T-hBMSC maintained characteristics of MSC and exhibited a proinflammatory phenotype

The surface markers of T-hBMSC detected by flow cytometry were positive for CD90, CD73, and CD29 (all > 95%) and negative for CD45, CD34, and CD79a (all < 1%), which were consistent with UT-hBMSC (n = 3 per group) (Fig. S1A, B). The cell counting kit-8 proliferation assay revealed that T-hBMSC exhibited no significant differences compared to UT-hBMSC (p > 0.05) at different time points (24, 48, 72, and 96 h), suggesting that training did not affect the proliferative capacity of hBMSC (Fig. S1C). The tri-lineage differentiation potential of hBMSC after training was also maintained (Fig. S1D). Transcriptome analysis demonstrated that the principal component analysis (PCA) of T-hBMSC(D1) group was distinct from UT-hBMSC(D1) group, and T-hBMSC(D3) group was close to UT-hBMSC(D3) group, indicating that the inflammatory response of T-hBMSC exhibited a recovery trend after the withdrawal of training (n = 4 per group) (Fig. 1B). At each time point, the number of up-regulated and down-regulated DEGs (log2|FoldChange|≥ 2, p < 0.05) and the top 10 genes with absolute FoldChange between T-hBMSC and UT-hBMSC were identified. Results revealed 1041 up-regulated genes and 827 down-regulated genes between T-hBMSC(D1) and UT-hBMSC(D1) groups, with representative immunoregulatory and chemokine genes (e.g., indoleamine 2,3-dioxygenase 1, IDO1; C-X-C motif chemokine ligand 11, CXCL11; C–C motif ligand 8, CCL8; and CXCL9) that were significantly up-regulated (Fig. 1C, D).

The up-regulated KEGG pathways in T-hBMSC(D1) were primarily immune-related, including NOD-like receptor, Janus Kinase-Signal Transducer and Activator of Transcription (JAK-STAT), TNF, NF-κB, and TLR signaling pathways, compared to UT-hBMSC(D1). In contrast, these pathways were absent from the top 20 up-regulated pathways in T-hBMSC(D3) (Fig. 1E). Representative plots of GSEA were shown in Fig. 1F. qRT-PCR analysis further confirmed that the expression levels of pro-inflammatory genes (interleukin (IL)1β and IL8, p < 0.0001; IL6, p < 0.001) and immunoregulatory genes (IDO1 and PDL1, both p < 0.0001) in T-hBMSC were significantly higher than those in UT-hBMSC (n = 6 per group) (Fig. 1G). Western blot also showed that the expression of IDO1 and PDL1 was significantly higher in T-hBMSC (n = 3 per group) (Fig. 1H). The above results showed that T-hBMSC characterized by multipotential differentiation exhibited a proinflammatory phenotype.

Trained immunity enhanced the immunoregulation of hBMSC

To simulate the inflammatory microenvironment, the same doses of the "trainers" were administered to both T-hBMSC(D3) and UT-hBMSC(D3) groups, followed by transcriptome sequencing after 6 h of restimulation. The PCA plot revealed a distinct separation between the T-Restimulation and UT-Restimulation groups (n = 4 per group) (Fig. S2A). Heatmap analysis of selected genes demonstrated that the expression levels of pro-inflammatory genes (e.g., IL1α, IL1β, IL6, IL8, and TNF) were significantly lower in the T-Restimulation group compared to the UT-Restimulation group. Conversely, immunoregulatory genes associated with macrophage polarization (e.g., IDO1; vascular endothelial growth factor A, VEGFA; VEGFC and TGFb2) showed increased expression in the T-Restimulation group, suggesting enhanced anti-inflammatory and immunoregulatory capabilities in T-hBMSC upon restimulation (Fig. 2A). qRT-PCR analysis further confirmed the above findings, showing significantly reduced expression of pro-inflammatory genes (IL1β, IL6, and IL8) and elevated levels of anti-inflammatory genes (IDO1, PDL1, and IL10) in the T-Restimulation group (n = 6 per group) (Fig. 2B). Western blot also showed that the expression of IDO1 and PDL1 was significantly higher in T-Restimulation (n = 3 per group) (Fig. 2C). DEGs analysis identified 633 significantly up-regulated and 611 down-regulated genes (Fig. 1C). Functional enrichment analysis of DEGs highlighted the top 15 up-regulated pathways, predominantly associated with immune disorders (e.g., systemic lupus erythematosus, autoimmune thyroid disease, and allograft rejection) and pathogen infections (e.g., Staphylococcus aureus, Epstein-Barr virus, and human T-cell leukemia virus 1 infection). The PI3K-Akt signaling pathway was significantly down-regulated in the T-Restimulation group, possibly contributing to the suppression of excessive inflammatory responses (Fig. 2D).

Fig. 2

The immunological responses of T-hBMSC and UT-hBMSC to mimic-restimulation. A Representative DEGs between UT-Restimulation and T-Restimulation groups. B qRT-PCR validation of significant pro-inflammatory and immunoregulatory genes (n = 6 per group). C The expression levels of IDO1 and PDL1 proteins in T-Restimulation and UT-Restimulation (n = 3 per group). D Top 15 up-regulated and down-regulated KEGG pathways in T-Restimulation versus UT-Restimulation. E Co-altered KEGG pathways between T-hBMSC and UT-hBMSC at different time points. F Dynamic expression patterns of T-hBMSC in response to mimic restimulation. G, H Time trajectory analysis revealed distinct responses to mimic stimulation in T-hBMSC and UT-hBMSC. **p < 0.01; ***p < 0.001; ****p < 0.0001. PCA, principal component analysis; KEGG, Kyoto Encyclopedia of Genes and Genomes

We performed KEGG enrichment analysis comparing the T-hBMSC and UT-hBMSC groups at various time points and identified several co-regulated pathways. Immune signaling pathways, including JAK-STAT, IL-17, NF-κB, and Retinoic Acid-inducible Gene-I (RIG-I)-like receptor signaling, were significantly enriched in the T-hBMSC(D1) and T-hBMSC(D3) groups, whereas these pathways exhibited greater activation in the UT-hBMSC group upon restimulation. These results indicate that T-hBMSC exhibited a reduced pro-inflammatory response and a superior capacity to modulate the immune microenvironment compared to UT-hBMSC upon restimulation (Fig. 2E). To explore the dynamic expression patterns of whole genomes, the Mfuzz package was used for time-series clustering analysis. Six distinct clusters were identified, and each cluster exhibited unique temporal expression trends. Genes within C4 (n = 2511) exhibited a gradual decline in expression over time, corresponding to the resolution or suppression of innate immune responses, mainly enriched in the regulation of cytokine production and immune response. In contrast, the C5 gene set (n = 1827) displayed a biphasic pattern characterized by an initial sharp increase in expression followed by a subsequent decrease, suggesting transient activation of immunomodulatory genes followed by the establishment of adaptive balance (Fig. 2F).

Time trajectory analysis revealed distinct responses to mimic stimulation in T-hBMSC and UT-hBMSC. T-hBMSC exhibited a rapid response after training, with a tendency toward recovery over time. Notably, T-hBMSC displayed distinct expression profiles, characterized by significantly lower expression levels of pro-inflammatory genes, including IL1α, IL1β, IL23α, IL6, and IL8, compared to UT-hBMSC. However, T-hBMSC maintained a consistently increased expression of immunoregulatory genes throughout the trajectory, with terminal expression levels surpassing those of UT-hBMSC (Fig. 2G, H). These findings suggested that T-hBMSC possessed enhanced immunoregulatory capacity and adaptability upon restimulation.

T-hBMSC responded to the inflammatory microenvironment in vitro

Serum collected from ACLF patients was utilized to mimic the lesion microenvironment. UT-ACLF and T-ACLF samples were collected after 6 h of restimulation to evaluate immune responses (Fig. 3A). qRT-PCR analysis revealed that the expression levels of IDO1 and PDL1 in the T-ACLF group were significantly higher compared to the UT-ACLF group (p < 0.0001). Moreover, the expression levels of immune-associated genes, including IL1β, IL6, IL8, and IL10, were notably elevated in the T-ACLF group, suggesting that T-hBMSC mounted a more robust immune response under liver failure conditions (n = 9 per group) (Fig. 3B). The expression of the two immunoregulatory factors at the protein level also showed the same trend, and the expression of IDO1 was significantly increased (n = 3 per group) (Fig. 3C), suggesting that T-hBMSC possessed enhanced immunomodulatory capabilities.

Fig. 3

The mimic-restimulation on T-hBMSC. A Study design of T-hBMSC stimulated with ACLF serum. B qRT-PCR validation of the expression levels of significant pro-inflammatory and immunoregulatory factors between T-ACLF and UT-ACLF (n = 9 per group). C The expression levels of IDO1 and PDL1 proteins in T-ACLF and UT-ACLF (n = 3 per group). D Schematic diagram of the co-culture system. E The morphology of macrophages co-cultured with UT-hBMSC or T-hBMSC. F Phenotypes of THP1-derived macrophages after 24 h co-culture with T-hBMSC or UT-hBMSC (n = 6 per group). G The response of T-hBMSC and UT-hBMSC after 24 h co-cultured with macrophages. H The effect of IL4Ra inhibitor Dupilumab on macrophages co-cultured with T-hBMSC. Bar = 100 μm. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001; ns, no significance

To assess the impact of T-hBMSC on the polarization of macrophages, THP1 cells were induced to differentiate into M1 macrophages, characterized by elevated expression of CD86, IL6, and IL8 (Fig. S2B). The morphology of THP1-derived M1 macrophages co-cultured with T-hBMSC (T-THP1) showed no significant difference compared to the UT-hBMSC co-culture group (UT-THP1) (Fig. 3E). Interestingly, T-THP1 exhibited a marked shift toward an anti-inflammatory M2-like phenotype, as evidenced by a significant reduction in the expression of M1-like markers (CD86, p < 0.01; IL1β, p < 0.0001; IL6, p < 0.001) and a concomitant increase in M2-like markers (CD206, p < 0.05) (Fig. 3F). T-hBMSC were found to express significantly higher gene levels of IDO1 and IL10 (both p < 0.0001), as well as IL4 and IL13 (both p < 0.01) (Fig. 3G), further promoting the differentiation of macrophages into the M2 phenotype (n = 6 per group). The IL4Ra inhibitor Dupilumab was supplemented into the co-culture system of T-hBMSC and THP1-derived M1 macrophages, but the flowcytometry results showed that the proportion of CD206+ cells increased following inhibition of the IL4/IL13 receptor (Figs. 3H, S2C).

T-hBMSC had the therapeutic potential in liver failure mice

To investigate the therapeutic efficacy of T-hBMSC, we established an FHF mouse model via tail vein injection of ConA (Fig. 4A). The ConA-induced FHF mice exhibited pronounced liver congestion, extensive and significant accumulation of erythrocytes within the small blood vessels and hepatic sinusoids, as evidenced by liver morphology and H&E staining (Fig. S3A, B). The 14-day survival curve indicated that the T-hBMSC and UT-hBMSC-treated groups exhibited significantly enhanced survival rates (p = 0.002; p = 0.003) compared to the NS sham-operated group, although there was no statistical significance between two treatment groups (Fig. 4B). Histological analysis demonstrated that the T-hBMSC-treated group exhibited significantly lower liver necrosis and inflammatory infiltration, as assessed by the Ishak scoring method (D1: 4.7 ± 0.5 vs. 6.7 ± 0.8, p < 0.01; D3: 3.8 ± 0.4 vs. 6.5 ± 1.0, p < 0.01), compared to the UT-hBMSC-treated group. T-hBMSC-treated significantly decreased serum alanine aminotransferase (ALT; 3669 ± 1680 U/L vs. 9370 ± 1619 U/L, p < 0.01) and aspartate aminotransferase (AST; 2586 ± 449.9 U/L vs. 6922 ± 2754 U/L, p < 0.01) levels in mice transplanted for one day, in comparison to UT-hBMSC-treated. Cytokine analysis revealed that serum levels of pro-inflammatory cytokines IL1a and IFN-γ were significantly elevated in NS(D1) sham-operated mice (all p < 0.05). Furthermore, T(D1) mice exhibited a lesser increase in MCP1 compared to UT(D1) mice (165.8 ± 185.0 pg/mL vs. 719.0 ± 1240 pg/mL, p < 0.05) (Fig. 4C, D).

Fig. 4

Transplantation of T-hBMSC to rescue FHF mice. A Schematic diagram of the establishment of FHF mice model and T-hBMSC transplantation to rescue FHF mice. B The 14-day survival curve of NS sham-operated mice, T-hBMSC-treated mice and UT-hBMSC-treated mice (n = 15 mice per group). C, D The appearance, H&E staining, and liver damage score of different treatment groups. E, F The levels of liver function (ALT, AST) and inflammatory cytokine (IL1a, IL6, MCP1, and IFN-γ) in different treatment groups (n = 6 mice per group). G, H Immunohistochemical staining (CD45, CD3, and F4/80) and TUNEL staining of liver tissues in different treatment groups. FHF, fulminant hepatic failure (n = 3 mice per group). Bar = 100 μm. NS, normal saline; ALT, alanine aminotransferase; AST, aspartate aminotransferase

Liver injury in FHF mice exhibited gradual improvement over time. The levels of ALT (58.17 ± 29.01 U/L vs. 142.5 ± 81.13 U/L, p < 0.05) and serum IL6 (75.09 ± 29.10 pg/mL vs. 148.5 ± 50.25 pg/mL, p < 0.05) were also improved on day 3 post-transplantation of T-hBMSC (Fig. 4E, F). Immunohistochemical staining revealed that T(D3) mice had the lowest percentage of CD45+ cells compared to NS(D3) sham-operated mice and UT(D3) mice (T(D3) vs. NS(D3): 1.44 ± 0.36 vs. 5.69 ± 0.90, p < 0.0001; T(D3) vs. UT(D3): 1.44 ± 0.36 vs. 2.89 ± 1.65, p < 0.05). In comparison to the NS(D3) group, the proportion of CD3+ cells in the UT(D3) and T(D3) mice was markedly diminished (both p < 0.0001), while there was no statistically significant distinction observed between the two treatment groups (Figs. 4G, S3F). Nonetheless, F4/80+ cells were more prevalent in the T(D3) group (Fig. 4G). TUNEL staining revealed fewer apoptotic hepatocytes and positive areas in T(D3) animals (Fig. 4H). Similar phenomena were observed in the liver of FHF mice one day post-transplantation (Fig. S3D, E). The aforementioned results indicated that T-hBMSC was superior to UT-hBMSC in reducing intrahepatic inflammation and repairing liver damage.

T-hBMSC restored tissue immuno-microenvironment in FHF mice

The ConA-induced FHF model exhibited features of immune-metabolism dysregulation, characterized by extensive infiltration of inflammatory cells and significant necrosis of liver tissue (Fig. S3A, B). PCA of liver tissues revealed distinct separation among two hBMSC-treated groups, NS group and NC group (Fig. 5A). The results of the differential expression analysis revealed that the T(D3) group had 34/245 (significantly up/down) DEGs in comparison to the UT(D3) group and 473/1867 DEGs in comparison to the NS(D3) group, with the top 20 significant DEGs highlighted (Fig. 5B). Transcriptome analysis revealed that the NS group exhibited activation of inflammatory signaling pathways (e.g., positive regulation of cytokine production, response to IFN) and suppression of hepatic metabolic functions (e.g., catabolic and metabolic processes) compared to the normal control (NC) group (Fig. S3C). Histological analysis confirmed the transcriptome findings, indicating that T-hBMSC-treated more efficiently ameliorated inflammation and restored metabolic balance in liver tissue compared to the UT-hBMSC group.

Fig. 5

Transcriptomic characteristics of FHF mice treated with T-hBMSC. A PCA of liver tissues from NC and three treatment groups on day 3 post-transplantation (n = 5 mice per group). B Volcano plots and the top 10 significantly up-regulated and down-regulated DEGs in comparisons of T(D3) versus UT(D3), T(D3) versus NS(D3), and UT(D3) versus NS(D3). C, D Top 15 KEGG or GO pathways from pairwise comparisons among the three treatment groups. Data were presented as mean ± SD. FHF, fulminant hepatic failure; PCA, principal component analysis; NC, normal control; NS, normal saline; KEGG, Kyoto Encyclopedia of Genes and Genomes; GO, gene ontology

To further investigate the mechanism of T-hBMSC in FHF mice, the GSEA of liver tissues identified the top 15 significantly up-regulated and down-regulated KEGG and GO pathways. The results revealed that inflammatory pathways (e.g., TNF, IL-17, NF-κB, and NOD-like receptor signaling) and immune cell activation processes (e.g., neutrophil chemotaxis and monocyte chemotaxis) were primarily enriched in the UT(D3) group compared to the T(D3) group (Fig. 5C(a-b)). In contrast, the T-hBMSC-treated group exhibited superior performance in biochemical indicators of liver function, reflecting improved metabolic capacity, including drug metabolism, steroid hormone biosynthesis, and linoleic acid metabolism (Fig. 6D(a)). Both the T(D3) group and the UT(D3) group demonstrated decreased inflammatory pathways, such as IL-17 signaling pathway, cytokine-cytokine receptor interaction, Th1 and Th2 cell differentiation, and improved metabolic pathways, such as those related to alcoholism (Fig. 5C, D). Notably, both the T-hBMSC and hBMSC-treated groups showed significant improvements in inflammatory responses and metabolic pathways compared to the NS group on day 1 and day 3 post-transplantation (Fig. S3). Figure 6A displays the co-altered KEGG pathways resulting from pairwise comparisons among the three groups. We developed gene regulatory networks to elucidate the regulatory differences in immunological balance and metabolic homeostasis between the T(D3) group and the UT(D3) group, highlighting critical signaling pathways and their interrelations. The NF-κB and TNF signaling pathways play central regulatory roles in orchestrating the inflammatory response, and the expression of phospho-p65 and TNF-α proteins was lower in T-hBMSC-treated mice compared to UT-hBMSC-treated mice (Fig. 6A-C). The metabolic network included pathways such as drug metabolism, steroid hormone biosynthesis, and linoleic acid metabolism, interconnected through various CYP450 and UGT family genes (e.g., Cyp3a11, Cyp2c55, and Ugt1a9), suggesting their importance in the regulation of hepatic metabolism (Fig. 6D, E). These results revealed that T-hBMSC rescued FHF mice by restoring hepatic immune balance and metabolic homeostasis.

Fig. 6

The patterns of inflammatory and metabolic pathways of FHF mice treated with T-hBMSC. A The co-altered KEGG pathways in comparisons of T(D3) versus UT(D3), T(D3) versus NS(D3), and UT(D3) versus NS(D3). B Immune balance was observed in T-hBMSC-treated mice on day 3 post-transplantation. C The expression level of p65, phospho-p65, and TNF-α proteins in UT(D3) and T(D3) mice (n = 3 per group). D Up-regulated KEGG pathways (mainly related to hepatic metabolism) in comparisons of T(D3) versus UT(D3) and T(D3) versus NS(D3). E Liver metabolic homeostasis was observed in T-hBMSC-treated mice compared to UT-hBMSC-treated mice on day 3 post-transplantation. n = 5 mice per group. FHF, fulminant hepatic failure; KEGG, Kyoto Encyclopedia of Genes and Genomes

T-hBMSC induced M2 macrophage polarization to treat FHF mice

The analysis of immune cell fractions revealed that T-hBMSC reduced the overall infiltration score compared to UT-hBMSC-treated group (T(D3) vs. UT(D3): 0.50 ± 1.72 vs. 1.94 ± 0.75, p < 0.01). The T(D3) group showed a lower proportion of T cell, NK cell (p < 0.05) and dendritic cell, and a higher proportion of granulocyte (p < 0.01), monocyte (p < 0.001), and macrophage (p < 0.01) (Figs. 7A, S4D). Flow cytometry further validated the proportions of myeloid cells (CD11b+) in the liver and blood, as well as Kupffer cells and Tregs. The results revealed that the ratios of CD45+CD11b+ cells in the liver and blood were significantly elevated in both the T(D3) and UT(D3) groups compared to NC mice (both p < 0.01). However, the NS(D3) group, which consisted of only two samples, was excluded from statistical analysis. The proportion of CD45+F4/80+ macrophages was significantly elevated in the T(D3) group (p < 0.0001) (Fig. 7B). Consistent with immunological profiling of transcriptome data, the ratios of M2 macrophages (F4/80+CD163+, p < 0.05) and Treg cells (CD25+FOXP3+) were increased in the T(D3) group, while the ratio of F4/80+CD86+macrophages was lower, although not significantly different (Fig. 7C, S4E). Immunohistochemical analysis revealed distinct spatial distribution patterns of YM1+ cells between two treatment groups. In UT-hBMSC-treated mice, YM1+ cells were predominantly localized to the necrotic core with sparse distribution at the injury periphery. Regional YM1+ cell accumulation was observed in the T-hBMSC-treated group, with prominent localization at the border regions of the injured area (Fig. 7C). mIHC staining of liver tissues confirmed similar results, showing fewer CD11b+ and iNOS+ cells and more F4/80+ and CD163+ cells in the T(D3) group (Fig. 7D).

Fig. 7

T-hBMSC induced M2 macrophage polarization to treat FHF mice. A The analysis of immune infiltration of liver tissues in T(D3) and UT(D3) group (n = 5 per group). B Flowcytometry detection of myeloid and Kupffer cells in liver tissues of the NC, NS(D3) (n = 2), UT(D3) and T(D3) groups (n = 3 mice per group). C Immunohistochemical staining of YM1 in UT(D3) and T(D3) groups. D The proportion of CD86+ and CD206+ cells among F4/80+ cells. E Multiplex immunohistochemical staining of F4/80, CD11b, iNOS, and CD163 in the liver of UT(D3) and T(D3) mice. Bar = 100 μm. *p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001. FHF, fulminant hepatic failure