Unique Early Gene Expression Patterns in Human Adult-to-Adult Living Donor Liver Grafts Compared to Deceased Donor Grafts


* Corresponding author: Kim M. Olthoff, kim.olthoff@uphs.upenn.edu


Because of inherent differences between deceased donor (DD) and living donor (LD) liver grafts, we hypothesize that the molecular signatures will be unique, correlating with specific biologic pathways and clinical patterns. Microarray profiles of 63 biopsies in 13 DD and 8 LD liver grafts done at serial time points (procurement, backbench and postreperfusion) were compared between groups using class comparisons, network and biological function analyses. Specific genes were validated by quantitative PCR and immunopathology. Clinical findings were also compared. Following reperfusion, 579 genes in DD grafts and 1324 genes in LDs were differentially expressed (p < 0.005). Many upregulated LD genes were related to regeneration, biosynthesis and cell cycle, and a large number of downregulated genes were linked to hepatic metabolism and energy pathways correlating with posttransplant clinical laboratory findings. There was significant upregulation of inflammatory/immune genes in both DD and LD, each with a distinct pattern. Gene expression patterns of select genes associated with inflammation and regeneration in LD and DD grafts correlated with protein expression. Unique patterns of early gene expression are seen in LD and DD liver grafts, correlating with protein expression and clinical results, demonstrating distinct inflammatory profiles and significant downregulation of metabolic pathways in LD grafts.


cold backbench biopsy


deceased donor


gene ontology


hepatitis C virus


living donor


Model of Endstage Liver Disease


postreperfusion biopsy


preprocurement biopsy


TaqMan Low-Density Array


The use of liver grafts from living donors (LD) for adult recipients was developed to help overcome the disparity between the number of deceased donors (DD) and the growing waitlist. In adult-to-adult LD liver transplantation grafts are smaller than the standard liver volume, with surgical technique and a set of attributes that are distinctive from whole DD grafts and unique risk factors for graft failure (1). Cellular processes for rapid restoration of liver mass in these grafts may compete with the need to maintain systemic metabolic homeostasis and impact immediate graft function. By comparison, DD grafts experience a degree of hepatocellular damage from brain death, donor events, procurement, preservation and reperfusion injury (2). Identifying these disparate responses to transplantation on a molecular level may provide clues to clinical interventions that can improve outcomes.

Rodent models of partial hepatectomy and transplant provide a large amount of information regarding molecular pathways of liver regeneration, inflammation and repair (3–6) and show that cellular proliferation causes a shift away from liver metabolic function (7–10). Several studies have used microarrays to investigate liver regeneration and injury in rodents (11–14), but few human studies have utilized whole genome expression data (15,16).

Adult-to-adult LD liver transplantation provides a unique opportunity to explore the recovery of a human liver not subject to the confounding effects of brain death and organ recovery. The cellular and molecular differences between DD and LD grafts with regard to injury, repair and function are not known. We used whole genome-wide expression analysis to assess the immediate phase of liver graft recovery after transplantation comparing human adult-to-adult right lobe LD and whole DD transplants. We hypothesized that molecular signatures of LD grafts will differ significantly from DD grafts, such that LD grafts will demonstrate early gene expression profiles linked to hepatic regeneration as well as unique proinflammatory profiles. Recognizing these distinctive differences at the molecular level provides an important basis for identifying potential targets for future studies and possible therapeutic intervention.

Methods and Materials

Patients and study design

All study protocols and consent procedures were approved by the Institutional Review Boards of the University of Pennsylvania, Virginia Commonwealth University and the Medical Advisory Committee of Gift of Life Donor Program. Core biopsies were taken in the LD and from the DD before manipulation of the liver (PRE). A second biopsy was taken on the backbench following cold preservation (COLD). The final biopsy (POST) was taken after reperfusion following completion of the bile duct anastamosis. Cold ischemic time was measured from time of donor cross-clamp to removal from ice. Warm ischemic time was defined as time of removal of liver from ice to reperfusion. Clinical data were collected from donor and recipient charts and from the electronic transplant data base at the University of Pennsylvania and Virginia Commonwealth University. These data included donor and recipient demographics, intraoperative details and postoperative liver function.

Isolation of RNA

Biopsies were placed immediately in RNA Later (Ambion, Austin, TX) and frozen. Total RNA, extracted using Trizol (Invitrogen, Carlsbad, CA) was cleaned on RNEasy columns (Qiagen Inc, Valencia, CA). RNA integrity was confirmed using the Agilent 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA) (17,18).

Preparation of labeled cRNA and hybridization

Five micrograms of mRNA was converted to double-stranded cDNA, then to cRNA using T7 RNA polymerase with biotinylated nucleotides (Ambion MessageAmp II Enhanced Biotin kit). Fifteen micrograms of fragmented cRNA was hybridized to each Affymetrix HG-U133Plus 2.0 GeneChip (Affymetrix, Santa Clara, CA). Chips were hybridized, washed and stained as described (20–22). Signal intensities were generated using the Affymetrix GeneChip® Operating System (GCOS).

Data analysis

CEL files were normalized with Robust Multichip Average (RMA Express 0.3 beta 1) (19). Class comparisons were done in BRB ArrayTools (http://linus.nci.nih.gov/BRB-ArrayTools.html). Two-class unpaired F-tests were performed using a random variance model. A False Discovery Rate of <0.1 was chosen to identify differentially expressed genes at a user-defined p-value of 0.005.

Biological function analysis was done with Expression Analysis Systematic Explorer (EASE 2.0; http://david.niaid.nih.gov/david/ease.htm) and the Data base for Annotation, Visualization and Integrated Discovery (DAVID; http://www.david.niaid.nih.gov/david). Functions were found using Gene Ontology (GO) (20). Pathway analysis and additional analysis of gene functions was done with Ingenuity Pathway Analysis (Ingenuity® Systems, Redwood City, CA). The resulting biological networks comprised up to 35 differentially expressed genes, called Focus genes.

RT-qPCR using TaqMan low-density arrays (TLDA)

TLDA microfluidic cards (384-well; Applied Biosystems, Foster City, CA) were used to validate differential gene expression of 63 genes, chosen for validation from the results of the class comparisons. Total RNA was extracted from the biopsies using TRIZOL (Invitrogen). Contaminating genomic DNA was removed using DNA-free (Ambion). For RT-qPCR, cDNA was transcribed from 2 μg total RNA using the cDNA archive kit (Applied Biosystems). All transcript assays consisted of two unlabeled PCR primers and a FAM™ dye-labeled TaqMan® MGB probe, prespotted on the TLDA card. All amplifications were done in triplicate and threshold cycle (Ct) scores were averaged to calculate relative expression. The Ct scores were normalized against 18S rRNA controls by the RQ Manager 1.2 software (Applied Biosystems).


Immunohistopathology was used to correlate gene expression with protein secretion. Paraffin-embedded PRE and POST liver sections from 4 LD and 4 DD grafts were stained for each of the chosen associated proteins. Sections were stained using an Envision immunoperoxidase kit (Dako) and antibodies to SOCS3 (Santa Cruz Biotech, Santa Cruz, CA), IL1RL1 (Sigma), IL-8 (Epitomics) plus cell-lineage markers and isotype-matched controls.

Supporting data

The CEL files for all the microarrays in this study are in the Gene Expression Omnibus (GEO): (http://www.ncbi.nlm.nih.gov/geo). Supporting data tables for all the class comparisons are at: (http://www.scripps.edu/mem/eht/salomon/livertxdata/).


Patient characteristics

A total of 21 patients, 13 DD recipients and 8 right lobe LD recipients had a full set of 3 serial liver biopsies (total 63 biopsies) and complete clinical information is presented in Table 1. Recipients of DD grafts had higher model for end-stage liver disease (MELD) scores than LDs, reflecting the need for a higher MELD in order to be offered a DD graft. There were no significant differences in recipient/donor gender, age, or primary liver disease. There were more patients with hepatitis C (HCV) receiving DD (n = 8) than LD grafts (n = 2), p = NS. There were no biopsy complications, all grafts functioned well, and graft and patient survival was 100% at 12 months posttransplant. Cold ischemia times were shorter in LD recipients where the donor and recipient operations are tightly coordinated. Anhepatic and warm ischemia times were also shorter in the LD recipients due to the differences in surgical techniques. In most of the DDs, portal and arterial reperfusion occurred within minutes of each other because both anastomses were performed during the warm ischemic period. In LDs the arterial anastomosis often required microsurgical techniques and surgeon preference was to perform these after portal reperfusion to keep warm ischemic time to a minimum. There were no significant differences in total operative time or blood replacement.

Table 1.  Patient and donor demographics and characteristics of the recipient surgical procedure
 Deceased donor transplants (n = 13)Living donor transplants (n = 8)p-Value
Recipient primary disease
 – Hepatitis C51 
 – Hepatoma1 
 – Hepatitis C + Hepatoma31 
 – Postalcoholic cirrhosis2 
 – Primary sclerosing cholangitis3 
 – Primary biliary cirrhosis11 
 – Acute liver failure1 
 – Other cholestatic disease11 
Recipient: Male/Female10/33/50.09
Recipient: Age (min–max, median)29–66 (49)37–63 (49)0.92
Recipient MELD (min–max, median)18–37 (25.5)6–18 (13.8)<0.001 
Liver donor
 – Gender (M/F)7/64/40.61
 – Age (min–max, median)19–74 (43)29–50 (40)0.70
Donor cause of death
 – Cerebrovascular accident/intracerebral bleed7N/A 
 – Anoxia sec. to cardiac arrest3  
 – Traumatic brain injury3  
Transplant surgery details
Total operative time (min)298 ± 10366 ± 390.22
Hepatectomy phase (min)125 ± 9 169 ± 280.79
Anhepatic phase (min)107 ± 8 79 ± 60.03
Packed red blood cell concentrate (units) 4.2 ± 0.7 2.3 ± 0.90.12
Cold ischemic time (min) (Cross-clamp to removal from ice)282 ± 1938 ± 7<0.001 
Warm ischemic time (min) (Removal from ice to portal reperfusion)59 ± 237 ± 3<0.001 
Time between base-line PRE biopsy and COLD biopsy at end CI (min)304 ± 12272 ± 730.68
Time between portal reperfusion and POST biopsy (min)53 ± 2 79 ± 130.10
Length of stay in hospital days (min–max, median)7–21 (10)9–39 (15)0.31

Class comparisons between time points and graft types

We compared differential gene expression as a function of the PRE, COLD and POST stages of the liver transplant. The total number of genes differentially expressed between all the class comparisons is illustrated in Figure 1A. There were significant differences between the PRE biopsy in the LD and in the DD, illustrating the effect of donor status and brain death compared to a healthy LD. Interestingly, cold storage of a DD graft results in little change in gene transcription as compared to the PRE biopsy (only 10 genes differentially expressed between COLD and PRE). In contrast, resection of the LD graft prior to the COLD biopsy resulted in the differential expression of 180 genes, most likely due to the complex process of parenchymal dissection. While these comparisons are important and illustrate underlying differences between brain dead and living donors, for the purpose of focus in this article we will not discuss these differences, but have included all comparisons in the Supporting online tables. To address our underlying hypothesis, a detailed analysis was focused on the following three class comparisons: (1) DD POST versus DD PRE, (2) LD POST versus LD PRE and, (3) LD POST versus DD POST. A large number of genes were differentially expressed in both graft types following reperfusion (POST) when compared to the PRE biopsies, but there were relatively few common genes expressed by both LD and DD grafts (Figure 1B).

Figure 1.

Class comparisons of differentially expressed genes. (A) Diagram of number of differentially expressed genes in each class comparison. The numbers of differentially expressed genes between groups are illustrated in the small boxes connecting the larger shaded boxes (at p-value ≤ 0.005). (B) Venn diagram of overlap of differentially expressed genes in LD and DD POST reperfusion, compared to PRE transplantation for each graft type.

Gene expression in DD grafts after reperfusion: DD POST versus DD PRE

The first class comparison compared DD POST biopsies to donor baseline biopsies. The number of differentially expressed genes POST was 579 (375 upregulated after reperfusion and 204 downregulated). Table 2 shows 20 selected genes that are present in the overrepresented biological processes based on GO classifications in the DD grafts after reperfusion (DD POST vs. DD PRE). The selection of these genes represents integration of two statistical metrics to rank genes for very high biological significance. First, our ANOVA-based tool for class comparison allows us to rank differentially expressed genes by their p-values, a standard and useful procedure in organizing these large data sets. Second, Ingenuity Pathway Analysis of the differentially expressed genes provides a statistical metric to confirm the overrepresentation (i.e. significance) of any functional gene networks that are identified as opposed to simply being found by random chance in any large list of gene expression values. So the 20 genes selected represent the highest ranked differentially expressed genes by p-values that are also identified as populating overrepresented functional gene networks. There are other important genes and pathways that are beyond the space limitations and results for both differential gene expression and Ingenuity functional networks are available as Supplemental Data (http://www.scripps.edu/mem/eht/salomon/livertxdata/).

Table 2.  Differential gene expression in overrepresented biological processes based on Gene Ontology classifications
Gene symbolDescriptionGO biological processGeom mean DD PREGeom mean DD POSTFold change POST vs. PREp-Value
  1. Twenty select genes from DD POST versus PRE class comparison and 20 select genes from LD POST versus PRE class comparison.

Differential gene expression for DD grafts: POST reperfusion vs. PRE transplant
ATF3Activating transcription factor 3(DNA dependent) transcription, regulation of nucleobase, nucleoside, nucleotide and nucleic acid metabolism62.41503.123.81p < 1e-07
JUNv-jun sarcoma virus 17 oncogene homolog (avian)(DNA dependent) transcription, regulation of nucleobase, nucleoside, nucleotide and nucleic acid metabolism221.32018.79.09p < 1e-07
TNFAIP3Tumor necrosis factor, alpha-induced protein 3 (A20)(DNA dependent) transcription, regulation of nucleobase, nucleoside, nucleotide and nucleic acid metabolism44.6166.63.73p < 1e-07
NFKBIZNuclear factor kappa B inhibitor, zeta (MAIL)Antiapoptosis, (DNA-dependent) regulation of transcription, regulation of nucleobase, nucleoside, nucleotide and nucleic acid metabolism176.7805.14.57p < 1e-07
CCL2Chemokine (C-C motif) ligand 2 (MCP-1)Antiapoptosis, inflammatory response, JAK-STAT cascade59.9706.311.76p < 1e-07
SOCS3Suppressor of cytokine signaling 3Antiapoptosis, JAK-STAT cascade79.9620.97.751.00E-07
HSPA1BHeat shock 70kDa protein 1BAntiapoptosis, negative regulation of biological and cellular process123.511619.433.00E-07
CDKN1ACyclin-dependent kinase inhibitor 1A (p21, Cip1)Apoptosis, cell cycle arrest, negative regulation of cell proliferation260.7707.12.714.00E-07
FOSv-fos FBJ murine osteosarcoma viral oncogene homologInflammatory response, (DNA dependent) transcription, regulation of nucleobase, nucleoside, nucleotide and nucleic acid metabolism111.51260.711.361.00E-07
CCL4Chemokine (C-C motif) ligand 4 (MIP1-beta)Inflammatory response, cell adhesion84.2568.56.76p < 1e-07
IL8Interleukin 8Inflammatory response, chemotaxis, cell cycle arrest40.2243.76.06p < 1e-07
IL1BInterleukin 1, betaInflammatory response, chemotaxis, negative regulation of cell proliferation39.8124.13.121.00E-07
IL6Interleukin 6 (interferon, beta 2)Inflammatory response, regulation of cell proliferation, negative regulation of metabolism23.433.21.426.22E-04
ICAM1Intercellular adhesion molecule 1 (CD54)Immune response, chemotaxis, cell adhesion45.1195.34.33p < 1e-07
IL10Interleukin 10Inflammatory response, antiapoptosis, regulation of T cell proliferation12.9201.552.35E-04
DUSP5Dual specificity phosphatase 5Dephosphorylation, protein metabolism41.3233.95.65P < 1e-07
HBEGFHeparin-binding EGF-like growth factorSignal transduction46.5165.73.56P < 1e-07
GSTA4Glutathione S-transferase A4Response to stress, metabolism2417.1−1.409.48E-04
EIF4E2Eukaryotic translation initiation factor 4E member 2Regulation of biosynthesis, protein metabolism90.964.7−1.401.96E-04
ALBAlbuminApoptosis, negative regulation of cellular process915115.2−7.943.00E-07
Differential gene expression for LD grafts: POST reperfusion vs. PRE transplant
SOCS3Suppressor of cytokine signaling 3Regulation of cell growth, antiapoptosis, JAK-STAT cascade, negative regulation of insulin receptor signaling pathway46.8443.19.432.51E-04
STAT3Signal transducer and activator of transcription 3(DNA-dependent) regulation of transcription, signal transduction, JAK-STAT cascade531.61511.82.840.0018449
NFKB1NF of kappa light polypeptide gene enhancer in B-cells 1 (p105)(DNA-dependent) regulation of transcription, antiapoptosis, inflammatory response, signal transduction102.5159.61.560.0014409
CDKN1ACyclin-dependent kinase inhibitor 1A (p21)Apoptosis, cell cycle arrest, negative regulation of cell proliferation164.8812.14.931.52E-05
JUNv-jun sarcoma virus 17 oncogene homolog (avian)(DNA-dependent) regulation of transcription, regulation of progression through cell cycle, leading edge cell differentiation76.7300.13.910.0015611
ODC1Ornithine decarboxylase 1Polyamine biosynthetic process471.63665.47.750.0003283
DHODHDihydroorotate dehydrogenaseDe novo’ pyrimidine base biosynthetic process222777.73.510.000586
HGFHepatocyte growth factor (hepapoietin A; scatter factor)Mitosis34.250.21.470.0010275
IL1RNInterleukin 1 receptor antagonistInflammatory response, insulin secretion104.9761.47.253.11E-05
IL4RInterleukin 4 receptorImmune response, signal transduction194.2458.12.360.0011693
ILF2Interleukin enhancer binding factor 2, 45kDa(DNA-dependent) regulation of transcription,immune response190.1423.92.230.0027839
PFKFB26-phosphofructo-2-kinase/fructose-2, 6-biphosphatase 2Fructose 2,6-bisphosphate metabolic process19.4281.440.0027338
CEBPACCAAT/enhancer binding protein (C/EBP), alpha(DNA-dependent) regulation of transcription, generation of precursor metabolites and energy, cytokine and chemokine mediated signaling pathway620.5140.1−4.430.0002454
PPARAPeroxisome proliferative activated receptor, alpha(DNA-dependent) regulation of transcription, lipid metabolic process, fatty acid transport, positive regulation of fatty acid beta-oxidation265.5114.8−2.319.90E-06
AMACRAlpha-methylacyl-CoA racemaseMetabolic process316.670.3−4.508.00E-07
GPD1Glycerol-3-phosphate dehydrogenase 1 (soluble)Gluconeogenesis, carbohydrate metabolic process, glycerol-3-phosphate metabolic process166.382.7−2.012.95E-05
ACOX2Acyl-Coenzyme A oxidase 2, branched chainBile acid metabolic process, electron transport, lipid metabolic process, fatty acid beta-oxidation1160.1271.1−4.288.77E-05
HNF4AHepatocyte nuclear factor 4, alphaLipid metabolic process, blood coagulation, (DNA-dependent) regulation of transcription448.5262.7−1.710.0019787
THRSPThyroid hormone responsive (SPOT14 homolog, rat)Lipid metabolic process, regulation of transcription from RNA polymerase II promoter1033.3118.9−8.690.0003317
PCK2Phosphoenolpyruvate carboxykinase 2 (PEPCK mitochondrial)Glucose metabolic process, gluconeogenesis1459.8799.7−1.830.0009141

The DD grafts demonstrated a large number of upregulated inflammatory genes after reperfusion, including cytokines, chemokines and other regulators of inflammation. Specific examples include the inflammatory chemokine IL-8, as well as CCL2 (MCP-1) and TNFAIP3/A20, described as hepatoprotective (21,22) and ICAM1, involved in leukocyte recruitment (23). Other overrepresented GO categories in deceased donors after reperfusion are transcription (DNA-dependent) and regulation of nucleobase, nucleoside, nucleotide and nucleic acid metabolism, with ATF3, induced by stress, as one of the most upregulated genes (23-fold increase).

Gene expression in LD grafts after reperfusion: LD POST versus LD PRE

The second detailed analysis compared gene expression in the LD liver grafts after reperfusion (POST) to the PRE biopsy taken in the donor. After reperfusion in the LD, the number of differentially expressed genes was 1324 (756 upregulated and 568 downregulated). Table 2 shows 20 selected genes that are present in the overrepresented biological processes based on GO classifications in the LD grafts after reperfusion (LD POST vs. LD PRE).

The gene lists generated support our hypothesis that activation of molecular networks for regeneration is an immediate event in partial LD grafts, associated with activation of proinflammatory pathways. There was a marked increase in genes of the IL-6/STAT3 pathway (24), including SOCS3, a feedback mediator that attenuates proinflammatory signaling and regulates STAT3 after partial hepatectomy (25), hepatocytes growth factor (HGF), a known stimulator of liver regeneration (3), and NFκB1, involved in regeneration and hepatoprotection (26). Other upregulated networks involve cell cycle progression and de novo biosynthesis of polyamines and pyrimidines.

A very significant finding in the LD POST biopsies was that most of the downregulated genes involve hepatic metabolic function: fatty acid metabolism (GPD1, AMCR), fatty acid oxidation and tissue oxidative metabolism (PPARA), lipogenesis (THRSP) and bile acid metabolism (ACOX2). Genes enhancing glycolysis (PFKFB2) were upregulated, while a key gluconeogenesis enzyme (PCK2) is downregulated. Downregulation of specific hepatic metabolism genes (HNF4α and C/EPBα) matches those seen in several animal models and also includes PEPCK, glutathione S-transferase and genes linked to steroid and hormone metabolism (14,27).

Similar to the DD grafts, there was upregulation of immune/inflammatory genes in LD POST grafts. However, they demonstrated a very different profile than in DD POST, with some downregulated as well. There were only 17 overlapping immune/inflammatory genes common to both DD and LD (Table 3). For example, CCL2 (MCP-1) and CCL20 (MIP-3α) play a role in immune response and were upregulated in both. LD POST grafts primarily demonstrated upregulation of interleukin-associated receptor genes such as IL1R and IL4R, as well as genes associated with innate immunity, such lipopolysaccharide binding protein (LPB), while the DD POST grafts demonstrated genes associated with cytokines; interferon gamma, IL1β, IL6, IL8 and IL10. Some MHC class II genes were upregulated in LD POST, a pattern not seen in the DD grafts.

Table 3.  Comparison of immune and inflammatory genes differentially expressed in LD and DD grafts in POST biopsies compared to PRE
Genes differentially expressed in LD POST vs. LD PREGenes differentially expressed in DD POST vs. DD PRE
Gene symbolDescriptionFold change LDPOST vs. LDPREGO categoryGene symbolDescriptionFold change DD POST vs. DDPREGO category
BCL3B-cell CLL/lymphoma 31.89immune responseBCL3B-cell CLL/lymphoma 32.75immune response
BCL6B-cell CLL/lymphoma 6 (zinc finger protein 51)1.42inflammatory responseBCL6B-cell CLL/lymphoma 6 (zinc finger protein 51)1.73inflammatory response
C5R1complement component 5 receptor 1 (C5a ligand)1.89immune responseC5R1complement component 5 receptor 1 (C5a ligand)2.48immune response
CALCAcalcitonin/calcitonin-related polypeptide, alpha5.99inflammatory responseCALCAcalcitonin/calcitonin-related polypeptide, alpha4.78inflammatory response
CCL2chemokine (C-C motif) ligand 2 (MCP-1)11.76inflammatory responseCCL2chemokine (C-C motif) ligand 2 (MCP-1)4.37inflammatory response
CCL4chemokine (C-C motif) ligand 4 (MIP-1-beta)6.76inflammatory responseCCL4chemokine (C-C motif) ligand 4 (MIP-1-beta)3.22inflammatory response
CCL18chemokine (C-C motif) ligand 181.47inflammatory responseCCL18chemokine (C-C motif) ligand 18 (pulmonary and activ1.46inflammatory response
CCL20chemokine (C-C motif) ligand 20 (MIP-3 alpha)5.81inflammatory responseCCL20chemokine (C-C motif) ligand 20 (MIP-3 alpha)5.49inflammatory response
IL1RL1interleukin 1 receptor-like 18.77immune responseIL1RL1interleukin 1 receptor-like 12.49immune response
IL1RNinterleukin 1 receptor antagonist7.25inflammatory responseIL1RNinterleukin 1 receptor antagonist4.88inflammatory response
NFKB2nuclear factor of kappa light polypeptide gene enhancer1.18immune responseNFKB2nuclear factor of kappa light polypeptide gene enhanc1.40immune response
RIPK2receptor-interacting serine-threonine kinase 21.67inflammatory responseRIPK2receptor-interacting serine-threonine kinase 21.62inflammatory response
THBS1thrombospondin 15.05inflammatory responseTHBS1thrombospondin 14.67inflammatory response
ELF3E74-like factor 32.79inflammatory responseELF3E74-like factor 31.94inflammatory response
FCN1ficolin (collagen/fibrinogen domain containing) 11.82immune responseFCN1ficolin (collagen/fibrinogen domain containing) 12.14immune response
FCARFc fragment of IgA, receptor for1.42immune responseFCARFc fragment of IgA, receptor for1.14immune response
ETS1v-ets erythroblastosis virus E26 oncogene homolog 11.29immune responseETS1v-ets erythroblastosis virus E26 oncogene homolog 11.25immune response
CRPC-reactive protein, pentraxin-related8.06inflammatory responseFOSv-fos FBJ murine osteosarcoma viral oncogene homol11.36inflammatory response
SAA1 /// SAA2serum amyloid A1//A27.30inflammatory responseIL8interleukin 86.06inflammatory response
LBPlipopolysaccharide binding protein3.16inflammatory responseSNF1LKSNF1-like kinase5.59immune response
RAB27ARAB27A, member RAS oncogene family3.07immune responseCCL3chemokine (C-C motif) ligand 3 (MIP-1-alpha)5.08inflammatory response
STAT3signal transducer and activator of transcription 32.84inflammatory responseNFIL3nuclear factor, interleukin 3 regulated3.75immune response
FPRL1formyl peptide receptor-like 12.50inflammatory responsePTGS2prostaglandin-endoperoxide synthase 23.58inflammatory response
IL4Rinterleukin 4 receptor2.36immune responseIL1Binterleukin 1, beta3.12inflammatory response
ILF2interleukin enhancer binding factor 2, 45kDa2.23immune responseRGS1regulator of G-protein signalling 12.91immune response
CHUKconserved helix-loop-helix ubiquitous kinase2.14immune responseZFP36zinc finger protein 36, C3H type, homolog (mouse)2.91inflammatory response
DKFZP564DKFZP564 J0863 protein1.96immune responseIRF1interferon regulatory factor 12.75immune response
F11RF11 receptor1.63inflammatory responseCD83CD83 antigen2.67immune response
I KB KGinhibitor of kappa light polypeptide gene enhancer in B-c1.63immune responseIRF4interferon regulatory factor 42.42immune response
NFKB1nuclear factor of kappa light polypeptide gene enhancer1.56inflammatory responseRGS1regulator of G-protein signalling 12.18immune response
HLA-DRB4major histocompatibility complex, class II, DR beta 41.24immune responseGBP1guanylate binding protein 1, interferon-inducible2.16immune response
LTAlymphotoxin alpha (TNF superfamily, member 1)1.23immune responseGEMGTP binding protein overexpressed in skeletal muscle2.15immune response
CD276CD276 antigen1.22immune responseCXCR4chemokine (C-X-C motif) receptor 42.01inflammatory response
NFE2L1nuclear factor (erythroid-derived 2)-like 11.36inflammatory responseANXA1annexin A11.98inflammatory response
C2complement component 2−1.47inflammatory responseFPR1formyl peptide receptor 11.88inflammatory response
ATRNattractin−1.58inflammatory responseCXCL3chemokine (C-X-C motif) ligand 3 (CINC/MIP-2b)1.86inflammatory response
F11RF11 receptor−1.59inflammatory responseEBI2Epstein-Barr virus induced gene 21.83immune response
FCGRTFc fragment of IgG, receptor, transporter, alpha−1.60immune responseCXCL1chemokine (C-X-C motif) ligand 11.74inflammatory response
DPP4dipeptidylpeptidase 4 (CD26)−1.62immune responseCIAS1cold autoinflammatory syndrome 11.56inflammatory response
AQP9aquaporin 9−1.79immune responseIL10interleukin 101.55inflammatory response
PXMP2peroxisomal membrane protein 2, 22kDa−1.99immune responseTREM1triggering receptor expressed on myeloid cells 11.54immune response
CNGA1cyclic nucleotide gated channel alpha 1−2.01immune responseCEBPBCCAAT/enhancer binding protein (C/EBP), beta1.52inflammatory response
CD4CD4 antigen (p55)−2.02immune responseBMP2bone morphogenetic protein 21.44inflammatory response
TLR3toll-like receptor 3−2.04inflammatory responseIL6interleukin 6 (interferon, beta 2)1.42inflammatory response
CXCL12chemokine (C-X-C motif) ligand 12−2.20inflammatory responseSQSTM1Sequestosome 11.41immune response
EPHX2epoxide hydrolase 2, cytoplasmic−2.26inflammatory responseFPRL1formyl peptide receptor-like 1 /// formyl peptide receptt1.38inflammatory response
FLJ20406hypothetical protein FLJ20406−2.73immune responseRELBv-rel reticuloendotheliosis viral oncogene homolog B,1.35immune response
IFIT3interferon-induced protein with tetratricopeptide repeats:−2.75immune responseC1SComplement component 1, s subcomponent1.28inflammatory response
KLKB1kallikrein B, plasma (Fletcher factor) 1−3.32inflammatory responseSLC11A1solute carrier family 11 (proton-coupled divalent metal1.27immune response
SCAP1src family associated phosphoprotein 1−3.42immune responseICEBERGICEBERG caspase-1 inhibitor1.14inflammatory response
IFIT1interferon-induced protein with tetratricopeptide repeats−5.54immune responseIFIH1Interferon induced with helicase C domain 1−1.10immune response
ZGPATzinc finger, CCCH-type with G patch domain−6.00immune responseADORA3adenosine A3 receptor−1.62inflammatory response
TNFSF10tumor necrosis factor (ligand) superfamily, member 10−6.44immune responseATRNattractin−1.75inflammatory response

Differences between deceased donor and living donor grafts following reperfusion: LD POST versus DD POST

The next analysis directly compared the expression profiles after reperfusion between LD and DD livers. One thousand five hundred and six genes were differentially expressed (726 upregulated and 780 downregulated, see online Supplemental Data). In this comparison, it was apparent that significant differences exist in expression of genes associated with biosynthesis, metabolism and energy pathways.

To better understand the functional significance of graft type on gene expression, we used an EASE analysis to assign p-values to each biological function rather than to individual genes (Table 4). The 12 functions that were increased in LDs are nearly all associated with cell proliferation and tissue regeneration. In contrast, all 17 downregulated processes in LDs represent metabolic liver functions, consistent with our hypothesis that recovery of the partial LD graft requires redistribution of energy in favor of regeneration.

Table 4.  Biological theme analysis by EASE comparing LD grafts POST reperfusion versus DD grafts POST reperfusion
 EASE score (p-Value)
Biological functions upregulated in LD POST
RNA metabolism9.86E-05
RNA processing2.28E-04
Nucleobase\Nucleoside\Nucleotide and Nucleic acid metabolism2.69E-03
Antigen processing1.23E-02
Transcription from Pol II promoter1.61E-02
rRNA processing2.15E-02
Protein amino acid phosphorylation2.16E-02
rRNA metabolism3.35E-02
Nucleobase biosynthesis3.52E-02
Protein modification4.59E-02
Nucleobase metabolism4.62E-02
Biological functions downregulated in LD POST
Lipid biosynthesis7.19E-03
Alcohol metabolism9.34E-03
Carbohydrate metabolism1.38E-02
Energy pathways1.61E-02
Lipid metabolism1.72E-02
Physiological process1.76E-02
Main pathways of carbohydrate metabolism2.14E-02
Macromolecule biosynthesis2.28E-02
Photoreceptor maintenance2.44E-02
Iron ion transport2.70E-02
Xenobiotic metabolism2.46E-02
Response to xenobiotic stimulus2.72E-02
Mismatch repair/maintenance of fidelity during DNA replication3.47E-02
Retinol metabolism3.47E-02
Amino acid and derivative metabolism3.64E-02

We wished to determine if HCV status impacted on the 1506 differentially expressed genes identified in LD POST versus DD POST. Only 23 (LD) and 7 (DD), respectively, were contained in the 1506 genes identified in earlier comparison of DD POST versus LD POST. Thus, HCV status did not appear to influence the selection of genes identified in this study (complete analysis in Supplemental Data).

Biological networks analysis

Ingenuity Pathway Analysis organizes gene expression data into molecular networks using algorithms based on literature, known molecular functions and compendia of annotated data bases. We utilized these networks to identify and graphically view biological mechanisms and intergene connectivity relevant to the data sets generated by our class comparisons and specific focus genes of interest. When all genes differentially expressed POST reperfusion in the LD to DD comparison were analyzed (N = 1506), 778 genes were found to be linked to 43 networks with 10 or more Focus Genes per network. In 8 networks, the maximum of all 35 Focus Genes were present. In each network, there is a ‘centered’ gene, which refers to the gene with the highest level of intergene connectivity. The 8 networks identified were: (1) cell growth and proliferation (centered around p53), (2) gene expression in cancer (centered around FOS), (3) connective tissue development (centered around JUN), (4) inflammation/immune responses (centered around NFκB), (5) lipid metabolism (centered around CEPBα/STAT3), (6) cell growth (centered around MAPK8), (7) cell cycle and DNA replication (centered around CDK4) and (8) cell death (centered around CAV1). These analyses further illustrated how cell growth and proliferation pathways are increased in LD grafts, and metabolism pathways are downregulated. The complete sets of 8 networks are in the online Supplemental Data (Figures S1–8). It is important to note that about 50% of the 1506 genes analyzed are not accounted for here. First, many genes in the human genome, even many with a known function, are not presently linked to discrete functional molecular networks in the literature and these account for the majority of these missing gene candidates. Second, we truncated our analysis for practical reasons at 43 different networks using a filter of >10 genes/network but another few hundred genes fall into smaller networks.

Validation of gene expression using TaqMan low-density arrays (TLDA)

Gene expression analysis requires independent validation of at least a subset of gene candidates. TLDA technology was used to validate 63 genes in triplicate plus controls. Validation genes were chosen from the results of the class comparisons: LD PRE versus LD POST (28 genes), DD PRE versus DD POST (22 genes), and LD POST versus DD POST (18 genes) based on their biological interest. There was nearly complete concordance for gene expression between the microarrays and TLDA with the exception of three genes in the last class comparison (Supplemental Figure S9). Because the same panel of genes was also measured for all samples profiled, we also validated expression of many genes in each sample that were not chosen for their biological interest or differential expression, with an 87% concordance for gene expression across all the samples studied. This amount of concordance of genes identified using Affymetrix arrays using a completely separate technology platform, and testing for many genes that were not differentially expressed in the group comparisons, is quite significant given the relatively high variability known for raw TaqMan PCR results and the fact that the fold-change differences for many of these comparisons were only 2-fold or less.

Relationship between gene expression and protein expression by immunopathology

To explore the relationship between our findings of early gene expression to protein expression, we identified several relevant cell proliferation and inflammatory genes that demonstrated significant expression differences in PRE and POST samples and used immunohistopathology to assess expression of their associated proteins in a representative sample of LD and DD biopsies. These genes were: SOCS3, a feedback regulator of regeneration; IL1RL1, a regulator of cell proliferation and member of the Toll superfamily; and IL-8, a proinflammatory CXC chemokine. We demonstrated protein expression patterns that matched gene expression. Immunoperoxidase comparison of biopsies from LD and DD showed comparable SOCS3 upregulation with gene expression in POST samples, which showed increased staining of hepatocytes, bile ducts and inflammatory cells postreperfusion regardless of the donor type (Figure 2A). When comparing POST biopsies to PRE, SOCS3 demonstrated a 7.75-fold change in gene expression in DD and a 9.43-fold change in LD grafts (Table 2). IL1RL1 showed minimal baseline PRE staining but significant upregulation postreperfusion primarily in LD tissues (Figure 2B), correlating with an 8.77-fold change seen in the array data (Table 4). For other genes, such as IL-8, minimal or no expression was detected in the PRE or POST LD biopsies, whereas significant staining was detected postreperfusion in the DD POST samples, correlating with a 6-fold increase seen in the array data (Figure 2C, Table 2).

Figure 2.

(A–C) Immunoperoxidase analysis of gene expression in liver biopsies collected from the donor liver (PRE) and postreperfusion (POST), showing representative data from four samples/group and comparing events using LD and DD liver tissues. Negligible staining for SOCS3, IL1RL1 or IL-8 was seen in the PRE biopsies. (Paraffin sections, hematoxylin counterstain, original magnifications ×250). (A) Postreperfusion, SOCS3 expression was upregulated in both LD and DD biopsies with staining of hepatocytes, bile ducts and inflammatory cells. (B) IL1RL1 demonstrates negligible expression in PRE LD biopsies but with upregulation by hepatocytes in postreperfusion biopsies of LD tissues. There was minimal staining in DD livers, either PRE or POST. (C) IL-8 was upregulated primarily in postreperfusion biopsies from DD livers, with minimal LD expression. IL-8 staining was detected in hepatocytes and also in conjunction with infiltrating polymorphonuclear neutrophils.

Postoperative course and clinical correlation

In order to determine if there may be clinical trends correlating with gene expression, we compared graft outcome, liver transaminases, bilirubin and coagulation factors in the first 4 weeks posttransplant in LD and DD recipients.

All LD and DD recipients had good early allograft function with no graft loss within the first year posttransplant, so these gene expression profiles were considered to correlate with good function. No significant differences were noted in transaminases levels (data not shown), however, bilirubin levels were elevated in LD compared to DD recipients, correlating to downregulation of genes associated with bile synthesis in LD, significant from POD 7 to 1 month after transplantation (Figure 3A). In DD recipients, the INR was higher just after surgery, but improved quickly, whereas LD recipients demonstrated significantly higher INR in the first week after transplantation and slower return to normal (Figure 3B). Despite highly downregulated albumin gene expression, no significant differences were seen for serum albumin, most likely due to albumin transfusions in both groups posttransplant (data not shown).

Figure 3.

Comparison of serum bilirubin and INR in LD and DD recipients. (A) Bilirubin serum levels in LD and DD liver transplant recipients in the perioperative period and the first month after transplant. POD = postoperative day. (B) Internationalized ratios in recipients of LD liver transplants (LDLT) and DD liver transplants (DD-OLT) in the first month after transplant.


In attempts to reduce adult waitlist mortality, right lobe LD transplantation was introduced in the adult population, with comparable postoperative outcomes (1,28). While it is agreed that both DD and LD livers can provide a suitable graft in the appropriate recipient, it must also be recognized that each is unique in surgical technique, the amount of damage incurred, the degree of metabolic demand and dysfunction, and the requirement for regeneration and recovery from injury. In DD liver transplantation, there is a loss of hepatocytes due to donor events and ischemic injury. In LD transplants, the liver parenchyma must be surgically divided, and a right lobe graft is required to regenerate a large amount of liver mass as well as recover from a short period of ischemia. Both graft types need to perform simultaneous basic and complex metabolic functions during this period of recovery. There is a necessary balance between the need for recovery and the need to maintain normal hepatic homeostasis posttransplant. We propose that DD grafts and LD grafts respond to these needs with markedly different molecular mechanisms. The type of graft transplanted has a significant effect on this balance and each, as a consequence, has a unique and informative profile at the level of early gene expression.

Several animal studies have examined regeneration and gene expression following partial hepatectomy using various cDNA array technologies, confirming the importance of cell cycle progression genes, cytokines and other novel immediate early genes (12–14,29). There is also rodent experimental evidence illustrating the shift in metabolic function, energy balance, acute phase response and lipid metabolism in regenerating livers (9,10,30). In the transplant setting, microarray analysis of small-for-size rodent liver grafts demonstrated upregulation of vasoconstrictive and adhesion molecule genes with increases in genes associated with inflammation and cell death, and downregulation of genes related to energy metabolism (13,31,32). One clinical study demonstrated expression of acute stress genes in human liver grafts immediately after living donor transplantation, and another identified several genes in postperfusion biopsies associated with early graft dysfunction (15,33). While these studies identified specific gene expression profiles in certain classes of liver grafts, whole genome expression in serial samples was not assessed. A recent study utilizing whole genome oligonucleotide microarray analysis of five DD grafts identified over 700 genes differentially expressed after ischemia-reperfusion (16).

In our study, we were able to perform a time course analysis of the same human graft from procurement to cold storage to reperfusion with a careful comparison of LD to DD grafts at each stage. There are inherent differences between these two types of grafts due to multiple variables: donor status and quality, surgical technique at each stage and transplanted liver mass. Each of these differences may have a significant impact on posttransplant recovery. Our analysis was directed toward the expression of genes that are known to be involved in metabolism, regeneration, and the proinflammatory response. The rationale to focus on these specific gene expression patterns is based on our working hypothesis that recovery is dependent on the graft's ability to sustain metabolic activity and energy levels, initiate cell repair and cycle pathways, maintain liver synthetic function and regulate the proinflammatory response.

There were significant differences noted in gene expression between LD and DD at all three time points, but the greatest disparities were noted after reperfusion. In the DD grafts, there was a large induction of genes, many inflammatory and others associated with cell cycle regulation, consistent with findings in experimental studies (4,34). A very different pattern was recognized in the LD graft recipients. There was differential expression of inflammatory, immune and cell stress genes in a pattern unique to the LD grafts, perhaps representing a mechanism of a small graft responding to the metabolic challenge created by size disparity. An interesting finding is the upregulation of MHC class II genes observed in the LD grafts, consistent with rodent model data where regeneration of reduced-size allografts is accompanied by accelerated alloreactivity (35,36). We also saw evidence for increased expression of genes associated with the innate immune system, which has been described to be important for initiation of liver regeneration (37). Weiss et al. recently noted a different pattern of inflammatory cytokines expression between DD and LD using RT-PCR (38). Our findings in the recipients of LD grafts also correlated well with previous results from Borozan et al., who described gene expression in LDs using a 19K-human microarray after reperfusion (15). Eleven of the 15 upregulated genes and 7 of the 10 downregulated genes, verified by RT-PCR in their study, were also found to be differentially expressed in our gene list.

The LD POST biopsies also showed upregulation of genes encoding purine, pyrimidine and structural protein synthesis, expected in a regenerating liver, while genes associated with higher metabolic liver functions such as bile acid metabolism and protein metabolism were markedly decreased. This shifting balance between metabolism and regeneration is further exemplified by the differential expression of genes encoding enzymes which play key roles in glycolysis and gluconeogenesis, correlating with previous observations in animal models describing downregulation of glucose metabolism, lipid metabolism, bile secretion and steroid and hormone metabolism (14,39). We saw the potential downstream effect of this with higher bilirubin and INR in the LD recipients.

The biological network analysis provides important insights into the processes initiated in early phases of liver recovery and evidence supporting our initial hypothesis. While these networks are computed pathways and associations, they identify similarities between the two graft types, as well as significant disparities. In the LD grafts, network analysis showed an increased role of cellular proliferation, a marked increase in RNA biosynthesis and structural protein metabolism, and downregulation of oxidative phosphorylation and global carbohydrate and lipid metabolism, none of which were seen in the DD grafts.

An important question for an analysis of functional molecular pathways is determining if genome-wide expression profiling studies reflect the actual molecular mechanisms driving states of health and disease. While it must always be presented in the correct context, these mechanistic connections drawn from unbiased genome-wide profiling are still of significant value. In the process of analyzing our genome-wide gene expression data we used tools such as EASE, DAVID, GO and Ingenuity to map highly significant differentially expressed genes to functional molecular pathways that were previously defined in the literature, resulting in mapping that is directly dependent on the value of the mechanistic data from individual genes and functional pathways. Therefore, the mechanistic insights drawn from genome-wide transcriptional analysis can also have significant value, similar to the gene-by-gene, pathway-by-pathway mechanistic data that it is based upon.

It is important to note that some of the differences described in gene and protein expression may be due to the contribution of variations in procurement and surgical technique, cold and warm ischemic times, and operative times, and not just due to the need for regeneration. These differences, however, emphasize the fact that these two types of transplants produce two very unique grafts. It is not possible to control many of these technical variables, but they clearly contribute to the different expression profiles and make the two types of grafts respond in a different way. In this limited sample size, we are not able to determine to what extent regeneration or ischemic injury plays in the differential gene expression compared to the other variables, but the patterns of gene expression seem to be illustrative of these two processes. Further work with more patients will be needed to determine the actual contribution of each element.

In summary, although this study was performed in a relatively small number of patients and is necessarily descriptive in nature, we have obtained detailed array data from serial biopsies of the same liver graft, which demonstrated a highly coordinated interplay of regenerative, metabolic and inflammatory pathways in the recovering human liver graft. The classical pattern of liver regeneration, known from rodent models, seems to be present in human partial liver transplantation, in association with a distinctive inflammatory profile. We were able to demonstrate correlation of relevant genes with protein expression, as well as clinical trends with serum levels in the patients. The downregulation of genes associated with metabolic pathways in human LD grafts is unique and may be a key in understanding how the recipient/graft combination determines where efforts and energy should be placed. With further validation, these initial data may be used to explore and identify areas where intervention may be possible to support metabolic activity, decrease inflammation or enhance regeneration.

Acknowledgments and Grant Support

KMO is supported by NIH grant RO1 DK07319202 and the Biesecker Center for Pediatric Liver Disease. We acknowledge the DNA Microarray Core of TSRI and funding for this project including NIH grant U19 AI063603 and the Molly Baber Research Fund. We also note the support of Applied Biosystems Incorporated (ABI) for a portion of the TLDA arrays. This research was also supported by The European Society for Organ Transplantation 2005 Research Grant and the Michael van Vloten Foundation by funding JdJ's research fellowship. We would like to acknowledge the efforts of Dr. Kun Ming Chang in clinical data collection. We are also grateful to the Gift of Life Donor Program of Philadelphia, PA and to Dr. Robert Fisher at Virginia Commonwealth University for providing additional clinical samples for our initial analysis.

Financial disclosures: The authors have no conflict of interest and no financial disclosures.

Transcript profiling: Gene expression data can be accessed via http://www.scripps.edu/mem/eht/salomon/livertxdata/