Animal study and dosage administration
As described previously [18], we used 24-week-old C57BL/6J male mice (Janvier SAS, Chassal, France) for our validated, centrally catheterized, fluid-resuscitated model of cecal ligation and puncture. Cecal ligation and puncture is the golden standard animal model for sepsis-induced critical illness [21, 22]. A timeframe of 5 days was used as this corresponds to the prolonged phase of human critical illness [23] and is highly suitable to investigate critical illness-induced muscle weakness [10, 19].
After induction of sepsis, mice received fluid resuscitation at 0.3mL/h (colloids/crystalloids, 1:4) for the first 20 h through the central catheter. Analgesics (0.3 mg/kg buprenorphine, Vetergesic, Patheon UK Ltd, Covingham, United Kingdom) and antibiotics (16.7 mg/kg imipenem/cilastatin, Aurobindo Pharma, Saronno, VA, Italy) were administered subcutaneously twice daily throughout the 5 days study period. From day 1 onward, standard mixed parenteral nutrition at a dose of 5.8 kcal/day (Olimel N7E, Baxter, Lessines, Belgium), equivalent to 40 % of normal caloric intake, was given and supplemented twice daily with either a subcutaneous bolus injection of glucose (an isovolumetric and isocaloric dose of 187.5 mg/day, further referred to as ‘placebo’) or 3HB-Na (Sigma-Aldrich, Saint Louis, MO, USA). A dose of 150 mg/day D,L-3HB-Na was used as the reference study dose, as is the dose that was previously shown to protect against muscle weakness in septic mice [10]. Animals were housed in individual house-made transparent swivel cages and placed in a temperature-controlled (27 °C) animal cabinet with 12 h light and dark cycles. Healthy control mice were pair-fed receiving standard chow (ssniff R/M-H, ssniff Spezialdiäten Gmbh, Soest, Germany) at a daily intake comparable to the daily PN dose in septic mice (5.8 kcal/day). Mice were randomly allocated to each group. Caretakers and data collectors were blinded for group allocation. The average start bodyweight of the animals was 28.9 g ± 0.2 g SEM and not different between groups (p = 0.96).
Pain/discomfort was assessed twice daily by means of the Mouse Grimace Score [24], and the summed score was used as the severity of illness score. Non-survivors were allocated the maximum severity of illness score + 1. To study the dose-responses up to the toxic/lethal 3HB-Na dose threshold, we used an up-and down dosing design starting by doubling the reference dose to 300 mg/day (Fig. 1a). If toxicity was observed, the dose was systematically reduced to identify the most effective non-toxic dose that protected against muscle weakness (Fig. 1a). In the absence of toxicity (increased severity of illness or lethality), the experiment was continued until 15 animals survived up to day 5 and ex vivo muscle force measurements could be obtained. Exclusion criteria were physical abnormalities present before surgery (n = 2), pre-randomization, death during surgery (n = 1, from PN + 300 mg/d 3HB-Na) or catheter-malfunction during experiment (n = 6, 1 from PN + 300 mg/d 3HB-Na, 1 from PN + 150 mg/d 3HB-Na, 4 from PN + 225 mg/d 3HB-Na).
Ex vivo measurement of muscle force
After 5 days of illness, surviving mice were anaesthetized (intraperitoneal injection of 100 mg/kg ketamine, Eurovet Animal Health BV, Bladel, The Netherlands, and 13 mg/kg xylazine, V.M.D. nv/sa, Arendonk, Belgium) and euthanized via cardiac puncture and decapitation. Immediately after euthanasia, the extensor digitorum longus (EDL) muscle was isolated and placed between a fixed clamp and lever-arm (300 C-LR Dual-Mode muscle lever, Aurora Scientific, Ontario, Canada) in a temperature controlled and continuously perfused organ bath (30 °C, 95 % O2 − 5 % CO2) filled with HEPES-buffered Krebs-ringer solution (0.57 mM MgSO4, 10 mM glucose, 4.5 mM KCl, 120 mM NaCl, 0.7 mM Na2HPO4 dibasic, 0.9 mM Na2H2PO4 monobasic, 5 mM MgCl2, 1.2 mM KH2PO4, 2 mM CaCl2, 10 mM HEPES, pH 7.3). Controlled pulses of 1 A were given through two platinum electrodes to stimulate the muscle. The highest produced twitch force determined the optimal muscle length (L0) for each muscle separately. The maximal isometric tetanic force was measured by averaging three consecutive tetanic stimuli (180 Hz, 200 ms duration, 0.2 ms pulse width; 2 min rest intervals). The specific maximal isometric tetanic force was determined by dividing the maximal isometric tetanic force by the muscle cross-sectional area (CSA). CSA was calculated by dividing the muscle mass by the product of the density of mammalian skeletal muscle (1.06 mg/mm3) and the optimal fiber length (Lf=0.44 x L0). Data collection was done with use of the Dynamic Muscle Analysis software (Aurora scientific).
Blood and plasma analyses
For surviving animals, after 5 days of illness, blood pH and blood Na+, K+, Cl−, HCO3−, and creatinine concentrations were measured at sacrifice with use of the Epoc® Blood Analysis System (Siemens Healthineers, The Hague, The Netherlands). Whole blood 3HB− concentrations were measured with the StatStrip Xpress®2 Glucose/Ketone meter (Nova Biomedical, Waltham, MA, USA) 30 min after injection of the study dose on day 1. In plasma collected at sacrifice, 3HB− was quantified with a commercial enzymatic kit (EnzyChrom™ ketone body assay kit, Bioassay Systems, Hayward, CA, USA). Impact of 3HB-Na on inflammation was assessed by quantification of plasma TNFα (Mouse TNF-alpha Quantikine HS ELISA Kit, R&D systems, Minneapolis, MN, USA). As ketone bodies can suppress lipolysis which may have detrimental effects during sepsis [10, 19], plasma free fatty acids (Free Fatty Acid Fluorometric Assay, Cayman, Ann Arbor, MI, USA) and glycerol (Glycerol Assay Kit, Sigma-Aldrich) were quantified. Plasma aldosterone (All species Aldosterone ELISA Kit, LSBio, Seattle, WA, USA) was quantified as a marker of the renin-angiotensin-aldosterone system, which is involved in fluid retention and can be affected by salt intake.
Tissue analyses
For surviving animals, after 5 days of illness, water content of liver, brain, lung and muscle biopsies was determined by a freeze-drying process. Tissue samples were weighed, dried at 95 °C for 6 h and weighed again. To determine gene expression of markers of tissues damage and inflammation, RNA was extracted from liver and kidney samples using the RNeasy mini RNA isolation kit (Qiagen, Hilden, Germany) and from hippocampi with the NucleoSpin RNA mini kit (Macherey-Nagel, Dueren, Germany). Liver and kidney were homogenized in Qiazol (Qiagen) and hippocampi in mercaptoethanol-supplemented RP1 buffer (Macherey-Nagel) at 6.500 rpm for 45 s with ceramic beads in a Precellys 24 machine (Bertin Technologies, Villeurbanne, France), followed by use of respective kits according to the manufacturer’s instructions. RNA concentrations were quantified by Nanodrop spectrophotometer (ND-1000, Nanodrop Technologies, Wilmington, DE, USA) and reverse-transcribed using Superscript III Reverse Transcriptase (Invitrogen, Merelbeke, Belgium) and random primers (Invitrogen). Real-time quantification of cDNA was performed with StepOne Plus (Applied Biosystems, Carlsbad, CA, USA) using commercial TaqMan assays (Applied Biosystems) for all gene expression analyses (supplementary Table 1). Data are shown normalized to hypoxanthine-guanine phosphoribosyltranferase (Hprt) and were expressed as a fold change of the mean of control mice.
Histological analyses
For animals surviving the 5 days of sepsis, hematoxylin and eosin stained formalin fixed paraffin tissue sections were used to semi-quantitatively assess changes in histological structure. Liver sections were scored for feathery appearance of cytoplasm, loss of structure, sinusoidal dilatation and infiltration of inflammatory cells [25]. To assess brain damage, hippocampal regions CA1, CA3 and dentate gyrus were scored for presence of damaged neurons, identified as neurons with shrunken eosinophilic cytoplasm and pyknotic nuclei [26]. The hippocampus was evaluated as it is the brain region most susceptible to damage during experimental sepsis and it is vulnerable to stress-induced damage [27,28,29,30]. Tissue sections were scored as 0, 1 or 2 for minimal (< 10 %), mild (10–20 %) or severe (> 20 %) abundance of aforementioned parameters. Scoring was performed by two independent observers who reached consensus in case of scoring discrepancies. Hippocampal microglia were stained with a rabbit anti-Iba1 polyclonal antibody (1:500, No.019-19741, Fuijifilm Wako chemicals, Richmond, VA, USA), followed by HRP-linked polyclonal goat anti-rabbit antibody (1:100, No.P0448, Dako, Glostrup, Denmark) and visualization with DAB (Dako). The number of microglia per mm2 were counted in the CA1, CA3 and dentate gyrus regions with ImageJ software.
Statistical analyses
Statistical analyses were performed with use of JMP Pro 14 (SAS Institute, Cary, USA). Data are presented as box plots with interquartile ranges and whiskers describing the 25th and 75th percentiles or as bars with standard error of the means. To compare differences between study groups, Student’s t-test, Wilcoxon, Log-Rank and Pearson’s chi-squared test were used, as appropriate. Two-sided p-values of ≤ 0.05 were considered significant.