Spontaneous and pharmacologically induced hypothermia protect mice against endotoxic shock

Abstract

Background and Purpose:

Despite the well-known occurrence of hypothermia during sepsis, its underlying biological nature and adaptive value remain debated.

Experimental Approach:

Using indirect calorimetry, telemetry, thermal gradient studies and pharmacological studies, we examined the thermal and metabolic responses of mice treated with a shock-inducing lethal dose of lipopolysaccharide (LPS).

Key Results:

We report that LPS-treated mice undergo spontaneous hypothermia, driven by hypometabolism and cold-seeking behaviours, even when animals approach the end of life. Conversely, rewarming LPS-treated mice at 30°C delayed hypothermia but worsened mortality, thus highlighting the adaptive importance of hypothermia. Additionally, we show that LPS-induced hypothermia was partly mediated by peripheral neurotensin expressed in response to vascular toll-like receptor 4 (TLR4) signalling. The administration of a neurotensin analogue (JMV449) induced pharmacological hypothermia and significantly ameliorated the clinical presentation and lethality rates in LPS-treated mice. Moreover, the therapeutic benefits of pharmacological hypothermia were prevented when LPS-treated mice were switched to 30°C. Lastly, these beneficial outcomes were attributed to a reduction in oxygen consumption, metabolic stress and cytopathic hypoxia, rather than the modulation of the cytokine storm.

Conclusion and Implications:

Collectively, our findings indicate that spontaneous and pharmacologically-induced hypothermia protect against endotoxic shock.

1 ∣. INTRODUCTION

Hypothermia, a marker of injury severity and poor prognosis in critically ill patients, including a subset of patients with sepsis and shock (Epstein & Anna, 2006; Kushimoto et al., 2014; Langhelle et al., 2012; Rumbus et al., 2017), remains a topic of debate regarding its harmfulness and necessity for treatment. While many clinicians consider hypothermia a deleterious symptom that requires rewarming (Harmon et al., 2020), the fact that hypothermic patients are usually sicker than those with fever or normothermia does not imply that hypothermia is causally related to a worsened outcome. Instead, hypothermia was suggested to be a defence mechanism, akin to regulated hypothermia (also known as torpor), aimed at conserving energy, minimizing harm resulting from hypoxia, and ultimately prolonging survival (Corrigan et al., 2014; Cuthbertson et al., 2001; Ganeshan et al., 2019; Stanzani et al., 2020; Steiner, 2015; Steiner & Romanovsky, 2019; Stoner, 1972). In support of this view, animals treated with bacterial endotoxins often display transient cold-seeking behaviours (Almeida et al., 2006a). At the physiological level, sepsis-related hypothermia in rodents is characterized by a wide interthreshold zone between cold and heat defence mechanisms, allowing core body temperature to fluctuate in response to ambient temperature (Romanovsky, 2018). While uncertainty remains over the relevance of prior animal studies in human biology, recent clinical evidence suggests that spontaneous bouts of regulated hypothermia occur during the early phase of sepsis in subsets of individuals with improved clinical outcomes (Fonseca et al., 2016; Stanzani et al., 2020). Importantly, these patients were not artificially rewarmed as is customary in hospitals.

On the other hand, conflicting results have emerged regarding the use of body cooling in sepsis and meningitis, with clinical trials revealing potential harm (Itenov et al., 2018; Mourvillier et al., 2013). These contradictory findings underscore the critical need to understand the fundamental biology of hypothermia in the treatment of critically ill animals and patients. Specifically, confusion arises from the distinction between forced body cooling, which prevents the body from maintaining normothermia, and the biological state of regulated hypothermia, where the body inhibits thermogenesis to achieve hypothermia (Geiser et al., 2014; Sunagawa & Takahashi, 2016). Therefore, regulated hypothermia can be described as a biological state entirely distinct from accidental hypothermia due to cold exposure. Whereas forced body cooling can easily be achieved with cold blankets and/or fluids, its clinical effectiveness is limited by shivering and cardiovascular complications upon recovery (Tveita & Sieck, 2022). In contrast, regulated hypothermia occurs without shivering and offers energy savings and protection against injuries and infections in hibernating animals (Dave et al., 2006; Geiser et al., 2014; Province et al., 2020; Quinones et al., 2014). Thus, it is likely that regulated hypothermia may offer a more marked protection against insults compared to body cooling alone.

Although significant progress has been made recently in understanding the brain circuits involved in regulated hypothermia in laboratory rodents (Ambler et al., 2022; Hrvatin et al., 2020; Takahashi et al., 2020; Tupone et al., 2013), the existing literature on sepsis primarily focuses on the mechanisms associated with fever rather than hypothermia (Almeida et al., 2006a; Lazarus et al., 2007). Additionally, the current methods available for artificially inducing regulated hypothermia in non-hibernating species are limited to a few select pharmacological treatments, genetically assisted neuronal manipulations and ultrasound stimulation (Cerri, 2017; Yang et al., 2023). However, these approaches are not easily applicable to human patients at this moment. Therefore, there is a need to further elucidate the adaptive value of hypothermia in sepsis and to identify additional practical strategies for its implementation in non-hibernating species. In this study, we provide evidence that regulated hypothermia naturally occurs in mice subjected to a lethal endotoxaemic shock. Furthermore, we investigate whether the induction of regulated hypothermia by pharmacological means provides protection against endotoxaemic shock.

2 ∣. METHODS

2.1 ∣. Animals

Wild-type (WT) male mice on a C57BL/6J background were used in this study. All mice were purchased from the Jackson Laboratory (Bar Harbor; Stock No. 000664; RRID:MGI:3028467). Mice were group housed in standard ventilated cages (with igloo and nestlets) in a barrier animal facility of the University of Texas Southwestern Medical Center (UTSW) with a temperature-controlled (22°C) and humidity-controlled (55%) environment, under a 12:12-h light–dark cycle. Procedures were approved by the UTSW Institutional Animal Care and Use Committee (IACUC) under Protocol Nos. 2017-101994 and 2016-101605.

Under the University of Texas at Dallas Animal Protocol Nos. 20-10 and 2023-0122, Tlr4^LoxTB mice were generated and validated as previously described (Jia et al., 2021; Ursino et al., 2022; Wickramasinghe et al., 2023). These mice were crossed with VE-Cadherin-Cre (Vcad-Cre; Jackson Strain No. 006137) to generate mice with toll-like receptor 4 (TLR4) restored only in endothelial cells (Tlr4^LoxTB × Vcad-Cre).

Animal studies are reported in compliance with the ARRIVE guidelines and with the recommendations made by the British Journal of Pharmacology. Procedures involving animals also complied with the principles of the US National Institutes of Health (NIH). Mice were fed ad libitum on a standard chow diet (Teklad 2016) and entered experimental studies when they reached approximately 25 g in body weight. Numbers of animals used in each study are indicated in respective graphs and/or figure legends. We only examined males, but acknowledge it would be interesting to study females in the future because female C57BL/6J are more tolerant to sepsis than males (Kuo, 2016).

2.2 ∣. Drug preparation, choice of dosage and treatment design

Lipopolysaccharide (LPS) (25 mg; Sigma L2880 Escherichia coli 055: B5) was prepared in 5 ml of sterile pyrogen-free 0.9% saline (Sigma), aliquoted and stored at −20°C. On the day of study, 600 μl of LPS was thawed and diluted in 400 μl of sterile saline. Each mouse received a dose equivalent to 15 mg·kg⁻¹ of body weight by the intravenous route (i.v.). Injected volumes never exceeded 120–200 μl. Control animals received 150 μl of 0.9% sterile saline. During injection, mice were restrained in a mouse tail illuminator (Braintree Scientific) and exposed to a heating lamp. A high dose of LPS was chosen based on the literature in laboratory mice and rats for the induction of severe endotoxic shock (Berger et al., 2017; Borovikova et al., 2000; Mahapatra et al., 2018). As an animal model of critical illness, LPS administration offers the advantages of simplicity and reproducibility (Dejager et al., 2011). Furthermore, unlike in polymicrobial puncture models of sepsis (Dejager et al., 2011), LPS models do not require surgery, sedatives, analgesics and post-surgical rewarming, all of which interfere with the thermoregulatory phenotypes of interest to us. LPS administration mimics the endotoxaemia seen in meningococcaemia, bacteraemia antibacterial therapy, and the early cytokine storm and blood pressure drop seen during vascular shock (Aziz et al., 2020; Borovikova et al., 2000; Lepper et al., 2002; Panayiotou et al., 2010). Nonetheless, mimicking the supportive care provided to critically ill patients (e.g., sedatives, fluids and antibiotics) is challenging in laboratory animals, thereby limiting the translational value of such studies. Note that LPS injections occurred during the light phase, usually between 10:00 am and 2:00 pm. Metabolic and thermal data were synchronized to the time of injection. Mice with a missed LPS injection were removed from studies (unless stated otherwise).

The neurotensin analogue, JMV449 (Tocris Bioscience; Cat. No. 1998; 1 mg; PubChem ID 164415), was dissolved in 200-μl distilled water, aliquoted and stored until needed. On the day of the experiment, aliquots were dissolved in sterile saline to achieve an injectable dose of 10 mg·kg⁻¹ (i.p.). JMV449 was given 30 min after the initial injection of either saline or LPS. The reason for choosing this analogue is its well-known hypothermic effects (Dubuc et al., 1992). A pilot assay in our lab determined that 10 mg·kg⁻¹ achieved a transient bout of hypothermia of a magnitude comparable to that observed after LPS. Control animals received 0.9% sterile saline by the same route.

We prepared a cocktail of neurotensin receptor antagonist consisting of SR142948A (Tocris Bioscience; Cat. No. 2309; 10 mg) and NTRC 824 (Tocris Bioscience; Cat. No. 5438; 10 mg; PubChem ID 5311451). SR142948A was dissolved in water, and aliquots were frozen. The stock solution was diluted further to achieve a dose of 2.5 mg·kg⁻¹ (i.p.). This dose was chosen based on previous literature to prevent the hypothermic effect of central neurotensin (Gully et al., 1997). NTRC 824 (Tocris Bioscience; Cat. No. 5438; 10 mg; PubChem ID 101873359) was dissolved in water, and aliquots were frozen. The stock solution was diluted in warm phosphate-buffered saline (PBS) to achieve a dose of 5 mg·kg⁻¹ (i.p.). Both solutions were mixed and injected 30 min after the initial injection of either saline or LPS. This dose of antagonist was chosen based on previous literature showing blocking of hypothermia and thermogenesis (Li et al., 2021; Phan et al., 2023). Control animals received 0.9% sterile saline.

2.3 ∣. Indirect calorimetry and telemetry

Metabolic cage studies were performed by the UTSW Metabolic Core Facility. Under isoflurane anaesthesia, telemeters (TSE) were implanted aseptically in the peritoneal cavity. Analgesia (buprenorphine HCl at 0.05–0.1 mg·kg⁻¹, i.p.) and anaesthesia (isoflurane 3%–4% for induction and 1%–3% for maintenance) were provided according to our approved animal protocol. After 1 week of recovery, mice were acclimated for 2 days to metabolic cages (TSE PhenoMaster and Stellar system). Thereafter, measurements were performed with ad libitum access to food and water and included O₂ consumption, core temperature, feeding behaviour and locomotor activity. Mice were only taken out of the cages for 5–10 min for drug administration. Data were recorded for 24 h after LPS, and all surviving mice were killed (anaesthetic overdose and decapitation). Body composition was assessed by NMR before entering metabolic cages. O₂ consumption was normalized to lean mass. Cohorts of mice also were studied by telemetry alone. Under isoflurane anaesthesia, telemeters (DSI) were implanted aseptically in the peritoneal cavity. After 1 week of recovery, mice received drug treatments as described before. Data were recorded for 24 h after initial injection, and all surviving mice were killed at the end of the experiment. Separate cohorts of mice were switched to 30°C (near-thermoneutrality) in a temperature-controlled chamber (Powers Scientific, Inc.). In such a case, DSI receivers were placed inside the chamber.

2.4 ∣. Thermal gradient studies

Mice were tested in a Thermal Gradient Ring (Ugo Basile, Italy) connected to the ANY-maze 7.4 advanced video tracking software (Stoelting Co., Wood Dale, USA). We manually set the thermal gradient to range from 10°C to 40°C dividing each half of the ring into 12 zones of 2°C increments. On the day of the behavioural test, typically around 10:00 am, each mouse received saline or LPS immediately before being placed in the ring. Data were continuously collected for a period of 20 h. Preferred time across 5-min increments was analysed, and the data were exported to GraphPad (see later). Heat maps were generated by ANY-maze. At the end of the test period, mice were returned to their home cages and monitored for illness and endpoints.

2.5 ∣. RNAscope multiplex fluorescent assays

RNAscope was performed to detect the transcripts for our genes of interest in fixed frozen sections of brain and liver, kidney and heart. Mice received an overdose of ketamine / xylazine (500/120 mg·kg⁻¹, i.p.) and received an intracardiac perfusion of 10% formalin (Sigma). Tissues were post-fixed for 24 h, incubated in 30% sucrose for another 24 h and frozen on dry ice. Brain and peripheral tissue sections (25 and 16 μm thick) were generated using a freezing microtome and cryostat, respectively. Slides were pretreated following the manufacturer's instructions (Bio-Techne and Advanced Cell Diagnostics, Inc. [ACD]). After boiling slides in a Target Retrieval Solution, slides were rinsed in distilled water and dehydrated in 100% ethanol. Tissue was digested at 40°C for 15 min with protease III. Hybridization was performed at 40°C for 2 h, and amplification was done following the recommended ACD procedure and reagents from the RNAscope^® Multiplex Detection Kit V2 (Cat. No. 323110). The following probes were used: Mm-Ntsr1-C2 #422411-C2; Mm-Ntsr2-C2 #452311-C2; Mm-Nts-C3 #420441-C3; Mm-Nefm # 315611; Mm-Gcgr-o5-C1 #1244271; Mm-fos #316921; Mm-Hif1a #313821; and Mm-Tlr4 #316801. Signal amplification was achieved using Opal dyes 520 and 570 (Akoya Biosciences). Slides were counterstained with 4′,6-diamidino-2-phenylindole (DAPI) before applying a mounting medium (ProLong^™ Gold Antifade Cat. No. P36930) and a coverslip over the tissue section. All images were obtained using a confocal Zeiss LSM 880 microscope (UTSW QLMC facility).

2.6 ∣. Blood chemistry

Lactate and glucose were measured from the tail blood of LPS-treated mice using a Lactate Plus Meter (Nova Biomedical) and a Contour Next EZ Glucometer (Bayer), respectively. Cytokines were measured in the plasma of LPS-treated mice using the Milliplex assay (EMD Millipore) at the UTSW Metabolic Phenotyping Core.

2.7 ∣. Quantitative polymerase chain reaction (qPCR) TaqMan assays

Total RNA from mouse tissues was extracted by RNeasy Mini Kit (Qiagen) or TRIzol (Invitrogen) according to the manufacturer's instructions and reverse transcribed to cDNA using the iScript cDNA Synthesis Kit (Bio-Rad, 170-8891). Gene expression differences were determined by qPCR using TaqMan validated assays (Thermo Fisher Assay Nos. Mm00435426_m1, Mm00444459_m1, Mm00481140_m1, Mm00443258_m1, Mm00446190_m1, Mm01288386_m1, Mm99999915_g1 and Hs99999901_s1). Values were normalized to reference genes (Gapdh or 18S) using the comparative Ct method.

2.8 ∣. Infrared (IR) imaging

Mice were injected with saline or LPS as described before. A total of six cohorts of three to four mice (three saline cohorts and three LPS cohorts) were left undisturbed in an open cardboard box at normal ambient temperature (~22°C). An IR camera (FLIR) was placed over the box, and images were recorded immediately, 30 min and every hour after injection for a period of up to 5 h. The huddling score was analysed with the FLIR ResearchIR software (Wilsonville, USA) by counting the ratio of mice huddling together at each time point.

2.9 ∣. Survival rates and murine sepsis score (MSS)

Survival rates were estimated in cohorts of mice after LPS administration combined or not with JMV449 treatment. A subset of mice was switched from 22°C housing temperature to 30°C (near-thermoneutrality) in a temperature-controlled chamber (Powers Scientific, Inc.). Mice were closely monitored for a modified MSS (appearance, eyes, responsiveness, etc.) (Shrum et al., 2014) and ability to right themselves over the next 72 h. Mice that were either unable to right themselves or had a maximal murine clinical score of 3 were immediately killed (by overdose of ketamine/xylazine at 500/120 mg·kg⁻¹, i.p., followed by decapitation) and censored as a mortality. Mice surviving to 72 h were considered censored as surviving animals and killed as described before.

2.10 ∣. Data and statistical analysis

RNAscope signals for Tlr4, Nts, Fos, neurotensin receptor 1 (Ntsr1) and neurotensin receptor 2 (Ntsr2), and hypoxia-inducible factor 1α (Hif1α) were evaluated on digital images using ImageJ Fiji (NIH). We employed several methods for analysing histology: calculating the percentage of RNAscope dots per field, determining the percentage of cells co-expressing two specific transcripts and measuring signal intensity with the ImageJ (NIH; RRID:SCR_003070) measurement tools. All data were expressed as the mean ± SEM and analysed either with an unpaired t-test when comparing two experimental groups or ordinary one-way analysis of variance (ANOVA) followed by Tukey's test when comparing three groups. We used GraphPad Prism to perform statistical analyses and generate graphs. No statistical methods were used to predetermine the sample size. P-values less than 0.05 were considered statistically significant. Survival curves were generated in GraphPad Prism and analysed with the log-rank (Mantel–Cox) test. qPCR data were analysed either with an unpaired t-test when comparing two experimental groups (saline vs. LPS) or ordinary one-way ANOVA followed by Tukey's test when comparing three groups. Area under the curve (AUC) data were analysed using an unpaired t-test when comparing two groups (metabolic cages data before and after LPS groups and huddling ratio). Oxygen consumption AUC data were analysed by ordinary one-way ANOVA followed by Dunnett's test. Blood chemistry data (glucose and lactate at 24 h) and MSS were analysed with ordinary one-way ANOVA. Thermal gradient data relevant to preferred temperature were analysed with an unpaired t-test. Blood chemistry longitudinal data (glucose and lactate over 5 h) were submitted to two-way ANOVA (group and time factors) followed by multiple comparison Dunnett's, Tukey's or Šidák's test depending on the recommendation made by the software. qPCR and blood cytokine data were analysed with ordinary one-way ANOVA followed by Tukey's multiple comparison. The number of animals (n) in each experimental group is included in legends and/or graphs. P-values under 0.05, 0.01, 0.001 and 0.0001 are indicated with one, two, three and four asterisks, respectively. Except for mice with missed LPS injections, as indicated by incomplete volume administration or no blood at the tail injection site, we did not remove animals or data points from studies. GraphPad Prism files with raw data and statistical tests are available upon request. The data and statistical analysis comply with the recommendations of the British Journal of Pharmacology on experimental design and analysis in pharmacology.

2.11 ∣. Materials

Details of materials and suppliers are provided in specific subsections in Methods.

2.12 ∣. Nomenclature of targets and ligands

Key protein targets and ligands in this article are hyperlinked to corresponding entries in https://www.guidetopharmacology.org and are permanently archived in the Concise Guide to PHARMACOLOGY 2021/22 (Alexander et al., 2021).

3 ∣. RESULTS

Mice treated with a lethal dose of LPS undergo regulated hypothermia and cold-seeking behaviours

We first aimed to characterize the thermoregulatory profile of mice treated with a shock-inducing dose of LPS with 100% lethality (Figure 1a). Core temperature assessed by telemetry started declining in all mice around 2 h after LPS administration and reached its lowest point at 10 h after LPS (Figure 1a). Mice reached euthanasia criteria as early as 22 h after LPS and as late as 48 h after LPS (Figure 1a). Of note, the severity of hypothermia varied between mice but correlated positively with the time of euthanasia (Figure 1b). However, when mice received a ‘missed’ LPS injection (i.e., partial injection), they exhibited hypothermia but spontaneously returned to normothermia within 24 h without succumbing to LPS (Figure S1), suggesting that hypothermia was not causal to mortality.

FIGURE 1.

LPS-induced regulated hypothermia assessed by telemetry and calorimetry. (a) WT mice (n = 9) with DSI implants were injected with LPS, and their core temperature was recorded for 48 h by telemetry. Each line represents the temperature curve of one mouse, and large crosses indicate when the mouse was euthanized. Studies were done at normal vivarium ambient temperature (~22°C). Hypothermia started around 2 h after LPS and peaked around 10 h after LPS. (b) Using the previous curves, area under the curve (AUC) was calculated in GraphPad and plotted against euthanasia time for each mouse. (c) Metabolic parameters were recorded 24 h before and after LPS inside indirect calorimetry chambers (TSE). Each dot on the curves represents mean cumulative feeding ± SEM (black dots and grey errors). Only the average curve was provided for rearing (green line) to avoid obscuring the graph. Average feeding over 24 h before and after LPS was calculated and analysed by unpaired t-test with P < 0.0001 and t = 11.30, d.f. = 8. Mice ate between 3 and 4 g of food before LPS but almost no food after LPS. Average rearing counts over 24 h before and after LPS were analysed by unpaired t-test with P = 0.0112 and t = 3.277, d.f. = 8. Note how mice become rapidly anorectic and lethargic soon after LPS administration. (d) Other metabolic parameters recorded included core temperature (black) and oxygen consumption (green). Each dot on the curves represents mean ± SEM. Average oxygen consumption over 24 h before and after LPS was analysed by unpaired t-test with P < 0.0001 and t = 26.11, d.f. = 8. Average temperature over 24 h before and after LPS was analysed by unpaired t-test with P < 0.0001 and t = 21.77, d.f. = 8. Note how oxygen levels fall in parallel to core temperature. (e) Temperature plotted against oxygen combustion (hysteresis). Each grey line represents one mouse, and the black line represents the average values. Note how the relationship between falling temperature and oxygen is highly reminiscent of active cooling seen in hibernating animals rather than passive cooling seen during accidental hypothermia (Geiser et al., 2014). Raw values and statistical tests are available as Prism files.

Regulated hypothermia (torpor) can be distinguished from other types of hypothermia by examining the relationship between temperature and energy expenditure (Geiser et al., 2014). Thus, to further gain insight into the nature of LPS-induced hypothermia, we employed simultaneous telemetry and indirect calorimetry (Stellar system) to assess metabolic changes (Figure 1c). Following LPS administration, all mice rapidly ceased eating and became immobile (Figure 1c). Notably, oxygen consumption sharply dropped starting at 2 h after LPS in a pattern closely resembling the decline in body temperature (Figure 1d). The relationship between energy consumption and temperature is consistent with an active cooling process reminiscent of what is observed during spontaneous torpor in hibernating animals (Figure 1e) (see also Geiser et al., 2014).

IR camera imaging revealed that, whereas saline-treated mice frequently huddled together during the day, LPS treatment caused mice to separate from each other (Figure S2A-C). The latter observations led us to investigate the role of behavioural thermoregulation in mitigating a lethal dose of LPS. Behavioural thermoregulation refers to the ability of an organism to seek a cool or warm ambient temperature to maintain homeostasis and cope with both environmental and bodily insults (Terrien et al., 2011). Using a thermal gradient apparatus (Ugo Basile, Italy) with a range from 10°C to 40°C (Figure 2a), we found that saline-treated mice typically preferred a temperature of ~31°C during a 20-h period (Figure 2b). In contrast, LPS-treated mice preferred a lower temperature of ~26.6°C on average. Preferred temperature started declining a few hours after LPS and continued declining through the experiment to be lowest at 20 h after LPS, when mice approached near death (Figure 2b). Interestingly, heat maps of average preferred temperatures revealed that LPS-treated mice spent more time than controls not only in the cool areas of the ring but also in the hot areas (Figure 2c). Indeed, individual curve analysis further demonstrated the decrease in preferred temperature in LPS-treated mice was typically polyphasic (Figure 2d). This trend indicates that LPS-treated mice favoured both fever- and hypothermia-inducing environments at different stages of their disease, even if their average preferred temperature irreversibly dropped. Specifically, long periods of low-temperature preference in LPS-treated mice (several hours) were typically interspersed by short bursts of high-temperature preferences above 35°C (Figure 2d). In addition, bouts of cold preference were not synchronized between mice and considerably varied in numbers, length and preferred temperature. In contrast, saline-treated mice only explored cold areas for brief periods of time and rarely stayed above 35°C.

FIGURE 2.

Behavioural thermoregulation, survival rates and thermoneutral switch. (a) WT mice (n = 7 saline; n = 9 LPS) were placed in a thermal gradient ring (TGR, 10–40°C), and their preferred ambient temperature was continuously recorded for 20 h. (b) Saline-treated mice (grey) spend most of their time above 30°C, while LPS mice (blue) progressively lowered their preferred temperature. Average preferred temperature over a 20-h period differed between groups by unpaired t-test with P < 0.0215 and t = 2.588, d.f. = 14. (c) Heat maps generated with ANY-maze representing the average temperature preference of saline versus LPS mice over a period of 20 h. Of note, LPS-treated animals explored both cold and hot areas of the ring more often than saline animals. (d) Individual preference temperature curves reveal the polyphasic nature of behavioural thermoregulation in LPS-treated mice. For instance, LPS-treated mouse #7 cycled through brief periods of heat-seeking and long periods of cold-seeking behaviour. (e) Mice were treated with LPS as described before and observed for euthanasia criteria over a period of up to 72 h. Each dot represents mean ± SEM. One cohort (orange) remained at 22°C the whole experiment. Another cohort was switched to 30°C (red) immediately after LPS. The last cohort (blue) was placed in the thermal ring immediately after LPS. Data were analysed using a log-rank (Mantel–Cox) test with results indicated on a graph. Housing at 30°C (rewarming) noticeably accelerated time of death. (f) WT mice (n = 4 saline; n = 4 LPS) with DSI implants were switched to 30°C housing immediately after injection. Data are presented as individual core temperature curves for saline (grey) and LPS mice (red). At 30°C, LPS-treated mice developed fever rather than hypothermia during the early stage of sepsis (<7 h). Thereafter, mice rapidly became hypothermic and lethargic and died early.

To assess the adaptive value of LPS-induced behavioural thermoregulation, we examined the mortality rates of LPS-treated mice at different ambient temperatures (Figure 2e). All mice housed at approximately ~22°C reached euthanasia criteria between 22 and 48 h. There was a trend towards a longer survival rate in mice kept for 20 h after LPS in the thermal gradient (10–40°C) before being returned to their vivarium, with one mouse surviving up to 72 h and another mouse alive at 72 h (Figure 2e). The latter data indicate that cold-seeking behaviours promote survival. In further support of this view, mice that were switched to 30°C immediately after LPS exhibited significantly earlier euthanasia criteria, with half of the mice meeting these criteria before 24 h (Figure 2e). Lastly, we used telemetry to determine the reason underlying the deleterious effect of 30°C rewarming (Figure 2f). Mice switched to 30°C immediately after LPS did not develop hypothermia until at least 7 h after LPS. In the period preceding hypothermia, mice exhibited fever (Figure 2f). However, following the onset of hypothermia, mice rapidly reached euthanasia criteria. In agreement with one previous study (Liu et al., 2012), fever is associated with a less favourable outcome than hypothermia in LPS-treated animals. Despite the well-known occurrence of hypothermia in laboratory rodents treated with a moderate dose of LPS (Corrigan et al., 2014; Ganeshan et al., 2019; Mei et al., 2018; Ogawa et al., 2016), here, we further establish that LPS-induced hypothermia during a lethal shock is a regulated and adaptive behavioural strategy with a survival benefit.

3.1 ∣. Neurotensin signalling partly contributes to LPS-induced hypothermia

Whereas the molecular mediators of hypothermic sepsis remain largely unknown, the fact that LPS triggers hypothermia only when administered peripherally suggests that molecules released from peripheral cells mediate hypothermia (Al-Saffar et al., 2013). Here, we investigated the possible role of peripheral neurotensin, a well-known potent hypothermic and thermolytic endogenous peptide (Li et al., 2021).

First, whereas peripheral neurotensin was barely detectable in control mice, we found that LPS induced neurotensin (Nts) expression across many organs (Figure 3a). Specifically, qPCR analysis demonstrated a significant elevation of Nts mRNA in the spleen, liver and lungs following LPS administration (Figure 3a). RNAscope analysis confirmed that induction of neurotensin transcripts specifically occurred in the vasculature of LPS-treated animals (Figure 3b,c). Although qPCR did not reveal altered neurotensin expression changes in the brain due to high basal levels, RNAscope unveiled a similar induction of Nts-positive cells around large brain blood vessels (Figure S3). Furthermore, we assessed the mechanisms governing LPS-induced neurotensin using transgenic models selectively expressing TLR4, the sole LPS receptor (Beutler, 2002), but only in vascular cells. As anticipated, mice lacking Tlr4 (mice carrying a loxP-flanked transcriptional blocker, TLR4lox-TB) showed little Tlr4 and Nts hepatic expression after LPS administration (Figure 3d,f-h). In contrast, mice expressing vascular Tlr4 (mice carrying a loxP-flanked transcriptional blocker, TLR4lox-TB, crossed with transgenic mice expressing Cre recombinase under the control of the Vcad promoter) showed both Tlr4 and Nts signals in the lining of hepatic blood vessels (Figure 3e-g). Approximately 40% of Nts-positive cells expressed Tlr4 (Figure 3h), suggesting that vascular TLR4 signalling exerts both direct and indirect influences on the production of peripheral neurotensin.

FIGURE 3.

LPS-induced vascular neurotensin (Nts). (a) qPCR for Nts in tissues harvested from WT mice (n = 4 saline; n = 5 LPS) at 3 h after injection. Data are presented as mean ± SEM and analysed with an unpaired t-test for brain (P = 0.4413, t = 0.8161, d.f. = 7), spleen (P < 0.0001, t = 7.34, d.f. = 8), liver (P = 0.0052, t = 3.810, d.f. = 8) and lungs (P < 0.0001, t = 10.14, d.f. = 8). LPS administration strongly stimulated Nts expression in peripheral tissues. Average Ct values are indicated next to each bar. (b, c) RNAscope assay for Nts mRNA (red) in the liver of one saline and LPS mouse. The tissue was counterstained with DAPI (grey). White arrows indicate representative vascular cells positive for Nts signals. Interestingly, Nts signals were absent from the parenchyma. (d, e) Multiplex RNAscope assay for Nts (green) and toll-like receptor (Tlr4) (red) in the liver of LPS mice. While fl/fl-TB-Tlr4 mice do not express endogenous Tlr4, mice carrying Vcad-Cre and fl/fl-TB-Tlr4 alleles expressed Tlr4 only in vascular cells. Tissue was counterstained with DAPI (grey). White arrows indicate representative vascular cells positive for Nts signals. (f) Mean number ± SEM of Tlr4-positive dots per blood vessel (P = 0.0028, F(4, 3)= 13.30). (g) Mean Nts signal intensity ± SEM (P = 0.0040, F(4, 3) = 53.08). (h) Mean percentage ± SEM of cells expressing both Tlr4 and Nts (P = 0.0004, F(4, 3) = ∞). bv, blood vessel; ns, not significant.

Second, we noted high expression of both Ntsr1 and Ntsr2 in the brain, with minimal expression in the spleen, liver and lungs (Figure 4a,b). Utilizing multiplex fluorescent RNAscope, we confirmed that Ntsr1 and Ntsr2 were predominantly produced in neurons and non-neuronal cells, respectively, but not in the liver (Figure S4). Furthermore, LPS induced Fos mRNA in forebrain Ntsr1-expressing neurons, notably those in the bed nucleus of the stria terminalis (BNST, 41% of Ntsr1-expressing cells co-expressed Fos in LPS mice, n = 3) (Figure 4c,c′,e,e′). Occasional co-expression of Ntsr1 and Fos was noticed in the forebrain including positive preoptic area, amygdala and midbrain but in relatively few cells (Figure S5A-E). Ntsr2-positive glial cells at the periphery of the brain (glia limitans and cortex) also were often Fos positive (Figure 4d,d′,f,f′). Hence, various neurotensin responsive cells are activated throughout the brain during endotoxic shock.

FIGURE 4.

Neurotensin receptor signalling and hypothermia. (a, b) qPCR for neurotensin receptors 1 and 2 (Ntsr1 and Ntsr2) in tissues harvested from WT mice (n = 5). Data are presented as mean ± SEM, and averaged Ct values for each tissue are added on top of each bar graph. (c, c′, d, d′) Multiplex RNAscope assay for Ntsr1 or Ntsr2 (red) and Fos (green) in the brain of saline-treated mice. Fos was minimally expressed by either Ntsr1- or Ntsr2-expressing cells. (e, e′, f, f′) Multiplex RNAscope assay for Ntsr1 or Ntsr2 (red) and Fos (green) in the brain of LPS-treated mice. Fos was induced in many cells that co-expressed Ntsr1- and Ntsr2-expressing cells (white arrows) at 2 h after LPS. Highlighted brain regions for Ntsr1 and Ntsr2 were the bed nucleus of the stria terminalis and cortex, respectively. (g) WT mice with DSI implants were administered with either saline only (n = 3, red line) or saline followed by a cocktail of neurotensin receptor antagonists (n = 4, grey line). (h) WT mice with DSI implants were administered with either LPS (n = 5, red line) or saline followed by a cocktail of neurotensin receptor antagonists (n = 6, grey line). Core temperature was recorded for 12 h by telemetry, and data were expressed as mean ± SEM. Data collected between 1 and 3 h after injection were converted to area under the curve (AUC) and analysed by t-test. Antagonists did not modify the temperature of saline-treated mice (P = 0.3109, t = 1.127, d.f. = 5). LPS-treated mice co-administered with antagonists had higher temperature (P = 0.0043, t = 1.127, d.f. = 5). Studies were done at normal vivarium ambient temperature (~22°C). nd, not detectable.

Third, we investigated the effects of NTSR1 and NTSR2 blockade on core temperature in response to a cocktail of antagonists previously shown to prevent neurotensin-induced hypothermia and inhibit thermogenesis (Li et al., 2021; Tabarean, 2020). In saline-treated mice, NTSR1/2 blockade did not produce a noticeable effect (Figure 4g). However, when mice were treated with LPS, NTSR1/2 blockade delayed the onset of hypothermia by approximately 2 h (Figure 4h). The above findings indicate that neurotensin of vascular origin participates, at least in part, in initiating LPS-induced hypothermia.

3.2 ∣. A neurotensin analogue causing regulated hypothermia protects against LPS-induced shock

Based on the above data, we further explored whether neurotensin signalling could artificially mimic regulated hypothermia, offering potential protective effects against shock. The neurotensin analogue JMV449 has known hypothermic properties (Craig et al., 2021; Dubuc et al., 1992; Kato et al., 2019). Telemetry studies demonstrated that the administration of JMV449 reversibly induced hypothermia without visual signs of shivering (Figure 5a). A single dose of JMV449 (10 mg·kg⁻¹, i.p.) caused a rapid reduction in core temperature, reaching its peak at 2 h after injection (Figure 5a). When JMV449 was administered to LPS-treated mice, a similar rapid decline in core temperature was observed (Figure 5b). Intriguingly, while the temperature of mice treated with LPS alone continued to decline, JMV449-treated animals showed partial recovery from hypothermia (Figure 5b). At the 24-h time point, mice treated with LPS alone exhibited more severe hypothermia, whereas mice treated with LPS and JMV449 were less hypothermic (Figure 5b,c), hence potentially less sick. Immediately following JMV449 administration, mice exhibited a higher MSS, likely due to lethargy associated with a state of ‘torpor’ (Figure 5e). However, consistent with its potential clinical benefit, JMV449 improved the MSS in LPS-treated mice at the 24-h time point (Figure 5d,f).

FIGURE 5.

JMV449-induced pharmacological hypothermia. (a) WT mice with DSI implants were injected with saline (n = 5, grey line) or saline followed by JMV449 (n = 4, green line), and their core temperature was recorded for 24 h by telemetry. Each line represents the mean core temperature ± SEM. A single injection of JMV449 caused a transient bout of hypothermia. (b) WT mice with DSI implants were injected with LPS (n = 9, red line) or LPS followed by JMV449 (n = 10, blue line). Animals treated with LPS and JMV449 became rapidly hypothermic but started recovering around 12 h after injection. Animals only treated with LPS gradually became hypothermic without ever recovering. All studies were done at normal vivarium ambient temperature (~22°C). (c) At 24 h after LPS, core temperature of animals treated with JMV449 was significantly higher. Data are the mean temperature ± SEM and analysed by ordinary one-way ANOVA (P < 0.0001; F(3, 24) = 4.477) followed by Tukey's test (P = 0.0127 between LPS + saline and LPS + JMV449). (d) Murine sepsis score (MSS) of WT mice injected with saline (n = 6, grey line) or LPS (n = 7, red line) or LPS followed by JMV449 (n = 9, blue line) over 24 h. Data are the mean score ± SEM. (e) Average MSS ± SEM at 2.5 h. Data points were analysed by ordinary one-way ANOVA (P < 0.0001; F(2, 19) = 1.397) followed by Tukey's test (P = 0.0079 for ‘saline + JMV449 vs. LPS + saline’; P < 0.0001 for ‘saline + JMV449 vs. LPS + JMV449’; and P = 0.0225 for ‘LPS + saline vs. LPS + JMV449’). (f) Average MSS ± SEM at 24 h. Data points were analysed by ordinary one-way ANOVA (P < 0.0001; F(2, 19) = 5.742) followed by Tukey's test (P < 0.001 or 0.0001 between groups).

We next assessed the influence of JMV449 on the early cytokine storm induced by LPS. Interestingly, we observed no significant effects on blood levels of tumour necrosis factor-α (TNF-α) or interleukin-10 (IL-10) (Figure 6a,b) but a notable reduction in interleukin-6 (IL-6) levels (Figure 6c). In general, qPCR analysis further confirmed that the impact of JMV449 treatment on peripheral expression levels of cytokines was modest and tissue dependent. For instance, JMV449 caused a slight elevation of Il-6, Tnf-α and Il-1β in the liver, spleen and adipose tissue, respectively (Figure 6d-l), all indicative of a slight proinflammatory bias. Notably, JMV449 treatment was associated with a significant reduction in adipose Il-6 expression (Figure 6k). Overall, our findings demonstrate that severely hypothermic animals are still capable of mounting a robust immune response. Furthermore, it is noteworthy that the benefits of JMV449 are not attributable to an anti-inflammatory mechanism.

FIGURE 6.

Assessment of blood and tissue cytokines. (a–c) WT mice were given ‘saline’ (n = 4, grey bar) or ‘LPS’ (n = 6, red bar) or ‘saline + JMV449’ (n = 5, black bar) or ‘LPS + JMV449’ (n = 6, blue bar). At 2 h after injection, blood was collected to measure circulating cytokines. Data are presented as mean ± SEM and analysed by two-way ANOVA. LPS treatment significantly stimulated the release of interleukin-10 (IL-10), tumour necrosis factor-α (TNF-α) and interleukin-6 (IL-6) (P < 0.0001 and F(1, 17) = 124.8; P < 0.0001 and F(1, 17) = 44.00; and P < 0.0001 and F(1, 17) = 505.6). However, Tukey's test revealed a difference between ‘LPS’ and ‘LPS + JMV449’ only for IL-6 (predicted LS mean diff. 11,728). (d–l) qPCR for indicated tissues and genes from WT mice were given ‘saline + JMV449’ (n = 4, grey bar) or ‘LPS’ (n = 6, red bar) or ‘LPS + JMV449’ (n = 6, blue bar). Data are presented as mean ± SEM. Data were analysed by ordinary one-way ANOVA. Raw values and statistical tests are provided as Prism files. Tukey's test revealed statistical differences between the ‘LPS’ and ‘LPS + JMV449’ groups for the following genes and tissues: liver Il-6 (P < 0.0001 and F(2, 13) = 25.09; post hoc adjusted P = 0.0139), spleen interleukin-1β (Il-1β) (P < 0.0001 and F(2, 13) = 37.33; post hoc adjusted P = 0.0040), epididymal white adipose tissue (eWAT) Tnf-α (P < 0.0001 and F(2, 13) = 51.37; post hoc adjusted P = 0.0117) and eWAT Il-6 (P < 0.0001 and F(2, 13) = 15.12; post hoc adjusted P = 0.0187). ns, not significant.

To gain deeper insights into the protective mechanisms of JMV449, we conducted combined telemetry and calorimetry studies (Figure 7a). As we previously showed, these investigations revealed that mice treated solely with LPS experienced a gradual onset of hypothermia and reduced metabolic activity without any recovery over a 24-h period (Figure 7b,c). In contrast, mice treated with JMV449 exhibited an initial episode of rapid hypothermia and suppressed energy expenditure indicative of torpor (Figure 7b,c). During the early shock phase (<5 h), JMV449-treated mice displayed lower oxygen consumption compared with mice treated only with LPS (Figure 7d). Interestingly, the torpid phase was succeeded by a recovery period extending beyond 12 h. As a result, at the 24-h post-LPS time point, the oxygen consumption levels in JMV449-treated animals were significantly higher than those in mice receiving only LPS (Figure 7d). These subsequent findings suggest that the energy conservation observed during the early stages of shock, facilitated by JMV449, contributed to its clinical benefits. In support of this idea, JMV449 treatment influenced markers of metabolic stress, such as hyperlactataemia and hypoglycaemia. Specifically, the results indicated a partial prevention of LPS-induced hyperlactataemia and hypoglycaemia during the initial phase of endotoxic shock (<5 h) (Figure 7e,f). Crucially, we conducted additional studies to ascertain whether the observed protective effects of JMV449 ultimately translated into survival benefits (Figure 7g). Administration of JMV449 led to a delayed mortality of approximately 24 h and a reduced overall mortality in LPS-treated mice (Figure 7g). Intriguingly, these benefits of JMV449 were reversed when mice were housed at 30°C upon LPS injection (Figure 7g). Telemetry studies further revealed that the hypothermic effects of JMV449 were diminished in mice switched to 30°C (Figure 7h).

FIGURE 7.

JMV449-induced suppression of metabolic rate and improved mortality. (a) WT mice with Stellar implants were injected with ‘saline’ (n = 4, grey) or ‘LPS’ (n = 6, red) or ‘LPS + JMV449’ (n = 5, blue), and their core temperature and oxygen consumption were recorded for 24 h with metabolic chambers (TSE). (b) Each line represents the mean temperature curve ± SEM over 24 h. Studies were done at normal vivarium ambient temperature (~22°C). (c) Each line represents the mean oxygen consumption curve ± SEM over 24 h. (d) Using the previous curves, area under the curve (AUC) was calculated in GraphPad and plotted for 0- to 5-h and 12- to 24-h periods. Each dot represents mean AUC ± SEM. Data were analysed by one-way ANOVA followed by Dunnett's test. For the 0- to 5-h period, all groups were significantly different (P < 0.0001 and F(2, 12) = 28.87; ‘LPS’ vs. ‘LPS + JMV449’ adjusted P = 0.0011; ‘LPS’ vs. ‘saline’ adjusted P = 0.0086). For the 12- to 24-h period, all groups were significantly different (P < 0.0001 and F(2, 12) = 180.9; ‘LPS’ vs. ‘LPS + JMV449’ adjusted P = 0.0180; ‘LPS’ vs. ‘saline’ adjusted P < 0.0001). (e) Average blood glucose ± SEM over 5 h after injection. Data points were analysed by two-way ANOVA followed by Tukey's test (treatment effect with P = 0.0069; ‘LPS’ vs. ‘saline’ post hoc adjusted P = 0.0123 at 3 h and P = 0.0197 at 4 h). (f) Average blood lactate ± SEM over 5 h after injection. Data points were analysed by two-way ANOVA followed by Dunnett's test (treatment effect with P = 0.0051; ‘saline’ vs. ‘LPS’ post hoc adjusted P = 0.0397 at 2 h). (g) Mice were treated with ‘LPS + saline’ or ‘LPS + JMV449’ as described before and observed for euthanasia criteria over a period of up to 72 h. Each dot represents mean ± SEM. One ‘LPS + saline’ cohort (n = 12, red) and one ‘LPS + JMV449’ (n = 13, blue) remained at 22°C the whole experiment. Another ‘LPS + JMV449’ cohort (n = 10, black) was switched to 30°C (red) immediately after LPS. Data were analysed using a log-rank (Mantel–Cox) test with results indicated on a graph. (h) WT mice were treated with ‘LPS + JMV449’ and either kept at their usual vivarium temperature (n = 4, blue line) or switched to a chamber at 30°C immediately after injection (n = 3, black line). Data are presented as mean core temperature ± SEM over 24 h. JMV449-induced hypothermia was greatly blunted at 30°C.

Hif1α expression is considered a marker of cellular hypoxia and inflammatory stress during sepsis (Peyssonnaux et al., 2007). To gain further insights into the molecular determinants of pharmacologically-induced hypothermia, we examined the expression of Hif1α by RNAscope across organs prone to failure during sepsis. Here, we report that saline-treated uniformly expressed low levels of Hif1α in all examined organs including the liver, kidney and heart (Figure 8a-c). At 3 h after LPS, robust Hif1α expression occurred in the parenchyma and vasculature of all examined organs (Figure 8d-f). The administration of JMV449 to LPS-treated mice visibly reduced Hif1α expression levels in all tissues (Figure 8g-i). However, reduced Hif1α reached significance in the liver and kidney, but not the myocardium (Figure 8j-l). Together, this aligns with evidence suggesting that JMV449-induced regulated hypothermia prevents cytopathic hypoxia.

FIGURE 8.

JMV449 reduced LPS-induced Hif1α expression. (a–i) WT mice received saline (n = 4), LPS (n = 4) or LPS combined with JMV449 (n = 4) before being sacrificed 3 h later. RNAscope assay for Hif1α (red) in the liver, kidney and myocardium was performed. Tissue was counterstained with DAPI (grey). The administration of JMV449 reduced Hif1α signals in select organs prone to failure during sepsis. (j) Mean Hif1α signal intensity ± SEM in the liver was analysed by one-way ANOVA (P = 0.0077, F(2, 11) = 7.810) followed by Dunnett's test (‘saline vs. LPS’ adjusted P = 0.0066; ‘LPS vs. LPS + JMV449’ adjusted P = 0.0448). (k) Mean Hif1α signal intensity ± SEM in the kidney was analysed by one-way ANOVA (P = 0.0039, F(2, 11) = 9.589) followed by Dunnett's test (‘saline vs. LPS’ adjusted P = 0.0049; ‘LPS vs. LPS + JMV449’ adjusted P = 0.0102). (l) Mean Hif1α signal intensity ± SEM in the myocardium was analysed by one-way ANOVA (P = 0.0090, F(2, 12) = 7.146) followed by Dunnett's test (‘saline vs. LPS’ adjusted P = 0.0074). ns, not significant.

4 ∣. DISCUSSION

Regulated hypothermia is a well-documented state of actively suppressed thermogenesis in hibernating animals (Cubuk et al., 2016; Geiser et al., 2014; Staples, 2016). However, regulated hypothermia can be artificially induced in non-hibernating laboratory mammals, revealing the existence of torpor-inducing neural circuits beyond hibernating species (Tupone et al., 2023). For example, prior research work has shown that rodents treated with moderate to high doses of LPS undergo regulated hypothermia (Almeida et al., 2006b; Ganeshan et al., 2019; Liu et al., 2012). Here, we further extend these observations by showing that regulated hypothermia naturally occurs during a lethal endotoxaemic shock in mice. The metabolic response of LPS-treated mice, even in mice approaching death, closely resembles that of torpid animals rather than a failure of the thermoregulatory system. Overall, our findings align well with the ‘thermometabolic hypothesis’ of sepsis postulating that animals enter a state of minimally regulated temperature as an adaptive strategy to cope with severe inflammatory insults (Ganeshan et al., 2019; Romanovsky, 2018; Stanzani et al., 2020; Steiner, 2015). It is particularly striking that LPS-treated mice placed in a thermal gradient preferred a cooler environment rather than a warm environment. Previous research has shown that LPS-treated animals seek cold environments (Almeida et al., 2006a; Wanner et al., 2017), although typically for short durations only. In contrast, our study observed that behavioural thermoregulation in LPS-treated mice involved multiple long bouts of cold-seeking, interspersed with brief periods of heat-seeking. This behaviour suggests that mice, even when facing life-threatening illness, utilize complex behavioural strategies. Conversely, LPS-treated mice deteriorated rapidly when body cooling was prevented by housing them at 30°C, highlighting the adaptive value of behavioural thermoregulation. Because the LPS model is sterile, the benefits of spontaneous hypothermia are likely due to enhanced tolerance to organ injury rather than improved bacterial control.

Due to a lack of temporal thermometabolic data in critically ill humans, especially in patients undergoing shock, it is difficult to ascertain that observations made in mice fully apply to human patients. While septic humans more often display fever than hypothermia (Rumbus et al., 2017), transient bouts of hypothermia were observed in septic patients, which may be equivalent to regulated hypothermia (Fonseca et al., 2016; Little & Stoner, 1981). Importantly, patients were not rewarmed to allow spontaneous hypothermia. As a result, investigators documented spontaneous hypothermic episodes in septic patients with a favourable clinical outcome. Moreover, hypothermia was commonly observed early during sepsis but rarely near death. This study implies, first, that hypothermic sepsis may be underdiagnosed because clinicians conventionally rewarm patients and, secondly, that rewarming should not always be recommended.

The exact molecular and neural pathways underlying LPS-induced hypothermia remain to be elucidated. Previous work has shown that LPS signalling in haematopoietic cells plays an important role in triggering hypothermia (Ganeshan et al., 2019). Here, we further found that vasculature-derived neurotensin may act as a cryogenic agent during sepsis. While identifying the site of actions of neurotensin was beyond the scope of the present study, previous research has indicated the preoptic area plays a crucial role in neurotensin-induced hypothermia (Tabarean, 2020). We also do not know which NTSR-expressing cells mediate LPS-induced hypothermia. Nonetheless, we observed that LPS activated Ntsr1-expressing neurons of the several brain regions, such as the BNST, but none are known to be associated with torpor-inducing neural networks (Hrvatin et al., 2020). It is may be more plausible that LPS-induced vascular neurotensin acts on glial NTSR2, which subsequently facilitates the release of cryogenic agents, thereby secondarily affecting the preoptic areas (Tabarean, 2020). Future studies are warranted to investigate the elimination of NTSR1 from specific brain areas and cell types to address these questions. It is also obvious from our pharmacology that neurotensin signalling alone is not solely responsible for LPS-induced hypothermia. Instead, it is likely that a cascade of peptides and inflammatory mediators may cause LPS-related hypothermia (Garami et al., 2018).

Our findings support the notion that pharmacologically-induced hypothermia contributes to a favourable outcome during endotoxic shock, by promoting energy savings and reducing cytopathic hypoxia rather than through the suppression of inflammation. One would expect Hif1α levels to be positively correlated with inflammation, but this was not observed in our JMV449-treated mice. This discrepancy suggests that JMV449-induced hypothermia may limit cellular hypoxia despite the presence of high levels of proinflammatory cytokines. The molecular mechanisms linking hypothermia to reduced hypoxia remain to be elucidated through high-throughput transcriptomic studies. The question arises of how to safely induce hypothermia for therapeutic purposes in human subjects. While therapeutic hypothermia using forced body cooling has shown promise in certain conditions (Froehler & Ovbiagele, 2010; Kochanek et al., 2018; Kurisu & Yenari, 2018; Lakhan & Pamplona, 2012), it does not fully replicate the benefits of natural torpor (Itenov et al., 2018; Kim et al., 2020; Li et al., 2015). In particular, a large randomized clinical trial demonstrated that external cooling is detrimental to septic patients (Itenov et al., 2018). Indeed, forced cooling engages thermogenesis, shivering and physiological stress (Blondin et al., 2017). Hence, pharmacologically-induced hypothermia mimicking a state of regulated hypothermia may be preferable. Notably, neurotensin drugs, if tolerated in humans, may offer a means to induce a torpor-like state, providing energy savings and reducing fever without compromising the ability to fight infection. Notably, JMV449 binds to the human NTSR1 (Kato et al., 2019), and neurotensin-induced hypothermia has been safely induced in non-human primates (Fantegrossi et al., 2005), suggesting that neurotensin drugs may work in humans. However, neurotensin currently has no application in humans. Instead, there are already commonly used drugs in humans, including sympatholytics, analgesics, sedatives, narcotics and hypnotics, which can cause hypothermia with hypometabolism (Briesenick et al., 2023; Dickerson & Roth-Yousey, 2005; Takahashi et al., 1997). Many clinical reports indicate that sedatives exert protective effects in sepsis (Aso et al., 2021; Ding et al., 2019; Reese et al., 2018). Thus, it is tempting to speculate that septic patients, if kept at a subneutral ambient temperature without rewarming, may benefit from sedation-induced hypothermia.

It is important to acknowledge potential contradictions in our findings with previous studies. Neurotensin is well-known to produce hypotension (Carraway & Leeman, 1973). In fact, one study using caecal ligation and puncture (CLP) in mice suggested a detrimental effect of neurotensin-induced hypotension (Piliponsky et al., 2008). However, our results using JMV449, an agonist known to cause a reduction in blood pressure (Vivancos et al., 2021), were without apparent harm. Differences in animal models, antagonist usage and the specific context of the studies may contribute to these discrepancies. Knockout models may be affected by compensatory changes in thermoregulatory pathways. For instance, the NTSR1 knockout mouse is slightly hyperthermic (Remaury et al., 2002). It also could be that hypotension, which may be harmful in certain circumstances, may not be universally detrimental, particularly in the context of suppressed oxygen consumption through pharmacological or spontaneous hypothermia. In fact, blood pressure and heart rate are dramatically low in hibernating animals (Horwitz et al., 2013). Also, hypotension is not harmful to all septic patients (Lavillegrand et al., 2023), and furthermore, a subset of septic patients with hypotension is at lower risk of mortality (Hernandez et al., 2012). Instead, pharmacological hypothermia may hypothetically be employed to mitigate the potential harm of hypotension and hypoperfusion.

In summary, our data demonstrate that both pharmacological and spontaneous hypothermia ameliorate the clinical profiles of mice undergoing a lethal endotoxaemic shock without modifying innate immunity. Future research on preclinical animal models should focus on identifying biomarkers of regulated hypothermia, factors that may potentially counteract the benefits of spontaneous regulated hypothermia, and designing pharmacological approaches to artificially induce or accompany regulated hypothermia in critically ill individuals.

What is already known

• Hypothermia in septic patients remains a topic of debate regarding its harmfulness and treatment.

What does this study add

• Pharmacologically-induced and spontaneous hypothermia protect mice undergoing a lethal endotoxic shock.

What is the clinical significance

• Pharmacologically-induced and spontaneous hypothermia are likely to enhance tolerance to sepsis.

Abbreviations:

LPS

lipopolysaccharide

NTSR

neurotensin receptor

TLR4

toll-like receptor 4

DATA AVAILABILITY STATEMENT

Raw data are available without restrictions as GraphPad Prism files upon request from the corresponding author.