Hypometabolism and hypothermia in the rat model of endotoxic shock: independence of circulatory hypoxia

Abstract

We tested the hypothesis that development of hypothermia instead of fever in endotoxic shock is consequential to hypoxia. Endotoxic shock was induced by bacterial lipopolysaccharide (LPS, 500 μg kg⁻¹ i.v.) in rats at an ambient temperature of 22°C. A β₃-adrenergic agonist known to activate metabolic heat production, CL316,243, was employed to evaluate whether thermogenic capacity could be impaired by the fall in oxygen delivery () during endotoxic shock. This possibility was rejected as CL316,243 (0.15 mg kg⁻¹ i.v.) evoked similar rises in oxygen consumption () in the presence and absence of endotoxic shock. Next, to investigate whether a less severe form of circulatory hypoxia could be triggering hypothermia, the circulating volume of LPS-injected rats was expanded using 6% hetastarch with the intention of improving tissue perfusion and alleviating hypoxia. This intervention attenuated not only the fall in arterial pressure induced by LPS, but also the associated falls in and body temperature. These effects, however, occurred independently of hypoxia, as they were not accompanied by any detectable changes in NAD⁺/NADH ratios. Further experimentation revealed that even the earliest drops in cardiac output and during endotoxic shock did not precede the reduction in that brings about hypothermia. In fact, and fell in such a synchrony that the / ratio remained unaffected. Only when hypothermia was prevented by exposure to a warm environment (30°C) did an imbalance in the / ratio become evident, and such an imbalance was associated with reductions in the renal and hypothalamic NAD⁺/NADH ratios. In conclusion, hypometabolism and hypothermia in endotoxic shock are not consequential to hypoxia but serve as a pre-emptive strategy to avoid hypoxia in this model.

Key points

• The hypometabolic, hypothermic response that often replaces fever in endotoxic shock might be consequential to hypoxia, but the available evidence is circumstantial.

• Here, this hypothesis was tested in an unprecedented experimental preparation that provides simultaneous measurements of oxygen delivery and consumption in rats whose ability to regulate body temperature has not been disrupted by anaesthetics.

• The results are striking for indicating that hypometabolism and hypothermia in endotoxic shock occur independently of global or tissue hypoxia.

• The results also demonstrate that a switch in thermal response from fever to hypothermia serves as a pre-emptive strategy to avoid hypoxia in endotoxic shock.

• These findings are potentially relevant to critical care, as interventions aimed at elevating oxygen delivery in patients with septic shock are based on the (perhaps erroneous) premise that circulatory hypoxia is the cause of hypometabolism in this patient population.

Introduction

A switch in thermal state from fever to hypothermia is associated with severe forms of sepsis and related systemic inflammatory syndromes (Clemmer 1992; Arons et al. 1999; Peres Bota et al. 2004). A similar phenomenon occurs in animal models of systemic inflammation, and has been best characterized in rats and mice challenged with bacterial lipopolysaccharide (LPS, also known as endotoxin); for review, see Romanovsky et al. (2005). However, the crucial event responsible for the fever–hypothermia switch in systemic inflammation remains unclear. While it is possible that the balance between fever and hypothermia is an inherent property of immune-to-brain signalling (Szekely & Romanovsky, 1998; Steiner & Romanovsky, 2007), it is equally possible that hypothermia is not a response to inflammation itself, but rather a reflection of the homeostatic disturbances that occur when the inflammatory reaction is too strong. In this context, circulatory shock deserves attention because of its frequent association with hypothermia in experimental endotoxaemia (Lang et al. 1985; Alexander et al. 1991; Romanovsky et al. 1996; Giusti-Paiva et al. 2002; Liu et al. 2012) as well as in clinical cases of sepsis (Clemmer et al. 1992; Arons et al. 1999; Peres Bota et al. 2004).

In resemblance to septic shock, endotoxic shock is characterized by a fall in arterial pressure that, at least in some occasions, is associated with a fall in cardiac output (CO) and hypoxia (Lang et al. 1985; Brackett et al. 1997; Miura et al. 2000; Levy et al. 2003). In theory, this state of hypoperfusion could promote a decrease in body temperature (T_b) by at least two mechanisms. The first mechanism is related to the fact that, to maintain aerobic metabolism, any tissue needs to receive a minimal, critical amount of oxygen. As oxygen delivery () is reduced below this critical limit, the rate of oxygen consumption () falls uncontrollably and becomes linearly dependent on (Torres-Filho et al. 2005). Such a supply-dependent metabolic failure would unavoidably impair thermogenic capacity, which might account for the suppression in thermogenesis known to drive hypothermia in endotoxic shock (Derijk et al. 1994; Romanovsky et al. 1996). The second mechanism by which circulatory shock could trigger hypothermia is related to the brain-driven, regulated suppression of thermogenesis that can be induced by less severe forms of hypoxia, in which the extent of the fall in is not drastic enough to force aerobic metabolism into a supply-dependent state (Saiki & Mortola, 1997; Tattersall & Milsom, 2009). Importantly, animals made febrile by a low dose of LPS retain their ability to develop hypothermia in response to hypoxia (Doherty & Blatteis, 1980). Furthermore, suppression of thermogenesis in response to hypoxia appears to account for the early hypothermia seen in another model of shock, namely, haemorrhagic shock (Haouzi & Van de Louw, 2013).

Despite this body of circumstantial evidence, the hypothesis that circulatory hypoxia is the trigger of the hypometabolic, hypothermic response in severe systemic inflammation remains untested. In the present study, two key corollaries of this hypothesis were tested in conscious rats subjected to endotoxic shock. First, we tested the notion that in case is insufficient to support aerobic metabolism during endotoxic shock, drug-induced increases in metabolic demand would not be met to result in full-blown rises. The drug of choice for this purpose was CL316,243, a β₃-adrenergic agonist known to raise by activating brown fat thermogenesis (Cannon & Nedergaard, 2004). Second, we sought to evaluate whether LPS-induced hypometabolism and hypothermia could be prevented or attenuated by an intervention that improves haemodynamics and, consequently, alleviates hypoxia during endotoxic shock. Circulating volume expansion by a combination of a colloid dispersion (6% hetastarch) and an isotonic electrolyte solution (saline) was the intervention of choice for its well-documented haemodynamic effects and metabolic inertness (Holbeck & Grande, 2002; Marx et al. 2002). Tissue hypoxia was evaluated by probing for decreases in NAD⁺/NADH ratios, whereas global circulatory hypoxia was assessed by probing for reductions in CO and relative to .

Methods

Ethical approval

Protocols involving animals were approved by Animal Care and Use Committees at the Albany College of Pharmacy and Health Sciences (Albany, NY, USA) and at the Institute of Biomedical Sciences (São Paulo, SP, Brazil). They complied with the US Guide for the Care and Use of Laboratory Animals (2011), the resolutions of the Brazilian Council for the Control of Animal Experimentation and the UK regulations on Animal Experimentation (Drummond, 2009).

Animals

The study was conducted in 159 male Wistar rats. The rats were obtained from Charles River Laboratories (Wilmington, MA, USA), with the exception of the subset destined to CO and measurements, which originated from the breeding facility of the Institute of Biomedical Sciences (University of São Paulo, São Paulo, SP, Brazil). Regardless of their source, the rats were caged under virtually identical conditions: they were grouped two to three per cage before surgery and singly after surgery; had free access to standard chow and filtered water; were under a 12:12 h light–dark cycle (lights on at 07.00 h); and were exposed to a room temperature of 23–26°C. The rats weighed an average of 290 g (190–360 g) at the time of the experiments. Each rat was used in an experiment once, and then killed with sodium pentobarbital (100 mg kg⁻¹ i.v.).

Surgical preparation

In preparation for an experiment, a rat was chronically implanted, as necessary, with one or more of the following: an intravenous (i.v.) catheter (later used for drug or fluid administration); an intra-arterial (i.a.) catheter (later used for arterial pressure measurement or blood collection); an abdominal telemetry transponder (later used for T_b measurement); and an ascending aorta flow probe (later used for CO measurement). The surgical procedures were performed under anaesthesia with ketamine–xylazine–acepromazine (80:8:1 mg kg⁻¹ i.p.), except for the implantation of the ascending aorta flow probe, which was done under anaesthesia with isoflurane (2%). All rats received a prophylactic dose of enrofloxacin (5 mg kg⁻¹ s.c.) and were maintained on a heated (37–39°C) operating board for the duration of the surgery.

The venous catheter was inserted into the left jugular vein, and advanced so its tip lay on the right atrium. The arterial catheter was inserted into the left carotid artery, and advanced so its tip lay on the descending aorta. Both catheters consisted of 3-Fr polyurethane tubing. They were secured in place by occlusive ligatures, exteriorized at the nape and locked with heparinized glycerol (500 U ml⁻¹). The telemetry transponder (G2 E-Mitter; Mini Mitter, Bend, OR, USA) was implanted via a midline laparotomy and sutured to the dorsolateral abdominal wall. The incision sites were closed in layers.

For implantation of the ascending aorta flow probe, the rats were tracheally intubated and mechanically ventilated (5 ml breath⁻¹; 65–75 breaths min⁻¹). After the thoracic cavity was accessed by a median sternotomy, the aortic arch was isolated and cleared from connective tissue by blunt dissection. An ultrasonic transit time flow probe (precision S-series; Transonic Systems, Ithaca, NY, USA) was then fitted around the ascending aorta, with its J-reflector facing away from the pulmonary artery. The probe's cable was passed through the sternal incision and tunnelled under the skin to the nape. Its four-pin connector was fixed to the skull with the help of a rigid cuff, microscrews and acrylic cement. The sternum was closed with stainless steel wires and the chest's negative pressure was restored through a pinhole sealed with a purse-string suture in the pectoralis minor muscle. The incision sites in muscle and skin were then sutured.

At the end of surgery, each rat received acetaminophen (150 mg kg⁻¹ s.c.) or carprofen (5 mg kg⁻¹ s.c.) for pain management. Rats subjected to the thoracic surgery received additional doses of carprofen on the 2 days postsurgery. The rats were allowed to recover for 7–10 days before an experiment. During the recovery period, the catheters were flushed with saline and re-locked with heparinized glycerol on the first day postsurgery and every third day thereafter.

Experimental setup

Between 07.00 and 08.00 h on the day of the experiment, each rat was placed inside a 5 litre open-flow respirometry box, which in turn was kept inside an environmental chamber (model NQ1; Environmental Growth Chambers, Chagrin Falls, OH, USA). The respirometry box was sealed, except for a 2.5 cm opening at the top and for three 6 mm hose connectors on the sides, from which air was pulled at a rate of ∼1800 ml min⁻¹ with the help of an adjustable pump coupled to a mass-flow controller (SS4 Gas Subsampler; Sable Systems, Las Vegas, NV, USA). The air entering and exiting the respirometry box was sampled continuously at 1 or 5 min intervals (depending on the experiment), and the fraction of oxygen in each sample was measured by a FOXBOX gas analyser (Sable Systems). The analog outputs of the mass flow controller and of the gas analyser were converted to digital by the UI-2 data acquisition interface, and acquired in a computer with the help of the Expedata software (Sable Systems); was estimated from these data (see ‘Data processing and analyses’). Beneath each respirometry box was an ER-4000 receiver (Mini Mitter), which captured the radio frequency of the pre-implanted telemetry transponder and conveyed it to a computer, where T_b signals were recorded and processed by the Vital View software. Saline-filled PE-50 extensions of the arterial and venous catheters were passed through the opening at the top of the respirometry box, connected to a two-channel swivel (Instech Laboratories, Plymouth Meeting, PA, USA), and then passed to the outside of the environmental chamber via a side port. An infusion harness worn by the rat and a spring coil protected the extensions from bites and scratches. The arterial catheter extension was connected to a differential pressure transducer, which conveyed pulsatile arterial pressure data to a computer via the Datamax analog-to-digital interface (Columbus Instruments, Columbus, OH, USA). The arterial catheters were also employed for collecting blood for the gasometry analysis (see below). The venous catheter extension was connected to a syringe filled with saline, and was employed for administering the drugs of interest as well as for anaesthetizing the rat immediately before tissue harvesting (see below).

When a rat was designated for CO measurements, an electrical swivel was fitted to the opening at the top of the respirometry box. This electrical swivel had a hollow shaft and did not impede the passage of the catheter extensions to the outside of the box. The electrical swivel served as a mobile relay for the extension cable connecting the flow-probe connector at the head of the rat to a TS420 perivascular flowmeter module (Transonic Systems). The analog output of the flowmeter module was converted to digital by the UI-2 data acquisition interface and acquired in a computer with the help of the Expedata software (Sable Systems).

The experimental interventions were initiated only after the physiological parameters indicated that the rats were well-habituated to the experimental conditions, usually about 12.00 h.

Drug administration

Phenol-extracted LPS from Escherichia coli 055:B5 was purchased from Sigma-Aldrich (St Louis, MO, USA). CL316,243 was purchased from Tocris Bioscience (Ellisville, MO, USA). Hetastarch (6% hydroxyethyl starch 670/0.75) and pyrogen-free isotonic saline were obtained from Hospira (Lake Forrest, IL, USA). LPS and CL316,243 were diluted in saline to the desired concentration (500 μg ml⁻¹ for LPS and 0.0625–1.0 mg ml⁻¹ for CL316,243), and administered in bolus over a volume of 1 ml kg⁻¹ via the pre-implanted i.v. catheters. The circulating volume expansion protocol consisted of a 1 h infusion of the hetastarch dispersion followed by a 2.5 h infusion of saline, both at a rate of 3 ml h⁻¹ via the pre-implanted i.v. catheters. This infusion rate was chosen based on pilot experiments aimed at minimizing the acute effects of circulating volume expansion on cardiovascular parameters while maintaining its effects on endotoxic shock. All preparations were brought to room temperature before administration.

Tissue harvesting and assays

An i.v. bolus injection of ketamine-xylazine (15:1.5 mg kg⁻¹) was employed to anaesthetize the rats immediately before specimen collection. The hypothalamus, left kidney and central lobe of the liver were collected, snap-frozen in liquid nitrogen and stored at −80°C for no longer than 3 months. Tissue hypoxia in these samples was assessed based on the NAD⁺/NADH ratio, which is known to fall during hypoxia because of the diminished oxidation of NADH to NAD⁺ in the respiratory chain (Mayevsky & Rogatsky, 2007). In addition to organ samples, blood samples were collected from the inferior vena cava into EDTA-coated tubes for determination of the plasma concentrations of total protein and tumour necrosis factor (TNF)-α. Like organ samples, plasma samples were stored at −80°C and used within 3 months.

The tissue contents of NAD⁺ and NADH were determined by a cycling enzyme method from BioVision (Montain View, CA, USA). In brief, the tissue samples were sonically disrupted in the extraction buffer at a ratio of 10 mg of tissue per 1 ml of buffer. The sample homogenate was cleared by centrifugation (12,000 g, 4°C, 20 min), and the supernatant was purified by passage through a 10 kDa spin column. The purified supernatant was then split into two fractions: one was heated (60°C, 30 min) to decompose NAD⁺, and the other remained undisturbed at 4°C and retained NAD⁺. Next, each fraction was reacted with a cycling enzyme mixture that converted all NAD⁺ to NADH and, in sequence, subjected to NADH quantification according to manufacturer's instructions. The coefficient of determination (r²) for the NADH standard curves (0.4–4.0 nmol ml⁻¹) ranged from 0.991 to 0.999 (linear fitting). The final NADH concentration in the NAD⁺-decomposed fraction reflected the actual content of NADH in the tissue, whereas the final NADH concentration in the NAD⁺-bearing fraction reflected the sum of the NAD⁺ and NADH contents. The NAD⁺ content was calculated from the difference between the two fractions.

In the plasma samples, total protein concentration was measured by the Bradford method using bovine serum albumin as standard (reagents from Cayman Chemical, Ann Arbor, MI, USA), whereas TNF-α was measured by sandwich ELISA using a kit from Thermo Scientific (manufactured by Pierce Biotechnology, Rockford, IL, USA). For both assays, endpoint absorbance was monitored on a Synergy 2 microplate reader and analysed using the Gen5 software (Biotek, Winooski, VT, USA). Samples were run in duplicate, simultaneously with a calibration curve (r² of 0.995 for the protein assay; r² of 0.999 for the TNF-α assay; both by four-parameter fitting).

Gasometry

Arterial partial pressure of oxygen (), arterial haemoglobin-O₂ saturation () and haemoglobin concentration (CH_b) were measured for the estimation of arterial oxygen content (). These measurements were made using the iSTAT-1 portable gas analyser and EG7⁺ cartridges (Abbott Point of Care, Abbott Park, IL, USA). The analysis required less than 100 μl of blood, which was dripped directly from the carotid catheter extension into the collection chamber of the EG7⁺ cartridge.

Data processing and analyses

T_b was sampled every 1 min using the Vital View software (Mini Mitter). Mean arterial pressure (MAP) and heart rate (HR) were computed every 10 s from the pulsatile arterial pressure data, whereas CO and stroke volume (SV) were determined every 60 s from the pulsatile aortic flow data, both acquired at a sampling rate of 200 Hz. was calculated from the oxygen fraction differential between the air entering () and the air exiting () the respirometry box, taking into consideration the actual air flow at the time of each measurement. The following formula was employed: = air flow × ( − )/[1 − ×(1 − respiratory exchange ratio)]/rat mass, where the respiratory exchange ratio was considered to be 0.71. was calculated as the product of CO and . The latter was in turn calculated from the following equation: = (1.36 × CH_b × /100) + (0.0031 × ).

A dose–response analysis for the effects of CL316,243 on was performed by the variable Hill slope method using OriginPro 9.0 (OriginLab, Northampton, MA, USA). Statistical comparisons were performed using Statistica Advanced 8.0 (StatSoft, Tulsa, OK, USA), with the level of significance set at P < 0.05. The T_b, , MAP, HR, CO and SV data were pooled over 1 min or 10 min intervals (depending on the experiment), and these parameters were evaluated for the effects of CL316,243, circulating volume expansion, and LPS over time by repeated-measures ANOVA. Blood gases, and / ratios were also evaluated by repeated-measures ANOVA. The NAD⁺/NADH ratios for each tissue, as well as the levels of total protein and TNF-α in the plasma, were evaluated for the effects of circulating volume expansion and LPS by two-way ANOVA or for the effects of LPS only by Student's t test. The post hoc Fisher least significant difference test was used when necessary.

Results

Experiment 1: Thermogenic capacity is not impaired during lipopolysaccharide-induced hypothermia

Experiments 1–4 were performed at an ambient temperature (22°C) that is subneutral and adequate to reveal the hypothermic response to LPS (Krall et al. 2010; Liu et al. 2012; Al-Saffar et al. 2013). In this thermal environment, CL316,243 (β₃-receptor agonist) promoted dose-dependent rises in over a relatively narrow dose range: the dose of 0.06 mg kg⁻¹ caused minimal or no change in ; the dose of 0.12 mg kg⁻¹ caused a half-maximal increase in ; and doses of 0.15 mg kg⁻¹ or higher caused maximal increases in (Fig. 1A). The peak of the response occurred 20 min after the injection of CL316,243 at the dose of 0.15 mg kg⁻¹ (Fig. 1B). This dose was selected for the thermogenic capacity test in LPS-injected rats. In this test, rats were injected with LPS at a shock-inducing dose (500 μg kg⁻¹ i.v.) or with its vehicle (saline), and, 30 min later, received CL316,243 or its vehicle (saline). This treatment schedule was chosen so that the peak of the metabolic response to CL316,243 would occur at a time when LPS was expected to be driving and T_b downwards. Indeed, as shown in Fig. 2A, both and T_b were declining at the expected time (50 min) post-LPS in the group that did not receive CL316,243. Such declines were statistically significant (F_15,225 = 1.99, P < 0.017 for VO₂; F_15,225 = 2.86, P < 0.001 for T_b). The decrease in preceded the decrease in T_b, as revealed by a phase-plane plot (Fig. 2B). CL316,243 produced significant elevations in not only in the absence of LPS (F_15,225 = 2.37, P = 0.004), but also during the course of the response to LPS (F_15,180 = 3.60, P < 0.001; Fig. 2A). An analysis of the intergroup differences in further demonstrated that the magnitude of the hypermetabolic effect of CL316,243 was comparable in LPS-injected rats and their saline-injected counterparts (Fig. 2C). Intriguingly, though, the hypermetabolic effect of CL316,243 did not affect T_b during the course of LPS hypothermia, even though the same effect was associated with a T_b rise in the rats not injected with LPS (F_15,225 = 3.40, P < 0.001; Fig. 2A). Presumably, heat loss effectors played a role in defending the T_b of the LPS-injected rats from the thermogenic effect of CL316,243.

Figure 1. Determination of the lowest dose of CL316,243 capable of maximally activating aerobic metabolism in rats not injected with lipopolysaccharide.

The rats were maintained at an ambient temperature of 22°C. A, peak changes in O₂ consumption induced by different i.v. doses of CL316,243. Data points are means ± S.E.M. of two to three rats per dose (17 rats in total). The data were fitted by the variable Hill slope method. B, typical recording from an individual rat in which CL316,243 induced a maximal rise in O₂ consumption.

Figure 2. The hypermetabolic effect of CL316,243 is preserved in rats injected with a shock-inducing dose of LPS.

In this experiment, rats kept at an ambient temperature of 22°C received two i.v. injections: the first injection consisted of LPS (500 μg kg⁻¹) or its vehicle (saline); the second injection consisted of CL316,243 (0.15 mg kg⁻¹) or its vehicle (saline). A, mean (±S.E.M.) values for O₂ consumption and deep body temperature over the course of the experiment. The times of injections are indicated. B, phase-plane plot of the relationship between deep body temperature and O₂ consumption in rats that received LPS followed by saline. C, intergroup difference in O₂ consumption between the groups injected with CL316,243 and their corresponding saline-injected counterparts. The number of animals in each group (n) is indicated. LPS, lipopolysaccharide.

Experiment 2: Attenuation of lipopolysaccharide-induced hypotension by circulating volume expansion is associated with suppression of lipopolysaccharide-induced hypothermia

The protocol for circulating volume expansion involved an infusion rate (3 ml h⁻¹) of hetastarch and saline that was low enough to avoid marked alterations in the resting levels of MAP, HR, and T_b (Fig. 3). Only a discrete increase in MAP (F_24,120 = 2.17, P = 0.003) and a moderate decrease in (F_24,72 = 1.79, P = 0.030) were observed. Both of these effects were transient and did not last for more than 20 min into the second step (saline) of the infusion protocol. The same protocol, however, appears to have promoted a more sustained elevation in CO lasting at least 120 min in three rats tested independently (CO peaked 21.8 ± 8.7% above baseline in this test). The effects of circulating volume expansion were then studied in LPS-challenged rats; LPS or its vehicle (saline) was injected at the changeover from the hetastarch infusion to the saline infusion. As shown in Fig. 4A, rats subjected to the volume expansion protocol responded to LPS with a less pronounced fall in MAP in comparison with rats not subjected to the volume expansion protocol (F_15,135 = 4.12, P < 0.001). Most importantly, their hypothermic response was similarly attenuated, as evidenced by less pronounced decreases in both (F_15,180 = 3.40, P < 0.001) and T_b (F_15,180 = 2.81, P < 0.001). These attenuated responses were observed while the tachycardic response to LPS remained unaffected (Fig. 4A). The locomotor activity (measured alongside T_b by the telemetry system) of the LPS-injected rats was also unaffected by the volume expansion protocol (data not shown).

Figure 3. The circulating volume expansion protocol employed in this study exerts minor cardiovascular and thermoregulatory effects of its own.

At an ambient temperature of 22°C, rats were infused with 6% hetastarch followed by isotonic saline, both at a rate of 3 ml h⁻¹. Data are shown as means ± S.E.M. The number of animals in each group (n) is indicated.

Figure 4. Circulating volume expansion with hetastarch and saline attenuates LPS-induced hypotension, hypometabolism and hypothermia, but such attenuation occurs independently of tissue hypoxia.

The experiment was performed at an ambient temperature of 22°C. In the circulating volume expansion group, LPS (500 μg kg⁻¹ i.v.) was bolus-injected at the transition from the 1 h hetastarch infusion to the 2.5 h saline infusion. Control rats received LPS in the absence of circulating volume expansion. The time courses of the cardiovascular, metabolic and thermoregulatory responses are shown in (A). As the volume expansion protocol exerted its greatest effect on the thermoregulatory response at 60 min post-LPS (see intergroup difference and derivative in B), this time point was selected for assessment of tissue hypoxia based on NAD⁺/NADH ratios. The results of this assessment are shown in (C). Data are means ± S.E.M. The number of animals per group (n) is indicated next to each time course curve (A) or in parenthesis (C). LPS, lipopolysaccharide.

Experiment 3: Attenuation of the hypothermic response to lipopolysaccharide by the volume expansion protocol is unrelated to tissue hypoxia

To verify whether an attenuation of tissue hypoxia could account for the effect of the circulating volume expansion protocol on LPS-induced hypothermia, tissue samples were analysed with respect to their NAD⁺/NADH ratios at 60 min after injection of LPS or its vehicle (saline). This time point corresponded to the maximal rate of separation between the hypothermic responses of those LPS-injected rats that were subjected to the volume expansion protocol and those that were not (Fig. 4B). To our surprise, though, no detectable decrease in the NAD⁺/NADH ratio was observed in the LPS-injected rats in relation to the saline-injected controls in any of the organs evaluated: hypothalamus, kidney and liver (Fig. 4C). This was the case regardless of the presence or absence of circulating volume expansion. Differences in NAD⁺/NADH ratios were noted across tissues, but such tissue-specific differences are normal and presumably reflect oxygen-unrelated differences in cellular biochemistry (Braidy et al. 2011).

In an effort to reveal possible hypoxia-unrelated, haemodilution-related effects of circulating volume expansion, the levels of TNF-α and total protein were measured in plasma samples collected concomitantly with the tissues designated for NAD⁺/NADH analysis. As presented in Table 1, the volume expansion protocol was associated with a significant reduction in the total protein concentration (F_1,26 = 108.59, P < 0.001), with the magnitude of the reduction being more pronounced in the LPS-challenged rats than in their saline-injected counterparts (44% versus 25%; P < 0.001 in post hoc analysis). In spite of that, the volume expansion protocol did not reduce the surge in plasma TNF-α induced by LPS (Table 1). On the contrary, by correcting the TNF-α level for the dilution in total plasma protein, it is possible to estimate that TNF-α production could be increased by ∼15% in the LPS-treated rats subjected to the circulating volume expansion protocol. It should be noted, though, that this is only a rough approximation based on three assumptions: (i) that hetastarch increases plasma volume without affecting interstitial volume; (ii) that TNF-α is evenly distributed in the interstitial and plasma compartments; and (iii) that the catabolism of TNF-α was not affected by the volume expansion protocol.

Table 1.

Effects of circulating volume expansion and LPS (500 μg kg⁻¹ i.v.) on the plasma levels of total protein and TNF-α

Experimental intervention		Total protein	TNF-α
Vol. expansion	LPS injection	(mg ml−1)	(ng ml−1)
No	No	65.8 ± 3.4	n.d.
		n = 7
Yes	No	49.3 ± 1.4*	n.d.
		n = 7
No	Yes	66.9 ± 2.7	16.5 ± 2.1*
		n = 8	n = 8
Yes	Yes	37.5 ± 1.4*,†	15.1 ± 1.5*
		n = 8	n = 8

Data are means ± S.E.M. Abbreviations: LPS, lipopolysaccharide; n.d., non-detectable levels; TNF, tumour necrosis factor. Plasma samples were collected 60 min after the bolus injection of LPS or its vehicle (saline). The rats were maintained at an ambient temperature of 22°C. The number of animals in each group (n) is indicated.

^*Statistically different from the group subjected to neither volume expansion nor LPS injection.

^†Statistically different from the group subjected to volume expansion but not injected with LPS.

Experiment 4: Lipopolysaccharide-induced hypothermia is triggered independently of any mismatch in the / relationship

To examine in more detail the possible lack of hypoxia at the onset of LPS-induced hypothermia, we performed a detailed analysis of the interplay among CO, and . The first step in the analysis was a minute-to-minute assessment of the changes in CO and . As in experiments 1–3, the rats were exposed to an ambient temperature (22°C) that is adequate to reveal the hypothermic response. LPS (500 μg kg⁻¹) induced a biphasic decrease in CO (Fig. 5A), with both phases being brought about by reductions in SV (data not shown). The decrease in CO was statistically significant from 5 min post-LPS to 150 min post-LPS, when the experiment was ended (F_37,185 = 11.35, P < 0.001). Despite its early onset, the CO response did not precede the LPS-induced reduction in , which had an equally early onset in this experiment (Fig. 5A). In fact, the CO and responses developed in remarkable synchrony until 40 min post-LPS, after which started to fall at a higher rate than CO (Fig. 5B).

Figure 5. There is no mismatch between O₂ delivery and O₂ consumption during the period corresponding to the onset of LPS-induced hypothermia.

As in the previous experiments, LPS (500 μg kg⁻¹ i.v.) was injected in rats kept at an ambient temperature of 22°C. A, minute-to-minute changes in cardiac output and O₂ consumption. B, phase-plane plot of the data shown in A. C, O₂ delivery and O₂ consumption at baseline and at 20 and 60 min post-LPS. The ratio between O₂ delivery and O₂ consumption at these time points is shown in (D). Data are shown as means ± S.E.M. The number of animals per group (n) is indicated next to each time course curve (A) or in parenthesis (C and D). *Statistically different from baseline. LPS, lipopolysaccharide.

The next step in the analysis was the determination of and at two crucial time points of the responses to LPS: 20 min (corresponds to the nadir of the initial drops in CO and ) and 60 min (corresponds to the time NAD⁺/NADH ratios were determined in experiment 3). Compared to the baseline (−60 min) period, LPS led to increases in both (F_2,10 = 9.57, P = 0.005) and (F_2,10 = 5.59, P = 0.023) at 20 and 60 min postinjection (Table 2), presumably reflecting tachypnoea without lung dysfunction at these time points. Such increases, however, were not robust enough to affect . Consequently, the abrupt reduction in CO at 20 min post-LPS was sufficient to drive downwards (P = 0.019 in post hoc analysis; Fig. 5C). However, this early reduction in was completely matched by a reduction in (Fig. 5C) to the extent that the / ratio remained unaffected (Fig. 5D).

Table 2.

Determinants of before and after injection of LPS (500 μg kg⁻¹ i.v.) in rats exposed to an ambient temperature (22°C) that allows the development of hypothermia

Time	CHb (g dl−1)	(%)	(Torr)	(%)
Baseline	12.0 ± 1.2	95 ± 1	75 ± 4	16 ± 1
20 min post-LPS	12.1 ± 1.2	97 ± 0*	90 ± 2*	16 ± 2
60 min post-LPS	11.5 ± 1.3	97 ± 0*	86 ± 2*	15 ± 2

Results are means ± S.E.M. (n = 6). Abbreviations: , arterial oxygen content; CH_b, haemoglobin concentration; LPS, lipopolysaccharide; , arterial partial pressure of oxygen; , arterial haemoglobin-O₂ saturation.

^*Statistically different from the baseline period.

Experiment 5: Prevention of hypothermia reveals a mismatch in the / relationship and tissue hypoxia in endotoxic shock

Intrigued by the lack of hypoxia when endotoxic shock was associated with hypothermia, we asked whether the hypothermic response itself could be protecting the LPS-injected rats from hypoxia. As in previous studies from our laboratory (Krall et al. 2010; Liu et al. 2012), exposure to a warm environment (30°C) was employed to prevent the development of hypothermia, a strategy based on the rationale that the thermogenic inhibition known to drive hypothermia in systemic inflammation cannot occur when the environment is warm enough to hold basal thermogenesis at a minimum. Accordingly, even though rats injected with LPS (500 μg kg⁻¹) in the warm environment continued to display decreases in MAP (F_37,111 = 5.04, P < 0.001) and CO (F_37,185 = 7.27, P < 0.001), their T_b and did not drop at all (Fig. 6A). On the contrary, a significant increase in T_b was observed between 45 and 150 min post-LPS (F_{37, 444} = 6.77, P < 0.001), and this increase in T_b was followed by an elevation in that became statistically significant at 80 min post-LPS (F_37,185 = 1.52, P = 0.038). Under this condition, rose at 20 min post-LPS because of a transient elevation in CH_b (P < 0.001 in post hoc analysis; Table 3). Owing to its rapid dynamics, this effect probably resulted from volume displacement from the circulation to the interstitium. The rise in , however, was not pronounced enough to prevent from falling (P < 0.0001 in post hoc analysis; Fig. 6B). As a result of these changes, the / ratio was reduced at 20 and 60 min when endotoxic shock was induced in the warm environment (F_2,10 = 8.32, P = 0.007; Fig. 6C). The observed fall in the / ratio seems to be of relevance for tissue oxygenation as clear decreases in NAD⁺/NADH ratios were observed in the kidney (t₁₅ = 3.06, P = 0.008) and hypothalamus (t₁₅ = 2.93, P = 0.010) at 60 min post-LPS in this experiment (Fig. 6D).

Figure 6. Elimination of hypothermia when endotoxic shock is induced in a warm environment reveals mismatches between O₂ delivery and O₂ consumption as well as decreases in NAD⁺/NADH ratios.

Rats were injected with the same dose of LPS (500 μg kg⁻¹ i.v.) employed in the other experiments, but at an ambient temperature of 30°C. A, minute-to-minute changes in mean arterial pressure, cardiac output, O₂ consumption and deep body temperature. B, O₂ delivery and O₂ consumption at baseline and at 20 and 60 min post-LPS. The ratio between O₂ delivery and O₂ consumption at these time points is shown in (C). D, NAD⁺/NADH ratios in tissues sampled at 60 min post-LPS or post-saline. Data are shown as means ± S.E.M. The number of animals per group (n) is indicated next to each time course curve (A) or in parenthesis (B–D). *Statistically different from baseline. LPS, lipopolysaccharide.

Table 3.

Determinants of before and after injection of LPS (500 μg kg⁻¹ i.v.) in rats exposed to an ambient temperature (30°C) that prevents development of hypothermia

Time	CHb (g dl−1)	(%)	(Torr)	(%)
Baseline	12.8 ± 0.4	97 ± 0	89 ± 3	17 ± 1
20 min post-LPS	15.3 ± 0.4*	98 ± 0	92 ± 1	21 ± 1*
60 min post-LPS	13.7 ± 0.6	97 ± 0	89 ± 2	18 ± 2

Results are means ± S.E.M. (n = 6). Abbreviations: , arterial oxygen content; CH_b, haemoglobin concentration; LPS, lipopolysaccharide; , arterial partial pressure of oxygen; , arterial haemoglobin-O₂ saturation.

^*Statistically different from the baseline period.

Experiment 6: Favouring hypoxia early in endotoxic shock by exposure to a warm environment does not affect the subsequent development of hypothermia in a cool environment

The finding that exposure to the warm environment resulted in an otherwise absent fall in the / ratio at 20 min post-LPS provided us with an unprecedented tool to directly test whether an early reduction in tissue oxygenation in endotoxic shock could exert any impact on the subsequent development of hypothermia in a cool environment. Rats were randomized into two experimental conditions: one condition consisted of holding the ambient temperature at 30°C until 20 min post-LPS, and then lowering it to 22°C as quickly as possible; the other (control) condition consisted of maintaining the ambient temperature at 22°C from the beginning of the experiment. Not surprisingly (Romanovsky et al. 2002), basal T_b was approximately 0.5°C higher in the warm environment than in the cool environment (t₁₁ = 2.47, P = 0.031). However, regardless of this difference, the hypothermic responses to LPS (500 μg kg⁻¹) followed a similar pattern in both groups (Fig. 7). In more detail, the T_b of the rats initially exposed to the warm environment started to decrease when the ambient temperature fell below 25°C (t′ in Fig. 7), and the rate of T_b decrease was similar in both groups from that time until the nadir of the hypothermic response. Moreover, T_b became virtually identical in both groups at the beginning of the recovery from hypothermia, which coincided with the time of equalization in ambient temperature (at 22°C) in the two experimental groups (t'' in Fig. 7).

Figure 7. Exposure to a warm environment early in endotoxic shock does not affect the subsequent development of hypothermia in a cool environment.

The ambient temperature was maintained at 30°C until 20 min post-LPS (500 μg kg⁻¹ i.v.), and then was quickly lowered to 22°C (warm to cool environment). In the control group, the ambient temperature was maintained at 22°C from the beginning of the experiment (cool environment only). The time courses of the changes in deep body temperature and ambient temperature are shown as means ± S.E.M. The number of animals in each group (n) is indicated. The times at which the ambient temperature of the warm-to-cool group dropped below 25.0°C and 22.5°C are indicated by t′ and t'', respectively. LPS, lipopolysaccharide.

Discussion

The interplay between and in septic shock and in LPS-induced (endotoxic) shock has been a matter of intense interest (for reviews, see Leach & Treacher, 2002; da Silva Ramos & Azevedo, 2010), but none of the studies published to date have investigated the relevance of thermoregulatory responses to this interplay. Furthermore, even though a temporal association between hypotension and hypothermia has been recognized in rats injected with moderate-to-high doses of LPS (Lang et al. 1985; Alexander et al. 1991; Romanovsky et al. 1996; Giusti-Paiva et al. 2002; Liu et al. 2012), the possible causal relationship between these two responses had never been investigated until now. Here, we tested the hypothesis that circulatory hypoxia is the trigger of the hypometabolic, hypothermic response that so often replaces fever in endotoxic shock. The reverse relationship—impact of naturally occurring hypothermia on tissue oxygenation in endotoxic shock—was also evaluated.

We initially tested whether the capacity for aerobic thermogenesis could be impaired in endotoxic shock. In this test, a β₃-receptor agonist (CL316,243) known to activate mitochondria-dependent thermogenesis in brown adipocytes (Cannon & Nedergaard, 2004) was employed to evoke acute increases in . The data obtained are indicative of an unimpaired capacity for thermogenesis in endotoxic shock, at least under the conditions studied in the present study: LPS dose of 500 μg kg⁻¹ (i.v.); mildly cool environment (22°C); and early (0–150 min) decreases in MAP, CO, , and T_b. In line with this preserved thermogenic capacity, Romanovsky et al. (1996) have been able to detect responses to forced body cooling during LPS-induced hypothermia in a study aimed at determining threshold temperatures for thermogenic activation in rats. However, unlike in the present study, the magnitudes of the cold-induced thermogenic responses were neither quantified nor compared in the study by Romanovsky et al. (1996). Another important consideration is that the non-impairment of thermogenic capacity implies that did not fall below the critical limit for aerobic metabolism during endotoxic shock. Indeed, dropped to ∼20 ml kg⁻¹ min⁻¹ during endotoxic shock in the present study, whereas a comprehensive estimate by Torres-Filho et al. (2005) indicates that the critical for rats lies at ∼10 ml kg⁻¹ min⁻¹. Although an upward shift in critical can occur in systemic inflammation, such a shift has been shown in late sepsis (Morita et al. 2003) and not in early endotoxaemia. Therefore, a -dependent state of metabolic impairment does not seem to explain the hypometabolic, hypothermic state that occurs early in endotoxic shock.

Another possibility is that a less severe form of hypoxia could be promoting a regulated suppression of thermogenesis in endotoxic shock. This possibility arises from the fact that hypothermia driven by suppression of thermogenesis is a well-established phenomenon in mammalian models of hypoxia (Wood, 1991; Mortola & Frappell, 2000; Steiner & Branco, 2002). In these models, thermogenic capacity is also unimpaired, as demonstrated by challenge with mitochondria-uncoupling drugs that unleash aerobic metabolism (Saiki & Mortola, 1997; Haouzi & Van de Louw, 2013). However, in spite of this knowledge, the relevance of hypoxia to the development of hypothermia in endotoxic shock had never been tested in direct experiments before the present study. The only previous observation serendipitously related to this matter was that pharmacological inhibition of cyclooxygenase-2 enhanced the hypothermic response to LPS while attenuating the hypotensive response (Steiner et al. 2009). However, that observation does not necessarily imply that hypothermia is independent of circulatory hypoxia, because in case attenuation of hypotension was achieved by means of arteriolar vasoconstriction, maintenance of a high afterload could have negatively impacted CO and worsened hypoxia (Errington & Rocha e Silva, 1974; Dahm et al. 1999). To obviate this experimental shortcoming, we have now employed an approach (circulating volume expansion) that is well known to promote preload-dependent elevations in CO and MAP, effects capable of alleviating hypoxia in experimental endotoxic and septic shock (Holbeck & Grande, 2002; Marx et al. 2002) as well as in clinical septic shock (da Silva Ramos & Azevedo, 2010).

Considering that preload drives primarily SV and not HR (Sanghvi et al. 1972; Cutilletta & Oparil, 1980), the finding that circulating volume expansion with hetastarch suppressed the hypotensive response to LPS without affecting the tachycardic response is not surprising. What is novel is the finding that the same volume expansion protocol resulted in attenuation of the hypometabolic, hypothermic response to LPS. This previously unrecognized phenomenon sheds new light into the relationship between circulating volume expansion strategies and in severe systemic inflammation, as previous studies on the subject have not accounted for thermoregulation-related changes in (Gilbert et al. 1986; Nimmo et al. 1992; Holbeck & Grande, 2002; Marx et al. 2002). However, as far as the hypothesis tested in the present study is concerned, the most relevant finding is that the effect of circulating volume expansion on LPS-induced hypothermia appears to have occurred independently of hypoxia, as assessed based on NAD⁺/NADH ratios in the hypothalamus, kidney and liver. Remarkably, even in the absence of circulating volume expansion, the decrease in NAD⁺/NADH ratio that is indicative of hypoxia was not observed during endotoxic shock when the environmental condition (22°C) allowed the development of hypothermia. Corroborating this observation, we found that the decrease in that brings about hypothermia in early endotoxic shock was not preceded by any drop in CO and, as was not reduced, by any drop in . The / ratio did not change. Therefore, the temporal association of shock and hypothermia in endotoxaemia does not seem to reflect a causal relationship between them; probably, the temporal association of these responses reflects their dependence on common inflammatory mediators, such as TNF-α (Ruetten & Thiemermann, 1997; Tollner et al. 2000).

At first glance, the results of these unprecedented experiments may strike one as controversial since previous studies did find evidence of hypoxia in endotoxic shock. However, a more detailed analysis reveals a different picture. First, previous studies frequently employed plasma lactate as an indicator of global hypoxia, but we now know that lactate production in systemic inflammation and shock-like states depends on many hypoxia-unrelated factors such as adrenergic tonus and pyruvate production (McCarter et al. 2002; Bundgaard et al. 2003; Levy et al. 2008). Second, another indicator of hypoxia commonly employed in previous studies is the fall in , but this alteration results from lung injury rather than circulatory shock, and is typically seen a few hours after LPS injection (Numata et al. 1998; Hou et al. 2014). Hence, it is not surprising that was not reduced at the earliest stage (20–60 min) of endotoxic shock in the present study. Last, but not least, none of the previous studies on the interplay between and in endotoxic shock considered the impact that thermoregulatory responses could have on this relationship. For example, an elegant study by Johannes et al. (2009) demonstrated the occurrence of renal hypoxia in early (60 min) endotoxic shock, but, in that study, anaesthetized rats had their T_b artificially maintained at 37°C. Accordingly, when rats in the present study were kept in an environment that was warm enough (30°C) to prevent the development of hypothermia, an imbalance between and became evident early (20 and 60 min) in endotoxic shock, and such an imbalance was associated with decreased NAD⁺/NADH ratios in the hypothalamus and kidneys. Notably, though, this early / imbalance when rats were kept in the warm environment until 20 min post-LPS did not exert any enhancing effect on the subsequent development of hypothermia in a cool environment (22°C). Therefore, even if hypoxia happens to be present to some extent early in endotoxic shock, it still does not contribute to the development of the hypometabolic, hypothermic response in this model. While this finding is novel and important, it should be pointed out that it does not necessarily apply to all forms of hypothermia that occur in the course of systemic inflammation. In this context, when lethal doses of LPS are administered, the early transient phase of hypothermia studied herein can be followed by a progressive fall in T_b (Romanovsky et al. 1996). Mechanistic comparisons of these phenomenologically distinct types of hypothermia have yet to be made.

The present study is also novel for providing the first demonstration that the risk of becoming hypoxic during endotoxic shock is largely reduced by the development of hypothermia instead of fever. Such demonstration addresses an important point raised by a previous study in which naturally occurring hypothermia was found to be more advantageous than fever in severe systemic inflammation despite the fact that it augmented the fall in MAP (Liu et al. 2012). It now appears that even if the fall in CO were to be similarly augmented by hypothermia, any shortage in oxygen supply is likely to be compensated by a decrease in oxygen demand, thus rendering tissues less liable to become hypoxic. Additionally, lessening of lung dysfunction by hypothermia might play a role in preventing or alleviating hypoxia at later stages of systemic inflammation (Liu et al. 2012).

Although the present study was not designed to answer the question as to which hypoxia-unrelated mechanism accounted for the effect of circulating volume expansion on LPS-induced hypothermia, potential mechanisms deserve comment. Based on the fact that circulating volume expansion resulted in haemodilution, we sought to determine whether it might have reduced the circulating concentration of TNF-α, a cytokine that is essential to both the hypotensive and hypothermic responses to LPS (Kozak et al. 1995; Ruetten & Thiemermann, 1997; Tollner et al. 2000). To our surprise, though, the LPS-induced rise in plasma TNF-α was not significantly reduced by the circulating volume expansion protocol. On the contrary, it is possible that TNF-α production might have been increased considering that the plasma concentration of TNF-α remained unchanged in spite of the fall in the plasma protein concentration. However, this observation does not exclude the possibility that other inflammatory mediators could have been reduced by the volume expansion protocol. Such selective changes could be caused by the recently recognized anti-inflammatory effects of hetastarch. It should be noted, though, that the amount (∼10 ml kg⁻¹) of the hetastarch dispersion infused in the present study was inferior to the amounts (≥15 ml kg⁻¹) shown to exert anti-inflammatory effects in rats (Lv et al. 2006; Feng et al. 2007). Moreover, the anti-inflammatory effects of hetastarch have not been universally found (Lee et al. 2005; Sossdorf et al. 2009). More detailed investigations on the pharmacology of hetastarch at low versus high doses will be necessary to clarify this issue.

In summary, the present study indicates that the hypometabolism and hypothermia that accompany endotoxic shock are not consequential to hypoxia. This conclusion is based on two primary findings: (i) thermogenic tissues receive enough oxygen to support full-blown metabolic responses during endotoxic shock, and (ii) there is no indication of hypoxia before or during the development of hypothermia in endotoxic shock. An important mechanistic implication of these findings is that by not supporting the hypothesis that hypothermia results from hypoxia in endotoxic shock, they favour the alternate hypothesis that the balance between fever and hypothermia in systemic inflammation is an inherent property of immune-to-brain signalling. The latter hypothesis has received some attention, and is partly supported by evidence suggesting the existence of distinct inflammatory pathways for the mediation of fever versus hypothermia in animal models of systemic inflammation (Paul et al. 1999; Dogan et al. 2002; Nautiyal et al. 2009; Steiner et al. 2009; Krall et al. 2010; Alexander & Fewell, 2011). None the less, it should be noted that even if hypothermia in endotoxic shock is not caused by hypoxia, it could still serve as a pre-emptive strategy to avoid the development of hypoxia. The efficacy of this strategy is underscored by the fact that we only found evidence of tissue hypoxia during endotoxic shock in rats that were not allowed to develop hypothermia by exposure to a warm environment. The applicability of this conceptual framework to clinical septic shock remains to be seen, and any effort along these lines should pay attention to the conserved versus divergent components of the LPS-dependent signalling pathways in rodents and humans (Schroder et al. 2012) as well as to fact that LPS is only one of the many bacterial constituents involved in the pathogenesis of septic shock (Esmon, 2004; Poli-de-Figueiredo et al. 2008).

Glossary

arterial oxygen content

CH_b

haemoglobin concentration

CO

cardiac output

oxygen delivery

HR

heart rate

LPS

lipopolysaccharide

MAP

mean arterial pressure

arterial partial pressure of oxygen

arterial haemoglobin-O₂ saturation

SV

stroke volume

T_b

body temperature

TNF

tumour necrosis factor

oxygen consumption

Additional Information

Competing interests

None declared.

Author contributions

A.A.S. conceived the study. J.J.C., M.T.F., E.A.F., K.L. and A.A.S. designed the experiments. J.J.C., M.T.F., E.A.F. and K.L. conducted the experiments. J.J.C., M.T.F., K.L. and A.A.S. analysed the data. A.A.S. wrote the manuscript, with the help of J.J.C. All authors approved the final version of the manuscript.

Funding

This study was supported, in part, by grants from Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP 2012/03831-8), American Heart Association (AHA 11SDG4880051) and Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq 305576/2013-5). MTF was the recipient of a CAPES doctorate fellowship. EAF was the recipient of a FAPESP technical training fellowship.