The reported data utilized material obtained during a recently published study investigating the effect of Na₂S₂O₃ in a long-term porcine model of hemorrhage (5). Experiments were conducted according to the National Institutes of Health Guidelines on the Use of Laboratory Animals and the European Union ‘Directive 2010/63 EU on the protection of animals used for scientific purposes’ after approval by the University of Ulm Animal Care Committee and the Federal Authorities for Animal Research (Regierungspräsidium Tübingen, Germany, Reg.-Nr. 1341, date of approval 02.05.2017). A total of 17 adult Bretoncelles-Meishan-Willebrand (BMW) pigs of both sexes (7 castrated males, 10 females) with a median weight of 62 kg (interquartile range (IQR) 56;67) and a median age of 15 months (IQR 13;16) were included in this study. The STS group included 4/5 male-castrated/female pigs and the vehicle control group consisted of 3/5 male-castrated/female animals. The BMW strain is characterized by a decreased activity of the von Willebrand factor resulting in a coagulatory state similar to that of human blood (45, 46).

Anesthesia and surgical instrumentation are described in detail in Messerer et al.; analyzing the impact of Na₂S₂O₃ on organ damage within the same cohort (5). Briefly, animals were randomly assigned to a control (n = 8) or Na₂S₂O₃-treatment group (n = 9), and the staff performing the experiment was blinded to the group assignment. Hemorrhagic shock was induced by passive removal of the blood with a titration of mean arterial blood pressure (MAP) to 40 ± 5 mmHg. After 3 h of hemorrhage, swine underwent resuscitation up to a maximum of 68 h, while noradrenaline was continuously administered intravenously to restore the MAP to pre-shock levels. At 2 h of shock, animals received either vehicle (NaCl 0.9%) or Na₂S₂O₃ (25 mg/kg/h for 2 h, thereafter 100 mg/kg/h until 24 h after shock, Dr. Franz Köhler Chemie GmbH, Bensheim, Germany) according to their group assignment. The choice of dose and time of administration is explained in detail in Messerer et al. (47). Briefly, it was performed in accordance with previous studies as well as established concentrations for treatment of cyanide poisoning (23, 47, 48).

Blood was drawn after the induction of anesthesia and a subsequent stabilization period (T1) as well as 24 h (T2) and 64 h (T3) after shock induction ( Figure 1 ). 14 lithium-heparin (LiHep) monovettes with a volume of 9 mL each (Sarstedt, Nümbrecht, Germany) were used to collect approximately 120 mL of arterial whole blood each time point. After 1:1 dilution with PBS (without CaCl₂, MgCl₂), the drawn blood was layered onto a two density gradient solutions (9 mL 1.119 and 8 mL 1.088 g/mL solution, Pancoll, PAN Biotech, Aidenbach, Germany). Centrifugation at 764 g without break at room temperature (RT) for 20 min resulted in a PBMC top layer and a bottom layer containing red blood cells (RBCs) and granulocytes. The cells were resuspended in water for short-time osmotic lysis to remove residual RBCs before stopping the reaction with 10 × PBS to avoid lysis of leukocytes. Osmotic lysis was applied to PBMCs only once, while three procedures were required for granulocytes due to the high fraction of RBCs in the bottom layer. After removal of all RBC contamination, cells were washed once with 1 × PBS and subsequently counted in a Neubauer counting chamber. On average, isolated cells could be used for experiments experiments 2 – 3 h after blood withdrawal was firstly initiated.

Mitochondrial respiration was measured by high-resolution respirometry using the Oroboros^® Oxygraph-2K (Oroboros Instruments, Innsbruck, Austria). This device allows for simultaneous recording of the O₂ concentration in two parallel chambers calibrated for 2 mL of mitochondrial respiration medium MiR05 (49). This medium contains 110 mM D-sucrose (Sigma Aldrich, St. Louis, MO, USA), 60 mM K-lactobionate (Sigma Aldrich, St. Louis, MO, USA), 0.5 mM ethylene glycol tetra acetic acid (Sigma Aldrich, St. Louis, MO, USA), 1 g/L bovine serum albumin free from essential fatty acids (Sigma Aldrich, St. Louis, MO, USA), 3 mM MgCl₂ (Scharlau, Hamburg, Germany), 20 mM taurine (Sigma Aldrich, St. Louis, MO, USA), 10 mM KH₂PO₄ (Merck, Darmstadt, Germany), 20 mM HEPES (Sigma Aldrich, St. Louis, MO, USA), adjusted to pH 7.1 with KOH and equilibrated with 21% O₂ at 37°C. Directly after cell isolation, 10 × 10⁶ PBMCs/granulocytes suspended in MiR05 were filled into a chamber and stirred at 750 rpm. Sealing the chambers of the device according to the manufacturers protocol started the continuous recording of mitochondrial respiration. Quantification of the oxygen flux (JO₂) was based on the rate of change in the O₂ concentration in the chambers and normalized for the cell number. Once the chambers were sealed, specific analysis of mitochondrial respiratory function was achieved by sequential injections of substrates and inhibitors into the respiration medium. Firstly, routine respiration was recorded once a stable JO₂-value was achieved after closing the chambers. Subsequently, 2.5 μM oligomycin was injected to block the ATP-synthase. This yielded the LEAK-state, which represents the respiratory activity required to maintain a stable membrane potential in absence of ATP-turnover. The titration of carbonyl cyanide p-(trifluoromethoxy)-phenylhydrazone (FCCP) in 1 µM steps allowed to achieve the maximum respiratory activity in the uncoupled state (ETS-state). The ETS state corresponds to state 3 as defined in Chance and Williams et al. (50) and is neither limited by substrate availability, cell energy demand, nucleotide availability or ATP synthase activity. Finally, 0.5 μM rotenone + 5 μM antimycin were added to block complex I and III respectively, yielding the residual (non-mitochondrial) oxygen consumption.

O 2 • − concentration in whole blood was determined immediately after blood removal. 25 µL of whole blood were mixed with an aliquot of 25 µL freshly thawed CMH (1-Hydroxy-3-methoxycarbonyl-2,2,5,5-tetramethylpyrrolidine) spin probe solution. The CMH solution contained 400 µM CMH spin probe, 25 µM deferoxamine, and 5 µM diethyldithiocarbamate to chelate transition metal ions in Krebs-HEPES-Buffer (KHB) (Noxygen, Elzach, Germany). After mixing whole blood with CMH spin probe solution, it was transferred to a 50 µL glass capillary, sealed, and measured with an EMXnano electron spin resonance (ESR) spectrometer (Bruker, Billerica, MA, USA) after 5 min incubation at 37°C (Bio-III, Noxygen, Elzach, Germany). The device settings are detailed in the Supplements. Radical concentration was quantified by comparison with a series of CP° (3-Carboxy-2,2,5,5-tetramethyl-1-pyrrolidinyloxy) radical standards solved in KHB. As a blank sample, KHB added to the respective amount of CMH spin probe solution was measured and subtracted from the sample value.

For determination of radical production by immune cells, 25 µL of a cell suspension containing 2.5 × 10⁶ cells/mL RPMI 1640 medium (Glucose 1.8 mg/mL, Glutamine 0.6 mg/mL, NaHCO₃ 100 µg/mL) were mixed with 25 µL of CMH spin probe solution. In contrast to whole blood, cell samples were measured over a 30 min interval to calculate the radical production rate. A sample of RPMI 1640 medium mixed 1:1 with CMH spin probe solution was used as a blank value for measuring cell suspensions and subtracted from sample values. Data were evaluated with the Xenon_nano software (version 1.3; Bruker BioSpin GmbH, Rheinstetten, Germany) and Microsoft Excel. Results regarding ROS determination by ESR were included in a dissertation by one of our co-authors (51). Additionally, the extracellular H₂O₂ concentration was determined in a suspension of 1 × 10⁶ PBMCs/granulocytes in 100 µL PBS after 30 min at RT. A three-electrode setup that has been previously thoroughly described was used for this purpose (52). The determination of the H₂O₂ concentration was not performed for each animal due to limited availability of the measurement device.

For investigating nutrient utilization of cells, we incubated three times 5 × 10⁶ freshly isolated cells in parallel in 1 mL RPMI containing one of the following tracers: 1,2-¹³C₂-labeled glucose, ¹³C₆-labeled glucose, and ¹³C₅-labeled glutamine (Cambridge Isotope Laboratories, Andover, MA, USA). Concentrations are specified in the Table S3 in the Supplements. The pH of the medium was adjusted to 7.4 before experimentation through addition of 1M HCl or NaOH. After incubation at 37°C for 2 h, cells were spun down and 850 µL of the supernatant was transferred to a crimp neck glass vial. The vial was frozen upside down at −20°C for later GC/MS analysis of ¹³CO₂ production and lactate released into the medium. The cell pellet was washed once with PBS and subsequently stored at −80°C after removal of all liquid. For MFA analysis, we required both the supernatant (¹³CO₂ production; mass isotopomer distributions (MIDs) of secreted lactate) and the cell pellet (MIDs of the selected metabolites lactate, glutamate, and aspartate). Samples were stored for 1 – 2 months until analysis.

The cumulative cellular ¹³CO₂ production was estimated by enrichment analysis from the spiked amount of CO₂ released from NaHCO₃ in the supernatant ( Table S1 ). The frozen supernatant was thawed, and 25 µL 1 M HCl were injected through the septum into the liquid to drive out CO₂ into the gaseous phase. For each vial, 10 replicates of 5 µL headspace gas each were injected into the GC/MS system (Agilent 6890 GC/5975B MSD, Agilent Technologies, Waldbronn, Germany) while analyzing the m/z of 44 and 45; corresponding to unlabeled and labeled CO₂, respectively. The average ratio of ¹³CO₂/¹²CO₂ amounted to 0.56% with an average standard error of 2% of the nominal value.

After ¹³CO₂ detection from the supernatant, we analyzed lactate secretion into the medium. Determination was performed by taking two aliquots of 100 µL supernatant and adding 500 µL acetonitrile. Samples were centrifuged at RT (13000 rpm for 5 min) and afterwards decanted into vials suited for derivatization. After drying in a Savant2010 SPD 2010 SpeedVac concentrator (Thermo Scientific, Waltham, MA, USA) (45°C, 14 mTorr) for about 50 min, derivatization was initiated with 100 µL acetonitrile and 25 µL N-(tert-butyldimethylsilyl)-N-methyltrifluoroacetamide (MTBSTFA). Samples were incubated at 80°C for 1 h with the lid closed and afterwards transferred into GC/MS vials. One of the initial 100 µL samples was incubated with 1 µg of internal standard (IS, corresponding to 20 µL of 50 µg/mL ¹³C₃ sodiumlactate solution) for 10 min beforehand to serve for quantification. A variety of essential calibration samples were prepared in replicates: 0.1 µg/0.2 µg/0.5 µg/0.75 µg/1 µg of lactate with an additional 1 µg of IS each, 1 µg of IS only, blank RPMI, and RPMI with 1 µg of IS. Details of calibration and quantification are specified in the Supplements.

For metabolite extraction from the cell pellet, 100 µL cold H₂O was added to the frozen pellets. The mixture was vortexed and sonicated for 10 min. Subsequently, 500 µl acetonitrile was added and the samples were centrifuged for 5 min at 13000 rpm. All samples were decanted into vials and dried for derivatization. Steps of derivatization follow those of lactate determined from medium. Standard mixes with 0.1 µg/0.2 µg/0.5 µg/0.75 µg/1 µg of analytes were prepared as control: the first with the respective amounts of lactate; the second with aspartate, glutamine, and glutamate.

For GC/MS detection, we used selected ion monitoring for optimal signal to noise ratios. Details of device settings and m/z of measured TBDMS derivatives are specified in the Supplements. Peak area integration was performed with our in-house program and MIDs were converted into carbon mass distributions (CMD) with a correction matrix approach (53, 54). MIDs were corrected for all isotopic interferences except for the natural ¹³C abundance, which is included in our CMDs.

For MFA, we established a combined model for glycolysis, the PPP and TCA cycle, which has been previously described and utilized in Stifel et al. (43). It predicts ¹³C mass distributions on metabolites based on flow rates of the metabolic system by utilizing the EMU concept (40, 55–57) and was implemented in RStan [R interface to Stan, a tool for Bayesian analysis (58)]. Comparing predictions for ¹³C mass distributions with the corresponding GC/MS measurements (section 2.6) using sampling-based Bayesian statistics allowed for identifying suitable fluxes within the network. It further estimated how the precision in measurements affects the precision of estimated fluxes, including standard deviations and confidence intervals. Conveniently, unidentifiable fluxes can be recognized by wide confidence ranges.

Our PPP estimation is built on the same method as the one used by Lee, Katz, and Rognstad (59, 60) that is based on the assumption that PPP utilization can be represented as a shift in the label (‘carbon scrambling’) of the top carbon atoms of PPP metabolites. For this approach, usually only the m+1/m+2 ratio on lactate would be used as a proxy for triose labeling using a 1,2-¹³C₂-labeled glucose input, but we expanded the method so that the complete CMD of the full metabolite as well as the CMD of the lactate fragment across carbon 2 and 3 were taken into account. The model firstly estimated relative fluxes from GC/MS measurements and subsequently utilized ¹³CO₂ production and the secretion of lactate into the medium to transform these relative fluxes into absolute values. The parallel tracer setup of 1,2-¹³C₂-labeled glucose, ¹³C₆-labeled glucose, and ¹³C₅-labeled glutamine enabled improved flux determination, as the estimated fluxes must apply to sets of measurements obtained from each tracer. The details of the metabolic model are available in the Supplements.

17 BMW pigs were included in this study. Animals were randomly assigned to a control vehicle (n = 8) or Na₂S₂O₃-treatment group (n = 9). Data are presented as median with IQR and the number of animals that could be included in the corresponding analyses are indicated in the respective figure legends. Missing values at later time points indicate animals with premature experiment termination in accordance with a list of predetermined criteria (5). Statistical and graphical presentation was performed with GraphPad Prism 9, version 9.4.1 (GraphPad Software Inc., La Jolla, CA, USA). Experimental data was considered to be non-parametric due to small sample sizes. We conducted the comparison between groups with Mann-Whitney U tests, while the effect of time within one group was analyzed with the Kruskal-Wallis rank sum test and a post hoc Dunn’s multiple comparisons test. Figure 1 uses templates provided by www.biorender.com.

To determine mitochondrial oxygen consumption, we evaluated routine respiration and ETS capacity; the former describing baseline respiration and the latter representing the maximum capacity of the mitochondria independent of factors like substrate availability, oxygen availability, ATP synthase activity and cell energy demand (50). Since the respective treatment was initiated at 2 h of shock, variances in groups caused by differential treatment should become apparent at the T2 (24 h after shock) or T3 (64 h after shock) time points, while baseline T1 was expected to be comparable ( Figure 1 ). Both routine and ETS mitochondrial oxygen consumption increased over the course of the experiment for PBMCs and granulocytes ( Figure 2 ). Generally, PBMCs demonstrated higher respiration than granulocytes. Interestingly, Na₂S₂O₃-treated PBMCs showed a significantly higher oxygen consumption compared to the vehicle-treated cells at T2, which was not present in granulocytes. By T3, this intergroup difference disappeared.

The capability to generate ROS was uncompromised, as determined by ESR spectroscopy, the gold standard for radical measurements. As depicted in Figure 3 , differences in production of O 2 • − of both cell types as well as O 2 • − levels in whole blood were non-significant between both groups at all times of measurement. When it comes to Regarding cell type specific effects, granulocytes demonstrated higher superoxide production levels than PBMCs. PBMCs slightly increased O 2 • − production over the course of the experiment, while granulocytes displayed significantly higher production levels at T2, which returned to baseline at the end of the experiment ( Figure 3B ). To further confirm the findings obtained by ESR, we determined H₂O₂ produced by PBMCs or granulocytes electrochemically and found it mirroring the ESR-measured pattern despite utilizing completely different means of detection and measuring different oxidative agents ( Figure 3C ). However, only few samples could be analyzed for their H₂O₂ concentration, making it difficult to obtain reliable statistical informative values.

We analyzed immune cell metabolism with a model formulated in RStan (visualized in Figure 4A ). Our ¹³C-based MFA method managed to successfully reproduce textbook knowledge regarding cell type specific preferences in metabolism such as granulocytes displaying higher glycolytic rates while PBMCs rather utilized the TCA cycle ( Figure 4B ) (32, 61, 62). Moreover, PBMCs displayed much higher glutamate metabolism (F8) in comparison to granulocytes: about 40 – 50% (Median 0.44; IQR 0.39; 0.48) of the main TCA cycle flux F3 originated from glutamine via glutamate in PBMCs, while it only added up to 13% for granulocytes (IQR 0.11; 0.19).

Although the glutamate input also slightly increased with shock progression for granulocytes, it only reached 22% at its highest at T3. PPP utilization was similar for both cell subsets, however, granulocytes showed completely divergent PPP utilization at the last measurement time point. So far, we could not find correlations with other parameters that could explain the two extremes.

Generally, cell metabolism reacted similarly in response to the intensive care treatment, inasmuch almost all fluxes gradually increased from T1 to T3 ( Figure 4B ). These changes were significant for PMBCs, while granulocytes only demonstrated a slight trend. The only notable exception were TCA cycle fluxes, which remained mostly constant, and PBMCs from Na₂S₂O₃-treated animals, where citrate synthesis significantly decreased at T2 ( Figure 5A ).

When it comes to the effect of Na₂S₂O₃ on immunometabolism, it seems like neither glycolytic nor PPP or TCA cycle fluxes of granulocytes were impacted by its administration ( Figure 4B ). While PBMCs also failed to show intergroup differences regarding glycolytic and PPP metabolism, there was a notable and significant difference in F4, the flux from oxaloacetate (OAA) over citrate to α-ketoglutarate, pertaining to the change over the course of the experiment ( Figure 5A ): Na₂S₂O₃-treated PBMCs significantly decreased their F4 flux from T1 to T2, while this was not the case for vehicle-treated PBMCs. Intriguingly, this difference in F4 between measurement time points highly correlated with the intergroup differences observed in routine respiration (R² = 0.8425, p < 0.0001). This negative linear regression was neither present in the control group nor in granulocytes. PBMCs from Na₂S₂O₃-treated animals further displayed a significantly increased loss in acetyl-CoA, representing the amount of acetyl-CoA exiting the described subsystems of Figure 4A , compared to vehicle-treated ones at T2 ( Figure 5B ). As an indicator of potential reductive carboxylation, we compared the m+3 ¹³C label on the complete aspartate fragment between vehicle and Na₂S₂O₃-treated animals at T2. There was no difference between groups (Vehicle: 1.94%, Na₂S₂O₃: 1.78%; p = 0.2925), implying that the higher amount of acetyl-CoA loss was not coupled to reductive carboxylation. The change in acetyl-CoA loss, like the change in mitochondrial oxygen consumption, correlated with the change in F4 ( Figure 5B ).

As we used density gradient centrifugation to isolate immune cells, it became increasingly difficult to obtain pure PBMC populations at later time points. Low density granulocytes are a population of granulocytes found in the PBMC layer of density gradients and tend to increase in number upon inflammation and, consequently, shock and resuscitation (74, 75). As they are metabolically not particularly well defined, this fraction within the PBMC population could potentially distort results at the later time points of blood withdrawal. Moreover, as PBMCs are not a metabolically uniform population, changes in one subpopulation might be masked by others. With our MFA routine, only conclusions about the whole population but not of their subsets can be drawn. Another factor to consider is potential immune cell activation during cell purification.

In summary, our novel MFA model is a suitable tool for detecting small scale metabolic alterations in immune cells. Further studies with more homogenic populations could elucidate which PBMC subset precisely is responsible for the treatment-induced metabolic effect.