Acute-on-chronic liver failure (ACLF) is a complex clinical syndrome that is featured by the acute deterioration of liver function in a patient with chronic liver disease and is associated with organ failure and a high short-term mortality rate.^1,2 Studies have shown that the progression of hepatitis B virus-(HBV) associated ACLF (HBV-ACLF) is closely associated with the immune pathway.³ Therefore, further exploration of immune dysfunction in ACLF may help improve the survival of ACLF patients.

The exhaustion of T cell is the essential component of immune dysfunction. It occurs in many chronic infections or cancers and shows a progressive reduction of T effector functions, increased expressions of multiple inhibitory receptors (e.g., lymphocyte activation gene 3 protein [LAG-3] and programmed cell death protein 1 [PD-1]), and altered metabolic functions.⁴ T-lymphocyte dysfunction has been found in ACLF patients.⁵ The effector function of CD8⁺ T lymphocytes is significantly reduced in HBV-ACLF patients.^6,7 Dirchwolf et al.⁸ found that the secretion capacity of peripheral blood T cell-associated cytokines was reduced in ACLF patients. Similarly, Li et al.⁹ found that genes involved in adaptive immunity were downregulated during the pathogenesis of HBV-ACLF, suggesting that adaptive immunity may be depleted in ACLF patients. Thus, T lymphocytes exhaustion may exist in patients with HBV-ACLF. However, none of these studies explored the gene expression characteristics of T lymphocytes in patients with HBV-ACLF.

The immune system consists of a complicated array of cells. Identifying the heterogeneity of immune cells contributes to the understanding of the immune system. Single-cell RNA sequencing (scRNA-seq) detects the transcriptome at the cell level, and is more informative than conventional methods in identifying cellular heterogeneity.¹⁰ Using this technique, in the present study, we found cellular heterogeneity among T lymphocytes from the peripheral blood of HBV-ACLF patients, suggesting differences in T lymphocyte differentiation during the pathogenesis of ACLF. It was further found that the number of exhausted T lymphocytes increased in ACLF, suggesting that the exhausted T cells may participate in the immune dysfunction of HBV-ACLF. This study explored peripheral T cells in HBV-ACLF patients using scRNA-seq, and our findings may help identify potentially useful target molecules for the treatment of T lymphocyte immune dysfunction in ACLF patients.

1. Subjects

This study included two groups: an experimental group and a verification group. The experimental group included six HBV-ACLF patients and three sex- and age-matched controls (Table 1). The validation group consisted of 30 HBV-ACLF patients. Two inclusion criteria were as follows: (1) HBV-ACLF was diagnosed according to the criteria of the Chinese Group on the Study of Severe Hepatitis B-ACLF;¹¹ and (2) HBV-DNA or hepatitis B surface antigen was positive. Four exclusion criteria were listed below: (1) with hepatitis caused by alcohol, drug, or other viruses; (2) with hepatocellular carcinoma and/or other tumors; (3) with immune system-related diseases; (4) receiving immunosuppressor drugs or glucocorticoids.

Table 1 . General Characteristics and ACLF Grade of the Subjects.

Characteristic	Age, yr	Sex	Baseline liver disease	TBil, μmol/L	INR	ALT, U/L	AST, U/L	AST/ALT	Alb, g/L	Glb, g/L	Alb/Glb	Cr, μmol/L	MELD score	PaO2/ FiO2, mm Hg	ACLF grade
Case 1	42	M	Cirrhosis	231.9	1.7	1,174	519	0.44	30.7	27.4	1.12	54	17.5	433	1
Case 2	39	M	Cirrhosis	251.2	2.2	101	105	1.04	31.8	32.3	0.98	54	20.7	452	1
Case 3	46	F	Cirrhosis	211.1	1.6	65	106	1.63	32.5	43.6	0.75	79	20.1	428	1
Case 4	52	F	Cirrhosis	406.2	3.0	74	130	1.76	29.5	24.6	1.20	55	26.2	563	2
Case 5	41	M	Cirrhosis	372.8	4.5	26	60	2.37	34.9	21.8	1.60	145	39.7	198	3
Case 6	40	M	Cirrhosis	255.8	3.3	30	23	0.77	45.5	31.5	1.44	79	29.0	477	2
HC 1	45	F	-	8.0	0.9	37	32	0.86	43	25	1.72	59	-	-	-
HC 2	47	M	-	7.9	0.8	27	25	0.93	52	32	1.63	62	-	-	-
HC 3	43	M	-	9.2	1.0	32	28	0.88	46	34	1.35	64	-	-	-

ACLF, acute-on-chronic liver failure; M, male; F, female; HC, healthy control; TBil, total bilirubin (normal range: 5–21 μmol/L); INR, international normalized ratio (normal range: 0.8–1.2); ALT, alanine aminotransferase (normal range: 9–50 U/L); AST, aspartate aminotransferase (normal range: 15–40 U/L); AST/ALT, De Ritis ratio (normal range: 0.8–1.5); Alb, albumin (normal range: 40–55 g/L); Glb, globulin (normal range: 20–40 g/L); Alb/Glb, albumin globulin ratio (normal range: 1.2–2.4); Cr, creatinine (normal range: 57–97 μmol/L); MELD, Model for End-Stage Liver Disease; PaO₂/FiO₂, a ratio of PaO₂ of arterial oxygen to FiO₂ (normal range: 400–500 mm Hg)..

According to the Chinese Group on the Study of Severe Hepatitis B-ACLF criterion,¹¹ the enrolled patients were classified into three categories. In the experimental group, three patients were categorized as grade 1, two patients were categorized as grade 2, and one patient was categorized as grade 3. In the validation group, 13 patients were categorized as grade 1, 12 patients were categorized as grade 2, and five patient was categorized as grade 3.

The study was conducted in accordance with the Declaration of Helsinki and approved by the Ethics Committee of Shanxi Bethune Hospital, Tai-yuan, China (Number: SBOKL_2020_048). Informed consent was obtained from all subjects involved in the study.

2. Experiments

Peripheral blood mononuclear cells (PBMCs) were extracted from healthy controls and HBV-ACLF patients. scRNA-seq was detected using 10X Genomics platform (10X Genomics, Pleasanton, CA, USA). Step 1: Unsupervised clustering of the cells was detected based on the gene expression profiles by the Seurat package and transmitted to the Uniform Manifold Approximation and Projection (UMAP) to analyze the clustering of immune cells and T cells. Cell subsets were detected using marker genes. Step 2: The differences in the expressions of exhausted T-cell subsets were compared between patients and healthy controls. Step 3: The developmental trajectories of T cells are inferred by pseudotime analysis. Step 4: The level of exhausted T cells was validated by flow cytometry, along with the differences in the secretion functions of cytokines (interleukin 2 [IL-2], interferon γ [IFN-γ], and tumor necrosis factor α [TNF-α]) was validated between exhausted T cells and non-exhausted T cells in 30 HBV-ACLF patients and 30 healthy controls.

3. 10X Genomics single-cell transcriptome sequencing of PBMC samples

PBMCs were extracted from 10 mL peripheral blood by density gradient centrifugation. Then the density and viability of PBMCs were detected by single-cell suspensions. PBMCs with cell viability greater than 90% were valid samples. The 10× Genomics library was carried out based on a single-cell automated preparation system (Chromium^TM Controller) and a Chromium^TM Single Cell 3´ Reagent Kit v.2 (10X Genomics). Sequencing was carried out based on the MGISEQ-2000 sequencing platform (BGI, Shenzhen, China).

4. Data analysis by scRNA-seq

Briefly, the following processes included: (1) Raw data quality control and filtering; (2) Data comparison and analysis: RNA reads were in line with the reference genome (refdata-gex-GRCh38-2020-A) by the STAR software; (3) Quantitative analysis of gene expression: each gene in each cell was defined by the Cell Ranger software according to UMI counts. A total of 106, 723 cells and 214, 627 genes were simultaneously detected; (4) Data quality controlling and filtering: cells with <200 genes or >90% of the maximum gene count or with a mitochondrial readout percentage >15% were excluded from the analysis. Thus, a total of 83, 577 cells reached the analytical criteria; (5) Hypervariable features were selected. In the study, the top 2000 hypervariable genes were selected for depth analysis; (6) The cell cluster analysis is as follows: genes were normalized using Seurat R package (version 4.2.0),¹² and genes with high variation were selected for further analysis. Reduce the dimension and use specific resolution parameters for clustering: 0.5 and 0.1 resolution parameters for clustering of all cells and T lymphocytes respectively. Visualize cells using Uniform Manifold Approximation and Projection; (7) Annotating of cell clusters: for each cluster, the marker gene was identified as a gene that is significantly over-expressed (being >0.25 log-fold higher than the mean expression value in the other sub-clusters), and with a determinable expression in >25% of all cells from the corresponding sub-cluster.¹³ Accordingly, the cell clusters were marked as known cell types based on the Cell Marker database;¹⁴ (8) Pseudotime analysis: the Monocle software (version 2.10.1) was used for cell trajectory analysis, which was based on the asynchronous nature of differentiation among different cells. When these cells were arranged on a trajectory describing the complete differentiation process, the pseudotime dynamics of cell differentiation and development could be reflected.¹⁵

5. Detection of the expressions of CD4⁺TIGIT⁺ T cells and CD8⁺LAG-3⁺ T cells using flow cytometry

Antibodies including phycoerythrin-labeled anti-CD4 antibody and anti-CD8 antibody as well as fluorescein isothiocyanate-labeled anti-T cell immunoreceptor with Ig and ITIM domains (TIGIT) antibody and LAG-3 antibody were used. Flow cytometry was performed on the BD FACSCalibur^TM flow cytometer (BD Biosciences, Franklin Lakes, NJ, USA). The results were analyzed by the Flowjo software, version 10.6 (BD Biosciences).

6. Secretion function assay for in vitro cultured CD8⁺LAG-3⁺ T lymphocytes and CD8⁺LAG-3^– T lymphocytes

The prepared PBMCs from HBV-ACLF patients were used. Miltenyi anti-human CD14 microbeads (Bergisch Gladbach, Germany) were used for multiple magnetic positive selection of CD8⁺LAG-3⁺ T lymphocytes. After 20 μL anti-CD8 beads were added, the mixture was incubated at 4℃ for 15 minutes, and then added with 500 μL bead buffer, and rinsed three times. Then the cell suspension was passed through an LS column (Miltenyi Biotec). The CD8⁺ cells were left in the column and then pressed out. After the beads were washed out, 20 μL anti-LAG-3⁺ beads were added, and then the above steps were repeated. The whole operation process was carried out under sterile conditions. The column was washed twice with normal saline to remove the beads. A small number of cells was added with CD8⁺LAG-3⁺ T-FITC-labeled antibody, and the purity of CD8⁺LAG-3⁺ T-cell was measured on a flow cytometer, and a purity of >99% was considered as meeting the experimental requirements.

The sorted CD8⁺LAG-3⁺ T cells and CD8⁺LAG-3– T cells were incubated with phorbol myristate acetate (Enzo, Farmingdale, NY, USA) and ionomycin (Enzo) for 6 hours. The fluorescently labeled antibodies were added and co-incubated for 30 minutes, followed by flow cytometry using the FACSCalibur^TM (BD Biosciences). The results were analyzed using the Flowjo software (version 10.6, BD Biosciences) and presented as mean fluorescence intensity (MFI).

7. Statistical analysis

An unpaired t-test was used to analyze the differences between two groups. The statistical analyses were performed in SPSS version 25.0 (IBM Corp., Armonk, NY, USA). Two-tailed statistical tests were conducted and a p<0.05 was considered statistically significant.

1. Single-cell transcriptional landscape of PBMCs from HBV-ACLF patients and healthy controls

A total of 106,723 cells and 214,627 genes were detected in PBMCs from the nine enrolled samples. After removing approximately 21.7% of cells that might represent empty or low-quality droplets, 83,577 cells, including 23,269 cells from three healthy controls and 60,308 cells from six HBV-ACLF patients, were analyzed (Supplementary Table 1). Nineteen different cell clusters were detected by unsupervised clustering using the Seurat software package (Fig. 1A). Clusters 0, 1, 2, 3, and 11 consisted of CD3D-expressing T cells; clusters 9 and 18 consisted of B cells expressing CD79A, CD79B, and MS4A1; clusters 4, 8, 12, and 13 consisted of KLRF1-expressing natural killer cells; cluster 6 consisted of dendritic cells expressing CD83; and clusters 5, 7, 10, 14, 15, and 16 consisted of monocytes expressing CD68 and CD14 (Supplementary Table 2). These subtypes were expressed in all nine samples, although in different proportions (Fig. 1B). Most clusters consisted of cells from patients with HBV-ACLF and healthy controls.

Figure 1. Nineteen different cell clusters and proportion of these clusters. (A) The clustering result of 83,577 cells from nine donors. Each dot represents a cell and each color represents one cell type. (B) The proportion of each sample in these 19 clusters. Each color represents one sample. (C) Proportions of various cell types from the healthy controls (HCs) and hepatitis B virus-related acute-on-chronic liver failure (HBV-ACLF) patients. Each color represents one cell type. (D) Box plots showing the ratio of clusters in the HBV-ACLF patients (green) and HCs (red). UMAP, Uniform Manifold Approximation and Projection; NK, natural killer.

Furthermore, scRNA-seq showed that PBMCs isolated from healthy controls consisted mainly of T cells (47.73%±9.43%), natural killer cells (22.96%±0.84%), and monocytes (18.69%±5.87%). PBMCs isolated from HBV-ACLF patients consisted mainly of T cells (50.37%±12.20%), natural killer cells (15.96%±7.57%), and monocytes (21.18%±9.68%). There was no statistical difference in the percentage of each cell between the two groups (Fig. 1C and D).

2. T-cell clusters and subtypes

T cells made up the largest proportion of peripheral blood PBMCs in patients with HBV-ACLF. Therefore, to elucidate the unique structures of the entire T-cell population and the potential functional subsets, we performed unsupervised clustering of the 40,318 T cells detected in all nine samples using hierarchical clustering. A total of eight stable clusters were identified (Fig. 2A). CD4⁺ T cells contained three clusters and CD8⁺ T cells contained five clusters, indicating a common immune profile among these patients (Fig. 2B). The representative markers in each cluster are shown in Fig. 2C.

Figure 2. T-cell subtype analysis based on single-cell gene expression. (A) Uniform Manifold Approximation and Projection (UMAP) projection of T cells, with eight major clusters shown in different colors. The functional description of each cluster is determined by the gene expression profile of each cluster. (B) The proportions of the three CD4⁺ T-cell subsets and five CD8⁺ T-cell subsets in each sample. (C) UMAP plots showing the marker genes in each cluster. (a-h) UMAP plots showing four representative markers in each cluster. These cell subsets include exhausted CD4⁺ T cells (exhausted CD4⁺T), effector CD4⁺ T cells, exhausted CD8⁺ T cells (exhausted CD8⁺ T), naïve CD8⁺ T cells, naïve CD8⁺ T cells, effector CD8⁺ T cells, naïve CD4⁺ T cells, and effector CD8⁺ T cells. TIGIT, T cell immunoreceptor with Ig and ITIM domains; LAG-3, lymphocyte activation gene 3 protein.

Based on the expressions of the classical markers, we identified three different CD4⁺ T clusters: clusters 0, 1, and 6. Cluster 0 represented exhausted CD4⁺ T cells, highly expressing the exhaust genes TIGIT, CTLA-4, and PD-1.¹⁶ Cluster 1 represented effector CD4⁺ T cells, with high expressions of effector genes CD40LG, TNFRSF4, and CXCR3.¹⁷ Cluster 6 represented naïve CD4⁺ T cells, with high expressions of genes including TCF7, LEF1, and CCR7.¹⁸ In addition, we identified five CD8⁺ T-cell clusters: clusters 2, 3, 4, 5, and 7. Cluster 2 represented exhausted CD8⁺ T cells, highly expressing the exhaust genes LAG-3, TIGIT, and PD-1.^19,20 Clusters 3 and 4 represented naïve CD8⁺ T cells, with high expressions of genes including CCR7, LEF1, and TCF7. However, cluster 3 had high expression of CCR7, whereas cluster 4 had high expression of TCF7. Both cluster 5 and cluster 7 represented effector CD8⁺ T cells. Cluster 5 was featured by the high expressions of CX3CR1, KLRC3, and IFNG, whereas cluster 7 had high expressions of KLRC3, FCGR3A, and CXCR3,^21-23 which were usually associated with T cells with specific effector functions.

3. Differences in the exhausted T-cell subsets between HBV-ACLF patients and healthy controls

scRNA-seq showed that T cells accounted for 47.73%±9.43% of the total PBMCs in healthy controls and 50.37%±12.20% in HBV-ACLF patients (Fig. 1C). There was no statistical difference between these two groups (p=0.718) (Fig. 1D). However, it was further found that the proportion of cluster 0 (exhausted CD4⁺ T cells) in T cells was significantly upregulated in the HBV-ACLF group (21.66%±2.33%) compared to that in the healthy controls (2.33%±0.71%, p=0.007) (Fig. 3). Similarly, the proportion of cluster 2 (exhausted CD8⁺ T cells) in T cells was also significantly upregulated in the HBV-ACLF group (18.36%±2.33%) compared to that in the healthy controls (0.20%±0.08%, p=0.010) (Fig. 3). Therefore, although there was no significant difference in the percentage of T cells between the healthy controls and HBV-ACLF patients, subset analysis revealed a significant elevated exhausted T cells in HBV-ACLF patients, suggesting that T cells in HBV-ACLF patients were partially differentiated into exhausted T cells with diminished functions.

Figure 3. Proportions of exhausted T cells in the HBV-ACLF patients and HCs. HBV-ACLF, hepatitis B virus-associated acute-on-chronic liver failure; HC, healthy control.

4. Results of pseudotime analysis of T-cell differentiation

The pseudotime analysis of CD8⁺ T cells (Fig. 4A and B) revealed that CD8-CCR7 and CD8-TCF7 (naïve CD8⁺ T cells) developed into CD8-CX3CR1 and CD8-KLRC3 (effector CD8⁺ cells). In addition, the transition became branched at a certain time point, with one branch remaining in the effector state and the other evolving into CD8-LAG-3 (exhausted CD8⁺ T cells), which was highly enriched in the later part of the pseudotime, indicating that the functions of effector T cells were distinctly differentiated. The pseudotime analysis also showed a naïve-effector-exhausted transition state of CD8⁺ T cells.

Figure 4. Analysis of the T-cell transition states. (A, B) Pseudotime analysis of CD8⁺ T cells. (A) The trend of T-cell transition from the beginning to the end, with darker colors representing the beginning stage and lighter ones representing the end stage. (B) Different colors represent different CD8⁺ T-cell clusters. (C, D) Pseudotime analysis of CD4⁺ T cells. (C) The trend of T-cell transition from the beginning to the end, with darker colors representing the beginning stage and lighter ones representing the end stage. (D) Different colors represent different CD4⁺ T-cell clusters.

The pseudotime analysis of CD4⁺ T cells (Fig. 4C and D) showed that CD4-TCF7 (naïve CD4⁺ T cells) and CD4-CD40LG (effector CD4⁺ T cells) were mainly distributed in the main state, while CD4-TIGIT (exhausted CD4⁺ T cells) were highly enriched in the later states. Pseudotime analysis revealed a naïve-effector-exhausted transition state of CD4⁺ T cells.

5. Results of flow cytometry in verifying the difference in expressions of CD4⁺TIGIT⁺ T cells and CD8⁺LAG-3⁺ T cells in HBV-ACLF patients and healthy controls

The level of TIGIT⁺ in CD4⁺ T lymphocytes showed a significant increase in HBV-ACLF patients (11.7±3.6) than in healthy controls (0.7±0.2) (p<0.001). The level of LAG-3 in CD8⁺ T lymphocytes in peripheral blood also showed a significant increase in HBV-ACLF patients (17.5±5.4) than in healthy controls (0.5±0.2) (p<0.001). Thus, the proportion of exhausted T cells increased in the peripheral blood of HBV-ACLF patients (Fig. 5A). Further, with the increase of ACLF grade, the expression level of LAG-3 in CD8⁺ T lymphocytes increased significantly (ACLF-grade 1: 12.15±3.4; ACLF-grade 2: 17.54±4.8; ACLF-grade 3: 22.92±5.2, p<0.001) (Fig. 5B), which suggested that exhaustion of CD8⁺T lymphocytes also became severe.

Figure 5. Differences in expressions of CD4⁺TIGIT⁺ T cells and CD8⁺LAG-3⁺ T cells by flow cytometry. (A) Differences in expressions of CD4⁺TIGIT⁺ T cells and CD8⁺LAG-3⁺ T cells between the HBV-ACLF patients and HCs. (B) The expression level of LAG-3 in CD8+ T lymphocytes in patients with different ACLF grades (ACLF-grade 1: 12.15±3.4; ACLF-grade 2: 17.54±4.8; ACLF-grade 3: 22.92±5.2, p<0.001). HBV-ACLF, hepatitis B virus-associated acute-on-chronic liver failure; HC, healthy control; TIGIT, T cell immunoreceptor with Ig and ITIM domains; LAG-3, lymphocyte activation gene 3 protein.

6. Difference in the secretion function between CD8⁺LAG-3⁺ T lymphocytes and CD8⁺LAG-3^– T lymphocytes

CD8⁺LAG-3⁺ T lymphocytes and CD8⁺LAG-3– T lymphocytes were screened with cell sorting techniques for in vitro culture. The results indicated that the capacity of CD8⁺LAG-3⁺ T lymphocytes (MFI: 1,688.8±142.2) in generating effector cytokines IL-2 was significantly diminished (p<0.001) when compared to CD8⁺LAG-3– T lymphocytes (MFI: 8,980.0±617.3). Furthermore, the capacity of CD8⁺LAG-3⁺ T lymphocytes (MFI: 1,140.8±93.2) in generating effector cytokines IFN-γ was significantly diminished (p<0.001) when compared with CD8⁺LAG-3– T lymphocytes (MFI: 7,678.4±423.6). CD8⁺LAG-3⁺ T lymphocytes (MFI: 1,272.9±71.5) also exhibited a significant decrease in their ability to produce effector cytokine TNF-α than CD8⁺LAG-3– T lymphocytes (MFI: 3,951.8±219.3) (p<0.001). These data suggested an immune exhaust state of CD8⁺LAG-3⁺ T lymphocytes (Fig. 6).

Figure 6. Flow cytometry analysis of the capacity of CD8⁺LAG-3^- T lymphocytes and CD8⁺LAG-3⁺ T lymphocytes in secreting cytokines. (A, B) Differences in the capacity of CD8⁺LAG-3^– T lymphocytes and CD8⁺LAG-3⁺ T lymphocytes in secreting cytokines (IL-2, IFN-γ, and TNF-α) (IL-2 MFI value: 8,980.6±617.3 vs 1,688.8±142.2, p<0.001; TNF-α MFI value: 3,951.8±219.3 vs 1,272.9±71.5, p<0.001; IFN-γ MFI value: 7,678.4±423.6 vs 1,140.8±93.2, p<0.001). IL-2, interleukin 2; IFN-γ, interferon γ; TNF-α, tumor necrosis factor α; MFI, mean fluorescence intensity; LAG-3, lymphocyte activation gene 3 protein.

In the present study, we provided transcriptomic data of peripheral blood T lymphocytes of HBV-ACLF patients by scRNA-seq. It was found that T lymphocytes of HBV-ACLF patients exhibited distinct differentiation trajectories and could be classified into eight clusters with different distributions and characteristics. Among these clusters, the CD4⁺ TIGIT⁺ subset and CD8⁺ LAG-3⁺ subset showed increased expression of exhaustion-associated genes. It was further confirmed by in vitro culture that the CD8⁺ LAG-3⁺ subset in the peripheral blood of ACLF patients had a reduced capacity to secrete cytokines (IL-2, TNF-α, and IFN-γ), suggesting that the exhausted T-cell subset CD8⁺LAG-3⁺ T lymphocytes may be a major component in the development of immune dysfunction in HBV-ACLF. Thus, our findings may provide information on the effective targets for the detection of T-cell exhaust status and the development of potential therapeutic options in HBV-ACLF.

There is heterogeneity in T lymphocytes from HBV-ACLF patients. By single-cell clustering, we identified different clusters of T lymphocyte subsets. These clusters were expressed in all six ACLF patients, indicating that most clusters consisted of cells from multiple patients and there were common immune characteristics among different patients. Notably, three exhaust genes TIGIT, CTLA-4, and PD-1 were highly expressed in cluster 0. In a previous study, all three of these genes were indicative of the exhausted status of CD4⁺ T lymphocytes.¹⁶ Among them, TIGIT had a higher expression level than CTLA-4 and PD-1, and therefore the corresponding T-cell subset was defined as CD4⁺ TIGIT⁺ T-cell subset. It has been shown that the binding of TIGIT to its ligand can transmit inhibitory signals, leading to a decrease in T-cell activation, IL-2 production, and T cell receptor-mediated proliferation.²⁴ Similarly, genes that are highly expressed in cluster 2 include LAG-3, TIGIT, and PD-1. In previous studies, all three of these genes were indicative of the exhausted status of CD8⁺ T lymphocytes.^19,20 LAG-3 has been demonstrated to be associated with CD3, and cross-linking of LAG-3 with CD3 inhibits T-cell proliferation and cytokine production.²⁵ Furthermore, the ratio of CD4⁺TIGIT⁺ T-cell subset to CD8⁺ LAG-3⁺ T cell subset was significantly higher in ACLF patients than in healthy controls, suggesting that T-cell expression is altered and the amount of exhausted T cells increases during the development of ACLF. As shown by pseudotime analysis, the CD4⁺ TIGIT⁺ T cells and CD8⁺ LAG-3⁺ T cells experienced a transformation from effector T cells to exhausted T cells. It was further demonstrated that T lymphocytes became heterogeneous during the pathogenesis of ACLF. Some distinct subsets gradually occur, among which some effector T cells gradually transform into exhausted T cells. Similarly, a gradual transformation of T cells into exhausted T cells under the influence of the external environment has also been found in infiltrating T cells in hepatocellular carcinoma.²³

In this study, the T lymphocytes of HBV-ACLF patients had increased expressions of exhaust genes and decreased capacity to secrete cytokines. The expressions of the known exhaust genes including PD-1 and Tim3 increased in T lymphocytes from ACLF patients, but not significantly. The exhaust genes with the highest expression levels were TIGIT and LAG-3, suggesting that T cells during the pathogenesis of ACLF may exhibit different exhaust profiles from those of tumor cells. TIGIT and LAG-3 are members of the Ig superfamily, both of which play a negative regulatory role in T-cell proliferation and activation.^26,27 It has been found that LAG-3 promotes CD8⁺ T-cell exhaust in a heavily inflammatory environment (e.g., chronic viral infection).¹⁹ Our results suggested that the expression level of LAG-3 in peripheral blood CD8⁺ T lymphocytes in HBV-ACLF patients increased with the progression of ACLF staging and that exhaustion became serious. In addition, the exhausted CD8⁺ T cells lose their effector functions in a progressive and stratified pattern, manifested as the reduced production of IL-2, TNF-α, and IFN-γ.^4,28,29 In our study, the in vitro culture of peripheral blood CD8⁺LAG-3⁺ T cells from ACLF patients also showed decreased levels of IL-2, IFN-γ, and TNF-α, suggesting that the CD8-LAG-3 T-cell subset was functionally in an exhausted status. Thus, we further confirmed that exhausted T cells take a significant role in the pathogenesis of immune dysfunction in ACLF patients.

However, our study had certain limitations. First, bias might exist due to the small sample size. Nevertheless, the identified T-cell subsets were present in each sample, suggesting that these subsets were representative. In addition, we identified the exhausted CD8⁺LAG-3⁺ T lymphocyte subset and demonstrated that their functions were restricted. Second, as a cross-sectional study, our study lacked dynamic detection of T-cell evolution. However, pseudotime analysis revealed the process from initial T cells to effector T cells and finally to exhausted T cells.

In summary, we used scRNA-seq to reveal the heterogeneity of peripheral blood T cells in HBV-ACLF. It was found the number of exhausted T cells, featured by CD4⁺TIGIT⁺ T lymphocyte subset and CD8⁺LAG-3⁺ T lymphocyte subset, was markedly increased during the pathogenesis of ACLF, suggesting that T-cell exhaustion participate in the immune dysfunction of HBV-ACLF patients. Thus, our findings may provide information on the effective targets for the detection of T-cell exhaust status and the development of potential therapeutic strategies in HBV-ACLF.

This research was funded by the National Natural Science Foundation of China (grant number 81700562), the Shanxi Province 136 Revitalization Medical Project, General Surgery Department (grant number: 2021YZ13), and the Shanxi Provincial Guiding Science and Technology Special Project (grant number 2021XM42).

We thank all the participants of this study for their important contributions.

No potential conflict of interest relevant to this article was reported.

Study concept and design: J.Y., X.Z. Data acquisition: H.W., R.Y., J.W. Data analysis and interpretation: Y.J., T.L., J.B. Drafting of the manuscript: J.Y., X.Z. Critical revision of the manuscript for important intellectual content: J.Y., Y.J. Statistical analysis: J.Y. Obtained funding: J.Y. Administrative, technical, or material support; study supervision: J.Y. Approval of final manuscript: all authors.

Supplementary materials can be accessed at https://doi.org/10.5009/gnl220449

The original contributions presented in the study are publicly available. Single-cell RNA-seq data can be found in the National Genomics Data Center (GSA) under accession number HRA002467.