1. Introduction
Liver transplantation (LT) represents the standard treatment for individuals suffering from end-stage liver diseases, with significant advancements in technology witnessed over the past few decades [
1]. However, apart from the organ shortage that remains a major limitation of LT [
2], incidence of post-LT early allograft dysfunction (EAD) is reportedly 20%-40%, causing critical complications that impact survival rate in allograft recipients [
3], [
4], [
5]. EAD is a multifactorial complication linked to various risk factors, including the donor risk index, surgery-related factors, and the model for end-stage liver disease score [
6], [
7]. Post-surgery liver ischemia-reperfusion injury is a complicated yet essential element in the EAD development that is regulated by multiple cell lineages and involves immune niche remodeling, inflammatory response, and cellular damage [
8], [
9]. Regrettably, the intricate cellular heterogeneity of the liver has limited our understanding of the cellular characteristics and molecular events contributing to EAD occurrence.
Recent strides in single-cell RNA sequencing (scRNA-seq) technologies have revolutionized the delineation of intricate tissues with exceptional resolution, leading to breakthroughs in the identification of key cell types and molecular events that underscore the physiological and pathological processes [
10], [
11], [
12]. For example, Shan et al. [
13] and Li et al. [
14] conducted scRNA-seq analyses to explore the heterogeneity of hepatic immune cells in transplanted livers, and elucidated the features of altered immune microenvironment in grafts, providing potential therapeutic targets for the treatment of liver transplant rejection. Moreover, based on single-cell profiling of transplanted rat livers under normal and fatty conditions, Yang et al. [
15] defined a pro-inflammatory Kupffer cell phenotype, characterized by high expression of colony stimulating factor 3 (
Csf3), and a subset of dendritic cells (DCs) highly expressed X-C motif chemokine receptor 1 (
Xcr1), demonstrating their role in mediating severe graft injury in transplanted steatotic livers. Nonetheless, very limited studies have investigated the hepatic heterogeneity between non-EAD and EAD cohorts at a single-cell resolution, further hampering our understanding of the mechanisms underlying the initiation and progression of EAD after LT.
Therefore, to bridge this gap in this study, we extensively analyzed single-cell transcriptomics on 12 human liver samples procured from three non-EAD and four EAD patients post-LT surgery, generating a total of 75 231 hepatic parenchymal and non-parenchymal cells extracted at the cold preservation and two hours portal perfusion stages. In comparison to non-EAD patients, we observed profound T/natural killer (NK) branch remodeling, specifically through mucosal-associated invariant T (MAIT) and granzyme B+ (GZMB +) granzyme K+ (GZMK +) NK cells, as well as the dominant activation and alteration of S100 calcium binding protein A12+ (S100A12 +) neutrophils in EAD patients. Notably, the transplanted livers of EAD patients exhibited a considerable increase in the presence of MAIT, GZMB + GZMK + NK cells, and S100A12 + neutrophils, which were together accompanied by EAD occurrence. We further confirmed the existence of the pathogenic immune niche and its correlation with EAD in two independent cohorts. Moreover, through our analysis of cell-cell communication mediated by ligand-receptor (LR) interactions, an essential role was identified for hepatocytes in facilitating communication with the pathogenic immune niche. Collectively, these findings shed light on a thorough understanding of the pathogenic cellular traits and molecular events associated with EAD onset, contributing valuable insights and presenting potential therapeutic strategies for preventing EAD after LT.
2. Methods
2.1. Collection of human liver samples
Human liver samples were procured from Shulan Hospital, Hangzhou, China. Samples were collected following the approval of the Institutional Review Board (2020065-77) and in strict conformity with the Helsinki Declaration (revised in 2013). The participating patients provided written informed consent. Twelve liver tissue samples were collected from seven healthy donors during the cold perfusion and portal reperfusion stages, approximately two hours after LT. The allograft was preserved and perfused with 0-4 °C Custodiol histidine-tryptophan-ketoglutarate (HTK) solution (China). Prior to performing the LT procedure, 6 mm × 6 mm × 6 mm graft blocks were collected and stored in tissue storage solution (Miltenyi, China) as preoperative samples. The allograft was then implanted via the piggyback technique involving the sequential procedures of inferior vena cava, portal vein, hepatic artery anastomosis, and biliary reconstruction. The postoperative samples were collected immediately prior to closing the abdomen and conserved in the same tissue storage solution. All liver samples underwent the tissue dissociation process in preparation for single-cell suspensions.
2.2. Classification of EAD patients
Blood samples were collected from transplant recipients on day 1 and 7 after surgery to evaluate their biomedical indicators, including alanine aminotransferase (ALT), aspartate aminotransferase (AST), total-bilirubin (T-Bil), and international normalized ratio (INR). Recipients with AST/ALT levels > 2000 U∙L−1 within the first week following surgery, or T-Bil levels ≥ 171 μmol∙L−1, or INR levels ≥ 1.6 on day 7 after surgery were classified as having EAD, whereas the remaining recipients were classified as non-EAD patients.
2.3. Preparation of single-cell suspensions
The liver tissues were sliced into pieces and placed into pre-prepared gentleMACSTM C tubes containing a dissociation enzyme mixture. The appropriate 37C_h_TDK_1 gentleMACSTM program was initiated for one hour using the gentleMACSTM Octo dissociator with heaters. Subsequently, the cell suspension was carefully filtered utilizing a 40 μm nylon cell strainer (Corning, China) and gently transferred to a 50 mL centrifuge tube for centrifugation at 500 g for 3 min. The erythrocytes were lysed with 5 mL ammonium-chloride-potassium (ACK) lysing buffer (Gibco, China). Dead cells were eliminated using a Dead Cell Removal kit (Miltenyi), according to manufacturer’s recommendations. Finally, the cell pellet was washed twice and resuspended in phosphate-buffered saline (PBS). Cell viability was calculated via a trypan blue assay (Gibco), after which it was properly placed on ice for subsequent use.
2.4. scRNA-seq
The liver single-cell suspensions were loaded onto the 10x Genomics Chromium chip (USA) to enthrall droplets for further processing. The resultant Gel beads-in-emulsion was then subjected to reverse transcription utilizing the ProFlex polymerase chain reaction (PCR) System (Thermo Fisher, USA). The resulting complementary DNA (cDNA) was then purified and amplified. Following quantification of the cDNA concentration by Qubit (Thermo Fisher), libraries were constructed utilizing a Chromium Single Cell 3′ Library & Gel Bead Kit v3 (10x Genomics) in line with the manufacturer’s instructions. Libraries were subsequently sequenced by Novogene (China) and LC-Bio Technology (China) on the Illumina NovaSeq 6000 sequencing system (paired-end multiplexing run platform).
2.5. Histological analysis
A portion of the flat tissue was sliced from the liver tissue sample kept in the preservation solution and expeditiously placed in a 10% formalin solution. Subsequently, tissue sections were paraffinized, cut into 5 μm slices, deparaffinized in xylene, and successively rehydrated in a graded series of alcohol. Next, the sections were incubated with 3% H2O2, and nonspecific binding blocking was performed with 5% bovine serum albumin for a duration of one hour. The tissue sections were stained with hematoxylin and eosin (H&E) for histological evaluation, and subjected to terminal deoxynucleotidyl transferase-mediated deoxygenation uridine triphosphate (dUTP) nick-end labeling (TUNEL) for apoptosis evaluation. Finally, images were captured utilizing an Olympus BX63 microscope (Japan) at a 200× magnification.
2.6. Immunofluorescence staining
The same liver sections were incubated with primary antibodies (Table S1 in Appendix A) overnight at a temperature of 4 °C. The sections were subsequently washed with PBS that contained Tween 20, after which they were incubated with the suitable fluorophore-conjugated secondary antibodies for 1.5 h at room temperature. The slides were mounted with antifade mounting medium and 4′,6-diamidino-2-phenylindole (DAPI) (Origene, China). The fluorescent images of MAIT and S100A12 + neutrophils were captured employing an Olympus BX63 microscope, and the fluorescent images of GZMB + GZMK + NK cells were scanned by the Olympus SLIDEVIEW VS200 system.
2.7. Data processing
The raw sequenced files were processed with Cell Ranger, based on the Genome Reference Consortium human genome build 38 (GRCh38) reference, to initially align the reads to generate the raw count data. These raw count data were further processed with Seurat [
16], wherein cells with < 200 or > 4000 unique features, or a mitochondrial percentage > 50% were filtered to exclude dead cells. Gene symbols were amended in accordance with the National Center for Biotechnology Information gene data, whereby unmatched genes and duplicate genes were removed.
2.8. Cell type annotation
The data matrix containing 12 liver tissue samples underwent normalization via LogNormalize and was integrated with the canonical correlation analysis (CCA) of Seurat. Principal component analyses (PCAs) were then conducted, followed by uniform manifold approximation and projection (UMAP) analysis for dimensional reduction and clustering analysis. To determine the cell types of each cluster using pre-computed data, scDeepSort [
17] and scCATCH [
18] were applied. Clusters expressing multiple known cell markers were manually removed, and each cluster was assigned a specific cell label based on the integration of highly expressed genes and markers in CellMatch [
18].
2.9. RNA velocity analysis
The two matrices of pre-mature (unspliced) and mature (spliced) abundances for liver samples were obtained from standard sequencing protocols using the velocyto [
19], followed by the filtration of T/NK cells and preparation of hierarchical data format version 5 annotated data (h5ad) files using SeuratWrappers R package. The RNA velocity analysis was performed based on the UMAP using scVelo [
20] with default parameters in Python.
2.10. Single-cell trajectory analysis
T/NK cells and neutrophils underwent pre-processing and were analyzed with Monocle 3 [
21], [
22] using default parameters to generate a single-cell trajectory and dissect cellular decisions. Cluster of differentiation (CD) 8
− T cells were defined as the root for T/NK cells, while peptidylprolyl isomerase F
+ (
PPIF +) neutrophils were considered the root based on the neutrophil activation score. Monocle 3 was used to track and rank the genes used to navigate these decisions over pseudotime, based on the significant correlations.
2.11. EAD-associated immune niche score
The EAD-associated immune niche score was calculated for each sample by summing the percentage of MAIT and GZMB + GZMK + NK cells among total T/NK cells as well as the percentage of S100A12 + neutrophils among the whole neutrophils of the corresponding sample.
2.12. Cell-cell communication analysis
The scCrossTalk tool was utilized to investigate cell-cell communication among various cell populations based on the highly expressed ligands of sender cells and receptors of receiver cells. For each cell type, human LR interaction pairs recorded in CellTalkDB were applied. Significantly highly expressed ligands and receptors were filtered with a percentage of expressed cells > 10% and P < 0.05, using the Z score for each gene. The interaction score, S( L i -R j ), for the ligand L of the sending cell type i and the receptor R of the receiving cell type j was defined as the product of the average expression of L and R in the sender and receiver cells, respectively.
2.13. Pathway and biological process enrichment
The Metascape [
23] web tool was used to carry out the enrichment of pathways and biological processes, wherein the top 100 highly expressed genes were chosen according to the average gene expression fold change. The AddModuleScore function in Seurat was utilized to compute the module score of Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways and Gene Ontology (GO) biological processes. To obtain the signatures, the Molecular Signatures Database (MSigDB) was accessed through msigdbr. Gene set enrichment analysis (GSEA) was conducted by clusterProfiler [
24] with the ordered gene list and MSigDB.
2.14. Verification in the human LT scRNA-seq dataset
The raw FASTQ files were retrieved from the Sequence Read Archive (SRA) database under the accession number SRP313633 [
25] and processed with Cell Ranger, utilizing the GRCh38 reference. The data matrix containing three liver samples collected from pre-procurement (PP), at the end of preservation (EP), and two hours post-reperfusion (PR) was normalized via LogNormalize and was then integrated with the Seurat’s CCA. PCAs were then conducted, followed by UMAP analysis for dimensional reduction and clustering analysis. Cell types in each pre-computed cluster were identified based on markers previously published in the original paper. Pearson’s correlation coefficient was used to analyze the correlation of MAIT,
GZMB + GZMK + NK cells,
S100A12 + neutrophils, and other cell types between our data and this independent data with the top 100 highly expressed genes for each subtype.
2.15. Verification in the human LT bulk RNA sequencing (RNA-seq) dataset
The raw bulk RNA-seq data matrix of liver cells obtained from eight donation after brain death (DBD) with subsequent post-implantation EAD in the recipient and eight DBD without EAD were downloaded from the Gene Expression Omnibus (GEO) database under the accession number GSE23649, wherein two samples were collected from each donor: Sample A was taken immediately before cold perfusion and sample B 2 h after portal reperfusion. Gene expression values less than 0 were transformed to 0 before normalization. The raw matrix was normalized by setting the median value to 1000 for each sample and transformed into a log2 matrix. To obtain the cell type composition of the bulk RNA-seq data, we used our scRNA-seq data as the reference for SpaTalk to deconvolute the bulk transcriptomics data by randomly selecting 1000 representative cells for each subtype while keeping all parameters as default. Robust cell type decomposition (RCTD) was used with the default parameters. The EAD-associated immune niche score was determined based on the deconvoluted weight of MAIT, GZMB + GZMK + NK cells, and S100A12 + neutrophils. The EAD module score for each sample was calculated using the AddModuleScore in Seurat, wherein the top 100 significantly highly expressed genes of the EAD-associated immune niche were selected as the signature by comparing the aggregated expression profile of MAIT, GZMB + GZMK + NK cells, and S100A12 + neutrophils in non-EAD patients and that in EAD patients.
2.16. Statistical analysis
R 4.1.3, Python 3.9, and GraphPad Prism 8.0.1 were used for the statistical analyses. Differences between two groups were displayed with mean and standard error of mean (SEM) and determined using Welch’s t-test. Results with P < 0.05 were considered statistically significant.
3. Results
3.1. Overview of the transplanted liver profiling for non-EAD and EAD patients
A total of 12 liver samples were obtained from 7 donors for scRNA-seq using the 10x Genomics platform (Table S2 in Appendix A), and blood samples of recipients were collected to assess biochemical indicators, namely, ALT, AST, T-Bil, and INR to categorize non-EAD and EAD patients (
Fig. 1(a); Table S3 in Appendix A). After eliminating dead and ambiguous cells, a total of 75 231 cells were included for integration, dimensionality reduction, and clustering analysis, resulting in 46 unique clusters for the 12 liver samples (Figs. S1(a) and (b) in Appendix A). Given the widespread batch effect, we applied the widely-used CCA algorithms in Seurat [
16] to integrate the 12 samples (Table S4 and Figs. S1(c)-(e) in Appendix A). Combining the predicted cell types by pre-trained scDeepSort [
17] and scCATCH [
18] with highly expressed genes for each cluster, these 45 clusters were categorized into 11 main cell types: B cells (
CD79A and membrane spanning 4-domains A1 (
MS4A1)), cholangiocytes (keratin 8 (
KRT8) and keratin 18 (
KRT18)), DCs (G protein-coupled receptor 183 (
GPR183) and
CD74), endothelial cells (ficolin 2 (
FCN2) and ficolin 3 (
FCN3)), hepatocytes (albumin (
ALB) and apolipoprotein A2 (
APOA2)), fibroblasts (biglycan (
BGN) and transgelin (
TAGLN)), mononuclear phagocytes (MPs) (complement component 1, q subcomponent, A chain (
C1QA) and complement component 1, q subcomponent, B chain (
C1QB)), neutrophils (S100 calcium binding protein A8 (
S100A8) and S100 calcium binding protein A9 (
S100A9)), plasma cells (immunoglobulin kappa constant (
IGKC) and immunoglobulin heavy constant alpha 1 (
IGHA1)), proliferating cells (marker of proliferation Ki-67 (
MKI67) and stathmin 1 (
STMN1)), and T/NK cells (interleukin (IL)-7 receptor (
IL7R) and natural killer cell granule protein 7 (
NKG7)) through analysis of known cell markers in CellMatch (
Figs. 1(b) and (c); Table S5 in Appendix A).
Compared to hepatic parenchymal cells, a significant proportion of hepatic non-parenchymal cells were detected in the livers collected before and after LT. Specifically, T/NK cells, neutrophils, and MPs constituted more than 80% of the cells obtained before and after LT (
Fig. 1(d)). This is attributed to the fact that liver ischemia-reperfusion injuries are intimately associated with diverse forms of inflammatory responses, such as immune cell activation, migration, and infiltration [
26]. Notably, the T/NK cell population exhibited a conspicuous decrease following LT, while neutrophils saw a substantial increase (
Fig. 1(d)). This implies the critical immune remodeling of T/NK cells and neutrophils during the liver ischemia-reperfusion process.
To analyze the difference in hepatic immune microenvironments between patients with non-EAD and EAD, the cohort of 7 patients was partitioned into two groups based on the globally established criteria for EAD (
Fig. 1(a)). Accordingly, Patient 1, 3, and 4 were categorized in the non-EAD group, while patient 2, 5, 6, and 7 were diagnosed with EAD (
Fig. 1(e), Table S3). Notably, the alterations in T/NK cells and neutrophils between the non-EAD and EAD groups exhibit comparable trends to those observed between the before and after LT groups (
Fig. 1(f); Fig. S1(f) in Appendix A). Dissociation of cells across each sample highlighted that T/NK cells and neutrophils remained steady after LT in non-EAD patients, whereas they displayed pronounced variations after LT in EAD patients (
Figs. 1(g) and (h); Fig. S1(g) in Appendix A), indicating that these critical immune cells have a decisive role in the onset of EAD, as they may predominantly induce more severe hepatic ischemia-reperfusion injury (
Fig. 1(i); Fig. S1(h) in Appendix A). Although the MP is also an important and large immune population in the transplanted livers, no potential MP sub-populations contributing to the EAD onset were observed (Fig. S2 in Appendix A). Therefore, we focused on the T/NK cell and neutrophil subpopulations in subsequent analyses to explore their associations with the occurrence of EAD.
3.2. T/NK branch remodeling through MAIT and GZMB + GZMK + NK cells in EAD patients
T/NK cells were filtered and integrated from 12 liver samples for dimensionality reduction and clustering analysis, creating 32 distinctive clusters for 44 799 cells (Fig. S3(a) in Appendix A). In light of well-established cell markers and highly expressed signatures for each cluster, T/NK cells were classified into 8 subtypes including
CD4 + naive T cells,
CD8 + GZMB + T cells,
CD8 + GZMB + GZMK + T cells,
CD8 − T cells, MAIT cells,
GZMB + NK cells,
GZMB + GZMK + NK cells, and NK T cells (
Figs. 2(a) and (b); Figs. S3(b) and (c) in Appendix A). Analysis of significantly highly expressed genes, enriched pathways, and GO biological processes through Metascape [
23] showed that these T/NK cell subtypes possessed similarities in functions despite their distinct gene signatures (
Figs. 2(c) and (d); Table S6 in Appendix A). MAIT cells, for example, presented a markedly elevated level of C-C motif chemokine ligand 20 (
CCL20), vimentin (
VIM), natural cytotoxicity triggering receptor 3 (
NCR3), and metallothionein 2A (
MT2A), alongside noteworthy positive regulation of cytokine production and regulation of IL-2 production. Contrarily,
GZMB + GZMK + NK cells demonstrated enhanced enrichment in the regulation of defense response and the positive regulation of response to external stimulus, as inferred from the heightened abundance of IL-1 receptor associated kinase 3 (
IRAK3), killer cell lectin like receptor C1 (
KLRC1), and phospholipase C gamma 2 (
PLCG2), among others.
Relative to
CD4 + naive T cells and
CD8 − T cells,
CD8 + GZMB + T cells,
CD8 + GZMB + GZMK + T cells, MAIT cells,
GZMB + NK cells,
GZMB + GZMK + NK cells, and NK T cells consistently exhibited stronger cell activation, adaptive immune response, positive regulation of IL-2 production, and regulation of defense response under ischemia-reperfusion stimuli (
Fig. 2(e)). As a result, the RNA velocity and single-cell trajectory analysis utilizing scVelo [
20] and Monocle 3 [
21], [
22] were conducted to dissect cellular decisions of T/NK cells, leading to the reconstruction of two main branches from the root towards the end states, namely the MAIT cells and
GZMB + GZMK + NK cells, indicating the gradient cell activation and gradually enhanced cytotoxicity along the pseudotime axis (
Fig. 2(f); Fig. S3(d) in Appendix A). Based on the reconstructed pseudotime trajectory, significantly correlated genes that cells use to navigate the decision over pseudotime were tracked, where
GZMB and
GZMK were found to be strikingly correlated with pseudotime (
Fig. 2(g)). It is commonly understood that the preproproteins encoded by
GZMB and
GZMK are secreted by NK cells and cytotoxic T cells and proteolytically processed to generate the active protease, which induces target cell apoptosis [
27], [
28], [
29]. These findings illustrate the crucial role of
GZMB and
GZMK in the activation and cytotoxicity of T/NK cells during the hepatic ischemia-reperfusion injury.
Next, we performed a comparison of the dynamic shift in T/NK cellular niche for 12 liver samples derived from non-EAD and EAD patients, from the stage of cold preservation to that of portal perfusion. Despite the absence of apparent differences between T/NK cell subtype compositions before and after LT, the T/NK cellular niches exhibited extensive heterogeneity across 12 liver samples among non-EAD and EAD groups (
Fig. 2(h); Fig. S3(e) in Appendix A). Nonetheless,
CD8 + GZMB + GZMK + T cells, MAIT cells,
GZMB + NK cells,
GZMB + GZMK + NK cells, and NK T cells emerged as the major T/NK cells, portraying a comparable percentage in both non-EAD and EAD groups (Figs. S3(f)-(h) in Appendix A). Notably, a slighter increase was recorded for
CD8 + GZMB + T cells,
CD8 + GZMB + GZMK + T cells,
GZMB + NK cells, and NK T cells within non-EAD patients as opposed to their EAD counterparts. Conversely, considerable percentages of MAIT cells and
GZMB + GZMK + NK cells were observed in EAD patients, displaying an overwhelming proportion in comparison to non-EAD patients both before and after LT (
Fig. 2(i)). These profound variations emphasize the occurrence of evident immune remodeling through MAIT cells and
GZMB + GZMK + NK cells in EAD patients, specifically branching from T/NK cells, highlighting their leading role in the ischemia-reperfusion injury of transplanted livers.
3.3. Dominant activation and alteration of S100A12+ neutrophils in EAD patients
We extracted 11 452 neutrophil cells from 12 hepatic tissue samples, which were subjected to dimensionality reduction and clustering analysis, producing 12 distinct clusters (Fig. S4(a) in Appendix A). Based on the identification of well-established cell markers and genes exhibiting high expression levels for each cluster, these neutrophil cells were stratified into five subtypes, namely C-C motif chemokine ligand 5
+ (
CCL5+ ), defensin alpha 3
+ (
DEFA3+ ), interferon induced protein with tetratricopeptide repeats 2
+ (
IFIT2+ ),
PPIF+, and
S100A12+ neutrophils (
Figs. 3(a) and (b)). These five subtypes of neutrophils within the hepatic neutrophil niche exhibited differential enrichment of pathways and GO biological processes, as discerned from the highly expressed genes for each subtype (
Fig. 3(c); Table S7 in Appendix A). For instance,
IFIT2+ neutrophils were marked by interferon-stimulated gene signatures, such as interferon induced protein with tetratricopeptide repeats 1 (
IFIT1), interferon induced protein with tetratricopeptide repeats 2 (
IFIT2), and interferon induced protein with tetratricopeptide repeats 3 (
IFIT3), and were significantly linked with the regulation of defense response and nucleotide-binding oligomerization domain (NOD)-like receptor signaling pathway.
PPIF+ neutrophils, on the other hand, displayed remarkable enrichment in the leukocyte migration and IL-18 signaling pathway. Meanwhile,
DEFA3+ and
S100A12+ neutrophils were found to perform optimally in cell killing, neutrophil degranulation, and neutrophil extracellular trap formation, as evidenced by the elevated levels of
DEFA3, elastase, neutrophil expressed (
ELANE),
S100A12, and cathelicidin antimicrobial peptide (
CAMP), among others.
Given the gradient distribution of the module score for neutrophil activation involved in immune response (
Fig. 3(d)), we performed a single-cell pseudotime analysis to delve into the cellular decision-making process of neutrophils in response to the ischemia-reperfusion stimuli. Subsequently, a pseudotime trajectory, starting from
PPIF+ neutrophils and progressing through
S100A12+ neutrophils, leading to
DEFA3+ neutrophils was reconstructed (
Fig. 3(e)). This trajectory was consistent with the pattern of neutrophil activation, leading from
PPIF+ neutrophils towards
S100A12+ neutrophils, as demonstrated by the strong correlation between the pseudotime and the corresponding module score (
Fig. 3(f)).
S100A12 participates in specific calcium-dependent signal transduction pathways and regulates cytoskeletal components, thus modulating various neutrophil activities [
30]. As anticipated, a notable correlation was observed between expression of
S100A12 and the pseudotime, implying that
S100A12+ neutrophils represent an immensely activated population.
Remarkably, we observed improved neutrophil migration, chemotaxis, and extravasation that commenced with the
DEFA3+ neutrophil, progressed through
S100A12+ neutrophils, and culminated in the
PPIF+ neutrophil. This process occurred in a direction that countered the pseudotime axis and was accompanied by an inverse correlation between the pseudotime and corresponding module score (
Fig. 3(g); Fig. S4(b) in Appendix A).
PPIF expression was also negatively correlated with pseudotime, as observed by the dramatic inhibition of cell migration upon gene silencing of
PPIF, thereby revealing the exceedingly migrated population for
PPIF+ neutrophils. It is noteworthy that
DEFA3+ neutrophils exhibited elevated levels of C-X-C motif chemokine receptor 4 (
CXCR4) expression, but negligible abundance of C-X-C motif chemokine receptor 2 (
CXCR2) and selectin L (
SELL) (
Fig. 3(h); Fig. S4(c) in Appendix A), thus suggesting that
DEFA3+ neutrophils were associated with the aged population, which is distinct from the mature neutrophil population exemplified by
IFIT2+,
PPIF+, and
S100A12+ neutrophils [
31].
Upon assessment of the hepatic neutrophil niche across a dozen liver samples obtained from both non-EAD and EAD patients,
DEFA3+,
PPIF+, and
S100A12+ neutrophils constitute the bulk of each sample, despite significant heterogeneity (
Fig. 3(i); Fig. S4(d) in Appendix A). By examining the neutrophil subtypes from the cold preservation stage to that of portal perfusion,
CCL5+ and
DEFA3+ neutrophils demonstrated a substantial decrease, whereas
IFIT2+,
PPIF+, and
S100A12+ saw varying degrees of increase after LT. These trends held true for non-EAD and EAD patients (
Fig. 3(j); Fig. S4(e) in Appendix A). Remarkedly, the proportion of
S100A12+ neutrophils in EAD patients exceeded that in non-EAD patients both before and after LT, with overwhelming scores in the regulation of neutrophil degranulation that is widely recognized to aggravate ischemia-reperfusion injury (
Fig. 3(k)). These observations revealed the dominant activation and alteration of
S100A12+ neutrophils in EAD patients, underscoring their crucial role in the ischemia-reperfusion injury of transplanted livers.
3.4. Identification of a pathogenic immune niche associated with EAD onset
In consideration of the prominent composition of T/NK cells and neutrophils in transplanted livers and their critical involvement in the hepatic ischemia-reperfusion injury, our quest delved into the differentiation of the T/NK and neutrophil niche between non-EAD and EAD patients (
Figs. 4(a) and (b)). Remarkably, both MAIT cells and
GZMB + GZMK + NK cells exhibited a marked rise in EAD patients in comparison to non-EAD patients, characterized by a percentage increase from 13% and 20% to 21% and 40%, respectively (
Fig. 4(c)). Conversely, other T/NK cell subtypes demonstrated trivial transitions or variable decline from non-EAD to EAD inception. In parallel, the
S100A12 + neutrophils illustrated a significant surge in the neutrophil niche, encompassing 55% of EAD patients, as against a mere 37% in non-EAD patients, while other neutrophil subtypes manifested inconspicuous alterations or a slight reduction (
Figs. 4(b) and (c)). Accordingly, our hypothesis posits that MAIT cells,
GZMB + GZMK + NK cells, and
S100A12 + neutrophils constitute a unique pathogenic immune niche closely affiliated with the incidence of EAD. Through an integrated analysis of T/NK cells and neutrophils, a stark upward trend from non-EAD to EAD patients was observed for the immune niche score incorporating MAIT cells,
GZMB + GZMK + NK cells, and
S100A12 + neutrophils (
Fig. 4(d)). Concordantly, EAD patients showed a significantly higher immune niche score than non-EAD patients (
Fig. 4(d)), underscoring the preeminent role of this pathological immune niche in driving EAD onset. Moreover, we employed immunofluorescent staining to validate the pathogenic immune niche using well-established markers. As a result, the dominant accumulation of MAIT cells,
GZMB + GZMK + NK cells, and
S100A12 + neutrophils forming the pathogenic immune niche was witnessed in EAD patients (
Fig. 4(e); Fig. S5(a) in Appendix A), reinforcing our findings.
To decipher the molecular characteristics underlying the onset of EAD, we conducted a comprehensive analysis of differentially expressed genes (DEGs) within the EAD-associated pathogenic immune niche (i.e., MAIT cells,
GZMB + GZMK + NK cells, and
S100A12 + neutrophils) between the non-EAD and EAD patients (Figs. 4(f)-(h)). We found that the immune niche in EAD patients exhibited a significant upregulation of multiple inflammatory and cytotoxicity-related cytokines and chemokines, including nuclear factor kappa B subunit 1 (
NFKB1), serum amyloid A1 (
SAA1),
GZMB,
GZMK,
S100A8,
S100A9, and
S100A12, which were commonly overexpressed among the three immune cell types in EAD patients (Fig. S5(b) and Tables S8-S10 in Appendix A). Specifically, our investigation in the MAIT cells of EAD patients revealed a remarkable elevation of niban apoptosis regulator 1 (
NIBAN1) (
Fig. 4(f)), whose encoded protein can govern p53-mediated apoptosis [
32]. Likewise, tumor necrosis factor receptor superfamily member 18 (
TNFRSF18) demonstrated augmented expression upon T cell activation [
33], with a marked surge in MAIT cells of EAD patients as compared to non-EAD patients (Fig. S5(c) in Appendix A). Notably,
GZMB + GZMK + NK cells from EAD patients displayed a greater abundance of arachidonate 5-lipoxygenase activating protein (
ALOX5AP) (
Fig. 4(g)), which is essential for the synthesis of leukotrienes and has been implicated in several types of inflammatory responses [
34]. Furthermore,
GZMB + GZMK + NK cells in EAD patients possessed a stronger migratory capacity, as evidenced by the heightened expression of adhesion G protein-coupled receptor G3 (
ADGRG3) (Fig. S5(d) in Appendix A), a gene involved in G protein-coupled receptor signaling and cell migration regulation. In the case of
S100A12+ neutrophils in EAD patients relative to non-EAD patients, we observed increased expression of immunoglobulin superfamily member 6 (
IGSF6),
ADGRG3, and triggering receptor expressed on myeloid cells 1 (
TREM1) (
Fig. 4(h); Fig. S5(e) in Appendix A), which promote the amplification of neutrophil-mediated inflammatory responses by stimulating the release of pro-inflammatory chemokines and cytokines and facilitate increased cell activation [
35], [
36], [
37]. These findings shed light on the molecular intricacies that underpin the pathogenesis of EAD, emphasizing the pivotal role of the pathogenic immune niche comprised of MAIT cells,
GZMB + GZMK + NK cells, and
S100A12 + neutrophils, and their associated signaling molecules, in the development of this debilitating condition.
3.5. Verification of the EAD-associated immune niche in two independent cohorts
Two human LT cohorts were acquired from publicly available platform, to further validate the EAD-associated immune niche. First, the human LT scRNA-seq dataset under accession number SRP313633 [
25] was downloaded and processed with CellRanger and Seurat, yielding nine main cell types (
Fig. 5(a); Figs. S6(a) and (b) in Appendix A), including T/NK cells and neutrophils, encompassing 16 714 cells from 3 liver samples collected from PP, EP, and two hours PR. T/NK cells and neutrophils exhibited a congruous pattern from EP to PR, thereby corroborating the implications deduced from our gathered data (Fig. S6(c) in Appendix A). By re-clustering the T/NK cells, we identified a total of 11 clusters (
Fig. 5(b)), two of which were annotated as MAIT cells due to their high expression of CD3 delta subunit of T cell receptor complex (
CD3D), CD8 subunit alpha (
CD8A), CD8 subunit beta (
CD8B), and solute carrier family 4 member 10 (
SLC4A10) (
Fig. 5(c)). Two additional clusters were classified as
GZMB + GZMK + NK cells, supported by the signatures of killer cell lectin like receptor F1 (
KLRF1),
GZMB, and
GZMK (
Fig. 5(c)). In parallel, we subjected neutrophils under three conditions to integrated clustering analysis (
Fig. 5(d)), identifying a cluster with high expression of
S100A12, designated as
S100A12 + neutrophils (
Fig. 5(e)). Concordantly, the MAIT cells,
GZMB + GZMK + NK cells, and
S100A12 + neutrophils demonstrated the most positively correlated transcriptomic profiles with the corresponding subtypes among T/NK and neutrophil subtypes in our study (
Fig. 5(f)), further corroborating the presence of the pathogenic immune niche comprised of MAIT cells,
GZMB + GZMK + NK cells, and
S100A12 + neutrophils in transplanted livers. Additionally, we confirmed the existence of
GZMB + GZMK + NK cells and
S100A12 + neutrophils in a rat LT dataset [
15]. In contrast, MAIT cells were not readily detected (Figs. S6(d)-(h) in Appendix A); these cells are reportedly enriched in humans and considerably less abundant in other species.
Second, we acquired a distinct cohort of human LT specimens, derived from bulk RNA-seq technology via the publicly available GSE23649 [
38], comprising 8 DBD with subsequent occurrence of post-implantation EAD in the recipient, juxtaposed with another 8 DBD donors without EAD (
Fig. 5(g)). Two tissue samples were extracted from each donor, immediately taken before cold perfusion and two hours after portal reperfusion, yielding 16 samples per group. After employing the SpaTalk [
39] and RCTD [
40] methodology to deconvolute bulk RNA-seq data and generate specific cell type proportions using our scRNA-seq data as the reference, pathogenic immune niche scores were obtained over samples of non-EAD and EAD patients (
Fig. 5(g)). Despite the variable distribution within samples, the pathogenic immune niche scores were significantly higher amongst EAD patients compared to non-EAD patients (
Fig. 5(h); Fig. S6(i) in Appendix A). By capitalizing on the top 100 EAD-associated genes identified in our study (Table S11 in Appendix A), we calculated the EAD module scores in EAD patients, which were markedly elevated in contrast to those amongst non-EAD patients (
Fig. 5(i); Fig. S6(j) in Appendix A). Furthermore, the analysis of DEGs in EAD patients, in comparison to non-EAD patients, revealed a multitude of high-level genes such as
NIBAN1,
ALOX5AP,
IGSF6,
TNFRSF18,
ADGRG3, and
TREM1 (
Figs. 5(j) and (k); Fig. S6(k) in Appendix A), all of which were identified within our pathogenic immune niche (Figs. 4(f)-(h)). We noted that
SIPA1L1, a gene associated with signal-induced proliferation, was commonly overexpressed in EAD patients within both the GSE23649 dataset and our data (
Fig. 5(k); Figs. S5(d) and (e)), indicating the heightened proliferative activity of the pathogenic immune niche in EAD patients, accounting for the accumulated percentage of MAIT cells,
GZMB + GZMK + NK cells, and
S100A12 + neutrophils in EAD patients when compared to non-EAD patients. These consistent findings substantiate a significant association between the pathogenic immune niche in transplanted livers and EAD incidence.
3.6. Uncovering cell-cell communications for the EAD-associated immune niche
Given the salient cellular composition and pivotal functions enacted by hepatocytes and endothelial cells in maintaining normal physiological hepatic functions, we proceeded to investigate cell-cell communications among hepatocytes, endothelial cells, and the EAD-associated immune niche, consisting of MAIT cells,
GZMB + GZMK + NK cells, and
S100A12 + neutrophils, in non-EAD and EAD patients (
Fig. 6(a); Fig. S7(a) in Appendix A). Leveraging the scCrossTalk algorithm and its associated CellTalkDB [
41] database, we identified a greater number of LR interaction pairs in EAD patients than in non-EAD patients (
Fig. 6(b)). The scores assigned to the communication network between pairwise cell types demonstrated significantly increased levels in EAD patients (
Figs. 6(c) and (d); Fig. S7(b) in Appendix A), indicating the heightened interconnectedness between hepatocytes, endothelial cells, MAIT cells,
GZMB + GZMK + NK cells, and
S100A12 + neutrophils in EAD patients. These observations suggest a perpetual communication network among these cellular entities, likely a consequence of the ischemia-reperfusion process. Further investigational insight has highlighted signals emanating from hepatocytes and endothelial cells comprising much of the cell-cell communication network (
Fig. 6(e); Fig. S7(c) in Appendix A), with a noteworthy intersection shared amongst LR pairs in non-EAD and EAD patients for a given sender-receiver pair (
Fig. 6(f); Fig. S7(d) in Appendix A). However, the LR pairs involved in mediating communication from hepatocytes and endothelial cells to additional cell populations exhibited considerable heterogeneity within each cohort, with
S100A12 + neutrophils emerging as the most responsive population towards the signals dispatched from hepatocytes and endothelial cells. Intriguingly, the pathogenic immune niche demonstrated weaker LR-mediated interaction with hepatocytes and endothelial cells (
Fig. 6(g); Fig. S7(e) in Appendix A), implying that these cells, as the primary sender, preferentially communicate with the EAD-associated immune niche during ischemia-reperfusion stimuli.
To assess the variegated cell-cell communications within non-EAD and EAD patients, we examined the top 10 LR interaction pairs that form the bedrock of hepatocyte-immune and endothelia-immune cross-talk (
Fig. 6(h)). Of particular note is the
SAA1-formyl peptide receptor 1 (
FPR1) interaction observed within the hepatocyte-
S100A12 + neutrophil communication, which exhibited the highest score in EAD patients. Widely disseminated literature attests to
SAA1 as a major acute-phase protein that demonstrates heightened expression in response to inflammation and tissue damage [
42], while
FPR1, as the receptor of
SAA1, mediates the response of phagocytic cells to host invasion by microorganisms and importantly plays a role in host defense and inflammation [
43]. Relative to non-EAD patients, EAD patients express significantly higher levels of both
SAA1 and
FPR1 within hepatocytes and
S100A12 + neutrophils, respectively (
Fig. 6(i)). Furthermore, the observed
SAA1-
FPR1 interaction is specific to hepatocytes and
S100A12 + neutrophils and is rarely observed in other cell populations (
Fig. 6(j)), consistent with the findings on the human bulk LT dataset (Fig. S7(f) in Appendix A), signifying its pivotal role in mediating the cross-talk between these two cell types during the ischemia-reperfusion process. Despite the broader range of LR interaction pairs underlying endothelial-immune communications in EAD patients, we noted reduced scores in the top mutually shared LR interaction pairs relative to non-EAD patients (
Fig. 6(h)), thus underscoring the critical communicative role of hepatocytes concerning the pathogenic immune niche.
Hence, a comparative assessment of DEGs was carried out to scrutinize the disparity in hepatocellular molecular signatures between non-EAD and EAD patients (
Fig. 6(k)). Notably, hepatocytes in the EAD cohort exhibited significantly elevated expression of multiple chemokines, including
S100A8,
S100A9, macrophage migration inhibitory factor (
MIF), C-C motif chemokine ligand 3 (
CCL3), and
CXCL8, in line with enriched pathways and biological processes anchored to acute inflammatory response, neutrophil chemotaxis, and positive cytokine production regulation. This compellingly positions hepatocytes as the key players in inducing chemotaxis while proffering numerous ligands, such as fibrinogen alpha chain (
FGA), fibrinogen beta chain (
FGB), and fibrinogen gamma chain (
FGG), which match the receptor of integrin subunit beta 2 (
ITGB2) in MAIT cells,
GZMB + GZMK + NK cells, and
S100A12 + neutrophils (
Fig. 6(h); Fig. S7(g) in Appendix A), in promoting leukocyte adhesion and facilitating cell surface-mediated immune responses [
44]. In contrast, the hepatocytes in non-EAD patients continued to execute normal metabolic functions alongside a weak acute inflammatory response, as discerned through the highly expressed mitochondrially encoded cytochrome c oxidase II pseudogene 12 (
MTCO2P12), glutathione S-transferase alpha 1 (
GSTA1), cytochrome P450 family 3 subfamily A member 4 (
CYP3A4), cytochrome P450 family 2 subfamily A member 7 (
CYP2A7), C-X-C motif chemokine ligand 10 (
CXCL10), and apolipoprotein A1 (
APOA1) (
Fig. 6(k)). Further GSEA concurred with this observation, notably signifying hepatocyte enrichment in EAD patients with hallmarks of tumor necrosis factor (TNFα) signaling via nuclear factor kappa B (NF-κB) and suppression of bile duct metabolism (
Fig. 6(l)). This highlights the impaired hepatic metabolic function in EAD patients. Upon the recruitment of injured hepatocytes, a communication network was established between MAIT cells,
GZMB + GZMK + NK cells, and
S100A12 + neutrophils, cooperatively forming the pathogenic immune niche via multiple signals such as beta-2-microglobulin (
B2M), transforming growth factor beta 1 (
TGFB1),
S100A8,
S100A9,
MIF, calmodulin 1 (
CALM1), and annexin A1 (
ANXA1) that correspond to
CD69,
ITGB2,
CXCR4, and
CD44 (Figs. S7(h)-(j) in Appendix A), which are crucially involved in enhancing leukocyte activation, adhesion, proliferation, and cytotoxicity [
45], [
46].
4. Discussion
In this study, we identified an EAD-associated immune niche that encompasses MAIT cells, GZMB + GZMK + NK cells, and S100A12 + neutrophils, through a comprehensive scrutiny of single-cell transcriptomic analysis of 12 human liver samples derived from non-EAD and EAD patients. In comparison to non-EAD patients, EAD patients exhibited a noteworthy surplus of MAIT cells, GZMB + GZMK + NK cells, and S100A12 + neutrophils with an ischemia-reperfusion boost in LT, which engenders a pathogenic immune niche, triggering the onset of EAD. Delving deeper into the research data involving two different cohorts of transplant patients, we have corroborated the prevalence of this pathogenic immune niche and its strong correlation with the development of EAD, highlighting the crucial role played by this pathogenic immune niche in uncovering critical cellular and molecular events underlying the occurrence of EAD following LT.
Despite the portal perfusion only lasting two hours, we observed a significant difference in the cellular composition and molecular expression between livers before and after LT. On the one hand, we found the neutrophils and MPs saw a substantial increase after LT, in line with the widespread immune infiltration during liver ischemia-reperfusion process (Fig. S1(g)). On the other hand, the abundance of endothelial cells and hepatocytes, as the major populations within the hepatic lobule, decreased after LT, in agreement with that endothelial cells and hepatocytes were remarkably impaired by ischemia-reperfusion injuries (Fig. S1(g)). Moreover, by analyzing the DEGs between hepatocytes before and after LT, the more significantly enriched pathways involving the inflammation and cell death were observed in hepatocytes after LT than that before LT (Fig. S8(a) in Appendix A). However, we noticed that the number of downexpressed and overexpressed genes in hepatocytes after LT are less than that in hepatocytes of EAD patients (Fig. S8(b) in Appendix A), and the expression profile of hepatocytes between non-EAD and EAD patients demonstrated the significant dissimilarity, as evidenced by the lowest correlation coefficient (Fig. S8(c) in Appendix A). Given the immune niche, no significant difference in the immune niche score was observed between the liver samples before and after LT in our cohort and the bulk cohort, either among non-EAD or EAD patients, except the slight increase existed in the livers after LT compared to the counterparts before LT in non-EAD patients of the bulk cohort (Figs. S9(a) and (b) in Appendix A). The analysis clearly supported the pulling of before LT and after LT samples for each of non-EAD and EAD groups. These findings indicate that the difference in hepatocytes and the pathogenic immune niche between non-EAD and EAD patients is more significant than between liver samples before and after LT.
Although differences in the immune niche score were observed between the non-EAD and EAD patients before LT, and between the non-EAD and EAD patients after LT, their
P values were higher than that obtained from the one-sided Welch test between non-EAD and EAD patients by integrating the liver samples before and after LT (Fig. S9(c) in Appendix A). This phenomenon was also observed in the bulk dataset and the integrated one of our cohorts and the bulk dataset (Figs. S9(d)-(f) in Appendix A), implying that the immune niche comprising MAIT cells,
GZMB + GZMK + NK cells, and
S100A12 + neutrophils in transplanted livers was indeed more significantly enriched in EAD patients than non-EAD patients. Indeed, the EAD grafts had higher immune niche scores than non-EAD ones even before LT in the bulk dataset, suggesting that such immune related differences are not induced by perfusion but by different conditions in the donors or grafts between EAD and non-EAD patients. Meanwhile, the immune niche score exhibited insignificant associations with conditions including the donor age, graft weight, recipient age, graft to recipient weight ratio (GRWR), procedure time, blood loss, and infusion (Fig. S9(g) in Appendix A). Expectedly, a significantly positive association between the immune niche score and the cold ischemia time (CIT) was observed with Pearson correlation coefficient (
r) reaching 0.94 with the longer CIT in EAD patients than in non-EAD patients, for the fact holds that the prolonged CIT is an independent risk factor for graft loss after LT [
47].
EAD has myriad etiologies, including rejection, ischemia-reperfusion injury, and small-for-size syndrome. Notably, the transplanted type of most recipients in our study is the whole LT
in situ, except the non-EAD patient 1 with the right LT
in situ (Table S12 in Appendix A). Specifically, the GRWR for all recipients are > 0.8%, which can rule out the small-for-size syndrome for recipients in our cohort [
48]. Given the rejection, all recipients in our cohort received routine anti-rejection drugs without significant adjustment, and no T cell-mediated rejection (TCMR) and antibody-mediated rejection (AMR) were observed among all recipients in our cohort, including the patient 3 with ABO incompatibility (Table S13 in Appendix A). The EAD occurrence rate and the subsequent survival rate may be related to the donor death type as the analysis performed by Mazilescu et al. [
49] showed the more frequent occurrence of EAD in the donation after circulatory death (DCD) versus DBD group based on the cohort of 1068 patients. However, we observed that half of recipients with DBD or DCD were accompanied by the occurrence of EAD with no significant difference in the EAD onset between the DBD and DCD groups (Table S2). The reason might be the limited number of patients in our cohort. Nevertheless, the majority of donors were observed to be DBD in our cohort, in line with the findings reported by Mazilescu et al. [
49].
Although ischemia-reperfusion injury is the major events in LT, regenerative events also occur. The hepatocyte growth factor (HGF)-methoprene-tolerant (MET) interaction is a critical LR pair for hepatic regeneration. In line with the previous findings [
50], the fibroblast emerged as the major source of secreted HGF, with no significantly differential expression in the livers before and after LT. However, a substantially differential expression was detected between livers of non-EAD and EAD patients in fibroblasts (Fig. S10(a) in Appendix A). Notably, the averaged log1p expression level of
MET in transplanted livers was as low as about 0.1, and the expression of
MET in hepatocytes exhibited a decrease in livers after LT and livers of non-EAD patients in comparison to the counterparts before LT and of EAD patients, respectively (Fig. S10(b) in Appendix A). These results indicated the regenerative capacity of hepatocytes were significantly impaired by the ischemia and ischemia-reperfusion injuries, especially in EAD-patients.
The preproproteins encoded by
GZMB and
GZMK are secreted by NK cells and cytotoxic T cells and are subsequently proteolytically processed to generate the active protease, which induces apoptosis in target cells via caspase activation [
27], [
28], [
29]. Despite GZMB and GZMK both being members of the granzyme subfamily executing similar functions, our observations revealed that
GZMB is predominantly expressed in NK cells, while T cells accounted for the majority of
GZMK abundance. Notwithstanding, we identified a distinct population of NK cells that simultaneously expresses high levels of
GZMB and
GZMK, both of which demonstrated a significant positive correlation with the cell activation along the pseudotime axis. This finding implies heightened cytotoxicity and defense response for
GZMB + GZMK + NK cells in comparison to
GZMB + NK cells. Furthermore, MAIT cells, a distinct subset of T lymphocytes that sit at a bridge between innate and adaptive immunity, enriched considerable levels of
GZMB and
GZMK, alluding to the highly activated characteristics of MAIT cells as reported in prior findings. These observations underscore the pivotal role of MAIT cells and
GZMB + GZMK + NK cells underlying the T/NK branch remodeling in sensing and selecting inflammatory cue.
S100A12 expression displayed markedly increased heterogeneity among neutrophils analyzed in our study, in comparison to
S100A8 and
S100A9, culminating in the emergence of a discrete subgroup of neutrophils that we henceforth referred to as
S100A12 + neutrophils. The ubiquitous recognition of
S100A12 being predominantly expressed and secreted by activated neutrophils, concomitant with a significant upregulation at inflammation loci, aligns with our observations of augmented activity and degranulation of
S100A12 + neutrophils. The canonical S100A8 and S100A9 proteins facilitate the characterization of neutrophils, given their pervasive expression across all neutrophil populations. Nevertheless, our findings demonstrate that
S100A8 and
S100A9 levels in
S100A12 + neutrophils surpass those in alternative neutrophil subgroups by a considerable margin. These outcomes accounted for the preponderant presence of
S100A12 + neutrophils in each tissue specimen, as the regulatory capacity of S100 proteins in various cellular processes such as cell cycle progression and differentiation is well-documented in the literature [
30].
While the cellular composition of T/NK cells and neutrophils revealed a marginal disparity before and after LT, this variation stems from the diminutive time interval between sample collection, which occurred during the cold preservation phase and, subsequently, after a two hours portal perfusion. Regardless of the inconspicuous differences in T/NK and neutrophil niches prior to and following LT, a profound dissimilarity between patients experiencing EAD and those who did not was observed within the T/NK and neutrophil subpopulations. Particularly with regards to MAIT cells, GZMB + GZMK + NK cells, and S100A12 + neutrophils, the investigations revealed a marked augmentation in EAD patients compared to their non-EAD counterparts in samples garnered before and after LT. Although the prevalence of such populations within MAIT cells, GZMB + GZMK + NK cells, and S100A12 + neutrophils in EAD patients is not ubiquitously greater than that observed in non-EAD patients—as exemplified by MAIT cells among patient 7, GZMB + GZMK + NK cells among patient 5, and S100A12 + neutrophils among patient 2—the overarching proportions nonetheless consistently exhibit a heightened abundance among EAD patients. These findings suggest the intricate and synergistic interplay among these immune populations, wherein they collectively construct a pathogenic immune niche conducive to the onset of EAD. Furthermore, the intimate cell-cell communication network, mediated through a plethora of LR interactions, underscores the close-knit relationship among MAIT cells, GZMB + GZMK + NK cells, and S100A12 + neutrophils. This network effectively perpetuates heightened leukocyte activation, adhesion, proliferation, and cytotoxicity, reinforcing the conclusions drawn from these immunological analyses.
EAD exemplifies a severe manifestation of ischemia-reperfusion injury within the hepatic graft, characterized by AST and ALT levels > 2000 U∙L
−1 within seven days post-surgery, T-Bil ≥ 171 μmol∙L
−1, or INR ≥ 1.6 on the seventh postoperative day [
4]. Serum ALT and AST levels serve as reliable biomarkers for hepatocellular damage, while T-Bil and INR are employed to ascertain graft metabolic function [
51]. Collectively, these indices provide insight into the extent of hepatocellular impairment and synthetic dysfunction, necessitating the establishment of definitive thresholds for each parameter. Remarkably, the pathogenic immune niche encompassing MAIT cells,
GZMB + GZMK + NK cells, and
S100A12 + neutrophils exhibits a similar cumulative effect in precipitating EAD onset, as evidenced by the significant diminution of immune niche scores among non-EAD patients. Consequently, future endeavors should be directed toward delineating the cutoff of the EAD-associated pathogenic immune niche, facilitating the prediction and prevention of EAD development during cold preservation. This could ultimately enhance survival rates for allografts and their recipients, contributing to a more favorable clinical outcome.
Acknowledgments
This work was supported by the National Natural Science Foundation of China (82200725), the Innovation Team and Talents Cultivation Program of National Administration of Traditional Chinese Medicine (ZYYCXTD-D-202002), the Fundamental Research Funds for the Central Universities (226-2023-00114, 226-2022-00226, and 226-2023-00059), the Key Program of National Natural Science Foundation of China (81930016), the Key Research and Development Program of China (2021YFA1100500), the Major Research Plan of the National Natural Science Foundation of China (92159202), and the Ningbo Top Medical and Health Research Program (2022030309). The authors thank Morphological Platform of Zhejiang University School of Medicine and High-Performance Computing Cluster of Zhejiang University Innovation Center of Yangtze River Delta for their technical support.
Data and code availability
The processed scRNA-seq counts data of 12 human liver samples collected from seven donors reported in this paper can be accessed through figshare†. The bulk RNA-seq data of liver samples for 16 non-EAD and 16 EAD patients were collected from GSE23649‡. The scRNA-seq data of 3 human transplanted liver samples are accessible at SRP313633††. The scRNA-seq data of six rat transplanted livers are accessible at CRA004061‡‡. Source codes for the scCrossTalk R package and the related scripts are available at Github†††.
Compliance with ethics guidelines
Xin Shao, Zheng Wang, Kai Wang, Xiaoyan Lu, Ping Zhang, Rongfang Guo, Jie Liao, Penghui Yang, Shusen Zheng, Xiao Xu, and Xiaohui Fan declare that they have no conflict of interest or financial conflicts to disclose.
PDF
(8677KB)
