Swiss Webster mice underwent standard cecal ligation and puncture (CLP, 18 gauge 2X) or sham operation, and received either normal saline (NS) or Zn protoporphyrin IX (ZN PP IX), a competitive HO inhibitor (n = 6-8/group). Hepatic tissue samples were collected at 3, 6, 12, and 24 hr from separate mice. Serum was collected at 3 and 24 hr. A semiquantitative reverse transcriptase polymerase chain reaction method was used to measure HO-1 mRNA levels. Hepatic cGMP levels were measured by ELISA. Groups were repeated (n = 10/group) to assess mortality. Serum was collected at 3 and 24 hr to measure serum aspartate aminotransferase (AST) levels. HO-1 mRNA expression increased significantly by 3 hr after CLP and with HO inhibition alone (P < 0.05 vs sham + NS). HO-1 mRNA remained elevated through 24 hr. CLP animals with HO inhibition showed a significant reduction of hepatic cGMP following CLP compared with CLP + saline at 24 hr (P < 0.05). Mortality was significantly increased in the CLP + ZN PP group at 24 hr (P < 0.05 CLP NS vs CLP ZN PP). CLP caused a marked increase in AST activity, which was increased further with HO inhibition. HO-1 mRNA expression was induced by CLP. AST levels following CLP were markedly increased with HO inhibition. HO-1 function appeared to contribute to elevation of hepatic cGMP during peritonitis and may be an important hepatic adaptive response to infection.

Differential intestinal microvascular dysfunction occurs during bacteremia.

Altered vascular responsiveness is the hallmark of septic shock. Recently, these changes have frequently been attributed to increased production of nitric oxide (NO). Continued exposure to high levels of NO may alter both endothelial and vascular smooth muscle cell function. Although ex vivo studies demonstrate hyporeactivity of large conduit arteries during established sepsis, it is unclear if the same phenomena exist during early sepsis. This is especially true in the small resistance arterioles of the viscera. We used in vivo microscopy of the rat small intestine to assess (1) endothelial-dependent relaxation and vasomotion (periodic contraction and relaxation of blood vessels) in response to acetylcholine (ACH; 10(-8) to 10(-5) M), (2) endothelial-independent relaxation to nitroprusside (NTP; 10(-5) M), and (3) vascular smooth muscle response to norepinephrine (NE; 10(-10) to 10(-7) M) in normal and bacteremic rats (Escherichia coli). There were no alterations in endothelial-dependent or -independent relaxation during bacteremia as measured by mean diameters. However, acute E. coli bacteremia severely impaired vasomotion in A1 (inflow) and A3 (premucosal) arterioles. Vasomotion was returned to baseline levels in A1 with low-dose ACH (10(-8) M) but only partially improved in A3 arterioles (P < 0.05). A1 response to NE was impaired, while A3 were minimally altered despite being more sensitive to E. coli-induced vasoconstriction. These data suggest that bacteremia causes a rapid, differential impairment of both endothelial-dependent (A3 vasomotion) and vascular smooth muscle cell (A1 constriction) functions. These microvascular impairments occur much earlier than previously described and may contribute to sepsis-induced mucosal ischemia of the intestines.

Preoperative saline loading improves outcome after elective, noncardiac surgical procedures.

Patients with multiple system disease undergoing elective noncardiac surgical procedures are at variable risk for developing postoperative complications and death. To determine whether preoperative expansion of plasma volume would improve outcome, 306 patients were admitted to the Surgical Intensive Care Unit of the Veterans Administration Center for Swan-Ganz catheter placement and measurement of hemodynamic responses to a 2 L infusion of normal saline over 2 hours. Intraoperative stability and postoperative outcome were assessed by chart review and compared with similar operative groups of patients who did not receive saline infusion. Eighty-eight per cent of the patients had a positive expansion of blood volume with saline infusion. In patients undergoing aortic reconstructive procedures, there was a reduction in the incidence of postoperative complications (52% to 28%) primarily attributed to a reduction in pulmonary complications. In all patients there was an improvement in intraoperative cardiovascular stability (57% saline vs 38% control), a reduction in the need for pharmacologic support of blood pressure (19% saline vs 30% control), and reduction in the amount of intraoperative fluid administration (hydration index: 5.12 saline vs 8.61 control). We therefore conclude that preoperative saline loading is associated with improved outcome in high risk elderly patients undergoing elective, noncardiac surgical procedures.

Nitric oxide mediates redistribution of intrarenal blood flow during bacteremia.

The normal or hyperdynamic circulatory response during the early phases of the systemic septic response is associated with renal microvascular constriction and can result in renal dysfunction. Intrarenal redistribution of blood flow from the outer cortex to the medulla appears to account for decreased glomerular filtration in spite of normal or elevated renal blood flow, but the mechanisms of this response are not well described. Nitric oxide is recognized as an important regulator of regional blood flow during both normal and pathologic conditions including sepsis, and we hypothesized that alterations in nitric oxide contribute to redistribution of renal blood flow during sepsis. The current study used laser Doppler fluximetry and clearance of p-aminohippuric acid (effective renal plasma flow, ERPF) to study intrarenal distribution of blood flow during basal conditions and during normodynamic Escherichia coli bacteremia, with and without inhibition of nitric oxide. Inhibition of nitric oxide in normal animals resulted in a decrease in ERPF (-19%) with a decrease in cortical flux (-39%) without alteration of medullary flux. Bacteremia resulted in a decrease in cortical flow (-17%), an increase in medullary flow (36%), and a modest reduction (-9%) in ERPF. Inhibition of nitric oxide synthase during bacteremia worsened cortical flow (-43%), reversed the increase in medullary flux (-42%), and further impaired ERPF (-28%). These data suggest that nitric oxide regulates renovascular tone during normal conditions and bacteremia, and indicate that it is a prime mediator of intrarenal redistribution of blood flow during sepsis.