Biliary atresia (BA) is one of the most common and fatal abnormalities of newborns. Most BA patients cannot survive with their native liver because of severe cholestatic cirrhosis, unless they receive Kasai portoenterostomy to reconstruct the bile drainage (1). However, more than half of the patients cannot survive with their native liver for a long term after receiving Kasai portoenterostomy, which contributes to 75% of pediatric liver transplantation under 2 years old in the world (2).

Although the etiology of BA is still not clear, some theories were related to genetics (3), immunology (4), and toxicology (5). Increasing evidence indicated that immunologic derangement was the core issue that caused BA. The animal model of BA was based on the immune reaction of rhesus rotavirus (RRV) infection in newborn Balb/c mice (6). The abnormality of B cells, T cells, and macrophages in BA patients was also revealed by single-cell RNA-seq in a small population recently, which provided strong evidence that the immune system took part in the occurrence and development of BA to a great extent (7). Therefore, research with a larger sample size were necessary to further prove these findings.

In the past, microarrays and RNA-seq data could only reflect the mRNA expressions of bulk liver tissues composed of hepatocytes, cholangiocytes, endothelial cells, and immune cells, making it impossible to demonstrate the potential functions of each kind of cell. With the development of bioinformatics, powerful deconvolution algorithms now enable us to calculate the immune microenvironment precisely by using the bulk mRNA expression data. Here, we used CIBERSORTx, a tool verified by fluorescence-activated cell sorting, to reveal the relationship between the immune microenvironment of BA liver and the progression of the disease by utilizing the present mRNA expressions data sets of BA (8). Since Rohr-Udilova et al. (9) first used it in exploring the immune cell landscape between healthy liver and hepatocellular carcinoma, CIBERSORT was widely applied in various liver diseases (10–12).

Biliary atresia is a fatal neonatal cholestatic liver disease with unknown etiology. However, sights were increasingly focused on the immune response to figure this puzzle out. Recently, Wang et al. (7) and Taylor et al. (29) used single-cell RNA sequencing to demonstrate the abnormality of immune cells in BA patients, which further proved the role of immune response in BA pathogenesis. Here, we utilized the public gene expression data set and deconvolution algorithm of BA liver to explore its immune-related genes and immune microenvironment. In this study, eight immune-related DEGs and six subtypes of immune cells were identified as critical factors in the progression of the disease.

Most of the immune-related DEGs that we found in this study were chemokines, which may greatly contribute to the pathogenesis of the BA. CXCL8, a chemotactic factor that guides neutrophils to the site of infection, was identified as the hub gene of BA in this study. The expression of CXCL8 was high in liver and peripheral blood in BA patients. In peripheral blood, it mainly originated from monocytes, neutrophils, T cells, and B cells (30). Meanwhile, it was primarily expressed in the hepatic parenchyma in liver (31). CXCL8 in BA liver was predominantly in cholangiocytes within areas of ductular reaction, which may play an essential role in BA progress by mediating the ductular reaction and liver fibrogenesis (19). It was believed that both expressions of CXCL8 in the liver and peripheral blood were a predominant part in BA progress and can be used as prognosis predictors of BA patients (32).

SLIT2 is an immune-related gene that regulates cell growth and migration (24). It can recruit immune cells, such as macrophages (33, 34). However, the mechanism may be different in the liver. Former studies found that liver fibrosis was mediated by activating hepatic stellate cells through the Slit2/Robo1 and Slit2/Robo2 signal pathway (24, 25). It has been proved that the hepatic expression of SLIT2 was significantly increased in patients with primary biliary cirrhosis and in bile duct ligation mouse model (35). CXCL6 is a chemokine which can recruit neutrophil, leading to tissue damage and prolonged inflammatory responses (36). Recently, CXCL6 was found to be upregulated in the serum and liver tissue of high-stage liver fibrosis patients (20) and was supposed to participate in fibrogenesis by stimulating Kupffer cells releasing TGF-β and thereby activating stellate cells. This study finds that the expression of SLIT2 and CXCL6 in BA liver is exceptionally high, and it is strongly related to liver fibrosis grade.

The microenvironment of BA was identified by CIBERSORTx. Natural killer cells, monocytes, macrophages, Treg cells, and plasma cells were significantly different in BA and normal livers, as reported in previous research. Moreover, a more specific fraction of immune cells was computed, and correlation analysis was applied to figure out the relationship between microenvironment and liver status of BA patients.

Natural killer cells were found to have adverse effects in BA livers, especially in experimental BA (37–39). Natural killer cells can be activated by dendritic cells in the RRV BA model, thus injuring the biliary trees (40). The deletion of NK cells further proved that the activation of NK cells plays an essential role in bile duct injury (38). Alexandra et al. (41) supposed that CXCL9 and CXCL10 were secreted by cholangiocytes after a virus infection. Natural killer cells were activated via NKG2D ligands expressed by infected cholangiocytes and migrated to the liver and biliary tree along the chemokine gradient of CXCL9 and CXCL10, causing an autoimmune response against biliary epithelium (41). In our research, although the proportion of NK cells activated is similar between normal liver and BA liver, the proportion of NK cells resting is significantly decreased in BA liver. This phenomenon may be caused by NK cells exhausting after immune response in BA liver. The correlation analysis of this research also revealed that NK cells resting was negatively correlated with chronic fibrosis of BA liver in humans, as reported formerly (42), which may give a clue in retarding the fibrosis procedure in BA patients. The inflammation grade of BA liver in our study reflects an acute inflammation status of BA liver. The positive correlation between inflammation grade and NK cell resting may reflect the dysfunction of NK cells in killing activated T cells and other liver-resident cells and promote the inflammation response (43).

The proportion of Treg cells was severely reduced according to our research. Treg cells play an essential role in regulating the immune microenvironment and maintaining immune homeostasis (44). In the murine BA model established by RRV, liver Treg cells decreased in both number and function (45), contributing to the progress of BA. It was proved that the proportion of Treg cells in peripheral blood was reduced in BA patients who were positive for cytomegalovirus (46). Although we found that the proportion of Treg cells was significantly reduced in BA liver, the correlation analysis did not detect the relationship between Treg cells and liver fibrosis grade or liver inflammation grade. On the other hand, the correlation analysis revealed that Treg cells had the widest connection with other immune cells. Both the strongest positive and negative correlation coefficients were related to the Treg cells, reflecting its powerful regulation role in BA liver.

The mononuclear phagocyte system, including macrophages and monocytes, was suppressed in BA liver. Although the total amount of macrophage was significantly high in BA liver, the subtypes of functional macrophages, such as M1 and M2 macrophage, did not show any statistical difference. Previous studies did not have a consensus on the effect of macrophage in BA progression (47–49). The proportion of M0 macrophage in BA liver was remarkably higher than the one in normal liver, indicating that macrophages were inactivated. Besides this, the proportion of monocytes was significantly decreased in BA liver, further demonstrating that the mononuclear phagocyte system was dysfunctional. As Wang et al. (7) have reported, the macrophages in BA liver were under a hypo-inflammation situation. Meanwhile, the Kupffer cell scavenger function was defective, which may explain why the proportion of M0 macrophages was extremely increased in BA liver.

In this research, we applied CIBERSORTx in the exploration of the microenvironment and correlation between immune cell infiltration and BA status. Meanwhile, several limitations inevitably existed in our exploration. Firstly, this result was calculated by the algorithm, which was an estimation of the actual situation. Despite the fact that LM22, the reference signature of immune cells, was widely used in CIBERSORT-related analysis, including liver diseases, no study has validated the reliability of LM22 in liver research, which may misrepresent the immune cell types present and active in BA or normal livers. Besides this, the present research cohort is limited; a prospective was required to validate the results. In any case, the correlation between immune cell infiltration and BA status not only deserves further research but also provides a potential prediction tool.

In conclusion, we depicted the microenvironment of BA liver and the correlation between the microenvironment and patient status. The expression of CXCL6, CXCL8, CXCL10, CCL20, IL32, TGFB2, SPP1, and SLIT2 was significantly different between BA and normal liver. The abnormal accumulation of six types of immune cells played an essential role in BA, such as the high proportions of M0 macrophages and resting memory CD4+ T cells and the low proportions of monocytes, resting NK cells, plasma cells, and Treg cells. Our results might provide a clue for further exploration of BA etiology and promote a potential prediction model based on immune infiltration features.

The original contributions presented in the study are included in the article/supplementary material, further inquiries can be directed to the corresponding author/s.

QX, YL, and JZ designed this research. JZ and MF collected data and wrote the manuscript. JZ and YL analyzed data. All authors read and approved the final manuscript.

This work was supported by grants from the National Natural Science Foundation of China (grant nos.: 81972205 and 92059205) and Shenkang three-year program (grant nos.: SHDC2020CR5012 and SHDC2020CR2003A).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.