Animal ethics (no human participants)
The animal experiments were conducted at the Volcani experimental farm (Rishon LeZion, Israel). All the procedures were approved by the Volcani Animal Care Committee (permit 885/20 IL), and the experiments were carried out by experienced personnel certified to work with sheep. This manuscript follows the ARRIVE reporting guidelines [32].
Handling of diabetic sheep
The choice of sheep as an experimental model for diabetes [33] was based on their similar body weight to that of humans, which enabled the utilization of similar drug doses and the collection of sufficient biological sample quantities for accurate metabolite measurements and analyses. For the same reasons, the results from this preclinical trial may be better translated to humans, and treatment doses of glycerol employed in this preclinical trial may serve as a good starting point to guide such future clinical trials. The selection of individual animals for the experiment was based on clinical veterinarian check-ups to ensure good health and physical status.
Six yearling ewes of the Afec-Assaf breed [30]approximately 11 months old with similar body weights ( ~ 54 kg), were selected from a cohort born in the same lambing period (around January). They were uniformly fed a maintenance diet to facilitate consistent administration of drug doses and to minimize the involvement of confounding factors. Unfortunately, one animal died a week after diabetes induction and was subsequently excluded from the final analysis. Insulin-dependent diabetes was induced by chemical destruction of pancreatic beta cells with alloxan, essentially as previously described [34,35,36]. Briefly, a filter-sterilized 5% alloxan-saline solution was administered intravenously as a single dose of 50 mg/kg body weight (BW) [36, 37]. As expected, the animals developed Type-1-like diabetes within 2-3 days, as confirmed by blood glucose values > 200 mg/dL [38] (Scheme 1), and their reliance on exogenous insulin administration to protect against hazardous metabolic values of blood glucose and β-hydroxybutyrate (BHB).
Timeline and experimental design for Experiments 1 and 2, for testing the capacity of glycerol, alone and as an insulin adjunct, to inhibit diabetic ketosis. Type-1-like diabetes was induced with alloxan infusions. Two days after confirming diabetes, the sheep were maintained on insulin therapy to manage the condition. The diabetic sheep were then randomized into GLY-SAL or SAL-GLY treatment sequences, and ketosis was induced through partial insulin withdrawal combined with empagliflozin administration. In Experiment 1, the treatments consisted of either 1 L saline (SAL) or 5% glycerol in saline (GLY). In Experiment 2, the treatments included 5 IU insulin supplemented with either 0.5 L SAL or GLY. Blood and plasma samples were collected repeatedly at indicated time points for up to 5 hours post-infusion. GLY: Glycerol. SAL: Saline. IV: intravenous.
If left untreated with insulin, these diabetic animals rapidly develop severe hyperglycemia and hyperketonemia within days (Fig. S1), closely resembling the presentation of DKA percipitation. Thus, strongly suggesting that virtually no residual endogenous insulin activity remains in the alloxan-treated sheep. These preliminary representative data from one such animal (as additional cases posed a high risk of lethality) indicate that the observed reduction in diabetic ketosis by glycerol (Fig. S1) is virtually independent of insulin.
As such, the animals were maintained on daily subcutaneous human insulin (Insulatard®) (Novo Nordisk A/S, Bagsvaerd, Denmark), according to the manufacturer’s instructions ( ~ 0.3–1.0 IU/Kg BW). The animals were fed a regular 100% maintenance diet based on the NRC guidelines [39] until the initiation of the experiments. Prior to the intravenous (IV) treatments, a single catheter (Delta Med s.p.a., Viadana, Italy) was installed into the jugular veins of each sheep, and as needed to prevent blood coagulation, the catheters were flushed with 3-5 mL of 20 IU/mL heparin (Merckle GmbH., Blaubeuren, Germany) in sterile saline (0.9% NaCl, Teva Medical Marketing Ltd., Ashdod, Israel).
Experimental design
To efficiently study the effects of the treatments, i.e., glycerol vs. control (saline or water), a 2 × 2 cross-over design was employed to minimize between-animal variation and confounding factors. For a complete randomized design, we estimated based on previous work [40]that 8 sheep would suffice to detect a difference of 20% in blood KB at a statistical power of 85% and a significance level of 0.05. However, we expected the cross-over design to be significantly more sensitive than a parallel trial and theoretically require half the number of animals to achieve the same statistical power. Therefore, five sheep (N = 5) were randomized, a priori, into two treatment sequences: [1] glycerol-control or [2] control-glycerol, where the individual animals served as the experimental units. Randomization was performed using the RAND function in Excel (Microsoft Office 2019). Therefore, each animal was treated with both interventions (glycerol and control) sequentially at a random order, allowing 12 h of “Washout” time (equating at least twice the time of the measured effects) between interventions to minimize carryover effects, as depicted in Scheme 1. Noteworthy, the carry-over effects, if any exist, are expected to be negligible for a treatment with energy substrates that are relatively rapidly metabolized, particularly as measured previously for glycerol [40].
Three (3) experiments were conducted with the diabetic sheep, each addressing a specific objective.
Experiment 1 aimed to study short-term effects of bolus glycerol treatments on diabetic ketosis, categorically defined as 1.5 mM < BHB < 3 mM. Ketosis was induced by partial insulin withdrawal ( ~ half the maintenance dose) and subcutaneous administration of 10 mg empagliflozin (dissolved in 4 ml of 50% ethanol) as a single dose per day. The IV-infused interventions consisted of 1 L filter-sterilized solution of saline (SAL) as a control, or of 5% glycerol (GLY) (Glycerin USP grade, 99.5% pure, C.S. Chemicals Ltd., Haifa, Israel) dissolved in saline; both administered sequentially to each animal at a random order (Scheme 1). The interventions were initiated only once sub-clinical ketosis was confirmed (typically ~ 8 h after empagliflozin injection). For each intervention, blood samples of 6 mL were collected into heparin vacutainers (BD., Plymouth, UK), immediately before the IV interventions, and at regular intervals for 5 hours post infusion (30, 60, 120, 180, 240 and 300 min, Scheme 1).
Experiment 2 aimed to evaluate the benefit of a combined treatment of insulin and glycerol on the resolution of severe diabetic ketosis, categorically defined as BHB > 3.0 mM, as typified in DKA. Current urgent care practice to resolve DKA employs initial treatment with insulin only, and, as needed, glucose may be supplemented a few hours later to prevent hypoglycemia. Therefore, Experiment 2 aims to shed light on the potential advantage of a combined hormonal-and-energetic therapy provided at once to target both DKA and hypoglycemia right from the start, for potentially safer and more efficient DKA resolution. The animals were IV treated sequentially with 5 IU of insulin supplemented with either 0.5 L of sterile saline only (SAL) as the control or 0.5 L of filter-sterilized 5% glycerol dissolved in saline (GLY), using the same timeline and diabetic ketosis induction procedures detailed for Experiment 1 (Scheme 1). For Experiments 1 and 2, the animals were housed as a group and fed a regular maintenance diet for ~2 weeks. To minimize diet-induced energetic fluctuations, a day before the initiation of the interventions and during the 5 h of monitoring time, the sheep were deprived of the concentrate (high energy feed) and served with straw only.
Experiment 3 aimed to investigate longer-term effects of dietary glycerol consumption on diabetic ketosis to explore the potential benefit in reducing the risk for DKA under SGLTi therapy. Diabetes was controlled with human insulin, as described above. For this experiment, insulin was administered mainly in the morning, whereas at around 5 pm, the animals were pretreated with subcutaneous 10 mg empagliflozin to induce moderate ketosis overnight. The diabetic sheep were housed individually to ensure precise control over their glycerol intake in the drinking water and their feed consumption, which consisted of a maintenance diet ration. The drinking interventions were provided immediately after empagliflozin administration by serving each sheep with 5 L of either water (control) or 5% glycerol in water (GLY) for overnight consumption. Each animal was treated with GLY or regular water for 7 consecutive nights, followed by one “Washout” day and a treatment swap for another week. Blood samples were collected daily at 8 am, before the morning feeding, for biochemical analysis (Scheme 2).
Timeline and experimental design for the nutritional study (Experiment 3). The diabetic sheep received daily insulin injections each morning and with empagliflozin each evening. Before the dietary interventions, the animals were randomized into one of two treatment sequences: GLY-WATER or WATER-GLY. Each intervention consisted of providing 5 L of 5% glycerol (GLY) or water) daily at approximately 5 pm for overnight consumption over 7 consecutive nights per period. Blood and plasma samples were collected each morning at around 8 am to measure BHB, glucose, and NEFA concentrations.
Blood biochemical analysis
Blood BHB and glucose concentrations were determined in real-time using the Freestyle Optium Neo glucometers (Abbot Diabetes Care Ltd., Oxfordshire, UK). The heparinized blood was centrifuged at 2000 × g at 4 °C for 15 min to isolate plasma samples, which were immediately stored at −80 °C until further biochemical analyses. Concentrations of plasma non-esterified fatty acids (NEFA) were determined using a NEFA analysis kit (Wako Chemicals, GmbH, Neuss, Germany).
Statistical analysis
All the statistical analyses were performed in JMP (VERSION 15.1.0, SAS Institute Inc., Cary, NC, USA). Continuous dependent variables (glucose, BHB, and NEFA), measured repeatedly over time, were analyzed using the JMP mixed-model approach via repeated measures ANOVA. The model included: Treatment (GLY vs. SAL) as a within-subject fixed factor, Sequence (GLY-SAL and SAL-GLY) as a between-subject fixed factor, Time as a within-subject nominal fixed factor nested within Treatment, Treatment x Sequence Interaction, and Individual Animal as a random factor nested within Sequence. Access to the raw data may be provided upon request.
To further quantify the effects of the Treatment factors on the response variables, in a manner that is independent of the sampling time, two additional statistics were employed – Area Under the Curve (AUC) and “Delta”. AUC was computed, using the trapezium rule, as the total area between the response curve and the linear line that connects the first and the last measured data points. The Delta statistic represents the difference between the baseline and the absolute maximum value obtained, and is expressed as a percentage of the baseline value. Statistical analyses of the treatment effects on the AUC and Delta statistics were performed using the standard least squares fit model in JMP at a significance level of 0.05. The model included: Treatment (GLY vs. SAL), Sequence (GLY-SAL and SAL-GLY), and Treatment x Sequence interaction as fixed factors, as well as Individual Animal as a random factor nested within Sequence.