SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES
  8. Supporting Information

Livers exposed to warm ischemia (WI) before transplantation are at risk for primary nonfunction (PNF), graft dysfunction, and ischemic biliary strictures, all associated with ischemia/reperfusion injury (IRI). Our multifactorial approach, Leuven drug protocol (LDP), has been shown to reduce these effects and increase recipient survival in WI/IRI-damaged porcine liver transplantation. The aim was the identification of the molecular mechanisms responsible for the hepatoprotective effects of the LDP. Porcine livers were exposed to 45 minutes of WI, cold-stored for 4 hours, transplanted, and either modulated (LDP group; n = 3) or not modulated (control group; n = 4). In the LDP group, the donor livers were flushed with streptokinase and epoprostenol before cold perfusion; the recipients received intravenous glycine, a-1-acid-glycoprotein, FR167653 (a mitogen-activated protein kinase inhibitor), a-tocopherol, glutathione, and apotransferrin. Liver samples were taken before WI and 1 hour after reperfusion. Gene expression was determined with microarrays and molecular pathways and key regulatory genes were identified. The number of genes changed between baseline and 1 hour after reperfusion was 686 in the LDP group and 325 in the control group. The extra genes in the LDP group belonged predominantly to pathways related to cytokine activity, apoptosis, and cell proliferation. We identified 7 genes that were suppressed in the LDP group. These genes could be linked in part to the administered drugs. New potential drug targets were identified on the basis of genes induced in the control group but unaffected in the LDP group and interactions predicted by the literature. In conclusion, the LDP primarily resulted in the suppression of inflammation-regulating genes in IRI. Furthermore, the microarray technique helped us to identify additional gene targets. Liver Transpl 18:206–218, 2012. © 2011 AASLD.

The demand for livers for transplantation has increased over the last decennium almost exponentially, but this increase has not been followed by a similar increase in the number of livers offered by conventional (brain-dead) donors.1 Because liver transplantation (LT) is the only effective treatment for irreversible liver failure, there is a strong demand for the optimal utilization of donor organs, including those from extended criteria donors and donation after cardiac death (DCD) donors.2, 3 A substantial number of donor livers are not being used because of the high risk for primary nonfunction (PNF), delayed graft function, and ischemic-type biliary strictures. Two factors contributing to these risks are the immediate ischemic damage in the donor and the additional damage during reperfusion in the recipient; altogether, these lead to ischemia/reperfusion injury (IRI).4 Warm ischemia (WI) enhances the severity of IRI and negatively affects the outcome.

The mechanisms of organ damage after IRI have been studied extensively and consist of complex interactions involving multiple inflammatory pathways. The major contributors to IRI include the production of reactive oxygen species, the release of proinflammatory cytokines and chemokines, and the activation of immune cells; altogether, these cause inflammation and result in tissue damage.5-8 Interventions for limiting this inflammatory cascade are considered beneficial for the outcome after transplantation. In small animal models, single drugs specifically targeting WI damage and IRI have been investigated, and these have shown beneficial effects on the liver.9-17 There have been only a few small controlled studies in humans that have investigated the effects of single compounds targeting IRI in LT.18-21

Our group has developed a preclinical pig model for LT to investigate tolerance for WI and the concomitant impact of IRI in this model. WI for 30 to 45 minutes has been found to be associated with an unacceptably high risk of PNF (>50%) in our model.22 This is in line with observations in humans, for whom 20 to 30 minutes of WI is considered safe by the American Society of Transplant Surgeons.23, 24 In our model, we have documented that the following factors play important pathogenic roles in IRI and PNF: Kupffer cell activation, the overproduction of proinflammatory cytokines, the depletion of antioxidants (α-tocopherol and glutathione), the high release of redox-active iron (which catalyzes the formation of cytotoxic reactive oxygen species),25 and enhanced serum phospholipase A2 activity.26 Accordingly, we designed the Leuven drug protocol (LDP), a multifactorial cocktail of selected drugs that specifically target these various mechanisms. The LDP eliminates the risk of PNF in livers exposed to 45 minutes of WI (which are otherwise destined to fail) and improves the biliary toxicity profile; this suggests a potential protective effect against the later development of ischemic-type biliary strictures.25 Recently, Moussavian et al.27 also used a combination of drugs in an isolated rat liver perfusion model and found a protective effect when it was applied as preconditioning for organ preservation. Except for this last work (which focused on the treatment of donors), no studies using a cocktail approach are available in the LT literature.

Although the impact of the LDP on clinical outcomes after LT is favorable, 2 important questions remain unanswered: (1) can we identify which complex molecular pathways are influenced by this LDP cocktail approach and which genes are successfully targeted, and (2) can we further improve the LDP by identifying the pathways that contribute to IRI but are still left unaffected?

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES
  8. Supporting Information

Experimental Transplantation

As a model of IRI, we set up a porcine model of LT as described previously.22, 25 Inbred female Landrace pigs that were weight-matched (30-35 kg) were used as donors and recipients for 2 groups of experimental animals. Under general anesthesia (isoflurane, atracurium, and fentanyl), the liver was first mobilized through a midline laparotomy, and the infrahepatic vena cava and the aorta were prepared for cannulation. Then, for the induction of a well-defined and reproducible period of WI, cardiac death was actively induced by ventricular fibrillation, and a clamp was then placed on the thoracic aorta.

Control Group (n = 4)

After exposure to in situ WI for 45 minutes (the period from the initiation of cardiac arrest to the start of cold perfusion), the donor livers were flushed out through the portal vein and through the hepatic artery with 10 L of a 4°C preservation solution (histidine tryptophan ketoglutarate), procured, cold-stored (4 hours), and transplanted. Before reperfusion, the grafts were flushed with cold Ringer's solution. Graft reperfusion was established immediately after the completion of the suprahepatic vena cava and portal vein anastomosis. Subsequently, the infrahepatic vena cava and the hepatic artery were reconstructed. Bile was drained externally and re-administered twice daily through a feeding jejunostomy. During the operative procedure, the arterial blood pressure and the central venous pressure were monitored. The intravenous infusion of fluids was individually guided by clinical signs of hypovolemia, hemodynamic parameters, and laboratory blood analysis. During the anhepatic phase, severe hypotension (mean arterial pressure ≤ 40 mm Hg) for up to 30 minutes was treated with the intravenous administration of 500 mL of oxypolygelatin. Postoperatively, the animals received tacrolimus by mouth (0.05 mg/kg twice daily).

LDP Group (n = 3)

These livers were exposed to in situ WI for 45 minutes, and this was followed by preflushing at gravity through the hepatic artery and through the portal vein with 1 L of warm (37°C) Ringer's solution containing 500,000 U of streptokinase and 250 μg of epoprostenol. Subsequently, the livers were flushed with 10 L of 4°C histidine tryptophan ketoglutarate, procured, cold-stored (4 hours), and transplanted (as described previously). The recipients received the following drugs. Glycine and α-1-acid glycoprotein were added to separate 500-mL infusions of Plasma-Lyte (Baxter, Lessen, Belgium) and were given as continuous intravenous infusions over 6 hours from the start of transplantation. The other reagents—FR167653 (Astellas, Osaka, Japan), glutathione (Tationil, Roche/Teofarma, Italy), α-tocopherol (Trolox, Calbiochem, Merck KGaA, Darmstadt, Germany), and apotransferrin—were administered as continuous intravenous infusions. These drips were started before the anhepatic phase and were given intravenously over 4 hours. The animals in the 2 groups received the same amounts of fluids. Aside from the administration of different drugs in the LDP group, similar anesthesiological and operative procedures were applied in the 2 groups.

The experiments were performed in accordance with international guidelines for animal welfare and were approved by the local ethics committee of the university.

Determinations in the Serum of Transplanted Animals

Blood samples were sequentially collected before laparotomy and 15 minutes, 60 minutes, 180 minutes, 12 hours, 36 hours, and 60 hours after reperfusion. The tubes were centrifuged for 10 minutes at 3000 rpm and room temperature, and the serum was stored at −20°C until the analysis.

Transaminase

Aspartate aminotransferase (AST) was measured with a standard ultraviolet absorption technique in serum samples sequentially collected after reperfusion so that hepatocellular damage could be assessed.

Liver Function

Lactate was measured in arterial blood samples as a reflection of the capacity for recovering from metabolic acidosis after reperfusion.

Serum Tumor Necrosis Factor α (TNF-α)

Serum concentrations of TNF-α, a proinflammatory cytokine generated in the liver mainly by Kupffer cells, were assessed with a porcine-specific enzyme-linked immunosorbent assay kit (R&D Systems, Abingdon, United Kingdom).

Survival and Graft Function

The clinical endpoints were PNF and recipient survival 60 hours after reperfusion. PNF was defined as persisting encephalopathy (the inability of recipients to wake up after LT despite the standard cessation of all anesthetics 60 minutes after reperfusion), irreversible metabolic acidosis (a lactate level > 7 mmol/L, pH < 7.1) within the first 180 minutes after reperfusion despite standard medical treatment, profound hypoglycemia (<30-50 mg/dL) despite repeated attempts to correct blood sugar levels, severe coagulopathy (massive hemorrhagic ascites and a prothrombin time < 20%-30% within the first 180 minutes after reperfusion), and very low (<2 mL) or no production of bile during the first 180 minutes after reperfusion.

RNA Extraction and Microarrays

Tissue samples were collected at the baseline (before WI) and 1 hour after reperfusion from the transplanted livers, snap-frozen in liquid nitrogen, and stored at −80°C. RNA was isolated by the TRIzol extraction method according to the manufacturer's instructions (Invitrogen, Carlsbad, CA).

From each sample, 2 μg of total RNA was reverse-transcribed into double-stranded DNA, which was converted into biotin-labeled target antisense RNA with the Affymetrix in vitro transcription labeling kit (Affymetrix, High Wycombe, United Kingdom). After purification (GeneChip sample cleanup module, Affymetrix), the yield and purity of the labeled amplified RNA were analyzed.

The labeled RNA was subsequently hybridized to the Affymetrix GeneChip porcine genome array according to the manufacturer's instructions. This array contains 23,937 probe sets, which represent 23,526 transcripts from the domestic pig (Sus scrofa). Image analysis was performed with Affymetrix GeneChip operating software.

Data Processing, Statistics, and Annotation

The results were analyzed with Affymetrix Microarray Suite 5.0 software in combination with the Bioconductor package.28 The significance of expression changes from the baseline was evaluated with a moderated t statistic (Limma 3.2.1). The resulting P values were corrected for multiple testing with the Benjamini-Hochberg method29 to control the false discovery rate.

Because the Affymetrix GeneChip porcine genome array is only partially annotated, we used the corresponding human annotation of the porcine microarray30 for the functional analysis. When genes were represented by multiple probes on the microarray, the average of the fold change (FC) values was calculated and used.

Real-Time Polymerase Chain Reaction (PCR)

For confirmation of the microarray findings, quantitative real-time PCR was performed for a selection of genes. Primers and probes were designed with Primer3 version 0.4.0 or were taken from published porcine nucleotide sequences (see Supporting File 1). Real-time quantification was performed with the Applied Biosystems 7500 fast real-time PCR system. The FC was calculated with the ΔΔCt method and was compared to the FC obtained by the microarray analysis.

Analysis of Molecular Interactions and Pathways

To explore the molecular pathways, we calculated the 2-log FC in the expression from the baseline to 1 hour after reperfusion for each gene in the LDP group and in the control group.

To determine the gene pathways and key regulatory genes, we used 2 different bioinformatics programs.

First, we used Gene Map Annotator and Pathway Profiler (GenMAPP) 2.1/MAPPFinder software analysis31 with reference pathways from the Gene data set Hs-20060526 to identify relevant molecular pathways. Among statistically differentially expressed genes, we included those genes that behaved similarly within a group (corrected P < 0.05 from the baseline); as an additional criterion, we used a 50% up or down FC (2-log FC >+0.58 or <−0.58).

The list of significantly differentially expressed genes generated by Bioconductor was loaded into the program and analyzed. In this way, we could identify processes [Gene Ontology Identification (GOID) numbers]32 in which a significant number of genes (3 or more genes that changed in the same direction resulted in P < 0.05 for that pathway) with changed expression (up-regulation and/or down-regulation) were present. Here we should note that the program can report only the number of genes that are differentially expressed and not the actual activity of the corresponding pathway; this requires a different technique such as enzymatic activity measurement or apoptosis detection.

Second, for the identification of key regulatory genes (those interacting with a high number of other genes), we used Search Tool for the Retrieval of Interacting Genes/Proteins (STRING) 8.3.33 We focused on the genes that displayed the strongest change [100% up or down FC (2-log FC >+1 or <−1)] and accepted a slightly higher variation (uncorrected P < 0.005).

STRING 8.3 weights and integrates information from numerous sources, including experimental repositories, computational prediction methods, and public text collections; it thus acts as a meta-database mapping all interaction evidence onto a common set of genomes and proteins.34 In addition to a graphical representation of the predicted interactions, the program provides a list of the most relevant gene-to-gene interactions (a confidence score). Using this score, we could determine the key regulator genes as those interacting with the greatest number of other proteins.

Statistical Analysis of the Serum Parameters

The data are presented as means and standard deviations. For the comparison of differences between the LDP group and the control group, an analysis of variance was applied. If the normality criteria were not met, the Mann-Whitney rank-sum test was applied. A P value < 0.05 was considered significant. The analysis was performed with SigmaStat 3.5 (Jandel Scientific Software, San Rafael, CA).

RESULTS

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES
  8. Supporting Information

Confirmation of Functional Characteristics of the Porcine Transplantation Model With 45 minutes of WI

This study is an extension of our previously reported LT study with the LDP.25 In that study, we examined serum parameters (the TNF-α, lactate, AST, and alanine aminotransferase levels and prothrombin times), bile flow and composition, and IRI by light microscopy. For the current study, we performed additional LT procedures to obtain a sufficient number of quality samples suitable for molecular analysis. We included 4 livers in the control group and 3 livers in the experimental (LDP) group, for which we present select functional characteristics.

Transaminase

After reperfusion, the AST level tended to be lower in the LDP group versus the control group at 15 and 60 minutes and almost equal at 180 minutes (Fig. 1A). This observation is in agreement with the findings of our previous study with a larger groups of animals.25

thumbnail image

Figure 1. Determination of the serum parameters for (A) liver damage (AST), (B) metabolic acidosis (lactate), and (C) inflammation (TNF-α). The serum parameters were determined sequentially at the baseline and after transplantation and restoration of the flow through the liver to monitor the early signs of WI/IRI. *P < 0.05. †Death of animals in the control group after PNF.

Download figure to PowerPoint

Liver Function

The lactate level was higher in the controls versus the LDP group 60 and 180 minutes after reperfusion. The lactate level returned to normal (1.5 ± 0.07 mmol/L) within the first 60 hours after reperfusion in the survivors of the LDP group only (Fig. 1B).

Serum TNF-α

The TNF-α level was determined during the first 3 hours after reperfusion. It remained almost the same in the LDP group in comparison with the baseline values. In sharp contrast, the TNF-α level increased significantly in the control pigs after reperfusion in comparison with the baseline values (Fig. 1C).

Survival and Graft Function

No PNF was seen in the LDP group; all 3 recipient animals survived until day 3 after reperfusion, at which time the experiment was terminated. Three animals in the control group died within the first 12 hours after reperfusion, and the fourth died on day 1 after LT; all deaths were due to PNF. These survival results are in full agreement with the findings of our previous study.25

Data Processing

We computed the changes in gene expression [corrected P < 0.05 with a 50% up or down FC (2-log FC >+0.58 or <−0.58)] in the LDP group and in the control group (1 hour after reperfusion versus the baseline at 0 hours). The results were computed from paired samples taken from the same animal at the baseline and 1 hour after reperfusion. This allowed us to exclude individual variations and to study a reduced number of animals. In the LDP group, we had 794 probes corresponding to 686 individual genes, and in the control group, we had 369 probes corresponding to 325 individual genes. Using a Venn diagram, we could identify 468 genes specific to the LDP group, 218 genes present in both groups, and 107 genes found only in the control group (Fig. 2); these 107 genes were further analyzed for regulatory genes with STRING (see the Discussion section). The top 20 genes that were up-regulated and the top 20 genes that were down-regulated in the LDP and control groups are shown in Supporting Table 1.

thumbnail image

Figure 2. Venn diagram of genes with changed expression that were specific to the LDP group or the control group or belonged to both. This graphic representation shows the overlap between the 794 probes corresponding to 686 individual genes in the LDP group and the 369 probes corresponding to 325 individual genes in the control group.

Download figure to PowerPoint

PCR Confirmation of the Array Results

We used pig-specific real-time PCR to verify the array results. Overall, there was a good resemblance between the results of the microarray and quantitative real-time PCR; we had already demonstrated this more extensively in our previously published microarray study of pigs.35 Genes that were up-regulated or down-regulated in the microarray experiment reacted in the same direction in real-time PCR. The dissimilarity between the expression levels of interleukin-6 (IL-6) and CXCL2 (chemokine [C-X-C motif] ligand 2) from the 2 techniques was probably due to the higher sensitivity of the real-time PCR technique for these genes (Supporting File 2).

Molecular Pathways

The lists of genes for the LDP and control groups (each with its corresponding expression data) were separately loaded into the GenMAPP program, and the significantly altered Gene Ontology (GO) pathways were calculated. The pathways found for both experimental conditions (control and LDP) are involved in DNA and RNA transcription and signal transduction. In the LDP group, there were changes in expression in processes related to cytokine activity, apoptosis, stress, and cell proliferation (Table 1).

Table 1. Significantly Altered Molecular Pathways
GO Pathways Found in Both Control and LDP-Treated Animals
GOIDGO NamePermuted P Value
UpDown
6512Ubiquitin cycle0.010.01
6950Response to stress0.03 
7165Signal transduction0.0070.004
7242Intracellular signaling cascade0.04 
8270Zinc ion binding0.050.04
45893Positive regulation of transcription, DNA-dependent0.03 
45941Positive regulation of transcription0.010.01
45944Positive regulation of transcription from RNA polymerase II promoter0.010.01
46872Metal ion binding0.040.04
50896Response to stimulus0.010.02
GO Pathways Found Only in Control Animals
GOIDGO NamePermuted P Value
UpDown
5829Cytosol0.050.04
8380RNA splicing 0.05
15630Microtubule cytoskeleton 0.05
GO Pathways Found Only in LDP-Treated Animals
GOIDGO NamePermuted P Value
UpDown
  1. NOTE: The data sets of the genes that were significantly changed for the 2 conditions (n = 686 for the LDP group and n = 325 for the control group; see Fig. 2) were loaded into GenMAPP. The results include these lists of GOID numbers, pathway names, and permuted P values, which have been corrected for the total number of genes within the pathways. Only pathways with a P value <0.05 have been included. The pathways are listed according to their GOID numbers.

5125Cytokine activity0.020.007
5198Structural molecule activity 0.03
5515Protein binding0.05 
5576Extracellular region<0.001<0.001
5615Extracellular space0.020.009
5737Cytoplasm<0.001<0.001
5975Carbohydrate metabolic process 0.05
6809Nitric oxide biosynthetic process0.05 
6915Apoptosis0.020.04
6952Defense response<0.001 
8083Growth factor activity0.050.03
8150Biological process0.001<0.001
8219Cell death0.005 
8283Cell proliferation0.0040.004
8289Lipid binding0.04 
8544Epidermis development0.04 
9058Biosynthetic process0.0070.03
16758Transferase activity, transferring hexosyl groups0.03 
19725Cell homeostasis0.05 
42127Regulation of cell proliferation0.030.02
42742Defense response to bacterium 0.04
43066Negative regulation of apoptosis0.05 

Key Regulatory Genes

From the differentially expressed genes with an FC <−1 or >+1 and an uncorrected P value <0.005, STRING could annotate 404 genes in the LDP group and 368 genes in the control group. After processing with the software, the genes reported to have the most interactions were identified. The program also assigned a confidence score (0-1) to each reported interaction, with the highest value indicating the most confirmation from the literature. We thus extracted the key regulatory genes from the STRING confidence score lists. Twenty key genes that interacted with 5 or more genes were identified, and these key genes were present for one or both of the conditions with a 2-log FC from the baseline >1 (Supporting File 3). The P values for this change from the baseline were corrected for multiple testing to control the false discovery rate, and this resulted in adjusted P values (see Tables 2 and 3). These genes were significantly changed from the baseline in the LDP animals, the control animals, or both; for the additional comparison of response we used a cut-off: a difference in 2-log FC > 0.58 or <−0.58).

Table 2. Genes Responsive to the Drug Treatment
Gene NameInteractions From STRING (n)Gene Expression From Baseline (Mean 2-log FC)*Change in Gene Expression: LDP to Control (2-log FC)
ControlLDP
  • NOTE: Key regulatory genes whose expression was affected by the LDP (determined with STRING 8.3 with an interaction term for expression) are listed. Differentially expressed genes of the LDP and control groups were loaded into the software, and this resulted in 2 separate lists of protein-protein interactions with a confidence score greater than 0.95. All genes that were reported to have 5 or more interactions were identified as key regulatory genes. The genes affected by LDP were defined according to the difference in the change in gene expression [(2-log FC for the LDP group at 1 hour versus the LDP group at 0 hours) − (2-log FC for the control group at 1 hour versus the control group at 0 hours) >+0.58 or <−0.58].

  • *

    Corrected P values are shown in parentheses.

IL-6153.48 (0.049)2.18 (0.003)−1.30
IL-8135.81 (0.02)4.78 (<0.001)−1.03
JUN222.36 (0.04)1.75 (0.03)−0.61
MMP171.10 (0.047)0.25 (0.25)−0.85
PTGS2113.24 (0.02)2.25 (0.03)−0.99
SERPINE183.83 (0.04)4.61 (<0.001)0.78
STAT3131.53 (0.049)0.62 (0.14)−0.91
Table 3. Genes Not Responsive to the Drug Treatment
Gene NameInteractions From STRING (n)Gene Expression From Baseline (Mean 2-log FC)*Change in Gene Expression: LDP to Control (2-log FC)
ControlLDP
  • NOTE: Key regulatory genes that were not modified by the LDP (determined with STRING 8.3 with an interaction term for expression) are listed. Differentially expressed genes of the LDP and control groups were loaded into the software, and this resulted in 2 separate lists of protein-protein interactions with a confidence score greater than 0.95. All genes that were reported to have 5 or more interactions were identified as key regulatory genes. Criterion was a change in gene expression between LDP and control (2-log FC) between +0.58 and −0.58.

  • *

    Corrected P values are shown in parentheses.

  • †New potential drug targets.

CCL293.73 (0.04)3.87 (0.002)0.14
CEBPB81.20 (0.06)1.36 (0.002)0.16
CSF150.97 (0.12)1.00 (0.049)−0.03
FOS†203.69 (0.047)3.37 (0.07)−0.32
ICAM1101.98 (0.049)1.75 (0.004)−0.23
IL-10†121.25 (0.02)1.17 (0.004)−0.08
IL-1A†91.33 (0.10)1.25 (0.02)−0.08
IL-1B†132.08 (0.06)1.94 (0.01)−0.14
MYC82.23 (0.08)2.29 (0.004)0.06
NFKBIA†71.87 (0.06)1.73 (0.001)−0.14
PLAU62.63 (0.02)2.93 (0.003)0.30
TIMP152.50 (0.03)2.22 (0.001)−0.28
VEGFA†151.22 (0.07)1.05 (0.005)−0.17

In Table 2, we present the 7 genes that were the most influenced by the LDP. Six genes were repressed by the drugs in the cocktail, and only 1 gene [serpin peptidase inhibitor E1 (SERPINE1)] was up-regulated; all the genes could be linked to inflammation-related processes.

Other regulatory genes whose expression changed between the baseline and 1 hour after reperfusion but for which clear differences were not found between the LDP and control groups are listed in Table 3. From this list, in accordance with their predicted interactions (Supporting File 4), we identified 6 potential drug targets [FOS, IL-1A, IL-1B, IL-10, nuclear factor of κ light polypeptide gene enhancer in B cells inhibitor α (NFKBIA), and vascular endothelial growth factor A (VEGFA)], which we suggest could be included in a new optimized formulation of the LDP.

Using the STRING bioinformatics program, we also investigated genes that were specific to the control group (107 genes; see Fig. 2). Here we identified 3 additional regulatory genes [dual specificity protein phosphatase 1 (DUSP1), early growth response protein 1 (EGR1), and secreted phosphoprotein 1 (SPP1)] that were strongly up-regulated in the control group between the baseline and 1 hour after reperfusion and were only slightly repressed by the LDP. We performed a literature search to investigate their possible involvement in IRI and LT. This survey confirmed their role in IRI processes, and the results are presented in the Discussion section.

DISCUSSION

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES
  8. Supporting Information

The porcine LT model used in this study was previously developed and extensively characterized by our group.22 The mechanisms of PNF and its associated complications in this model have been reported in detail.26, 36, 37 On the basis of these findings and the literature, a number of potential targets for drug intervention were identified. Subsequently, the LDP was selected and applied to our LT model, and it was shown to dramatically improve outcomes.25 The molecular changes induced by the LDP have not been investigated previously. An understanding of these processes is important to (1) better understand the mechanisms of hepatoprotection in general and those mechanisms associated with the LDP in particular, (2) attempt to link the compounds in the LDP to observed genetic changes, and (3) determine whether additional gene targets (unaffected by the LDP) can be identified.

To address these questions, we performed a microarray study with our porcine LT model. We performed new LT experiments to obtain sufficient tissue samples suitable for molecular analysis. The clinical outcomes of these additional experiments were in agreement with our previous observations, and this confirmed that the transplantation of livers exposed to 45 minutes of WI resulted functionally in profound tissue damage (AST; Fig. 1A), metabolic acidosis (lactate; Fig. 1B), and up-regulation of inflammatory cytokines (TNF-α; Fig. 1C). As also observed previously, the LDP decreased these parameters, eliminated PNF, and dramatically improved survival.

Expression data from the microarray showed that only a small portion of the 23,526 genes that were present on the array were significantly changed between the baseline and 1 hour after transplantation/reperfusion (325 genes in the control group and 686 genes in the LDP group). The reasons for these limited changes are probably multifactorial. First, the study period was short because we were mainly interested in identifying the first steps in the organ response and the early pathways and genes that could initiate the subsequent WI/IRI-induced damage. In favor of that explanation, the numbers of genes that were differentially expressed were considerably higher (a 5-fold increase) in a previously published porcine model of acute fecal peritonitis when specimens were analyzed 21 hours after the onset of sepsis.35 Second, genetic changes may have been limited by the absence of blood in the organ during preservation because lymphocytes and other blood components are known to play roles in the inflammatory response.6-9 Third, the period between the baseline and postreperfusion biopsy included 4 hours of cold storage, and hypothermia halts the enzyme activity needed for gene expression or messenger RNA and protein breakdown.

The reliability of the pig microarray results from the Flanders Institute for Biotechnology MicroArray Facility (Leuven, Belgium) was validated in our previous sepsis model with real-time PCR.35 We acknowledge that enlarging the number of animals in our experiments may result in the unmasking of more genes, but this would be probably limited to genes around the statistical threshold for up- or down-regulation and would not affect the already significantly changed genes.

GenMAPP compares a list of input genes (from the microarray) with more than 1500 GOID pathways; in general, there are 30 to 100 genes in each GOID. A pathway is identified only when a statistically defined fraction of genes is changed.31 GenMAPP analysis mapped the genes in the LDP group to 32 GOIDs and the genes in the control to 13 GOIDs. Ten GOIDs were present in both the LDP and control groups; 3 were specific to the control group, and 22 were identified only in the LDP group (Table 1). This analysis shows that most of the 22 pathways affected by the LDP are involved in cytokine activity, (anti-)apoptosis, cell proliferation and related processes, and processes related to metabolism.

Nearly all the GOID numbers in the control group (10 of 13) were also found in the LDP group. This indicates that 45 minutes of WI does not induce specific molecular pathways (1 hour after reperfusion) that would necessarily trigger PNF. This suggests that more time is required for WI to have an effect on gene expression. When we looked at the liver histology at this time, the number of neutrophils and apoptotic hepatocytes was low and did not differ between the control and LDP-treated animals25; this also supports the hypothesis that gene and protein expression after WI/IRI needs more time to develop. In contrast, the effect of the LDP seems to take place at the transcriptional level very early after reperfusion, and the expression of a large number of molecular pathways is affected. Therefore, we speculate that the LDP acts by activating mechanisms preventing injury caused by inflammation and apoptosis rather than by repairing early damage inflicted during WI.

Using STRING, we further aimed to identify the key regulatory genes playing a central role in the WI/IRI response that are affected by the LDP and those that are not affected by the LDP and thereby identify new potential targets for an improved/optimized protocol. Starting from lists of significantly changed genes from paired observations, we tried to go a step further by comparing 2 sets of experiments (control LT versus LDP LT). This last step somewhat reduced the statistical power of the analysis, but this was regained when we evaluated the pattern of combined genes (as we also did in the GenMAPP/pathway analysis).

We found that central to the early gene expression response were 20 genes in the control and LDP groups that interacted with 5 or more other genes from the list of differentially expressed genes (Tables 2 and 3). These 20 genes were divided into 2 groups: 7 that were significantly affected by the LDP (Table 2) and 13 that were only marginally affected (Table 3). The functions of these 20 genes were investigated with the National Center for Biotechnology Information Entrez gene database,38 which summarizes all available information for each specific gene (see Supporting File 3). Because their individual functions have been reviewed recently, we refer readers to the literature for further reading.39-43 Briefly, the 7 most significantly changed genes were inflammation-related [IL-6, IL-8, and cyclooxygenase 2 (COX2)/prostaglandin-endoperoxide synthase 2 (PTGS2)], were transcription factors/regulators [JUN (jun proto-oncogene) and signal transducer and activator of transcription 3 (STAT3)], or were related to enzymatic activity on the extracellular side of the cell membrane [eg, tissue remodeling; matrix metalloproteinase 1 (MMP1) and SERPINE1]. These 7 genes were all up-regulated in the controls (Table 2), but 6 of the 7 genes were repressed by 1 or more components of the LDP. Streptokinase and epoprostenol can be linked to MMP1 and SERPINE1. Glycine, apotransferrin, and the 3 antioxidant compounds (glutathione, α-1-acid glycoprotein, and α-tocopherol) are expected to be active against IL-6, IL-8, and COX2/PTGS2. A direct target for the mitogen-activated protein kinase (MAPK) inhibitor FR167653 could not be clearly identified within these 20 genes. It could be that MAPKs are mostly active at a later stage of the inflammatory cascade or in WI/IRI-induced processes. An argument in favor of the inclusion of an MAPK inhibitor in the LDP comes from STRING. Indeed, several of the 13 nonresponsive genes (Table 3) have been reported to interact with MAPKs (FOS with MAPK1 and MAPK3 and IL-1B with MAPK14). In addition, DUSP1 (identified among the 107 WI-specific genes and discussed later) can interact with several MAPKs (MAPK1, MAPK3, MAPK8, MAPK9, MAPK10, MAPK11, MAPK12, and MAPK14), and EGR1 (also among these 107 genes) acts on FOS and in that way indirectly acts on MAPK1 and MAPK3.

Among the 107 differentially expressed genes found only in the controls (Fig. 2), 3 genes were identified as potential drug targets with STRING. The first target is MAPK phosphatase 1/DUSP1, which is a regulator of the human liver response to transplantation44 that signals endoplasmic reticulum stress and cell survival.45 DUSP1, JUN, EGR1, FOS, and PTGS2 have been identified as genes associated with transcriptional responses in the mouse hippocampus after transient forebrain ischemia.46 The second target is represented by the transcription factor EGR1, which mediates inflammatory gene expression and brain damage after transient focal ischemia.47 On the other hand, it has been shown that EGR1 antisense oligodeoxyribonucleotide has a protective effect against myocardial injury induced by ischemia/reperfusion and hypoxia/reoxygenation.48 Furthermore, the suppression of EGR1 may contribute to the protective mechanisms in hepatic IRI.49 Also, EGR1 has been suggested as a potential target of cardioprotective calcium channel blockers for IRI in rats.50 The third target is SPP1/osteopontin (OPN), which was initially considered a proinflammatory cytokine but has now been found to play a broader role in immune regulation and stress responses.51 OPN, which is expressed in tubular epithelial cells, regulates natural killer cell–mediated kidney IRI. Inhibiting OPN expression at an early stage of IRI may thus be protective and preserve kidney function after transplantation.52 OPN has also been found to be protective against cardiac IRI through late preconditioning.53 Thus, DUSP1, EGR1, and SPP1/OPN are strongly up-regulated after WI/cold storage and LT; they are only slightly suppressed by the drugs used in the current LDP, and they are known from the literature to be involved in WI/IRI and, therefore, are potential candidates for additional drug targeting.

The 13 unaffected key genes can be classified as cytokines [IL-1A, IL-1B, IL-10, chemokine (C-C motif) ligand 2 (CCL2), and colony-stimulating factor 1 (CSF1)], transcription factors [FOS, CCAAT/enhancer-binding protein β (CEBPB), MYC, and NFKBIA], or extracellular active/cell adhesion genes [intercellular cell adhesion molecule 1 (ICAM1), plasminogen activator urokinase (PLAU), tissue inhibitor of metalloproteinase 1 (TIMP1), and VEGFA]. On the basis of the mutual interactions of the 7 + 13 + 3 genes in STRING (see Supporting File 4), we propose 6 genes (FOS, IL-1A, IL-1B, IL-10, NFKBIA, and VEGFA) as new/additional candidates to be targeted for improving the hepatoprotective effect of the LDP. All 6 of these genes (see Supporting File 3) are involved in the proinflammatory nuclear factor kappa B, proapoptotic c-Jun N-terminal kinase, or hypoxia-inducible factor 1α hypoxia pathways. JUN and FOS have been shown to play roles in liver ischemic preconditioning.54, 55 Idebenone56 and curcumin57 target these genes, and this probably accounts for their protection against IRI. Other hepatoprotective drugs that target the production of pro- or anti-inflammatory cytokines (ie, IL-1A, IL-1B, and IL-10) include adrenomedulin,58 transresveratrol,59 tacrolimus,60 and erythropoietin.17 This last molecule is part of the hypoxia-related mechanisms that also involve VEGF. Alternatively, the nuclear factor kappa B pathway can be targeted with carbon monoxide–releasing molecule 2,61 epigallocatechin-3-gallate,62 oxymatrine,63 or nuclear factor kappa B decoy oligodeoxynucleotides.11, 64

How Can We Position Our Model in the Clinical Context of LT, and What Are Its Limitations?

For this study, we transplanted porcine livers exposed to WI before cold preservation. With this model, we attempted to mirror the clinical setting of DCD LT. However, we acknowledge that our model is not truly a controlled Maastricht category 3 DCD model. We simulated WI by actively inducing cardiac arrest via ventricular fibrillation. This is different from the clinical situation of controlled DCD. Indeed, in controlled DCD, an agonal phase follows the withdrawal of ventilatory support, and this results in a variable period of progressive reduction in oxygen delivery and blood flow and eventually leads to cardiac arrest. Although there is still not a uniform and widely accepted definition of WI, we elected to induce a well-defined and experimentally reproducible length of WI because the length of WI on its own can bias the outcome. Recently, Rhee et al.65 reported on a porcine model mimicking the situation of human uncontrolled DCD. The withdrawal of ventilatory support was initiated under general anesthesia (isoflurane and thiopental infusion) and resulted in passive progression to cardiac death. Interestingly, substantial variations in the times between the withdrawal of support and circulatory arrest (mean = 26 minutes 19 seconds, range = 12 minutes 30 seconds to 57 minutes 30 seconds) and electrical standstill (mean = 48 minutes 38 seconds, range = 24 minutes 1 second to 76 minutes) were observed. Consequently, very different outcomes ranging from immediate graft function to PNF could be expected when these livers were to be transplanted. Perhaps our model more closely mimics the uncontrolled DCD scenario, although in some clinical cases after abrupt cardiac standstill, there is an attempt at cardiopulmonary resuscitation, and a partial restoration of some degree of oxygenation and perfusion may take place before the eventual cardiac and circulatory standstill. Most importantly, the livers in our model are exposed to substantial normothermic ischemic injury, which provokes substantial IRI.

In support of our model is the fact that similar models are also being used by other research groups in the field.66-68

As described in our model, the type and length of WI can serve as guidelines for both controlled and uncontrolled DCD settings despite the absence of an agonal phase; the abrupt cessation of flow should, however, be taken into account when our data are extrapolated to the clinic.

What Are the Limitations of the Microarray?

An important aspect of this study is that our LT model is highly reproducible and is strictly controlled with respect to the status of the animals (health, age, sex, and genetics) and the experimental procedure. In this setting along with the different subsequent statistical validation steps used in the microarray analysis, the pathway analysis led to the set of genes reported here. As we pointed out earlier, increasing the sample size may have led to the identification of new genes, but our main object was to explore the combination of genes present in relevant pathways; for this purpose, the software analysis with its repeated statistical tests should have overcome a possible limitation inherent to the sample size.

The use of a cocktail strategy has already been repeatedly advocated by our group,25, 69 and this is based on the multitude and redundancy of the mechanisms at stake in IRI and liver graft failure. We acknowledge that the combination of several drugs makes it difficult to speculate about what the effects of single drugs on gene expression will be. Although we could link specific drugs to particular molecular pathways (on the basis of their established working mechanisms), we do not propose that any observed effect is solely the result of a single component. On the contrary, the observed changes have to be interpreted as secondary to the entire LDP. We are now preparing a prospective randomized clinical trial to test the effects of the LDP in a large cohort of LT recipients. The information obtained from this microarray study of pigs has helped us to select molecular markers that will be investigated in these patients. However, we realize that the molecular changes induced by the LDP in humans will not necessarily mirror those observed in pigs.

The data reported here have unmasked new target genes and opened new avenues for improving the LDP and hepatoprotection against IRI in general. Both livers exposed to WI and extended criteria donor livers (which are increasingly being used) could potentially benefit from this type of protocol. Biliary complications are important causes of morbidity after LT.70 In our experiment, pigs were sacrificed early after transplantation, and this precluded the study of biliary strictures. However, a favorable effect of the LDP was observed on the biliary salt/phospholipid ratio, which is a surrogate for the later development of biliary strictures.25 The cocktail approach does not interfere with the preservation phase of the transplant process. Here too, promising strategies have been described: oxygenated machine perfusion, oxygen persufflation, and new preservation solutions or additives to preservation solutions.68, 71-73

An advantage of the bioinformatics/data mining approach that we have used is that it searches for relevant and known drug targets. Many of the corresponding drugs have been developed. Information on their working mechanisms and interactions is available. Often, these drugs are registered for use in humans and are commercially available (eg, oxymatrine, carbon monoxide–releasing molecule 2, erythropoietin, and curcumin), and this allows experimental results to be translated to clinics.

In conclusion, we have studied the molecular mechanisms involved in the protective effect of the LDP after the transplantation of WI/IRI-damaged livers. The effect of the LDP seems to take place at the transcriptional level very early after reperfusion. We have found that the LDP induces gene expression early (1 hour after reperfusion) and predominantly in molecular pathways related to cytokine activity, (anti-)apoptosis, and cell proliferation. The observed changes can be linked in part to the known biological effects of single components of the LDP. This protective effect of the LDP appears to be based on the induction of pathways that prevent inflammation-induced injury rather than on the repair of damage caused by WI.

Using bioinformatics, we have identified 7 key regulatory genes (IL-6, IL-8, JUN, MMP1, PTGS2/COX2, SERPINE1, and STAT3) specifically induced by WI and transplantation and repressed by the applied drugs. Six genes unaffected by the LDP (FOS, IL-1A, IL-1B, IL-10, NFKBIA, and VEGFA) have been proposed as potential drug targets because of their literature-reported regulatory functions. On the basis of the gene expression and literature evidence, 3 additional drug targets (DUSP1, EGR1, and SPP1/OPN) have been identified. Compounds targeting these new genes could be included in a new and perhaps more efficient formulation of the LDP.

Acknowledgements

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES
  8. Supporting Information

The authors thank Dr. Tine Wylin for assisting with the animal experiments at their laboratory and Dr. Wouter Van Delm and the staff of the Flanders Institute for Biotechnology MicroArray Facility for performing the microarray experiments.

REFERENCES

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES
  8. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. MATERIALS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES
  8. Supporting Information

Additional Supporting Information may be found in the online version of this article.

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.