Abstract

   Liver transplantation is the definitive treatment for end-stage liver
   disease, yet T-cell mediated rejection (TCMR) remains a major
   challenge. This study aims to identify key genes associated with TCMR
   and their potential biological processes and mechanisms. The
   [44]GSE145780 dataset was subjected to differential expression
   analysis, weighted gene co-expression network analysis (WGCNA), and
   machine learning algorithms to pinpoint key genes associated with TCMR.
   Gene Set Enrichment Analysis (GSEA), immune infiltration analysis, and
   regulatory networks were constructed to ascertain the biological
   relevance of these genes. Expression validation was performed using
   single-cell RNA-seq (scRNA-seq) data and liver biopsy tissues from
   patients. We identified 5 key genes (ITGB2, FCER1G, IL-18, GBP1, and
   CD53) that are associated with immunological functions, such as
   chemotactic activity, antigen processing, and T cell differentiation.
   GSEA highlighted enrichment in chemokine signaling and antigen
   presentation pathways. A lncRNA-miRNA-mRNA network was delineated, and
   drug target prediction yielded 26 potential drugs. Evaluation of
   expression levels in non-rejection (NR) and TCMR groups exhibited
   significant disparities in T cells and myeloid cells. Tissue analyses
   from patients corroborated the upregulation of GBP1, IL-18, CD53, and
   FCER1G in TCMR cases. Through comprehensive analysis, this research has
   identified 4 genes intimately connected with TCMR following liver
   transplantation, shedding light on the underlying immune activation
   pathways and suggesting putative targets for therapeutic intervention.

Supplementary Information

   The online version contains supplementary material available at
   10.1038/s41598-024-74874-8.

   Keywords: T-cell mediated rejection, Liver transplant rejection,
   Single-cell RNA sequencing, Enrichment analysis, Immune analysis

   Subject terms: Transplant immunology, Allotransplantation,
   Bioinformatics

Introduction

   In 1963, Dr. Starzl pioneered human liver transplantation, marking a
   significant milestone in medical history^[45]1. While advancements in
   surgical techniques and perioperative care were made, early
   postoperative rejection resulted in one-year survival rates of only
   about 30% before the 1980s^[46]2. The introduction of immunosuppressive
   drugs, particularly cyclosporin A, revolutionized the field by
   significantly improving patient outcomes^[47]3. Subsequent drugs like
   tacrolimus and mycophenolate further enhanced survival rates. However,
   15-35% of liver transplant recipients still face T-cell-mediated
   rejection (TCMR) within two years post-transplant, highlighting the
   need for further molecular research to improve early diagnosis and
   treatment^[48]4,[49]5.

   Recent advancements in next-generation sequencing, machine learning,
   and bioinformatics tools have dramatically enhanced our ability to
   analyze complex post-transplant data. These technologies allow for a
   detailed examination of immune cell diversity and gene expression
   changes associated with transplant rejection. Integrating omics
   research with clinical findings provides a deeper understanding of
   liver transplant rejection mechanisms and facilitates the development
   of personalized treatment strategies^[50]6. High-throughput sequencing
   technologies have enabled extensive studies on gene and protein
   expression alterations in both human and animal models, shedding light
   on the immune processes involved in rejection^[51]7–[52]10. Research
   has identified specific transcriptional markers in peripheral blood as
   predictive indicators of rejection. For instance, proteomic analysis
   has highlighted Heme Oxygenase-1 (HO-1) as a potential biomarker for
   predicting acute rejection after liver transplantation^[53]11.
   Additionally, RNA sequencing (RNA-seq) has revealed significant
   increases in CXCL8 in liver tissues of pediatric patients with
   subclinical rejection post-transplantation. Clinical studies have shown
   that serum CXCL8 levels can diagnose subclinical rejection with high
   accuracy^[54]12.

   In conclusion, identifying novel biomarkers for liver transplant
   rejection and elucidating their molecular mechanisms is crucial for
   early diagnosis and effective treatment. This study analyzes liver
   biopsy transcriptome data from post-liver transplant patients using
   publicly available databases. By integrating bulk and single-cell RNA
   sequencing (scRNA-seq) with machine learning algorithms, we identified
   five key genes associated with TCMR. We further explored the pathways
   and biological processes enriched by these genes, providing insights
   into the molecular mechanisms of TCMR and potential therapeutic targets
   in liver transplantation.

Results

Research workflow

   We identified LTR-DEGs by intersecting DEGs between the TCMR and NR
   groups with TCMR-related genes elucidated in the WGCNA. Following that,
   GO and KEGG analyses were conducted to explore the functions and
   pathways of these LTR-DEGs. A PPI network and machine learning
   algorithms were subsequently employed to pinpoint 5 key genes: ITGB2,
   FCER1G, IL-18, GBP1, and CD53. Single-gene GSEA analysis further
   revealed their biological function. Additionally, immune infiltration
   analysis was conducted to assess fluctuations in immune cells during
   TCMR and to examine the associations between these key genes and immune
   cells. Furthermore, we also constructed a lncRNA-miRNA-mRNA network
   pertinent to key genes and predicted targeted drugs. The scRNA-seq data
   validated the differential expression of these key genes in distinct
   cell types. Lastly, we confirmed the expression levels of these genes
   in liver biopsy samples from post-transplantation patients
   (Fig. [55]1A).

Fig. 1.

   [56]Fig. 1
   [57]Open in a new tab

   Identification of differentially expressed genes (DEGs) and key module
   genes. (A) The workflow of the study. (B–C) Volcano plot and heatmap of
   DEGs; Orange represents upregulated genes, gray represents genes with
   no significant difference, and green represents downregulated genes.
   (D) The sample clustering diagram shows an outlier sample; red
   represents T-cell-mediated rejection (TCMR) samples and white
   represents no rejection (NR) samples. (E) Re-cluster after removing
   outlier samples. (F) Analysis of network topology for various
   soft-threshold powers. (G) Clustering dendrogram of DEGs, genes are
   divided into different modules. (H) Heatmap of module-trait
   correlations. Each gene depicts the correlation coefficients and
   p-values. Genes are colored according to correlation intensity: red for
   positive and blue for negative, as per the color legend.

Identification of DEGs and key module genes

   A total of 180 DEGs were obtained between the TCMR and NR groups,
   including 171 upregulated genes and 9 downregulated genes
   (Fig. [58]1B–C, Supplementary Table [59]1). To further identify genes
   associated with TCMR, WGCNA was conducted. Sample clustering results
   showed the presence of an outlier sample ([60]GSM4332835), which was
   removed and the samples were re-clustered (Fig. [61]1D–E). When the
   soft threshold was set to meet a scale-free distribution (Fig. [62]1F),
   a total of 10 modules were obtained through the dynamic tree cutting
   algorithm (Fig. [63]1G). Among them, the MEblue module (cor = 0.79,
   p = 5e-36) showed the highest correlation with TCMR (Fig. 1[64]1H).
   Therefore, this module was considered the key module, and the 5,498
   genes within this module were defined as key module genes for
   subsequent analysis.

Acquisition and functional enrichment of LTR-DEGs

   Based on the intersection of DEGs and key module genes, 119 L-DEGs were
   identified (Fig. [65]2A). Enrichment analysis revealed that these
   LTR-DEGs are involved in 525 GO terms and 38 KEGG pathways. The GO
   analysis showed that the biological processes (BP) predominantly relate
   to cytokine-mediated signaling pathways, granulocyte chemotaxis,
   myeloid cell migration, and antigen processing and presentation. The
   cellular components (CC) are mainly associated with the secretory
   granule membrane, MHC protein complex, and MHC class II protein
   complex. The molecular functions (MF) are chiefly related to chemokine
   activity, chemokine receptor binding, MHC protein complex binding, and
   G protein-coupled receptor binding (Fig. [66]2B, Supplementary
   Fig. [67]1A, Supplementary Table [68]2). KEGG enrichment analysis
   included pathways such as antigen processing and presentation,
   chemokine signaling pathways, and cytokine-cytokine receptor
   interaction (Fig. [69]2C, Supplementary Fig. [70]1B, Supplementary
   Table [71]3). This suggests that these LTR-DEGs are primarily involved
   in activating and regulating immune cells, as well as processing and
   presenting antigens, thereby initiating specific immune responses.

Fig. 2.

   [72]Fig. 2
   [73]Open in a new tab

   Definition and functional analysis of DEGs associated with liver
   transplant rejection reactions (LTR-DEGs). (A) Venn diagram illustrates
   LTR-DEGs by overlapping DEGs and key module genes. (B) Lollipop diagram
   of LTR-DEGs’ Gene Ontology (GO) enrichment analysis. BP: biological
   process. CC: cellular components. MF: molecular functions. (C) Lollipop
   diagram of LTR-DEGs’ Kyoto Encyclopedia of Genes and Genomes (KEGG)
   pathway enrichment analysis, displaying the 20 most significantly
   different pathways^[74]13–[75]15.

Screening and expression analysis of key genes

   A PPI network constructed from the 119 L-DEGs consisted of 114 nodes
   and 735 edges (Fig. [76]3A). Ten candidate genes (CYBB, ITGB2, FCER1G,
   IL-18, CXCL11, GBP1, PLEK, CD53, LAPTM5, and C3AR1) were ultimately
   selected for further analysis from hub gene networks derived through
   four algorithms (Fig. [77]3B, Supplementary Fig. [78]2). LASSO
   regression analysis then selected 5 LASSO feature genes (ITGB2, FCER1G,
   IL-18, GBP1, and CD53) (Fig. [79]3C–D). The SVM model had the lowest
   error rate with six genes (GBP1, FCER1G, ITGB2, CYBB, IL-18, and CD53)
   (Fig. [80]3E). Overlapping LASSO and SVM-RFE feature genes, five key
   genes were identified (ITGB2, FCER1G, IL-18, GBP1, and CD53)
   (Fig. [81]3F). Expression analysis showed significantly higher
   expression of these genes in the TCMR group than in the NR group
   (Supplementary Fig. [82]3). The ROC curves showed that the AUC values
   of the five genes were greater than 0.9, indicating that the five genes
   had good diagnostic performance. It also indicated the good performance
   of both SVM-RFE and LASSO methods (Supplementary Fig. [83]4). These
   genes, as novel biomarkers and therapeutic targets, provide deeper
   insight into TCMR mechanisms and may guide future diagnostic and
   therapeutic strategies.

Fig. 3.

   [84]Fig. 3
   [85]Open in a new tab

   Construction of the Protein-Protein Interaction (PPI) network and key
   gene screening. (A) The nodes indicate proteins, and the letters
   represent gene symbols. (B) A Venn diagram illustrates candidate genes
   by overlapping the hub genes of the 4 algorithms. (C, D) The results of
   least absolute shrinkage and selection operator (LASSO) COX regression
   analysis. The dotted line on the left indicates the position with the
   smallest cross-validation error. At this position (Lambda.min), one
   identifies the corresponding log (Lambda) value on the horizontal axis.
   The upper horizontal axis displays the number of feature genes to find
   the optimal log (Lambda) value, identifying the relevant genes and
   their coefficients, and explaining the proportion of residuals in the
   model. (E) When the gene count is six, the error rate is at its lowest.
   (F) Venn diagram illustrates key genes by overlapping the results of
   two machine algorithms.

Single-gene GSEA analysis

   Single-gene GSEA analysis identified that key genes are mainly enriched
   in antigen processing and presentation, cell cycle, T cell receptor
   pathway, chemokine signaling pathway, NK cell-mediated cytotoxicity,
   Toll-like receptor signaling pathway, etc. (Supplementary Fig. [86]5,
   Supplementary Tables [87]4–[88]8). These findings indicate complex
   functions of key genes in processes like antigen recognition,
   activation of immune cells, and regulation of immune responses.

Immune-related analysis of key genes

   Bar graphs displayed the proportions of 28 immune cells in each sample
   (Fig. [89]4A). Significant differences in 27 immune cells were observed
   between the TCMR and NR groups, with a greater degree of differential
   immune cell infiltration in the TCMR group (Fig. [90]4B). Correlation
   analysis showed that CD53 had the strongest positive correlation with
   Th1 cells. There was a significant positive correlation between key
   genes and all differential immune cells (Fig. [91]4C, Supplementary
   Tables [92]9–[93]10).

Fig. 4.

   [94]Fig. 4
   [95]Open in a new tab

   Immune cell profiling in TCMR and NR groups. (A) Bar graph of immune
   scores for 28 immune cell types between TCMR and NR groups. (B)
   Comparative scoring of 28 immune cell types in two groups of samples.
   (C) Correlation between key genes and immune cells, the x-axis
   represents immune cells, and the y-axis represents biomarkers.
   *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001, ns: p > 0.05.

Drug prediction for key genes and construction of lncRNA-miRNA-mRNA network

   We have identified 26 therapeutic drugs targeting three central genes
   by the DGIdb database (Fig. [96]5A). The network included 19 drugs
   targeting ITGB2 (ERLIZUMAB, LIFITEGRAST, cyclophosphamide, EFALIZUMAB,
   etc.), 7 drugs targeting IL-18 (IBOCTADEKIN, THYROXINE, etc.), and one
   drug targeting FCER1G (ASPIRIN). To identify the lncRNA-miRNA-mRNA
   network of key genes in TCMR, several online databases (mirTarbase and
   starBase) were used. The network comprised 4 mRNAs (ITGB2, FCER1G,
   IL-18, and GBP1), 8 miRNAs (hsa-miR-26b-5p, hsa-miR-1225-3p,
   hsa-miR-335-5p, etc.), and 43 lncRNAs ([97]AC008040.1, HAGLR,
   [98]AL158206.1, FGD5-AS1, [99]AC069281.2, etc.) (Fig. [100]5B,
   Supplementary Tables [101]11–[102]12).

Fig. 5.

   [103]Fig. 5
   [104]Open in a new tab

   Networks of key genes-drug interaction and lncRNA-miRNA-mRNA. (A)
   Drug-target network diagram for key genes, with green rectangles
   representing drugs and red shapes representing key genes. (B)
   lncRNA-miRNA-mRNA network diagram, where red triangles represent mRNAs,
   green circles represent miRNAs, and blue rectangles represent lncRNAs.
   Red lines in the diagram indicate interactions between miRNAs and
   mRNAs, while grey lines indicate interactions between lncRNAs and
   miRNAs.

Expression analysis of key genes in different cell clusters

   After quality control, scRNA-seq data were available for subsequent
   analysis (Supplementary Fig. [105]6A). Cell screening and standard data
   processing yielded 2000 highly variable genes (Fig. [106]6A).
   Subsequently, we further analyzed the top 20 PCs of the inflection
   point through principal component analysis (PCA) (Supplementary
   Fig. [107]6B). To assess the effectiveness of the clustering results,
   cell flow Sankey diagrams at different resolutions were drawn
   (Supplementary Fig. [108]6C). Unsupervised clustering divided these
   cells into 11 cell clusters, visualized using t-SNE (Fig. [109]6B).
   Based on the expression patterns of marker genes in each cell cluster,
   these cell clusters were divided into 8 cell subtypes (T cells, B
   cells, myeloid cells, etc.) (Fig. [110]6C–D). The clustering results
   for the NR and TCMR groups were presented separately (Fig. [111]6E–F),
   with the highest proportion of T cells in both the TCMR and NR groups
   (Fig. [112]6G). The proportions of endothelial cells, hepatocytes, and
   neutrophils were significantly higher in the NR group than in the TCMR
   group. Expression validation results showed key genes were primarily
   expressed in T cells, myeloid cells, and NK cells (Fig. [113]7A–E).
   Significant differences in the expression of key genes between T cells
   and myeloid cells were observed between NR and TCMR groups.

Fig. 6.

   [114]Fig. 6
   [115]Open in a new tab

   Cell clustering analysis of single-cell RNA sequence (scRNA-seq) data.
   (A) Red dots represent high-variability genes, and black dots represent
   invariant genes; the greater the height on the y-axis, the larger the
   variance and difference of the genes. The names of the top 10
   high-variability genes are also displayed. (B) t-distributed stochastic
   neighborhood embedding (t-SNE) plot colored by different cell clusters.
   (C) Bubble chart of classic marker genes for each cell group. (D) t-SNE
   plot of cell clustering annotation results. (E) Cell clustering
   annotation results (NR group). (F) Cell clustering annotation results
   (TCMR group). (G) Proportion of each cell group among all cells.

Fig. 7.

   [116]Fig. 7
   [117]Open in a new tab

   Expression differences of key genes across various cell groups. (A)
   Expression level differences of CD53 across different cell groups. (B)
   Expression level differences of FCER1G across different cell groups.
   (C) Expression level differences of GBP1 across different cell groups.
   (D) Expression level differences of IL-18 across different cell groups.
   (E) Expression level differences of ITGB2 across different cell groups.

Key genes expression validation

   Immunohistochemical staining confirmed the expression levels of five
   key genes in TCMR patients and controls. Results showed FCER1G, IL-18,
   GBP1, and CD53 were upregulated in TCMR liver tissues compared to
   controls, while ITGB2 expression did not differ significantly
   (Fig. [118]8). These findings indicate FCER1G, IL-18, GBP1, and CD53
   have potential as predictive indicators of immune rejection post-liver
   transplantation and may play significant roles in immune activation
   during rejection.

Fig. 8.

   [119]Fig. 8
   [120]Open in a new tab

   Immunohistochemical staining of key genes in liver biopsy tissue.

Discussion

   Liver transplantation is a proven treatment for end-stage liver
   disease. However, TCMR substantially risks graft failure and elevates
   patient mortality, highlighting the imperative for investigating its
   molecular determinants^[121]16. Such inquiries are fundamental to
   enhancing early detection and therapeutic interventions. While the
   original study utilized the dataset to develop a diagnostic system for
   rejection, this study delves deeper into the molecular mechanisms of
   TCMR and identifies potential drug targets, thus providing a valuable
   complement to the original research^[122]17.

   Recent studies have increasingly integrated machine learning algorithms
   with bioinformatics methods to create decision models that aid in
   identifying potential diagnostic or prognostic biomarkers and
   therapeutic targets, as well as elucidating the molecular mechanisms of
   disease pathogenesis. Several studies on kidney transplant TCMR have
   employed machine learning algorithms in the analysis of transcript
   expression data to explore diagnostic biomarkers and potential
   immunotherapeutic targets for TCMR^[123]18,[124]19. However, similar
   research has not been conducted for liver transplant TCMR. This study
   builds on these approaches by combining Single-cell RNA-seq analysis to
   identify genes closely associated with TCMR in liver transplantation
   and further clarifies the differential expression of key genes across
   various cell populations, exploring potential immune activation
   pathways in the context of liver transplant TCMR. In this study, we
   utilized SVM-RFE and LASSO for feature selection, as both methods are
   highly effective in handling high-dimensional datasets and identifying
   key genes in complex biological processes. SVM-RFE combines the
   nonlinear classification capabilities of the SVM model with a recursive
   feature elimination strategy, enabling the automated selection of the
   most performance-contributive feature subset, particularly excelling in
   handling high-dimensional data and nonlinear relationships. LASSO,
   through L1 regularization, achieves sparse feature selection, which not
   only reduces model complexity but also enhances model interpretability,
   making it especially suitable for high-dimensional data scenarios. Both
   methods do not require assumptions about the specific functional form
   between features and the target variable, with LASSO performing
   particularly well in the presence of multicollinearity^[125]20,[126]21.
   Therefore, in the context of optimizing model performance, automating
   feature selection, and processing high-dimensional data, choosing
   SVM-RFE and LASSO as feature selection methods offers distinct
   advantages. We also recognize that future research could further
   explore other feature selection techniques to more comprehensively
   evaluate their performance and optimize our analytical processes.

   In the initial phase of our research, we identified DEGs between the
   TCMR and NR groups from the [127]GSE145780 dataset and combined a WGCNA
   network to select LTR-DEGs. GO and KEGG enrichment analyses connected
   these LTR-DEGs with immune functions such as cytokine activation and
   antigen processing. Following this, we constructed a PPI network and
   used 4 algorithms to further identify the candidate genes. Machine
   learning further validated the reliability of our findings, pinpointing
   five key genes (ITGB2, FCER1G, IL-18, GBP1, and CD53) linked to TCMR.
   Further interrogation using scRNA-seq confirmed that the five key genes
   were upregulated in both T cells and myeloid cells within the TCMR
   group.

   ITGB2/CD18 encodes an integrin beta chain protein, which is
   specifically expressed by leukocytes and can form corresponding β2
   integrin heterodimers with four different known alpha chains,
   participating in cell adhesion and cell surface-mediated signal
   transduction^[128]22. Mac-1 (CD11b/CD18) and LFA-1 (CD11a/CD18), as
   members of the β2 integrin family, play critical roles in mediating
   leukocyte adhesion to target structures and other immune cells^[129]23.
   Research found that their interaction with ICAM-1 facilitates adhesion,
   crucial for lymphocyte regulation and inflammation control during early
   liver transplant rejection^[130]24; this finding is consistent with our
   research results. Additionally, in mucosal biopsies from lung
   transplant recipients, ITGB2 has been identified as a key gene
   associated with lung transplant rejection^[131]25. And in a rabbit
   model of heterotopic heart transplantation, antibody therapy targeting
   ITGB2 has shown significant efficacy in preventing transplant
   rejection^[132]26.

   Leukocyte surface antigen CD53, a member of the tetraspanin family, is
   a transmembrane signaling and pro-inflammatory protein widely expressed
   in various types of immune cells^[133]27,[134]28. Literature reports
   that CD53 can induce homotypic cell adhesion by activating β2 integrin
   LFA-1 in NK, T, and B cells, promoting adhesion and migration of immune
   cells^[135]29,[136]30. Additionally, studies have shown that CD53
   stimulation enhances T cell proliferation. It drives the transition of
   naive T cells to effector/memory phenotypes. These effects have been
   observed in vitro in human T cells and in genetically modified mice.
   Further investigation revealed CD53’s role in regulating migration rate
   and stability of CD45RO on T cell surfaces. This regulation impacts TCR
   signal transduction. The absence of CD53 results in a loss of CD45RO
   expression on T cells. Consequently, it alters the expression of CD45
   isoforms and diminishes T cell activation^[137]31. In a transcriptomic
   analysis of post-kidney transplant recipients, CD53 was identified as a
   biomarker for acute rejection in kidney transplantation^[138]32.

   FCER1G encodes the γ subunit of the high-affinity immunoglobulin E
   (IgE) Fc receptor, widely expressed in various types of immune cells.
   The FCER1G gene is involved in various biological processes, including
   neutrophil activation, T cell differentiation, immunoglobulin-mediated
   immune responses, and Fc receptor-mediated signaling
   pathways^[139]33–[140]35. Similar to this study, FCER1G has been proven
   to be an upregulated gene highly representative of acute rejection in
   heart, liver, and kidney transplants in a transcriptome data analysis
   of various solid organ transplants^[141]36.

   GBP1 is a GTPase of the dynamin superfamily, involved in the regulation
   of membrane, cytoskeleton, and cell cycle progression dynamics^[142]37.
   Furthermore, GBP1 is key to host cell immunity and antibacterial
   protection, recognizing infection and inhibiting bacterial
   proliferation by activating inflammasomes and regulating
   pyroptosis^[143]38. GBP1 is highly expressed in macrophages,
   endothelial cells, and epithelial cells, with continued high expression
   levels in T cells after IFN-γ stimulation induced by interferon-γ
   (IFN-γ)^[144]39. In several studies of renal transplant rejection,
   researchers have confirmed through transcriptomic analysis and tissue
   biopsies that GBP1 may serve as a biological characteristic and
   predictive model for acute rejection^[145]40–[146]42. Recent research
   have found that GBP1 binds to lipopolysaccharide (LPS) during bacterial
   infection, mediating the recruitment and activation of inflammatory
   caspase-4 through cleavage of GSDMD to induce
   pyroptosis^[147]43,[148]44.

   IL-18 is a member of the IL-1 family of pro-inflammatory cytokines,
   originally identified in the cytoplasm of macrophages as an inactive
   precursor (pro-IL-18). This precursor is transformed into an active
   mature form by proteolytic cleavage^[149]45. Recent proteomic studies
   have revealed IL-18 plays a significant biological role in regulating
   innate and adaptive immunity^[150]46. It also effectively induces IFN-γ
   production, exerting various immunoregulatory functions in the presence
   of different cytokines. These functions include regulating the
   transformation of T cells to regulatory T cells, modulating Th1 and Th2
   responses, and participating in Th17 responses^[151]47–[152]49. It has
   been documented in existing literature that IL-18 plays a significant
   role in the rejection reactions of various solid organs. In a rat
   orthotopic liver transplantation model, blocking the binding of IL-18
   to its receptor significantly alleviated graft rejection^[153]50.
   Elevated levels of IL-18 have also been observed in acute rejection
   reactions following kidney and heart
   transplantation^[154]51,[155]52.Historically, the maturation of IL-18
   has been ascribed primarily to Caspase-1-mediated cleavage within the
   NLRP3 inflammasome complex^[156]53,[157]54. However, recent studies
   indicate that in Caspase-4, as part of the non-canonical pyroptosis
   pathway, can recognize and cleave the same site on pro-IL-18, leading
   to its activation. This interaction establishes a Caspase-4–IL-18 axis
   that links non-canonical pyroptosis to adaptive immunity^[158]55. Our
   findings lend support to the hypothesis that both GBP1 and IL-18 levels
   increase substantially in liver tissues following TCMR. This suggests
   that post-transplantation dysbiosis and the aberrant translocation of
   lipopolysaccharide (LPS) into the liver through the gut-liver axis
   might provoke non-canonical pyroptotic pathways in liver macrophages
   and other cells, mediated by GBP1. This cascade potentially triggers
   TCMR via cytokine release subsequent to cell death.

   TCMR is characterized by a complex, multifactorial immune response,
   with extant experimental research underscoring the pivotal role of
   various immune cells in this process^[159]56. In our study, immune
   infiltration analysis of two groups of data revealed a significant
   increase in the majority of immune cells in TCMR tissues, with
   scRNA-seq clustering analysis further showing the most significant
   upregulation in T and myeloid cells. Additionally, correlation analysis
   of the five selected key genes with 28 types of immune cells showed
   strong correlations with T cells, B cells, myeloid cells, etc.
   scRNA-seq revealed that the expression of the five key genes in the
   TCMR group was significantly increased in T and myeloid cell clusters
   compared to the NR group. Integrating bulk RNA-seq and scRNA-seq
   analyses, these key genes appear to be pivotal in immune activation
   during TCMR, significantly influencing antigen presentation and
   lymphocyte activation between Antigen-Presenting Cell (APC) and T
   cells.

   To deepen the understanding of the specific mechanisms of TCMR, we
   explored differences in miRNA and lncRNA levels based on mRNA
   differences and constructed an entire ceRNA network, which may be
   involved in important biological pathways related to TCMR. Currently,
   some studies have found that these miRNAs affect the progression of
   certain immune-related diseases by regulating corresponding mRNAs. This
   holds important guiding significance for transplant rejection, which
   also involves aberrant activation and dysfunction of immune cells.
   Research has found that miR-346 expression is present in synovial cells
   activated by LPS induction of pattern recognition receptors (PRRs) in
   rheumatoid arthritis patients. This miRNA negatively regulates the
   IL-18 response in fibroblast-like synoviocytes by inhibiting the
   transcription of Bruton’s tyrosine kinase. This mechanism has been
   validated in THP-1 cells, demonstrating miR-346’s inhibitory effect on
   IL-18^[160]57. Researchers have found through dual-luciferase reporter
   analysis that miR-130a negatively regulates IL-18, thereby
   participating in the regulation of primary immune
   thrombocytopenia^[161]58. The miR-124-3p, which regulates GBP1, acts as
   a suppressor in various cancers and also plays a role in cell apoptosis
   and inflammatory lesions^[162]59. This approach potentially offers a
   new perspective for researching TCMR mechanisms, discovering novel
   therapeutic targets, and developing biomarkers. Moreover, drug
   prediction for the five key genes identified 26 therapeutic drugs
   corresponding to three key genes, with tacrolimus, mycophenolate
   mofetil, methylprednisolone, and cyclosporine already widely used in
   the treatment of liver transplant rejection^[163]56,[164]60,[165]61.
   Previous research has confirmed that colchicine inhibits T cell
   proliferation by blocking the expression of the IL-2R gene, damaging
   the antigen recognition process^[166]62. Lifitegrast, an integrin
   antagonist, has been proven in dry eye disease research to block the
   antigen transfer process from APC to T cells^[167]63. These targeted
   drugs may provide new insights for the prevention and treatment of TCMR
   after liver transplantation.

   Finally, immunohistochemistry was used to validate the upregulated
   expression of key genes in human liver biopsy tissues. The significant
   upregulation of four key genes: GBP1, IL-18, CD53, FCER1G, in the TCMR
   group further supports their potential as predictive indicators and
   therapeutic targets for liver transplant TCMR. This finding also
   provides a theoretical basis for further exploration of their molecular
   mechanisms. However, we acknowledge the limitations of our study,
   primarily due to the scarcity of post-transplant liver biopsy samples,
   leading to insufficient validation. Our data were derived from a public
   database, which may not capture all possible TCMR case types or patient
   subgroups, potentially limiting the generalizability of our
   conclusions. Additionally, there may be unknown systematic biases or
   technical variations in the data collection and processing that could
   affect the accuracy of the results. To address this, we incorporated
   additional datasets for single-cell analysis and immunohistochemistry
   to enhance the study’s comprehensiveness and validation, but these
   datasets may also share similar limitations. Furthermore, the specific
   functions and potential mechanisms of key genes involved in liver
   transplant rejection require further investigation. Predictions related
   to miRNA, lncRNA, and targeted drugs were also based on public
   databases and have yet to be experimentally validated. In future
   research, we plan to explore more comprehensive, diverse, and
   high-quality datasets while implementing corresponding foundational
   experiments to gain deeper insights into gene expression changes and
   their biological significance in the TCMR process.

Conclusions

   To sum up, this work provides insights into the molecular mechanisms of
   TCMR through a series of bioinformatics methods such as differential
   expression analysis, WGCNA, enrichment analysis, immune infiltration
   analysis, GSEA enrichment analysis, and single-cell analysis.
   Additionally, drug analysis targeting key genes helps provide more
   references for the treatment of post-transplant rejection.