StemCell Therapy

ACLF is a systemic inflammatory disease accompanied by immune dysfunction and disturbances in energy metabolism. ACLF has a high short-term mortality, which increases with the incidence of failing organs. Although many studies regarding ACLF have been performed, its underlying mechanisms remain to be fully explored. Meanwhile, it has been demonstrated that a strong immune response is a key mechanism of ACLF18,19. Therefore, we explored the intrinsic mechanisms of immune cell infiltration in ACLF using an effective informational biology approach.

Herein, we identified macrophage-associated co-expressed gene modules in ACLF for the first time using a combination of WGCNA and CIBERSORT. We identified immune-related key genes and provided new pathways for future studies on effective targets for ACLF treatments. After bioinformatics and qRT-PCR experiments, 10 immune-related hub genes were identified and mir-16-5p and mir-26a-5p were validated. Altogether, these results might provide new strategies for understanding the pathogenesis of ACLF and developing targeted therapeutic molecules.

In the present study, we evaluated potential pathways and biological processes of ACLF using enrichment analyses. GSEA is characterized by the analysis of collections of genes rather than individual genes, which helps to avoid the inability to reproduce individual high-scoring genes due to poor annotation. In the GSE142255 dataset, the GSEA indicated that immune response, inflammatory pathways, and metabolic pathways were mainly involved in ACLF. Then, we found that the downregulated DEGs were mainly engaged in immune response and inflammatory reaction, while upregulated ones regulated biosynthetic and substance metabolism pathways. These results reflected two major biological processes that co-occurred during the progression of ACLF regulated by different genes: imbalance of immune-inflammatory response and energy metabolism. According to the BP analysis, immune cell activation, differentiation, proliferation, and migration were the major biological processes in ACLF, leading to an expanding inflammatory response. Recently, it was reported that excessive activation of the immune response not only causes a systemic inflammatory response, which subsequently mediates immune-related tissue damage, but also leads to high energy demand. Consequently, the immune system competes with peripheral organs for energy, triggering an immune-related energy crisis in the organism and increasing the risk of organ failure18,20,21. Overall, it was suggested that the hyperimmune response and dysfunctional energy metabolism in ACLF are biologically coupled processes, largely influencing ACLF progress.

CIBERSORT is a widely used deconvolution machine algorithm for estimating the composition of immune cells. It shows superior performance in the identification and fine delineation of immune cells when processing highly noisy mixture data8,22. Here, the CIBERSORT results showed that the population of M0 and M1 macrophages was significantly increased in ACLF patients compared to healthy subjects. We also labeled markers on the surface of macrophages by immunofluorescence and validated their increase in M1 macrophages in the liver of ACLF rats. Macrophages can be polarized into M1 or M2 phenotypes. M1 macrophages can release significant influxes of inflammatory factors and induce cytokine storms with pro-inflammatory effects. On the other hand, M2 macrophages secrete tissue repair factors and exhibit anti-inflammatory and reparative properties23,24. Kupffer cells, a type of macrophage that resides in the hepatic sinusoids, mainly perform innate immune and inflammatory responses25. To search for highly related gene modules, WGCNA identifies similar gene clusters and gene modules by hierarchical clustering. WGCNA also supports the analysis of correlations between gene modules and phenotypic traits26,27. To identify gene clusters associated with macrophages, we performed WGCNA and identified gene modules closely related to M1 macrophage polarization, including the coral1 (containing 3631 genes) and darkseagreen4 (containing 307 genes) modules. Based on the WGCNAs gene modules, we screened immune-related DEGs and constructed a PPI network to find ACLF immune-related hub genes.

Ten hub genes were screened using CytoHubba: RSL1D1, RPS5, CCL5, HSPA8, PRKCQ, MMP9, ITGAM, LCK, IL7R, and HP (Table 3). The differential expression of hub genes was further confirmed by qRT-PCR in ACLF rats. Overall, MMP9, ITGAM, and IL7R were highly expressed during ACLF. Furthermore, ACLF has high 28-day mortality that is closely related to the degree of organ failure in patients. Hence, we used the GSE168048 microarray containing gene expression data of ACLF patients who survived or died at 28days for further investigation. We verified that the expression of RPS5, PRKCQ, MMP9, LCK, ITGAM, IL7R, and CCL5 differed between surviving and deceased patients, suggesting that these genes might be closely related to ACLF progression and could be used to predict ACLF survival status at 28days. Notably, downregulated genes were mostly involved in the promotion of immune response, while the upregulated gene, MMP9, was associated with hepatocyte necrosis. These results suggested that the coexistence of immune paralysis and cell necrosis is a potential ACLF mechanism leading to poor prognosis.

Moreover, miRNAs are potential targets in numerous diseases and control various biological processes. As short-chain RNAs with a coding length of only about 22 nucleotides, miRNAs cannot directly be translated into proteins, but rather regulate protein synthesis by disrupting the stability of target mRNAs and inhibiting their translation through complementary pairing28. Studies have explored the relationship between miRNAs and diseases and proposed the use of miRNAs as a biomarker for disease diagnosis and prognosis as well as a small molecule drug target29. Considering the time and cost of experimental studies, we adopted a database approach combined with experimental validation to study miRNAs that were significantly altered in ACLF. The miRNet 2.0 integrates data from 15 prediction databases and provides visual analytics to enable a more comprehensive and convenient evaluation of the interactions between miRNAs, mRNAs, lncRNAs, and transcription factors15. Herein, we used miRNet 2.0 to construct a miRNA-hub genes network to explore potential miRNAs related to ACLF. During the validation, two miRNAs were significantly altered in ACLF rats: mir-16-5p presented increased expression and mir-26a-5p showed decreased expression. M1 macrophages can transfer mir-16-5p to gastric cancer (GC) cells by secreting exosomes and triggering a T-cell immune response to suppress tumor formation by decreasing the expression of PD-L130. It has been demonstrated that mir-26a-5p decreases with ACLF progression and is associated with worsening liver function and increasing liver disease severity31. However, further studies are needed to validate the potential association between miRNA regulatory networks and ACLF.

Predicting potential disease-associated miRNAs is very meaningful and challenging. Thus, researchers have developed several computational methods and models to perform those predictions. These models can be classified into four categories: score functions, complex network algorithms, machine learning, and multiple biological information29. For example, Chen et al.32 proposed an inductive matrix filling model (IMCMDA) for miRNA-disease association prediction. By integrating miRNA and disease similarity information into the matrix-populated objective function, a low-dimensional representation matrix of miRNAs and diseases was obtained, which was finally combined into a miRNA-disease association score matrix. Chen et al.33 improved the HGIMDA model and further provided the MDHGI model. This model first decomposes the miRNA-disease association matrix to remove data noise, then uses the topological information implied to make predictions through heterogeneous graph inference. It combines machine learning with network analysis methods to make effective predictions for new disease-miRNA associations. Further, Chen et al. proposed an Ensemble of Decision Tree-based MiRNA-Disease Association prediction (EDTMDA) model34 based on the construction of multiple decision trees by randomly selecting negative samples, miRNA features, and disease features, and by dimensionality reduction of the features. The mean of the predicted values from these decision trees is used as the miRNA-disease association score. This model incorporates feature dimensionality reduction into integrated learning to remove noise and redundant information in the learning process and reduce the computational complexity of the model with higher prediction accuracy. Moreover, Liu et al.35 proposed a DFELMDA-based deep forest integrated learning approach to infer miRNA-disease correlations. This model trains a random forest by constructing two auto-encoders based on miRNAs and diseases, extracting low-dimensional feature representation, and finally predicting potential miRNA-disease associations through the random forest. This model combines feature and deep forest-integrated learning models to enhance the prediction accuracy. Bioinformatics-based prediction methods are constantly evolving. Nevertheless, different models have almost different predictive performance for the same datasets. Hence, it is not only necessary to collect large-scale experimental data but also consider other algorithms to improve predictive performance for specific diseases.

Besides the methods covered in this study, the multi-field predictive research of bioinformatics offers a unique perspective on the exploration of diagnostic and therapeutic tools for diseases, not only for ACLF. Currently, with the development of genome-wide technologies, there is an increasing need to explore models that detail the exact mechanisms in which genes and proteins interact to form complex living systems. A gene regulatory network (GRN) is a network of interactions between gene molecules. An improved Markov blanket discovery algorithm based on IMBDANET has been proposed and can effectively distinguish between direct and indirect regulatory genes from GN and reduce the false-positive rate in the network inference process36. Additionally, RWRNET is an algorithm of Random Walk with Restart (RWR) modified by restart probability, initial probability vector, and roaming network applied to GRN that continuously maps the global topology of the network and estimates the affinity between nodes in the network through circular iterations until all nodes are traversed37. In contrast, IMBDANET uses a Markov blanket discovery algorithm for network topology analysis and processing, identifying direct and indirect regulatory genes while solving the problem of isolated nodes. On the other hand, RWRNET focuses on global network topology information but it cannot handle isolated nodes. Finally, the integration of different methods can be more beneficial for the prediction of gene regulatory relationships.

Here, we combined WGCNA and CIBERSORT algorithms and employed GSEA, KEGG, and GO enrichment analyses to explore immune-related hub genes and potential biological mechanisms in ACLF. The hub genes and miRNAs involved in ACLF regulation were also further validated. Since there are few studies regarding ACLF mechanisms, adopting bioinformatics analyses provided valid information and guidance for our research. However, our current study also has some limitations. First, we used an animal model rather than samples from humans to validate the ACLF immune-related hub genes, and the results from animal studies should be treated with caution. Furthermore, although these hub genes and miRNAs were altered and might be involved in the development of ACLF, whether these genes can be new therapeutic targets for ACLF still needs to be explored. Therefore, further experiments are required to validate our findings and explore potential ACLF mechanisms.

Original post:
Bioinformatics analyses of potential ACLF biological mechanisms and identification of immune-related hub genes and vital miRNAs | Scientific Reports -...

This entry was posted on August 19, 2022 at 2:08 am and is filed under Immune System. You can follow any responses to this entry through the RSS 2.0 feed. Both comments and pings are currently closed. |