Twenty-four hours after sepsis induction, 24 animals (n = 8 per group) were then re-anesthetized and tracheotomized 29. Intravital microscopy of the left liver lobe was performed as follows: median laparotomy was extended by a left subcostal incision and the hepatic ligaments of the left liver lobe were carefully dissected. The animal was placed in a 110° position on its left side onto the microscope (Eclipse 300, Nikon, Düsseldorf, Germany). The left liver lobe was exteriorized and the lower surface was placed on a microscope slide in a tension free position. Intravenously, 2 μmol/kg sodium fluorescein and 0.2 μmol/kg rhodamine 6 G (Sigma, Deisenhofen, Germany) were used for contrast enhancement.

In each experiment, 10 randomly chosen acini and 10 postsinusoidal venoles were recorded for 30 seconds both with sodium-fluorescein and rhodamine contrast enhancement. Offline image analysis was performed by a blinded investigator (CG) using a computer-assisted image analysis system (AnalySIS, OSIS, Muenster, Germany). Hepatic microcirculation was assessed by the periportal sinusoidal diameter of 10 sinusoids per acinus and the loss of sinusoidal perfusion, defined as the number of non-perfused sinusoids divided by all visible sinusoids of the acinus.

The leukocyte adhesion was evaluated separately in sinusoids and postsinusoidal venoles. Temporary adherent, that is, slowly moving or adhering at the sinusoidal wall for less than 20 seconds, and permanently adherent, that is, adherent for more than 20 seconds, leukocytes were counted in each acinus and expressed as cells/μm². Accordingly, temporarily and permanently adherent leukocytes in the venoles were counted as cells/μm² venolar endothelium.

In another set of 21 animals, cardiac output was determined applying the thermodilution technique 24 hours after sepsis. In these animals, a thermocouple catheter (IT21, Physitemp, Clifton, NJ, USA) was introduced into the aortic arch via the left carotid artery during instrumentation. For measurement of cardiac output, the area under temperature curve after injection of 0.3 ml cold saline solution (8°C) was recorded (Cardiac Output pod and Powerlab 4/20, ADInstruments, Spechbach, Germany). The results of three measurements were averaged in each animal. To reduce bias of emotional stress the procedure was mimicked every hour for four hours before measurement.

Liver cell injury was assessed 24 hours after induction of sepsis by measuring serum activities of aspartate aminotransferase and alanine aminotransferase. Blood was withdrawn via aortic puncture and plasma enzyme activity was determined by means of standard enzymatic techniques (Ektachem, Kodak, Stuttgart, Germany).

Specimens of the left liver lobe were collected immediately after death and fixed by immersion in 4% formaldehyde solution. Subsequently, they were dehydrated and embedded in paraffin wax to cut sections at a thickness of 5 μm. Slides were stained with H&E and assessed by an experienced pathologist.

Twenty-four hours after CLP and sham-procedure respectively, serum concentration of TNF-α was measured by a commercially available anti-rat TNF-α ELISA (BD OptEIA, Cat No. 550734, Becton Dickinson, Heidelberg, Germany) according to the manufacturer's instructions and read-out by a fluorometric plate reader (EL808, BioTek, Bad Friedrichshall, Germany).

Sigmastat 3.0 (Systat Software, Richmond, CA, USA) was used for statistical analysis. Normal distribution and equal variance tests were performed. Sepsis-induced and TEA-related effects were evaluated by one-way analysis of variance (ANOVA) with post-hoc Student Newman Keuls test or ANOVA on Ranks with post-hoc Dunn's test as appropriate. A P < 0.05 was defined as the level of significance. Data are presented as mean ± 95% confidence interval or as median (25%/75% percentiles) as appropriate.

Sinusoidal blood flow increased in the Sepsis group, whereas in the Sepsis + TEA group flow returned to Sham levels (Figure 2). The numbers of perfused sinusoids did not differ between groups. However, in the Sepsis group sinusoidal constriction was induced, which was not influenced in the Sepsis + TEA group (Figure 3).

Temporary leukocyte adhesion increased in sepsis both in the sinusoids and in the postsinusoidal venules. TEA reduced the temporary venolar leukocyte adhesion significantly, whereas it did not affect the increased sinusoidal adherence (Figure 4). The permanent sinusoidal and venolar leukocyte adherence was neither affected in the Sepsis group, nor in animals subjected to Sepsis + TEA.

In the untreated Sepsis group, serum activity of aspartate aminotransferase rose from 275 ± 101 U/l to 454 ± 108 U/l and alanine aminotransferase activity increased from 97 ± 81 U/l to 185 ± 58 U/l (P < 0.05 vs. Sham). These increases were not significantly affected by TEA. Similarly, histopathologic examination revealed only mild edema formation and patchy pericentral necrosis in sepsis.

Hepatic macrovascular inflow, although not directly measured in this study, most likely remained constant because cardiac output was not altered. This assumption is supported by numerous studies showing a consistent correlation of cardiac output and macrovascular hepatosplanchnic inflow in sepsis. In human sepsis, macrovascular hepatic inflow rose with cardiac output after therapeutic interventions 394041. Both in early and late CLP-sepsis macrovascular hepatic inflow was reduced in parallel with cardiac output 4243. Furthermore, in the presence of unchanged or increased cardiac output, hepatic macrovascular inflow also paralleled these changes in cardiac output 424445. Consequently, the sepsis-induced increase in sinusoidal blood flow and its reversal by TEA are most probably not caused by changes of macrovascular hepatic inflow.

Microvascular tissue perfusion in sepsis, however, is often uncoupled from the systemic circulation. In the clinical therapy of critical illness this dissociation might contribute to the persistent organ failure after hemodynamic stabilization 46. Earlier studies demonstrated unchanged or even decreased microvascular blood flow in the presence of two to three-fold increased regional blood flow 4245. In our study sinusoidal blood flow was increased in the Sepsis group whereas cardiac output was not altered. Intravital microscopy and cardiac output, however, needed to be performed in different sets of animals to minimize interaction between both techniques.

The increase in hepatic microvascular blood flow occurred despite sinusoidal vasoconstriction and consequently was related to increased sinusoidal blood flow velocity. These findings are consistent with an increased arteriolar inflow and are thus the first hint to an impaired hepatic arterial buffer response (HABR) in late polymicrobial sepsis. In HABR, liver arterial blood flow is adapted in response to changes in portal blood flow. Intrahepatic vasodilatation occurs at the level of the preterminal branches of the hepatic artery and is regulated by hydrogen sulfide and adenosine washout by portal blood flow 4748. Our results are supported by earlier findings demonstrating impaired HABR and a selective increase in hepatic arterial blood flow in endotoxemia 4950. In the Sepsis + TEA group, the sepsis-related increase in liver microvascular blood flow was blunted. It is therefore most likely that the continuous TEA restored HABR. However, this line of interpretation of the presented data is limited by the fact that we did not measure hepatic and portal flow, pressure and resistance separately in this study. Further investigations on the influence of TEA on hepatic blood flow regulation in sepsis and other clinically relevant conditions such as major liver resections are warranted.

In this study, loss of sinusoidal perfusion was not present in the Sepsis group whereas in earlier studies intravital microscopy revealed sinusoidal vasoconstriction and up to 30% reduction in sinusoidal perfusion both in early and late rodent CLP-sepsis 515253. Differences in volume resuscitation might partly explain the differing results. In the previous studies an initial bolus of 20 to 60 ml/kg saline solution was administered during 20 to 24 hour CLP-sepsis 5354. In our study, volume was infused continuously to a total dose of 144 ml/kg/24 hours. Detection of hypovolemia is often difficult in critical care. In experimental rat sepsis, volume depletion is even harder to exclude. In the present study, mean arterial pressure and heart rate were not significantly affected. Cardiac output remained stable both in untreated rats and in the sepsis + TEA group. The stable systemic hemodynamic parameters combined with stable acid-base-balance and a well-maintained microvascular perfusion in the liver suggests a sufficient resuscitation in this model. This clinically relevant infusion regimen might have prevented loss of perfused sinusoids and contributed to increased tissue blood flow in our study.

In the Sepsis + TEA group, sepsis-induced temporary leukocyte adhesion to the venolar endothelium decreased, whereas temporary sinusoidal leukocyte adhesion was not prevented. In contrast, the permanent leukocyte adhesion after 24 hour of sepsis was neither affected in the Sepsis group nor in the Sepsis + TEA group. This observation is in line with prior findings of time-dependent pattern of leukocyte recruitment in rat CLP-sepsis with increased leukocyte adhesion after seven hours and normalized values after 20 hours 55. In a recent study hepatic neutrophil recruitment declined after eight hours 56. In contrast to sinusoidal temporary adhesion, hepatic venolar temporary leukocyte adhesion is initiated by selectins 4575859.

Reduced venolar rolling in the Sepsis + TEA group might have been related to immune regulatory consequences of splanchnic sympathetic block. The technique of continuous TEA used in this study induced a sympathetic block including hepatic and intestinal sympathetic nerve roots as demonstrated by thermography 26. This block can be induced as long as 72 hours after catheter placement. There is some evidence of hepatic sympathetic immune regulation supporting this interpretation. Increased sympathetic activity in acute urinary retention results in hepatic inter-cellular adhesion molecule-1 expression 60. Increased portal norepinephrine in sepsis trigger hepatic release of TNF-α 1718, which is in turn a prerequisite for venolar temporary leukocyte adhesion 6162. Furthermore, TEA has already been shown to reduce temporary adhesion in mesenteric venoles in hemorrhagic shock 63. Therefore, the abdominal sympathetic block associated with TEA also might have reduced the venolar temporary leukocyte adhesion in sepsis.

Finally, intestinal injury and portal inflow of inflammatory mediators induce secondary hepatic inflammation 16465. Consequently, the decreased hepatic leukocyte adhesion may be related to the intestinal protection provided by TEA 293066. The systemic inflammatory response as measured by systemic release of TNF-alpha was not influenced in the Sepsis + TEA group.

Thoracic epidural infusion of local anesthetics is related to a segmental sensoric block, analgesia, and sympathetic block. Each of these aspects, as well as systemically resorbed bupivacaine might contribute to the observed effects of TEA on hepatic microvascular perfusion and leukocyte adhesion. However, the present study does not allow to distinguish or weight these potential mechanisms or to separate primarily hepatic effects from those secondary to intestinal effects of TEA.