Assessment of liver graft quality during hypothermic oxygenated perfusion: the first international validation study

Abstract

Background and Aim

While it is currently assumed that liver assessment is only possible during normothermic machine perfusion (NMP), there is uncertainty regarding a reliable and quick prediction of graft injury during ex situ hypothermic oxygenated perfusion (HOPE). We therefore intended to test, in an international liver transplant cohort, recently described mitochondrial injury biomarkers measured during HOPE before liver transplantation.

Study design

Perfusate samples of human livers from 10 centers in 7 countries with HOPE-experience were analyzed for released mitochondrial compounds, i.e. flavin mononucleotide (FMN), NADH, purine derivates and inflammatory markers. Perfusate FMN was correlated with graft loss due to primary non-function or symptomatic non-anastomotic biliary strictures (NAS), and kidney failure, as well as liver injury after transplantation. Livers deemed unsuitable for transplantation served as negative control.

Results

We collected 473 perfusate samples of human DCD (n=315) and DBD livers (n=158). Fluorometric assessment of FMN in perfusate was validated by mass spectrometry (R=0.7011,p<0.0001). Graft loss due to primary non-function or cholangiopathy was predicted by perfusate FMN values (c-statistic mass spectrometry 0.8418 (95%CI 0.7466–0.9370,p<0.0001), c-statistic fluorometry 0.7733 (95%CI 0.7006–0.8461,p<0.0001). Perfusate FMN values were also significantly correlated with symptomatic NAS and kidney failure, and superior in prediction of graft loss when compared to conventional scores derived from donor and recipient parameters, such as the donor risk index and the balance of risk score. Mitochondrial FMN values in liver tissues of non-utilized livers were low, and inversely correlated to high perfusate FMN values and purine metabolite release.

Conclusions

This first international study validates the predictive value of the mitochondrial co-factor FMN, released from complex I during HOPE, and may therefore contribute to a better risk stratification of injured livers before implantation.

Graphical Abstract

Introduction

Ex-situ machine perfusion is increasingly recognized as a strategy to optimize the utilization of donor livers for transplantation [1, 2]. This approach targets two key applications: first, machine perfusion provides a dynamic platform to mitigate ischemia-reperfusion injury (IRI), reducing post-transplant complications. Secondly, machine perfusion allows assessment of organ quality prior to transplantation [3, 4]. Machine perfusion has therefore gained interest worldwide, as collection and analysis of circulating perfusate and/or bile during organ perfusion may allow for quick recognition of organ injury, which in turn may aid in the selection of suitable donor liver grafts for subsequent transplantation. However, due to variation in protocols and different machine perfusion approaches by transplant centers, an objective, reliable and universally accepted method for assessment of organ viability is currently lacking.

Two main ex-situ perfusion concepts have been tested in the last five years by randomized clinical studies and are increasingly being adopted into clinical practice. Normothermic machine perfusion (NMP; 37°C) aims to provide ‘near-physiological’ conditions to induce and maintain full metabolic activity of the organ [5]. With active metabolism, biochemical parameters of perfusate and bile can be assessed for signs of cellular damage resulting from IRI, and allows insight into the injury of the biliary tree [4, 6, 7]. In contrast, (dual) hypothermic oxygenated machine perfusion ([D]HOPE) provides oxygen under cold conditions (8–12°C), using acellular preservation fluids [3, 8]. Despite the increasingly demonstrated protective ‘HOPE-effect’ in human liver transplantation, showing improved graft survival and fewer post-transplant complications [3, 9], the lower metabolic activity of the liver under hypothermic conditions has often been claimed as the main disadvantage for assessment of liver injury and function during (D)HOPE. We have, however, shown in previous work that mitochondrial function through analysis of flavin mononucleotide (FMN), a cofactor of mitochondrial complex I, is detectable during (D)HOPE, and was predictive of graft function after liver transplantation [10, 11]. Yet, these results were found in a single center cohort and are in need of external validation by other centers. As the next logical step, we performed an international observational cohort study to analyze the release of FMN during (D)HOPE, validate FMN fluorometry values with mass spectrometry analysis, and investigate the utility of FMN to perform risk stratification as a useful predictor of liver graft quality.

Methods

Study design

In this international, observational cohort study we included donor livers from 10 international liver transplant centers between 2019 and 2022. End-of-life care and donor management were according to national legislation and in accordance with ongoing, standard clinical practice for each individual center. Both DBD (donation after brain death) and DCD (donation after circulatory death) donor livers were included, respecting the considerable variability in donor management, particularly with DCD donors, between countries [1]. Both HOPE (perfusion via portal vein only) and DHOPE (dual perfusion via portal vein and hepatic artery) were used for a minimum of 2 hours prior to transplantation, using either the commercially available, pressure-regulated perfusion device Vitasmart^® (Bridge to life Ltd., Chicago, IL, United States) or the LiverAssist device (XVIVO B.V. Groningen, the Netherlands). Oxygenated (pO2 ≥ 80kPa) University of Wisconsin machine perfusion solution (MPS) was recirculated at temperatures between 8–12°C, with a maximum perfusion pressure of 3–5mmHg. Perfusate samples were collected at 60 min after start of (D)HOPE, stored in −80°C, and shipped per batch to University Hospital Zurich, Switzerland, for further analysis. All DCD livers in Zurich underwent routine perfusate analysis for FMN since 2018, and the decision whether to utilize DCD livers or not was exclusively dependent on perfusate analysis with however varying cutoffs.

An additional cohort of 38 perfused livers from the University Hospital Zurich which underwent HOPE were included, but were not transplanted after FMN assessment (further referred to as ‘non-utilized’). The non-utilized donor livers underwent a further estimation of graft quality by additional ex-situ evaluation of liver injury and function using a normothermic perfusion system, as previously reported [11]. These non-utilized livers were considered as high-risk liver grafts not suitable for transplantation, and were therefore used as a reference group for poor liver graft quality.

The study protocol relates to the ethical guidelines of the 1975 Declaration of Helsinki and was approved by the following regulatory bodies in the individual centers: Zurich University Hospital, Switzerland: 2019–01000; Bologna, Italy: RF-2016–02364732: Turin, Italy: 739/2019; Milan Policlinico, Italy: 1792/22; Berlin/Aachen, Germany: EA2/270/20; Erasmus MC Rotterdam, the Netherlands: METC 2014–060; University Medical Center Groningen, the Netherlands: METc 2014/077; Santiago, Chile: 000021/Code: D/UTH/520; Sao Paulo 0: 4.740.772/C: 45630421.0.0000.0071, Vienna, Austria: 2209/2018. When writing the manuscript, STROBE guidelines were adhered to [12].

Perfusate fluorometric spectroscopy analysis

Perfusate fluorometric spectroscopy was performed with a Cytation 3 imaging microplate reader (BioTek Cytation Hybrid Multimode Reader, Agilent Crosslab, Santa Clara, USA), as established previously [8, 10, 11, 13]. Briefly, FMN values were measured on a 96-well microplate (350 μl, Merck, Switzerland) using the following settings, set at excitation at 485nm; emission at 528nm and gain of 130. NADH values were measured, set at excitation at 360nm; emission at 460nm and gain of 100. Fluorometric measurements were calibrated using FMN standard concentrations (Sigma-Aldrich F2253, Saint Louis, USA).

Metabolite extraction and profiling by targeted liquid chromatography-mass spectrometry (LC-MS/MS)

Fifty μL perfusate per sample was used for metabolite profiling. Proteins were precipitated by the addition of two volumes of acetonitrile and an internal standard mixture, containing chloramphenicol and C13-labeled L-glutamine, L-arginine, L-proline, L-valine, and uracil, added to each supernatant (10mM final concentration), followed by vortexing for 5 min, and centrifugation (16,000 × g) at 4C° for 10 min. Pure FMN (Sigma-Aldrich F2253, Saint Louis, USA) was used to generate the LC-MS method. The protein pellet gained after centrifugation was used in a bicinchoninic acid based protein assay for normalization among samples. Metabolites in the supernatants were lyophilized, dissolved in 15 μL water, sonicated in an ice water bath for 5 minutes, and centrifuged again (16,000 × g). Five microliters of the supernatant were mixed with 45 μL water for metabolite analysis on a reversed-phase column (Reprosil-PUR C18-AQ (1.9 μm, 120 Å, 150 × 2 mm ID; Dr. Maisch; Ammerbuch, Germany) or with 45 μL acetonitrile for metabolites separated on a zicHILIC column (3.5 μm, 100 Å, 150 × 2.1 mm ID; di2chrom; Marl, Germany). Subsequent separation of the metabolites was performed on an LC instrument (1290 series UHPLC; Agilent, Santa Clara, CA, USA), online coupled to a triple quadrupole hybrid ion trap mass spectrometer QTrap 6500 (Sciex, Foster City, CA, USA), based on multiple reaction monitoring (MRM) method, as reported previously [14]. Three transitions per selected metabolite were measured. Relative quantification was performed using MultiQuantTM software v.2.1.1 (Sciex, Foster City, CA, USA). Peak integrations were reviewed manually, and peak intensities were normalized, first against the internal standards, and subsequently against protein abundances obtained from the bicinchoninic acid assay. The first transition of each metabolite was used for relative quantitation between samples. Metabolomics data have been deposited in the publicly available repository PeptideAtlas with the identifier PASS05861 and can be downloaded via http://www.peptideatlas.org/PASS/PASS05861.

Study endpoints

The main outcome parameters for this study were (1) graft loss due to either primary non-function (PNF, defined as organ failure resulting in death or retransplantation within 7 days) or symptomatic NAS (2) incidence of symptomatic (defined as need for either antibiotic treatment, radiologic, endoscopic, or surgical intervention, retransplantation or patient death) non-anastomotic strictures (NAS, defined as cholangiopathy outside the biliary anastomosis region, and in presence of a patent hepatic artery), and (3) kidney injury (defined as the need for new onset renal replacement therapy [RRT] after transplantation). Secondary outcome parameters include liver lab values during the first week after transplantation, as a surrogate marker for ischemia-reperfusion injury. The median follow up for all transplanted patients was 2.3y (1.1–6.9y).

Tissue mitochondrial analysis

To further investigate mechanistic details and the link between perfusate FMN values and tissue mitochondrial FMN content, a subset of paired perfusate and tissue samples from human donor livers that underwent HOPE-treatment at the University Hospital Zurich were analyzed. Perfusates, liver tissues and isolated mitochondria of both implanted and non-utilized livers were further analyzed for mitochondrial injury and function, and were tested in correlation with perfusate and mitochondrial FMN content. Isolation of mitochondria was done according to previous studies (11).

Perfusate injury marker analysis

Perfusate samples obtained 60 min after start of (D)HOPE were further quantified for their content of mitochondrial DNA and ctDNA, as well as proinflammatory molecules, including danger associated molecular patterns (DAMPS) such as high-mobility-group-box-1 (Hmgb1, IBL International GmbH; ST51011)), Toll-like receptor 4 and 9, inflammasome (NLRP-3), interleukins (IL-1β, IL-18) and liver transaminases (AST).

Staining

A novel staining procedure for the NDUFS1 subunit of human mitochondrial complex I (Complex I-75kD (CI-75kD) also known as the NADH-ubiquinone oxidoreductase 75 kDa subunit) was developed according to the staining protocol provided by Akoya biosciences for multiplex staining. Paraffin-embedded tissue sections (5um) were used. Heat-mediated antigen retrieval was performed with citrate buffer pH6, blocked and incubated with primary antibody overnight at 4°C (ABN302) in antibody diluent and blocking buffer. The specimen was incubated with the secondary antibody (Akoya, OPAL polymer HRP Ms+Rb) and fluorophore (Akoya, OPAL 520) at room temperature for 10min (each). Additional stainings (NFkb, CD 68 (macrosialin; MCA341R, RRID:AB_2291300)) and vWF. Quantification was done in a blinded fashion.

Statistical analysis

Data are presented using the median and interquartile range (IQR) for continuous variables and in numbers (n) and percentage (%) for dichotomous variables. The non-parametric Mann-Whitney U-test was used to test for differences between 2 groups. One-way ANOVA (Kruskal-Wallis) was used for testing of multiple non-parametric data. A p value of <0.05 was considered statistically significant. Pearson’s parametric correlation was used to assess the relationship between two metric variables. Receiver-operating characteristic (ROC) curves were used to determine the threshold of highest sensitivity and specificity (Youden index) in human perfusate FMN samples for liver graft loss, symptomatic NAS, and kidney injury requiring RRT. Metabolite activities were measured in duplicates or quadruplicates at 25° C. All activities are shown in mmol substrate/min/mg protein. All data were analyzed using SPSS (IBM v.24.0; Armonk, NY: IBM Corp.) and Graph Pad Prism, version 10.0 (San Diego, CA, USA).

Results

International comparison of perfusate FMN by fluorometry and mass spectrometry

We included a total of 473 perfusate samples collected during (D)HOPE-treatment of 315 DCD and 158 extended criteria DBD livers (Figure 1A) performed at 10 centers from 7 countries (Figure 1B). Of these, 435 livers were transplanted (91.7%), whereas 38 livers were considered not suitable for transplantation (non-utilized) after FMN assessment. Twenty DCD livers (13 in Milano, and 7 in Turin) underwent mandatory NRP prior to HOPE (20/435, 4.6%). Relevant donor and recipient information of implanted livers is provided in Table 1.

Figure 1. Study design, correlation of fluorescence and mass spectrometry.

The study refers to 473 machine liver perfusates, 435 livers were implanted and 38 livers were discarded. All perfusates were analyzed fluorometrically for FMN, 163 perfusates were additionally analyzed by mass spectrometry. (A). Ten centers from 7 countries participated in this study (B). Calibration of fluorometric measurements allowed calculation of μg FMN/ml perfusate (C). Next, MRM transitions for FMN are shown with overlay of 3 chromatograms for FMN specific ion transitions in positive ionization mode: m/z 457→ (439, 359, 243)(D). The corresponding fragment structures are shown on the right. The areas above the red lines were used for peak integrations (Gaussian smoothing width: 2 points. MRM, multiple reaction monitoring; m/z, mass over charge). Mass spectrometry for FMN correlated with fluorometric values (E)(Pearson’s parametric correlation). The single values of all centers are visualized with medians and IQR in scatter plots for fluorescence and mass spectrometry (F, G, respectively). P values are two taild and refer to discarded livers (Mann-Whiney U test, levels of significance p < 0.01.

Table 1:

Donor and recipient parameters of implanted livers from 10 international centers (n=435)

	DBD n=158	DCD n=277
Donor age (y)	61 (51–75)	58 (49–65.3)
Donor BMI (kg/m2)	27 (24–31)	25 (23–28)
Cold ischemia time (CIT)(min)	360 (290–459)	294 (233–344)
Functional warm ischemia time (min)	-	29 (26–33)
HOPE perfusion time (min)	183 (116–330)	154 (120–218)
Donor risk index (DRI)	1.94 (1.56–2.30)	2.48 (2.16–2.85)
Recipient age (y)	59.5 (51–64.3)	60 (51–65.3)
Recipient MELD	15 (9–25)	14 (10–19)
Primary transplant (n/%)	3/158 (1.9)	6/277 (2.2)
Peak AST (U/l)	865 (500–1501)	2177 (1006–4310)
Peak ALT (U/l)	586 (308–1077)	1176 (610–2414)
Peak Creatinine (μmol/l)	123 (75–186)	182 (91–276)
Intensive care unit stay (days)	3.5 (2–6)	2 (1–4)
Hospital stay (day)	15 (10–21.8)	15 (12–24)
Primary non function (PNF)(n/%)	3/158 (1.9)	8/277 (2.9)
Non anastomotic stricture (NAS)(n/%)	8/158 (5.1)	28/277 (10.1)
Graft loss due to PNF, NAS (n/%)	10/158 (6.3)	31/277 (11.2)
Kidney failure post TPL (RRT)(n/%)	15/158 (9.5)	72/277 (25.9)
5y Graft survival (%)	95.8 ± 2.1	79.8 ± 4.6

First, we aimed to identify the presence of FMN in (D)HOPE perfusates among different international centers. Fluorometric absorbance measurements were strongly correlated with FMN standard concentrations within the linear range (R²=0.99) (Figure 1C). Secondly, we identified FMN abundance in the perfusate samples through mass spectrometry analyses (Figure 1D). In a subset of 163 perfusate samples, fluorometric measurements of perfusate FMN values correlated with FMN abundance measured through mass spectrometry (R=0.70, p<0.0001) (Figure 1E). Next, we compared FMN perfusate values of implanted livers from all involved centers with non-utilized livers from Zurich, separated for DBD and DCD livers. FMN fluorometric measurements were significantly different in implanted livers compared to non-utilized livers, with the exception of Milan, Turin, Rotterdam, and Sao Paulo (Kruskal-Wallis; p>0.99, p=0.45, p>0.99, and p=0.26, respectively), (Figure 1F), possibly related to the presence of free hemoglobin in samples from livers after NRP pretreatment (Millan, Turin), or to a sample size effect (Rotterdam, Sao Paulo). In the subset of 163 perfusate samples of which FMN values were measured by mass spectrometry, all centres showed significantly lower FMN values for transplanted livers when compared to non-utilized livers (Kruskal-Wallis; all p<0.05; Figure 1G).

Perfusate FMN predicts graft loss, cholangiopathy and kidney failure

Next, for all livers that were subsequently transplanted, we sought to estimate the predictive value of FMN levels for clinically relevant outcomes, related to any injury the graft may have suffered due to the donation, preservation, or the ischemia and reperfusion process upon transplantation. We analyzed the predictive value of perfusate FMN values for graft loss due to either primary non-function (PNF) or cholangiopathy. We found a c-statistic of 0.77 (95% CI 0.70–0.85, p<0.0001) for perfusate FMN by fluorometry (Figure 2A) and a c-statistic of 0.84 (95% CI 0.75–0.94, p<0.0001) for perfusate FMN analysis by mass spectrometry (Figure 2B). Likewise, perfusate FMN analysis was also predictive for the occurrence of symptomatic NAS with a c-statistic of 0.75 (95% CI 0.69–0.80, p<0.0001) by fluorescence (Figure 2C), and with a c-statistic of 0.77 (95% CI 0.65–0.88, p<0.0001) by mass spectrometry (Figure 2D). In addition, perfusate FMN values had a strong predictive value for post-transplant kidney failure requiring RRT, with a c-statistic of 0.76 (95% CI 0.71–0.81, p<0.0001) by fluorometry (Figure 2E), and with a c-statistic of 0.71 (95% CI 0.59–0.83, p<0.0014) by mass spectrometry (Figure 2F). Detailed results of the ROC analysis is available in Supplementary Table 1.

Figure 2. Predictive capacity of perfusate FMN by fluorescence and mass spectrometry.

Receiver operating characteristic (ROC) curves based thresholds for clinical decisions on perfused liver grafts are illustrated for fluorometric perfusate FMN measurements or mass spectrometry for graft loss (A and B), non-anastomotic strictures (C and D) and renal replacement therapy (E and F). The area under the curve (AUC) represents the concordance statistic (C statistic), quantifying the overall ability of the test to discriminate between positive and negative values. The reported p values test the null hypothesis that the AUC equally 0.5, according to Berrar D et al (supplementary reference 1). Confidence intervals are calculated by Prism using the hybrid Wilson/Brown method (supplementary reference 2–3).

Accordingly, we calculated the maximal Youden index of the ROC curves to define safety thresholds for accepting donor livers by fluorometry (Figure 3A) as wells as mass spectrometry (Figure 3B). Of note, perfusate FMN values of DBD and DCD livers were relatively homogenously distributed (Figures 3A,B). Low risk livers were defined as those that did not result in graft loss, cholangiopathy or kidney injury post-transplantation. Intermediate risk livers were defined as those that did result in cholangiopathy and/or kidney injury, but not in graft loss post-transplantation. High-risk livers were defined as those that result in graft loss, and/or cholangiopathy, and/or kidney injury. Perfusate FMN threshold values, as measurered by fluorometry for low to medium risk livers, was set at 5165 A.U. (0.017 μg/ml), whereas for medium to high-risk livers this was set at 6415 A.U. (0.021 μg/ml). For FMN values measured by mass spectrometry, cut off values for low to medium risk livers was 63.100 A.U. and for medium to high-risk livers this was 110.000 A.U., respectively. These results show that perfusate FMN values, detected during (D)HOPE, by either fluorometry or mass spectrometry analysis is related to key outcome parameters after transplantation.

Figure 3. Determination of thresholds for accepting or discarding liver grafts during HOPE.

Proposed acceptance, risk and discard thresholds determined by the maximal Youden index of the receiver operating characteristics (ROC) curves by fluorescence (A) and mass spectrometry (B). ).

Perfusate FMN correlates with first week lab values

We proceeded in relating perfusate values of FMN with injury markers and liver function after transplantation. For example, INR levels at day 1, 2 and 7 correlated significantly with perfusate FMN values, as well as liver transaminases (Supplementary Figure 1A–F, I), while serum bilirubin and peak creatinine showed no correlation (Supplementary Figure 1G, H) with perfusate FMN values. These results indicate that the degree of mitochondrial injury occurring during procurement and preservation, is associated with impaired liver graft function after implantation.

Perfusate FMN correlated with mitochondrial function and purine metabolites

In another step, we analyzed machine liver perfusates, collected in the first hour of (D)HOPE, to assess the effect on other mitochondrial processes, including direct complex I FMN content, citric acid and purine metabolism. Initially, we observed that high perfusate FMN values inversely correlated with the FMN content of mitochondrial complex I in several exemplary livers (Figure 4A). The key role of complex I in proton pumping, and thus ATP synthesis, was further supported by the strong correlation of mitochondrial FMN values with mitochondrial ATP (R=0.93, p<0.0001, Figure 4B). Consistently, complex I bound mitochondrial FMN was prerequisite for complex I function (R=0.68, p<0.0001, Figure 4C), and NADH presence was associated with FMN levels in perfusate (R=0.29, p<0.0001, Figure 4D) pointing to impaired oxidation of NADH at increased FMN release, i.e. complex I injury. These results suggest that perfusate FMN is released by mechanism of mitochondrial injury during HOPE. With regard to other metabolites, we observed a significant increase of succinate compared to fumarate, and a significant increase of adenylate breakdown products AMP, IMP, inosine, hypoxanthine, compared to ADP and ATP (Figure 4E, F), which was significantly more pronounced in DCD livers (Supplementary Figure 2). Accordingly, perfusate FMN values also showed moderate correlation with perfusate NAO⁺, succinate, ADP and IMP (Figure 5A–D), as well as with other purine and citric acid metabolites (Supplementary Figure 3A–D). Of note, xanthine and uric acid values were less increased, possibly due to inhibition of xanthine oxidase through the presence of allopurinol in the machine perfusion solution. These findings point to increased hypoxic conditions before or during organ procurement, and consequently to a high need of ATP during the first hour of ex-situ perfusion. Perfusate FMN values did not correlate with hepatocyte enzyme release (AST, ALT), underlining its mitochondrial origin (Supplementary Figure 4).

Figure 4. Correlation of mitochondrial FMN and perfusate FMN, correlation of perfusate FMN and complex I function.

Perfusate FMN inversely correlated with mitochondrial FMN (A), and mitochondrial FMN correlated with mitochondrial content of ATP and IMP (B). Consistently, complex I bound mitochondrial FMN correlated with complex I activity (C), and perfusate FMN was associated with perfusate NADH (D). Perfusates collected within the first hour of HOPE showed an increasing level of purine metabolites, underlining mitochondrial activity under hypothermic conditions (E, F). P values are two taild (Mann-Whiney U test, levels of significance p < 0.01 (A,F). Pearson’s parametric correlation was used in B, C, D, levels of significance p < 0.01, two-tailed p values.

Figure 5. Perfusate FMN and purine metabolites.

Perfusate FMN measure by mass spectrometry shows positive correlations with perfusate (A) NAD, (B) succinate, (C) ADP and (D) IMP. Pearson’s parametric correlation was used in A-D, levels of significance p < 0.01, two-tailed p values.

Mitochondrial injury and downstream inflammation in discarded livers

As a proof of concept, in a subset of perfused livers we compared paired liver perfusates and tissues of implanted livers with samples of non-utilized livers. For this purpose, we investigated mitochondrial injury and downstream inflammation, i.e. mitochondrial DNA release, toll like receptor activation, inflammasome activation and cytokine release. These parameters were correlated with stainings for the FMN docking site at complex I (NDUFS-1), release of NFκB, Kupffer cell activation (CD 68), and endothelial cell activation (vWF). All tested parameters showed significant differences between discarded and implanted livers (Figures 6 and 7). These findings confirmed that non-utilized livers displayed high levels of mitochondrial injury and inflammation.

Figure 6. Mitochondrial injury and inflammation in implanted and discarded livers (liver tissue samples).

Liver tissue of discarded and implanted livers were compared for changes at the NDUFS1 region of complex I, for NFκB, CD 68, and for vWF (A) with respective quantifications (B). P values are two taild (Mann-Whiney U test, levels of significance p < 0.01 (B).

Figure 7. Mitochondrial injury and inflammation in implanted and discarded livers (perfusate analysis).

Schematic of mitochondrial-related inflammatory cascades (A). Inflammatory markers differed significantly between perfusates of discarded and implanted livers (B). P values are two taild (Mann-Whiney U test, levels of significance p < 0.01 (B).

Donor and recipient parameters are inferior compared to perfusate analysis

Finally, we investigated the predictive value of conventional donor and recipient scores in the same cohort. We used the donor risk index (DRI) and the balance of risk core (BAR), which however both showed only weak prediction of graft loss (c-statistics 0.5997 and 0.5488), compared to perfusate FMN values (Supplementary Figure 5A, B). These results underline that donor and recipient parameters are not useful for evaluation of liver graft quality.

Discussion

In this observational cohort study involving 10 international centers, we demonstrate that perfusate FMN values, measured during (D)HOPE, enable to stratify liver grafts into different risk categories correlating to patient outcomes. This offers a new and highly needed tool to select individual grafts for transplantation, where the final decision remains in the hand of the medical team taking into account the risk they are willing to take.

Mitochondria are regarded as the primary trigger of ischemia reperfusion injury and play therefore an important role in organ transplantation [15]. Machine perfusion before transplantation modifies mitochondrial metabolism by delivery of both, oxygen and metabolites. At the same time, machine preservation perfusates are an easily obtainable and highly informative source of organ cellular and subcellular injury, due to analysis of perfusate compounds, i.e. biomarkers. While such assessment of liver mitochondrial injury by FMN release during (D)HOPE has previously been reported [10, 11], it remained to be demonstrated whether these results could also be replicated in other, independent, international transplant centers.

Our results show that perfusate FMN values, measured during the first hour of (D)HOPE of human liver grafts, correlated with graft loss, with symptomatic NAS, and with kidney failure requiring RRT after liver transplantation. In addition, perfusate FMN values inversely correlated with mitochondrial FMN levels, underlining its mitochondrial origin. Third, increased perfusate FMN values were associated with markers of downstream inflammation, including mitochondrial DNA, danger associated molecular patterns (DAMPs), and cytokines. Fourth, and importantly, perfusate FMN values had higher predictive performance than conventional risk scores based on donor or recipient factors, and have also been shown to be superior than other conventional perfusate markers as for example AST, ALT or lactate [10, 11]. Consistently, we found in this study that DCD livers can have low perfusate FMN signals, while DBD livers potentially range at a very high level (Figure 1F,G). This underlines the strength of an objective perfusate assessment apart from the conventional donor information, including graft type.

Whether FMN analysis in liver machine perfusate can not only predict PNF or kidney injury but also later occurring biliary injury, is currently debated. For example, several researchers rely only on testing biliary viability under normothermic and therefore more physiologic conditions, as livers produce minimal-to-no bile in the cold [16]. Mitochondrial aerobic energy production plays, however, an important role in the development of biliary injury, as it generates ATP necessary for bile salt transporters [17]. Accordingly, mitochondrial injury leads to ATP depletion, and secondly triggers an inflammatory status of cholangiocytes by DAMP signalling [18]. Third, mitochondrial ATP also controls the cystic fibrosis transmembrane conductance regulator (CFTR) protein on cholangiocytes, which in turn regulates bicarbonate secretion [19]. Consecutively, defective CFTR causes changes in bile composition and mucinous obstruction [20]. For all these reasons, mitochondrial injury is likely to have many consequences, not only for hepatocytes but also for cholangiocytes, which explains why monitoring mitochondrial injury during (D)HOPE is also informative for assessment of biliary injury. This has important clinical implications, as many centers currently perform HOPE only for treatment, but not for assessment of livers. We show here, in contrast, that assessment of livers during HOPE is possible by a relatively simple and rapid fluorometric measurement of perfusate, which is even feasible by real time monitoring, as previously reported [21]. Consistent with our analysis, several recent reports have been published on the key role of mitochondrial function for evaluation organ quality in livers and kidneys before implantation. This includes for example high resolution respirometry [22], real time confocal microscopy [23], and circulating cell free mitochondrial DNA in perfusates or flush solutions [24]. We propose therefore further evaluation of perfusate FMN together with additional mitochondrial biomarkers, such as oxidative phosphorylation function, micro-RNA, proteomics, or lipidomics.

This study has limitations. We focused on one biomarker only, despite good prediction of graft loss (c statistic 0.8418 and 0.7733). Yet, while sensitivity, and the specificity values are useful and widely accepted for the general accuracy, methods for optimal cutoff findings are debatable [25]. We relied here on the maximum Youden index for threshold determination, but discrimination against other thresholds was relatively weak. These values are therefore not absolute but should be interpreted more as a guideline threshold rather than as a sharp cutoff. In addition, while the predictive value of FMN outperforms conventional risk scores used for organ acceptance at the time of donor offer, no perfusate measurement accounts for subsequent intraoperative and recipient risk factors that may also result in NAS, kidney injury requiring RRT, and PNF. We recognize that every liver assessment to date, including ours, is a snapshot of a dynamic process. As a methodological limitation, perfusate fluorometry can result in higher values in the presence of free hemoglobin despite probe centrifugation, which we observed in some of the HOPE perfusate samples of DCD livers pretreated by NRP (centers Milan and Turin). Mass spectrometry analysis appears more accurate in this respect, resulting also in better ROC curves (Figure 2). Another limitation is the lack of proof of clinical non-function in discarded livers. Non-utilized livers in Zurich were therefore either tested on normothermic ex situ perfusion systems for ischemia reperfusion injury, or underwent biochemical analysis for mitochondrial injury (Figure 6,7), with solid confirmation of the predicted injury. Finally, all perfusates were collected and shipped from perfusion centers with dry ice to either Zurich (fluorometric and ELISA measurements) or Berlin (mass spectrometry). We regard this strategy, however, as an important step for centrally testing a new method before establishing center derived own perfusate assessment. In future, real time prospective assessment of machine perfusates for released FMN will be possible by specific sensors, providing centers calibrated fluorescence values in perfused livers. Such a trial is planned and will be registered.

We envision that FMN can be used as a biomarker to accept a liver graft when the perfusate FMN values are in the low risk area. We suggest furthermore, that intermediate or high-risk perfusate FMN values may be a reason to add further graft evaluation by for example controlled oxygenated rewarming and NMP after HOPE/DHOPE, with functional assessment of biliary injury by bile analysis [4,7]. Our study also underlines, that machine perfusion omitting assessment, for example by FMN release, does not ensure a good outcome.

In conclusion, this first international study validates the predictive value of the mitochondrial cofactor FMN, released from complex I during (D)HOPE. It may therefore contribute to a better risk stratification of injured livers before implantation.

Highlights.

• Fluorometric perfusate analysis for flavin mononucleotide (FMN) correlates to mass spectrometric analysis

• Perfusate FMN values are predictive for graft loss, cholangiopathy, and kidney failure

• The release of mitochondrial FMN relates to the degree of inflammatory activation during hypothermic oxygenated perfusion (HOPE)

Impact and Implications.

Analysis of 473 perfusates, collected from 10 international centers during hypothermic oxygenated perfusion (HOPE), revealed that mitochondria derived flavin mononucleotide (FMN) values in perfusate is predictive for graft loss, cholangiopathy, and kidney failure after liver transplantation. This result is of high clinical relevance, as recognition of graft quality is urgently needed to improve the safe utilization of marginal livers. Ex-situ machine perfusion approaches, such as HOPE, are therefore likely to increase the number of useable liver grafts.

Abbreviations

ALT

Alanine Aminotransferase

AST

Aspartate Aminotransferase

DBD

Donation after Brain Death

DCD

Donation after circulatory death

PNF

Primary Non Function

NMP

Normothermic Machine Perfusion

NAS

non anastomotic strictures

HOPE

Hypothermic Oxygenated Perfusion

RRT

renal replacement therapy

Data availability statement:

Metabolomics data have been deposited in the publicly available repository PeptideAtlas with the identifier PASS05861 and can be downloaded via http://www.peptideatlas.org/PASS/PASS05861.