Based on a previously established technology with RTCA [19, 21, 22], a prospective, single arm, observational clinical trial was conducted at the Medical University of Innsbruck between October 2017 and October 2019. The study was approved by the institutional review board of the Medical University of Innsbruck (vote number 1025/2017). All patients participating in the trial signed the respective informed consent form.

All liver grafts stemmed from donors after brain death and none of the livers underwent machine perfusion.

Forty-three consecutive patients were included in this trial. Recipient, donor, and transplant characteristics were collected and collated. Early allograft dysfunction (EAD) served as primary endpoint, Model for Early Allograft Function (MEAF [23]) Liver Graft Assessment Following Transplantation (L-GrAFT [24, 25]), graft and patient survival, length of stay and biliary complications served as secondary endpoints. EAD was defined as the presence of one or more of i) bilirubin ≥10 mg·dL⁻¹ on day seven after transplantation, ii) international normalized ratio (INR) ≥ 1.6 on day seven, and iii) alanine (ALT) or aspartate aminotransferases (AST) > 2000 IU·L⁻¹ within the first 7 days after liver transplantation [26].

Liver wedge biopsies were taken during the back-table preparation. All biopsy samples were placed in HTK solution (Custodiol^®, Dr. Franz Köhler Chemie GmbH, Bensheim, Germany) on ice for transportation prior to analysis.

Real-time live confocal microscopy assessment was performed using the following live stains: Wheat germ agglutinin conjugate (WGA; Molecular Probes, Eugene, OR, United States; 10 μg·mL⁻¹ final concentration) visualizes the tissue morphology, SYTO^®16 (Molecular Probes; final concentration 5 µM) visualizes all nuclei and propidium iodide (PI) (Molecular Probes; final concentration 500 nM) the nuclei of dead cells [20]. Incubation time was 15 min at 37°C. Real-time live confocal imaging was performed in eight-well chambered cover glasses (Nalge Nunc International). Images were acquired with a spinning disk confocal system (UltraVIEW VoX; Perkin Elmer, Waltham, MA) connected to a Zeiss Axio Observer Z1 microscope (Zeiss, Oberkochen, Germany) and visualized employing the Volocity software (Perkin Elmer) using a ×10 objective. Time for readout was approximately 5 min per sample.

High-resolution respirometry (HRR, O2k, Oroboros Instruments, Innsbruck, Austria) was applied to assess mitochondrial respiration. All measurements were carried out in O2k-chambers of 2 mL at 37°C under constant stirring at 750 rpm [27]. Data were acquired at intervals of 2 s and analysed with the DatLab software (Datlab 7.4, Oroboros Instruments, Innsbruck, Austria). Besides monthly instrumental background calibrations, before each experiment, air-calibration was performed with MiR05 mitochondrial respiration medium (MiR05-Kit, Oroboros Instruments, Innsbruck, Austria). The finally prepared medium consists of 0.5 mM EGTA, 3 mM MgCl₂ • 6 H₂O, 60 mM lactobionic acid, 20 mM taurine, 10 mM KH₂PO₄, 20 mM HEPES, 110 mM D-sucrose, 1 g·L⁻¹ essentially fatty acid free bovine serum albumin. Twenty mg of liver tissue was dissected on a cooled plate at 4°C, weighted, and subsequently homogenized in 4°C MiR05 using a PBI-Shredder O2k-Set (Oroboros Instruments, Innsbruck, Austria) according to the manufacturer’s instructions. 2-mL tissue homogenate with a final concentration of 1 mg wet mass·mL⁻¹ was immediately added into each of the O2k-chambers. Chemicals for the pre-defined substrate-uncoupler-inhibitor titration (SUIT) protocols were titrated using glass microsyringes (Oroboros Instruments, Innsbruck, Austria). The SUIT protocols (i; corresponding SUIT-025 ¹ ) and (ii; corresponding SUIT-006 O2 mt D047 ² ) are defined in the Supplementary Tables S1, S2. Each titration step was carried out after respiration reached a steady state. Measurements were performed in technical duplicates.

Respiration rates were expressed as O₂ flux per wet mass tissue [pmol O₂·s⁻¹·mg⁻¹].

Three substrate pathways delivering convergent electron flow to the electron transport system were investigated. The fatty acid oxidation (FAO)-pathway F was determined in the presence of octanoylcarnitine and a low concentration of malate, the NADH-pathway N with the substrates pyruvate, glutamate, and malate. The succinate-linked pathway S was assessed after inhibiting the mitochondrial Complex I with rotenone and adding succinate. In addition to studying these pathways separately, the combined pathways FNS feeding electrons into the coenzyme Q-junction were investigated to reconstitute the tricarboxylic acid cycle function of the living cell and determine possible additive effects [28]. For further details, see Supplementary Tables S1, S2.

Respiratory capacities were normalized to an internal reference rate for each measurement to determine flux control ratios (FCR) for evaluation of SUIT protocol (i). For SUIT protocol (ii) the coupling states LEAK (L), OXPHOS (P), and OXPHOS(c) (Pc), were evaluated [28]. LEAK, a dissipative component of respiration, was measured in the presence of the mitochondrial Complex I inhibitor rotenone and reducing substrate succinate without ADP (rate S _L ). The respiratory capacity of oxidative phosphorylation (OXPHOS) was assessed in the presence of succinate, 5 mM ADP, and 10 mM inorganic phosphate in MiR05 (S _P ). Finally, cytochrome c was added to test the integrity of the mitochondrial outer membrane, obtaining the rate S _Pc . Based on the above, the following control efficiencies were calculated for the succinate pathway: P-L control efficiency (1-L/P), to evaluate the efficiency of ATP production in the succinate pathway, and cytochrome c control efficiency (j _c = 1-P/Pc) to evaluate the damage to the mitochondrial outer membrane [28].

In each liver biopsy, 10 optical sections of 1 µm were analysed. Cell viability and matrix architecture of the liver were quantified by counting events (one event is either a viable or a non-viable cell) and groups which comprise i) total count of cells, irrespective of the localization; ii) cells from the central vein area; iii) cells from the portal triad area.

For each group, the number (total count) of viable cells was divided by the number of non-viable cells (total count) with following possible results: (+1) for highly viable biopsies/areas with more viable than non-viable cells; (0) for biopsies/areas in which the number of viable cells equals the one of non-viable cells; (−1) for those in which the number of non-viable cells outnumbers the one of viable cells. For each biopsy, a score was calculated consisting of two central vein areas and one portal triad area resulting in a maximum of +3 points in the best or −3 in the worst-case scenario.

After completion of live confocal imaging, the liver biopsy was placed and fixed in Millonig’s solution and processed for paraffin embedding. Four µm thick sections were stained by haematoxylin and eosin as per standard protocols. Light microscopy observations were carried out on a Nikon Eclipse 50i microscope (Nikon Corporation, Japan). Histological assessment was performed according to a modified Suzuki score [29] based on necrosis, steatosis, inflammation, fibrosis and vascular changes (Supplementary Table S3). An overall histopathologic score indicated i) normal liver tissue or only mild histopathologic alterations—score 3 (subscores 0 or 1), ii) moderate histopathologic alterations—score 2 (at least one subscore 2), or iii) severe histopathologic alterations—score 1 (at least one subscore 3). Slide scanning was performed on an Olympus VS120 microscope and evaluated using Olympus OlyVIA software.

The statistical testing was done with Graph Pad Prism 9 and IBM^® SPSS^® Statistics Version 25. A p-value of <0.05 was considered as statistically significant. Biopsy results (RTCA, histology scores and HRR), recipient, donor and transplant factors were analysed using parametric and non-parametric tests (including Spearman rank correlation). The RTCA score and P-L control efficiency were adjusted for clinically relevant parameters and evaluated in uni- and multivariate logistic regression analyses.

Table 2 depicts the demographics and the transplant data of 43 liver transplants stratified for EAD (N = 22, 51.2%) and initial function (IF, N = 21) following liver transplantation. The proportion of ECD was numerically, but not statistically higher in the cohort developing EAD (18/22, 81.1%) compared to patients with IF (13/21, 61.9%), p = 0.27. Recipients with EAD received livers from donors with a significantly higher BMI (28.05 ± 6.29 kg·m⁻² in EAD vs. 24.6 ± 4.16 kg·m⁻² in IF, mean ± SD; p = 0.031). The Liver and the Eurotransplant donor risk indices (DRI) were comparable between the groups. The anastomosis time was significantly longer in EAD-patients compared to patients with IF livers (47.64 ± 9.86 min in EAD vs. 40.57 ± 5.92 min in IF-patients, p = 0.02). Patients developing EAD had significantly higher mean MEAF-scores (6.68 ± 1.3, compared to liver recipients with IF, 4.77 ± 1.41, p < 0.0001). The mean L-GrAFT score was −0.63 ± 1.13 and corresponded with EAD (−0.26 ± 1.21 in EAD vs. −1.11 ± 0.82 in IF, p = 0.015).

RTCA and scoring in fresh liver wedge biopsy samples collected from donor livers after static cold storage was completed in approximately 30 min. HRR took 90 min including sample preparation. Hence, the two methods which were carried out simultaneously proved to be feasible for immediate assessment albeit requiring availability of staff and respective expertise at the point in time. Both assessments required <40 mg tissue sample (wet mass), and, if the technology is available, have a cost per analysed sample of ca. 200 EUR.

In a first step, we analysed the capacity of mitochondrial oxidative phosphorylation (OXPHOS) in liver crude homogenates for the NADH-linked, fatty acid oxidation (FAO), and succinate pathways (Figure 1A; Table 1). In the biopsies after SCS in the whole cohort (N = 43), respiration was highest for succinate-linked OXPHOS (40.4 ± 12.6 pmol·s⁻¹·mg wet mass⁻¹), with a considerable variation between grafts. In contrast, respiration was markedly lower for the FAO and NADH pathways (10.5 ± 4.3 and 4.3 ± 3.0 pmol·s⁻¹·mg wet mass⁻¹, respectively). No difference was found between the EAD and IF groups.

Next, we calculated the flux control ratios (FCR) as the single pathway capacities relative to the maximum OXPHOS respiration reached with the combination of all substrates. As shown in Figure 1B; Table 1, succinate-linked respiration alone was sufficient to saturate OXPHOS capacity. This pattern of pathway control reflects an incomplete additivity [28]. Thus, our detailed analysis of mitochondrial function and calculation of the coupling control efficiencies focused on the S pathway.

The assessment of RTCA and P-L coupling control efficiency revealed significant differences between EAD and IF livers: The mean RTCA score was significantly lower in the EAD cohort (0.75 ± 2.27 compared to 0.70 ± 2.08 in the IF cohort; p = 0.01), indicating a decreased cell viability. In agreement with the RTCA results, the P-L control efficiency was significantly better and predictive of IF (mean P-L control efficiency of 0.76 ± 0.06 in IF-livers vs. 0.70 ± 0.08 in EAD-livers; p = 0.02; Table 2; Figures 2, 3). The MEAF score correlated negatively with the RTCA score; p = 0.01, Spearman’s rho correlation coefficient −0.407; a lower viability correlated with a higher risk of liver dysfunction. Nonparametric correlation analysis showed that RTCA and P-L control efficiency are closely linked: p = 0.005, Spearman’s rho correlation coefficient was 0.493. When RTCA score and P-L control efficiency were adjusted for recipient and donor age as strongest confounders, the significance of both RTCA and OXPHOS coupling was confirmed (Table 3).

In contrast, histology did not differ between EAD and IF, although there was a trend towards better overall scores in the IF group (Table 2; Figure 4). Accordingly, none of the individual histopathological features such as necrosis, steatosis, inflammation, and vasculitis correlated with the outcome.

The 90-day mortality was 7.0% (3/43; 2 after EAD); 90-day graft loss was 4.7% (2/43; 1 after EAD). Eight patients (8/43, 18.6%, 6 after EAD) died during the follow up; three patients (3/43, 7.0%, 2 after EAD) had to undergo a re-transplant within the first year after transplant. The succinate-linked OXPHOS capacity was predictive for patient survival in the univariate Cox regression analysis.

Overall graft loss and patient death were numerically higher in the EAD group, but not significantly different in comparison to IF livers; graft loss after EAD 4/22 (18.2%) vs. IF 1/21 (4.8%), p = 0.18; death after EAD 6/22 (27.3%) vs. 3/21 (14.3%), p = 0.31. Re-transplantation was the only risk factor independently predictive for graft survival in the univariate Cox regression analysis; p = 0.04, HR 19.3, Wald 4.4. Univariate and multivariate analyses for patient survival are displayed in Tables 4, 5. The most important and independent factor for patient survival was also re-transplantation; p = 0.001, HR 105.2, Wald 10.279.