Main

Recently, several lines of research have converged on the notion that sepsis alters vascular responsiveness to physiologic stimuli (1–5). In particular, sepsis has been postulated to alter the ability of the peripheral microcirculation to match blood flow to metabolic needs during periods of experimentally induced tissue hypoxia (6–9). We wondered whether this abnormal vascular response to hypoxia would be manifest on a larger scale. In particular, we wondered whether global systemic circulatory responses to hypoxia might be disrupted by sepsis.

To address this question, we studied the hemodynamic responses of piglets to acute hypoxia [induced by breathing 7% fraction of inspired oxygen (Fio2) for 7.5 min] before and during sepsis (induced by intravenous administration of live GBS). We found, as we had hypothesized, that piglets that under nonseptic conditions were capable of tolerating hypoxia (that is, maintaining BP and CO within normal ranges as Pao2 fell from ∼130 to ∼40 mm Hg) were no longer able to maintain stable BP and CO when the same hypoxic stimulus was induced during GBS infusion.

We next wondered what might underlie this sepsis-induced disruption of hemodynamic homeostasis during hypoxia. We tested three distinct hypotheses.

  • 1. GBS reduced systemic CO and QO2 below a critical value (7), at which point the superimposition of hypoxia became intolerable. To test this hypothesis, we reduced CO and QO2 in a matched group of nonseptic piglets by inflating an LAB and then superimposed the hypoxia stimulus.

  • 2. GBS inhibited nitric oxide (NO) synthesis (10–13), which is necessary for piglets to tolerate hypoxia. To test this hypothesis, we augmented NO synthesis in both septic and nonseptic piglets (by infusing an NO synthase substrate, l-ARG) and then superimposed the hypoxia stimulus. In addition, we inhibited NO synthesis in a matched group of nonseptic piglets (by infusing NNLA, a competitive inhibitor of NO synthase) and superimposed the hypoxia stimulus.

  • 3. GBS induced the production of pathologic prostaglandins (PG) (14, 15), which impaired the ability of piglets to tolerate hypoxia. To test this hypothesis, we inhibited PG synthesis by infusing a cyclooxygenase inhibitor, INDO, in both septic and nonseptic piglets and then superimposed the hypoxia stimulus.

We compared the physiologic effects of these interventions in an attempt to gain insight into the inability of piglets to tolerate hypoxia during GBS sepsis.

METHODS

Anesthesia and Ventilation

Piglets (2–3 wk old;n = 29) received ketamine intraperitoneally (20 mg/kg) and were intubated endotracheally. An ear vein was cannulated for venous access. The piglets were anesthetized with sodium pentobarbital (20 mg/kg i.v.), and muscle relaxation was achieved with d-tubocurarine (1 mg/kg). Throughout the experiment, the level of anesthesia was assessed by determining the response of heart rate and BP to noxious stimuli and was adjusted by intermittent bolus of pentobarbital, ∼2 mg·kg−1·h−1. Mechanical ventilation (Harvard Small Animal Ventilator, Harvard Medical Supplies, Dover, MA) was used with initial ventilator settings of Fio2 = 0.50, positive end-expiratory pressure (PEEP) = 4 cm H2O, tidal volume = 15 mL/kg, rate = 10 breaths per minute. Throughout the experiment, tidal volume was adjusted to maintain Pco2 in the range 35–40 mm Hg. Warming blankets and overhead heating lamps were used to maintain body core temperature at 37–38°C. A femoral artery catheter was placed to monitor BP, and a suprapubic cystostomy catheter was placed to establish urinary drainage.

Thoracic Surgery

Polyethylene catheters were surgically introduced into the RA via the left external jugular vein and into the pulmonary artery (PA) via left lateral sternotomy. In four piglets, an 8F Foley balloon catheter was placed into the left atrium through a small incision in the left atrial appendage. After our previously described methodology, an external electromagnetic flow probe (EP 425; Carolina Medical Electronics, King, NC) was placed around the PA proximal to the tip of the PA catheter. During each experiment, the PA flow probe was zeroed frequently using the approximation that diastolic blood flow in the PA is zero. This diastolic zero assumption has been shown to be appropriate when PA pressure is normal as well as during periods of elevated PA pressure (such as occurred during hypoxia or GBS sepsis). The possibility of vascular shunts through either the foramen ovale or ductus arteriosus was assessed by comparison of oxygen contents in blood samples obtained simultaneously from the RA, PA, and aorta at frequent intervals during each experiment. In the documented absence of vascular shunts between the systemic and pulmonary circulations, PA blood flow was taken to reflect total body CO. The PA and RA catheters were connected to pressure transducers to determine PAP and central venous pressure (CVP), respectively.

Blood specimens were taken from the femoral artery (arterial) and PA (mixed venous) at specified intervals throughout the experimental protocols (see below). From these samples, pH, Po2, Pco2, and base excess were determined using a blood gas analyzer (IL 1303, Instrumentation Laboratories, Lexington, MA), and Hb concentration and oxygen saturation were determined using a cooximeter (IL 282, Instrumentation Laboratories, Lexington, MA). This cooximeter had previously been calibrated for piglet blood in our laboratory by using a carbon monoxide scrubbing technique (16). QO2 and oxygen extraction were calculated in the standard manner.

All piglets were allowed to stabilize for approximately 1 h after surgery before the initiation of any experimental protocol.

Preparation of GBS Organisms

GBS organisms, serotype Ib (graciously provided by Dr. Lawrence Madoff, Channing Laboratories, Harvard Medical School), were incubated in Todd-Hewitt media overnight to late log phase (∼3 × 109 organisms per milliliter). Immediately before infusion into the piglet, the organisms were centrifuged (20 min at 3000 ×g), the supernatant discarded, and the pellet of organisms resuspended in 0.9% NaCl to one fourth its original volume. Quantitative assessments of the infusate concentrations were determined retrospectively by serial dilution.

Induction of Hypoxia

To measure the hemodynamic response to acute hypoxia, we devised a technique to produce extremely rapid alteration in Pao2 without simultaneously altering either Paco2 or mean airway pressure. Piglets were hyperventilated (40 breaths per minute; 25 mL/kg tidal volume; 1000 mL·kg−1·min−1 minute ventilation) with an inspired gas mixture of 5% CO2, 21% O2, 74% N2 to fix their Pao2 at ∼130 mm Hg and Paco2 at ∼40 mm Hg. The piglets were allowed to stabilize for 10 min on this ventilation protocol, during which time steady state was assured and an arterial blood gas sample was obtained. The inspired gas circuit was then instantly switched to 5% CO2, 7% O2, 88% N2 without any other change of ventilator settings. We have shown elsewhere (17) that, by use of this technique, the half-time of change of arterial Po2 from ∼130 to ∼40 mm Hg is 17 s, whereas Paco2 and mean airway pressure are unaffected. Continuous observations of CO, BP, PAP, and CVP during hypoxia were obtained for 7.5 min, a time previously determined by us as sufficient to achieve steady state when hypoxia is induced in nonseptic piglets. A second arterial blood gas sample was obtained at the completion of the hypoxia trial.

Experimental Protocols

Twenty-nine piglets were used in two or more experimental protocols, as described below.

Baseline/hypoxia.

For each of the 29 piglets, after postsurgical stabilization, baseline hemodynamic measurements and arterial blood gas samples were obtained. Hypoxia was then induced, and hemodynamic outcomes after hypoxia were observed.

GBS/hypoxia.

For nine piglets, after the baseline/hypoxia challenge, intravenous infusion of GBS (∼4 × 107 organisms per kg/min) was begun and continued sufficiently to reduce systemic CO to a level ∼60% of baseline values (cumulative dose of ∼1 × 109 organisms per kilogram; mean duration of GBS infusion, ∼2 h). Hemodynamic observations were determined continuously during GBS infusion, and blood gas samples were obtained at half-hour intervals. For each of the nine piglets, with GBS infusion continuing, hypoxia was again induced as described above, and the hemodynamic responses during hypoxia in septic piglets were observed.

GBS/INDO/hypoxia.

For five of the nine piglets that received the GBS/hypoxia protocol, after the hypoxia challenge during the GBS condition had been completed, the piglets were returned to normoxia with the GBS infusion continuing at the same rate. After 20 min of normoxia, 5 mg/kg INDO was administered intravenously over 10 min. This dose of INDO has previously been documented to eliminate PG synthesis for several hours in both septic and nonseptic piglets (14). Twenty minutes after the completion of the INDO infusion, with the GBS infusion continuing, hypoxia was again induced as described above.

GBS/l-ARG/hypoxia.

For the other four of nine piglets that received GBS, after the hypoxia challenge during the GBS condition had been completed, the piglets were returned to normoxia with the GBS infusion continuing at the same rate. After 20 min of normoxia, 300 mg/kg l-ARG was administered according to the following regimen: 100 mg/kg was infused via slow bolus over 5 min, followed by 10 mg·kg−1·min−1 × 20 min. This dose of l-ARG has been described previously by us to reverse the competitive effects of the NO synthase inhibitor NNLA when administered under similar experimental conditions (18). Twenty minutes after the completion of the l-ARG infusion, with the GBS infusion continuing, hypoxia was again induced as described above.

LAB/hypoxia.

For four piglets, after the baseline/hypoxia challenge, an LAB catheter was inflated in an attempt to induce stepwise reductions of systemic CO and QO2 that matched the reductions noted during the GBS protocol (approximately 60% of baseline over 2 h). After CO and QO2 had fallen to the desired levels, hypoxia was again induced as described above.

INDO/hypoxia.

For five piglets, after the baseline/hypoxia challenge, INDO (5 mg/kg) was infused over 10 min. Twenty minutes after the completion of the INDO infusion, hypoxia was again introduced as described above.

l-ARG/hypoxia.

For three piglets, after the baseline/hypoxia challenge, l-ARG (300 mg/kg) was infused over 30 min. Twenty minutes after completion of the l-ARG infusion, hypoxia was again introduced as described above.

NNLA/hypoxia.

For eight piglets, after the baseline/hypoxia challenge, NNLA was infused (10 mg/kg over 5 min followed by infusion at 1 mg·kg−1·min−1 continuously). This dose of NNLA has been shown by us and others to significantly impair NO synthesis under similar experimental conditions (18). Twenty minutes after completion of the bolus NNLA infusion, hypoxia was again introduced as described above.

NNLA/l-ARG/hypoxia.

For six of the eight piglets that received the NNLA/hypoxia protocol, after the hypoxia challenge during the NNLA condition had been completed, the piglets were returned to normoxia with the NNLA infusion continuing at the same rate. After 20 min of normoxia, 300 mg/kg of l-ARG was administered as described above. Ten minutes after the completion of the l-ARG infusion, hypoxia was again induced as described above.

In summary, all 29 piglets received an initial baseline/hypoxia trial. Next, each piglet received one of the following interventions: GBS (9), NNLA (8), LAB (4), INDO (5), or l-ARG (3) and a subsequent hypoxia trial. Finally, each of the nine GBS piglets received a second intervention [either GBS/INDO (5) or GBS/l-ARG (4)], and six of the eight NNLA piglets received a second intervention (NNLA/l-ARG) followed by a final hypoxia trial. Data from these 73 separate hypoxia trials were analyzed and are presented here.

Statistical Analysis

The experimental protocols were designed to compare responses to hypoxia under several experimental conditions. Consequently, two aspects of the response to hypoxia were determined under each condition:a) hemodynamic values immediately before the onset of hypoxia (i.e. values at 0 min of hypoxia), and b) hemodynamic values at the end of the hypoxia challenge (i.e. values at 7.5 min of hypoxia). For each analyzed variable, values at 0 and 7.5 min of hypoxia were compared using ANOVA. When indicated by ANOVA, subsequent pairwise comparisons were performed using the Newman-Keuls test.

For each experimental intervention (e.g. GBS, INDO), hemodynamic and blood gas values during normoxia (immediately before the onset of the hypoxia stimulus) were compared with the baseline normoxia values by using ANOVA with subsequent pairwise comparisons performed using the Newman-Keuls test.

Statistical significance was accepted at the p < 0.05 level.

RESULTS

Effects of GBS infusion on hemodynamics during hypoxia.

Figure 1 presents the central finding of this series of experiments. A beat-to-beat tracing of BP, PAP, CVP, and CO is displayed for a single piglet during GBS infusion before and after the onset of acute hypoxia. Within 90 s of the hypoxia stimulus, BP fell precipitously as the piglet became bradycardic and CO collapsed. This phenomenon was robust; in six of the nine GBS/hypoxia trials, the hypoxia stimulus was terminated before the 7.5-min challenge could be completed for fear that the piglet would die. In these trials, the hemodynamic values taken for subsequent data analysis were the last (and lowest) values recorded before the piglet was returned to normoxia. We recognize that in so doing, we have probably underestimated the effect of GBS on tolerance to hypoxia;e.g. the values of BP would probably have fallen even further had we persisted in the hypoxia trial. However, pilot studies had shown us that the piglets that reacted in this way to hypoxia during sepsis would indeed die were they not restored to normoxia.

Figure 1
figure 1

A beat-to-beat tracing of BP, PAP, CVP, and CO for a single piglet during GBS infusion before and after the onset of acute hypoxia (black arrow). Within 90 s of the hypoxia stimulus, BP fell precipitously as the piglet became bradycardic and CO collapsed. When the piglet was restored to normoxia (white arrow), both BP and CO improved.

Full size image

Figure 2 presents this phenomenon for nine piglets displaying normalized values of CO during hypoxia trials before and after GBS infusion. During the baseline/hypoxia protocol, as Pao2 fell from 132 ± 14 to 38 ± 8 mm Hg, CO remained stable [667 ± 72 (SD) to 685 ± 113 mL/min]. In contrast, during GBS/hypoxia, CO fell by almost half (397 ± 146 to 223 ± 142 mL/min;p < 0.001 versus ΔCO during baseline/hypoxia). In addition (not shown), during baseline/hypoxia, BP was well preserved (105 ± 9 to 97 ± 14 mm Hg), whereas, during GBS/hypoxia, BP fell markedly (94 ± 17 to 49 ± 25 mm Hg;p < 0.001 versus ΔBP during baseline/hypoxia). As expected, PAP rose significantly during baseline/hypoxia (from 18 ± 3 to 33 ± 7 mm Hg;p < 0.001). In contrast, PAP actually fell during GBS/hypoxia (42 ± 7 to 32 ± 7 mm Hg;p < 0.001 versus ΔPAP during baseline/hypoxia), attributable entirely to the concurrent drop in CO.

Figure 2
figure 2

Normalized values of CO during hypoxia trials before (baseline) and after GBS infusion for nine piglets. During the baseline/hypoxia protocol, CO remained stable. In contrast, during GBS/hypoxia, CO fell by more than half for this group of piglets.

Full size image

Interaction of INDO or l-ARG with GBS duringhypoxia.

Figure 3 displays the effects on CO of inhibition of PG synthesis (INDO) or augmentation of NO synthesis (l-ARG) during GBS/hypoxia protocols for the same piglets displayed in Figure 2. Two points are apparent. Despite continuing GBS infusion, INDO significantly improved the ability of the piglet to tolerate hypoxia. Each of the five piglets in the GBS/INDO/hypoxia protocol was able to endure the full 7.5-min hypoxia protocol with no significant decrement in CO (464 ± 100 to 550 ± 57 mL/min) despite the fact that in these same five piglets before INDO, CO had fallen significantly (394 ± 112 to 200 ± 165 mL/min;p < 0.001) and four of five GBS/hypoxia trials had been aborted for fear of imminent death. BP was also significantly more stable during GBS/INDO/hypoxia (119 ± 18 to 92 ± 17 mm Hg) than during GBS/hypoxia before INDO (97 ± 17 to 45 ± 22 mm Hg;p < 0.001).

Figure 3
figure 3

The effects of inhibition of PG synthesis (GBS/INDO;n = 5) or augmentation of NO synthesis (GBS/l-ARG;n = 4) on CO (normalized) during hypoxia protocols for the same septic piglets displayed in Figure 2. INDO significantly improved the ability of the piglet to tolerate hypoxia. In contrast, l-ARG did not significantly improve the ability of septic piglets to tolerate hypoxia.

Full size image

Figure 3 also reveals that, in contrast to the restorative effect of INDO, l-ARG did not improve the ability of GBS piglets to tolerate hypoxia. CO fell from 417 ± 82 to 200 ± 142 mL/min and BP from 81 ± 22 to 42 ± 29 mm Hg during the GBS/l-ARG/hypoxia protocols (both p < 0.001 versus GBS/INDO/l-ARG). Indeed, three of the four piglets receiving GBS/l-ARG tolerated hypoxia so poorly that the protocols were again curtailed before the designed 7.5 min.

Effects of LAB inflation, PG inhibition, NO augmentation, and NOinhibition during hypoxia in nonseptic piglets.

Figure 4 presents CO during hypoxia in three groups of nonseptic piglets: LAB, four piglets in which CO and QO2 had been reduced to levels equivalent to those noted during GBS infusion by using inflation of a LAB catheter; INDO, five piglets in which PG synthesis had been inhibited by INDO; and l-ARG, three piglets in which NO synthesis had been supplemented by pharmacologic doses of l-ARG. The baseline/hypoxia response of CO for these same 13 piglets are included for comparison. As Figure 4 reveals, none of these protocols significantly augmented or impaired the effects of hypoxia on CO in nonseptic piglets.

Figure 4
figure 4

CO (normalized) during hypoxia in three groups of nonseptic piglets: LAB, four piglets in which CO and QO2 had been reduced to levels equivalent to those noted during GBS infusion by using inflation of an LAB catheter; INDO, six piglets in which PG synthesis had been inhibited by INDO; l-ARG, three piglets in which NO synthesis had been supplemented by pharmacologic doses of the substrate l-ARG. The baseline response of these same 13 piglets to hypoxia is included for comparison. None of these protocols significantly augmented or impaired the effects of hypoxia on CO in nonseptic piglets.

Full size image

Figure 5 presents CO during hypoxia for eight nonseptic piglets in which endothelium-derived relaxation factor synthesis had been inhibited with NNLA. Pretreatment with NNLA dramatically impaired the ability of nonseptic piglets to tolerate hypoxia. Indeed, the fall of CO during the NNLA/hypoxia protocol (from 412 ± 105 to 285 ± 60 mL/min) resembled, at least qualitatively, the fall of CO during GBS/hypoxia protocols. Figure 5 also reveals that this NNLA-induced intolerance to hypoxia was completely reversed by l-ARG in the six piglets receiving the NNLA/l-ARG/hypoxia protocol.

Figure 5
figure 5

CO (normalized) during hypoxia for eight nonseptic piglets in which endothelium-derived relaxation factor synthesis had been inhibited with NNLA. Pretreatment with NNLA dramatically impaired the ability of piglets to tolerate hypoxia. This NNLA-induced intolerance to hypoxia was completely reversed by l-ARG (n = 6).

Full size image

Correlation of steady state hemodynamics and response tohypoxia.

Table 1 displays systemic and pulmonary hemodynamics and blood gas values immediately before hypoxia for each of the nine experimental protocols. No steady state hemodynamic or blood gas value was successful in predicting tolerance or nontolerance of subsequent hypoxia. There were no significant differences in final Pao2 (39 ± 6 mm Hg) or arterial oxygen saturation (47 ± 9%) among the hypoxia protocols.

Table 1 Hemodynamic and blood gas values before hypoxia * Denotes significantly different from baseline.
Full size table

DISCUSSION

The central hypothesis of this series of experiments was that GBS infusion would impair the previously existent capacity of piglets to tolerate acute hypoxia. This hypothesis was convincingly confirmed. Before the onset of GBS infusion, a reduction of Pao2 from ∼130 to ∼40 mm Hg was well tolerated in this experimental protocol: BP fell slightly (likely a response to hypoxia-induced peripheral vasodilation), CO rose slightly (presumably an attempt to offset the reduction in QO2 brought about by the fall in arterial oxygen content), and PAP rose significantly (the well-recognized hypoxia-induced pulmonary vasoconstriction). During GBS sepsis, these homeostatic responses to hypoxia were overwhelmed. The most remarkable aspect of this phenomenon was its severity; the hypoxia protocol induced complete circulatory collapse in nine of 13 trials during GBS and GBS/l-ARG infusion (as well as several more in pilot series that yielded no usable data because the animals died). To our knowledge, this is the first report of the striking interaction between GBS sepsis and superimposed hypoxia.

Mediators of intolerance to hypoxia.

What might mediate this phenomenon? PG inhibition with INDO completely restored the ability of septic piglets to tolerate hypoxia. Yet, INDO in nonseptic piglets had no impact on the hemodynamic consequences of hypoxia. We draw two conclusions from these observations. First, intact metabolism of PG is apparently unnecessary for nonseptic piglets to tolerate hypoxia, as INDO-treated nonseptic piglets tolerated hypoxia nicely. Second, abnormal PG synthesis, induced by GBS infusion and inhibited by INDO, must underlie the inability of septic piglets to tolerate imposition of hypoxia, as GBS/INDO piglets tolerated hypoxia that GBS piglets did not. Apparently, pathologic GBS-induced PG are centrally involved in this phenomenon, and administration of INDO to septic piglets completely reverses it.

In contrast, NO plays a very different role from PG in the tolerance of hypoxia. Even in nonseptic piglets, intact metabolism of NO appears necessary for tolerance to hypoxia; when NO synthesis was inhibited by NNLA in nonseptic piglets, hypoxia was not tolerated, and, when NO synthesis was restored in nonseptic piglets by NNLA/l-ARG, tolerance to hypoxia was restored as well.

In septic piglets, intolerance to hypoxia is almost certainly not mediated by reduction of NO synthesis because adding l-ARG did not restore tolerance to hypoxia during the GBS/l-ARG condition (as it did after inhibition of NO synthesis in nonseptic piglets during the NNLA/l-ARG condition). Moreover, we suggest that GBS-mediated elevation in NO synthesis is also highly unlikely to play any role in GBS-induced intolerance to hypoxia. We have previously demonstrated that much lower doses of NNLA than administered here are very poorly tolerated by septic piglets, even during normoxia (10). Superimposing hypoxia on a GBS/NNLA protocol is technically not possible; the piglets all die. In summary, intact metabolism of NO appears necessary but not sufficient to permit tolerance of hypoxia in septic piglets.

Mechanism of intolerance to hypoxia.

What is it that the pathologic PG, induced by GBS and inhibited by INDO, do to impair the piglet's capacity to tolerate hypoxia? Although we cannot be certain, our data provide some insight into the possibilities. We can effectively rule out the critical QO2 hypothesis that GBS lowers systemic CO and QO2 below a threshold value to the extent that hemodynamic homeostasis during superimposition of hypoxia is no longer possible (7). Piglets subjected to comparable reductions in CO and QO2 during the LAB protocol tolerated superimposed hypoxia without cardiovascular collapse.

What of the cor pulmonale of sepsis hypothesis? Might GBS-induced thromboxane elevate PAP or PVR to such a level that superimposition of hypoxia with attendant hypoxia-induced pulmonary vasospasm causes the right ventricle to fail acutely? Perhaps. The tendency for CVP to rise immediately before the collapse of CO (cf. the beat-to-beat tracing in Fig. 1) is consistent with this hypothesis. However, elevated PAP does not appear to be an absolute prerequisite for intolerance to hypoxia as NNLA-treated nonseptic piglets also did not tolerate hypoxia, yet their PAP values were not outside the range of the other protocols that did (although their PVR values were elevated). We were unable to explore the cor pulmonale hypothesis further, in large part because none of our other protocols could model the severe elevation of PAP or PVR induced by GBS.

An alternative hypothesis regarding the etiology of GBS-induced intolerance to hypoxia takes into account the vasoconstrictor effect of thromboxanes on the coronary circulation. This view begins with recognition that the heart is particularly sensitive to a combination of hypoxia and diminished vasodilator capacity. We and others have previously demonstrated in piglets that even under normal conditions, the heart extracts ∼70% of all delivered oxygen, approximately twice as much as the systemic circulation overall (19–21). This extraction rises to 80–90% during periods of low CO and QO2, as induced by either the GBS or LAB protocols here (20, 21). Consequently, in response to the added stress hypoxia, a further increase in oxygen extraction is not feasible, leaving the heart particularly dependent on coronary vasodilation to increase myocardial blood flow and preserve coronary QO2. According to this hypothesis, during the LAB protocol, compensatory coronary vasodilation remains possible. However, GBS-induced thromboxanes vasoconstrict the coronary arteries pathologically (22), precluding the necessary increase in coronary blood flow required during subsequent hypoxia. Myocardial ischemia and heart failure rapidly follow.

Completing the syllogism, NNLA also induces abnormal coronary vasoconstriction and precludes compensatory hypoxia-induced vasodilation normally mediated via endogenous NO (23, 24). Consequently, myocardial blood vessels are again unable to vasodilate sufficiently in response to the superimposed stress of hypoxia, and, during NNLA/hypoxia protocols, the heart becomes ischemic and fails. l-ARG restores endogenous NO synthesis during NNLA infusion sufficiently to permit coronary vasodilation and hemodynamic stability during the NNLA/l-ARG/hypoxia protocol.

Methodologic considerations: GBS infusion, PG production, and NOmodulation.

Pulmonary hypertension is the most consistent of all hemodynamic effects elicited by GBS infusion in experimental laboratory animals. PAP rises within minutes of the onset of GBS infusion and remains elevated throughout GBS infusion protocols. GBS-induced pulmonary hypertension is mediated by thromboxane; comparable elevations of PAP can be mimicked by infusion of the thromboxane analog U46619, and PAP can be reduced during GBS infusion either by administration of nonspecific PG inhibitors (INDO, ibuprofen) or specific thromboxane synthetase inhibitors (dazmegrel) (14, 15, 25–28). INDO at the dose used in these protocols has been shown to reduce thromboxane levels during GBS infusion to baseline levels, an effect which lasts for hours (considerably longer than required to complete the protocols reported here) (14, 15). Later during GBS infusion, other mediators (tumor necrosis factor, platelet activating factor, and IL) play more important roles and are particularly associated with the development of pulmonary edema (29–31). These concerns were diminished by the relatively short course of the protocols reported here.

NNLA has been shown to inhibit NO production in cultured endothelial cells and macrophages, an effect that could be overcome by administration of excess l-ARG (32–34). In an in vivo study, Fineman et al. (35) demonstrated that pretreatment with l-ARG (but not d-arginine) blocks the vasoconstrictive effects of NNLA in the systemic and pulmonary circulations of lambs. We subsequently extended these observations in piglets, demonstrating a dose-dependent hierarchical constriction of systemic, pulmonary, and cerebral circulations to NNLA and the reversal of this constriction by pharmacologic doses of l-ARG (11, 36). Finally, Fineman et al. (35) demonstrated in lambs that infusion of NNLA at doses comparable to those used in these protocols selectively attenuates the endothelium-dependent vasodilation associated with infusion of acetylcholine and ATP but has no effect on the endothelium-independent actions of isoproterenol and sodium nitroprusside. These observations all support the conclusion that NNLA competitively inhibits the synthesis of NO from l-ARG, both in vitro and in vivo.

Metabolic acidosis develops during GBS infusion. Might acidosis have played a role in the GBS-induced intolerance to hypoxia? Our data suggest this is not likely. The piglets in our longest GBS protocols (GBS/INDO/hypoxia, GBS/l-ARG/hypoxia) did develop a statistically significant degree of metabolic acidosis. However, GBS/hypoxia piglets were not significantly acidotic when they did not tolerate hypoxia. Moreover, there was no significant difference in metabolic acidosis between NNLA/l-ARG and GBS/INDO piglets, both of which tolerated hypoxia, and GBS/l-ARG piglets that did not.

The nonlinear effects of time during experimental sepsis protocols are important to acknowledge. At some GBS infusion rates (lower than those used here), piglets will remain relatively healthy for many hours with little hemodynamic or metabolic derangement. At rates considerably higher than those reported here, piglets will die within 2 h. We have previously published data that GBS infusion at the intermediate rates used in these protocols will maintain relatively stable (although reduced) CO and BP for up to 4 h in piglets, longer than any of these protocols lasted (8, 9). Moreover, the LAB protocol was explicitly designed to match both the extent and duration of reductions in CO and QO2 to the GBS protocol, yet LAB piglets tolerated hypoxia.

In summary, we have demonstrated that piglets that were previously able to tolerate the reduction of Pao2 from ∼130 to ∼40 mm Hg were unable to do so during GBS infusion. Septic piglets experiencing hypoxia became hypotensive, hypoperfused, and often approached cardiac arrest. The mechanism of this GBS-induced intolerance to hypoxia remains incompletely elucidated, but the phenomenon can be reversed by PG inhibition with INDO although not by NO augmentation with l-ARG. Further evaluation of this phenomenon may provide insight into the incompletely understood mechanism by which sepsis induces circulatory collapse in experimental animals and in humans.