Introduction
Autoimmune hepatitis (AIH) is a severe and chronic progressive inflammatory liver ailment characterized by an enduring autoimmune response directed against liver autoantigens [
1,
2]. Primary cholestatic hepatitis (PBC) manifests as a chronic, cholestatic autoimmune liver condition characterized by the destruction of biliary epithelial cells and the presence of antimitochondrial antibodies (AMA) [
3,
4]. PBC/AIH variant syndrome (VS) represents a variant form of either AIH or PBC, encompassing AIH features with cholestatic attributes or elevated aminotransferases in the context of PBC, accompanied by autoantibodies and elevated Immunoglobulin G (IgG) levels [
5‐
8]. The manifestation of variant syndromes occurs often sequentially over time with no overlapping features at the initial diagnosis. PBC/AIH VS remains an uncommon disorder, affecting about 5–20% of patients with diagnostic features of either AIH or PBC [
9]. This wide range reflects the lack of standardization of criteria used to define PBC/AIH overlapping disease. From a biological perspective, it remains unclear whether these syndromes should be considered unique autoimmune entities causing damage to bile ducts and hepatocytes or whether they represent presentations of two distinct diseases (PBC and AIH).
Based on the histologic, biochemical, and immunologic features of each parent disorder, the Paris criteria represent the most widely accepted diagnostic method for PBC/AIH VS [
10]. The majority of investigators have opted for a first-line treatment regimen that combines ursodeoxycholic acid (UDCA) and corticosteroids, with or without azathioprine (AZA) [
10‐
17]. Nevertheless, patients with PBC/AIH VS often experience more unfavorable prognoses compared to those with PBC or AIH alone [
18]. A retrospective cohort study in Korea demonstrated that overlap syndrome patients exhibited a lower remission rate to UDCA and steroid combination therapy, coupled with a significantly shorter time-to-progression of liver disease than AIH patients [
19]. Our previous study, encompassing 432 PBC patients, revealed that the biochemical remission rate of PBC patients with AIH features was markedly lower than that of patients with PBC alone, with 88.9% of liver biopsy specimens from non-responders showing interface hepatitis [
20]. In a retrospective study of a substantial PBC/AIH patient cohort, UDCA alone failed to induce a biochemical response in the majority of patients with interface hepatitis [
13]. Despite this, there is a dearth of studies concerning the response in this patient subgroup. Our earlier investigation, which involved 134 PBC-AIH VS patients, indicated a significantly higher incidence of liver-related adverse events in non-responders compared to responders, with high cholesterol levels, reduced histologic bile ducts, and cirrhosis being potential risk factors for poor response [
21]. The development of PBC/AIH VS is a multifaceted, multistep process involving various molecular pathways and factors [
5]. Identifying more detailed regulatory molecules is imperative to better predict drug response in these patients, facilitating early intervention.
The advancement and decreasing costs of molecular measurement technologies have enabled the diverse profiling of disease molecular features, spanning the genome, transcriptome, proteome, metabolome, single-cell sequencing, or cytokinome [
22,
23]. Whole-transcriptome analyses have revealed that a substantial portion of the human genome undergoes transcription, generating non-coding (nc) transcripts, such as long noncoding RNAs (lncRNAs), microRNAs (miRNAs), and circular RNAs (circRNAs) [
24]. MiRNAs execute their biological roles by binding to miRNA response elements (MREs) on target mRNA, inducing gene silencing post-transcription [
25]. LncRNAs have been implicated in modulating various cell phenotypes by influencing miRNA and mRNA expression and stability [
26]. CircRNAs, owing to their resistance to exonucleases and enhanced stability compared to linear RNAs, can continuously accumulate within cells [
27]. A comprehensive understanding of the molecular mechanisms and intricate network of RNA-level interactions could unveil innovative diagnostic and prognostic biomarkers [
28,
29]. However, integrating this knowledge with other "omic" studies may provide a more profound understanding of disease processes [
30,
31]. Metabolic profiling, or metabolomics, provides information-rich insights into metabolic alterations reflective of genetic, epigenetic, and environmental influences on cellular physiology [
32,
33]. Cytokines, contributing to disease development and progression, exhibit deregulated serum levels in various liver diseases, correlating with patient outcomes [
34,
35]. The integration of transcriptomics with metabolomics or cytokinomics holds the potential to offer deeper insights into disease pathogenesis compared to either approach in isolation [
36].
In this study, we utilized whole-transcriptome sequencing to discern driver genes exhibiting notable transcriptional variations among PBC/AIH VS patients classified as good responders and poor responders. This thorough analysis sought to identify differentially expressed genes and non-coding RNAs (ncRNAs) linked to distinct metabolic and cytokine shifts in PBC/AIH VS. The discerned genetic, metabolic, and cytokine modifications were subsequently authenticated in an independent cohort. Additionally, a combined model integrating transcriptomic, metabolic, and cytokine data was formulated to predict an inadequate biochemical response to drug treatment in PBC/AIH VS patients.
Discussion
PBC/AIH VS, a morbid condition within autoimmune liver diseases, poses challenges for rigorous study due to its rarity in the general population. Despite numerous small-scale studies over recent decades, uncertainty persists regarding optimal treatment strategies for this syndrome. While UDCA and corticosteroids, with or without AZA, have been acknowledged as the first-line treatment for PBC/AIH VS, 40%-60% of patients exhibit inadequate biochemical responses, remaining at risk of progressing to advanced disease stages, such as liver fibrosis and cirrhosis [
20,
21]. Second-line therapies, including tacrolimus [
13], mycophenolate mofetil (MMF) [
59,
60], or cyclosporine [
13,
15], are anticipated to enhance the prognosis and survival of patients within this subgroup. However, evaluating the biochemical response necessitates a wait of 6 months or more after treatment. Thus, there is an urgent need for predictive indicators to identify patients more likely to exhibit insufficient biochemical responses, facilitating early access to additional treatment. In this preliminary study, we observed significant alterations in transcript levels associated with metabolic and immune responses in the livers of patients in the PR and GR groups via whole transcriptome analysis. Furthermore, we elucidated specific metabolic pathways, metabolites, and immune cytokines through metabolomic and cytokinomic analyses. Lipid species (PC (18:2/18:2) and PC (16:0/20:3)), cytokines (IL-4), and genes (CACNA1H and ACAA1) were identified as indicative of insufficient biochemical responses in PBC/AIH VS patients. The combined model, incorporating these five indicators, demonstrated proficiency in predicting the risk of insufficient response in PBC/AIH VS.
In our transcriptome analysis, the target genes CACNA1H and ACAA1 exhibited significant expression differences between the PR and GR groups. Functional analysis revealed that CACNA1H was primarily associated with T cell differentiation, activation, and aggregation, while ACAA1 was enriched in fatty acid metabolism. CACNA1H encodes a protein for the α1 subunit of voltage-gated calcium channels, playing a regulatory role in calcium ion entry into cells. Its presence spans all CD4
+ T cell subsets, including Th1, Th2, Th17, and Tregs [
61], modulating T cell expansion and apoptosis through voltage-gated calcium channel function. Notably, CACNA1C and CACNA1G, encoding α1 channel subunits, have been linked to Th2 and Th17 cell function, respectively, through voltage-activated calcium influx [
62,
63]. AIH and PBC are characterized by T cell-mediated autoimmune responses against liver autoantigens with distinct patterns of destruction [
1,
2]. Hence, CACNA1H may contribute to the patient response process by modulating T cell activity, although more in-depth studies are warranted. ACAA1, or acetyl-CoA acyltransferase 1, serves a key role in lipid metabolism, specifically in the beta-oxidation of fatty acids within mitochondria [
64]. Dysregulation of ACAA1 is associated with disturbances in lipid metabolism, implicating its role in various metabolic disorders [
65]. Altered ACAA1 expression leads to hepatic lipid metabolism abnormalities in our PBC/AIH VS patients, influencing their drug response. The ceRNA network constitutes a sophisticated regulatory network encompassing diverse RNA molecules vying for binding to shared miRNAs. Within this intricate interplay, lncRNAs, circRNAs, and mRNAs harbor MREs, enabling them to sequester miRNAs and consequently exert influence on each other's expression levels [
25]. This competitive binding gives rise to a network wherein changes in one RNA type can impact the abundance and function of others within the network [
66]. The ceRNA hypothesis posits that these molecules engage in crosstalk, thereby contributing to the regulation of gene expression and cellular processes. Our findings showcase the involvement of a ceRNA network in the regulation of target genes, encompassing lncRNAs, circRNAs, and miRNAs. However, owing to the limitations of the sample size, further validation of these regulatory mechanisms was unattainable in this study, underscoring the need for future investigations in this direction.
The liver plays a central role in lipid metabolism, encompassing functions ranging from lipid synthesis and storage to lipid breakdown and energy release. Our metabolomic data unveil dysregulated lipid metabolism between the PR and GR groups, suggesting its implication in the response of our patients. Levels of ALP and GGT in our patients were significantly higher in the PR group than in the GR group. ALP and GGT, enzymes primarily localized in the liver, are involved in bile acid metabolism. Dysregulation of bile acid metabolism affects the digestion and absorption of lipids in the intestinal tract, thereby contributing to the differences in lipid molecules between the two groups. These lipid molecules have been reported to participate in the progression of various diseases and are regarded as biomarkers [
67,
68]. Our results identified PC (18:2/18:2) and PC (16:0/20:3) as suggestive predictors for predicting the risk of insufficient response in PBC/AIH VS, indicating their involvement in disease development. Additionally, lipid species have been reported to participate in various cellular activities, providing energy and serving as signaling molecules. Our previous study found that lipid species, including lysophosphatidylcholine (LPC) 16:0, 18:1, 18:2, and 18:3, were associated with monocyte activation in AIH [
69]. However, whether PC (18:2/18:2) and PC (16:0/20:3) can regulate the immune response to participate in disease development requires further investigation.
PBC/AIH VS is characterized by a T cell–mediated autoimmune response against liver autoantigens, yet its immunological foundations are largely unexplored. Our findings revealed a significant Th1/Th2 and Th17/Treg cell immune imbalance in both PR and GR groups. Th1 and Th17 cells, secreting IFN-γ and IL-17, contribute to the inflammatory response, while Th2 cells, secreting IL-4, and Tregs play a critical role in maintaining immune tolerance and preventing autoimmunity by suppressing excessive immune responses. The balance and coordination of these Th cell subsets are crucial for a well-regulated immune system, and their dysregulation can lead to various immunological disorders and diseases [
68]. The majority of VS patients exhibited AIH-like signatures, characterized by a more prominent inflammatory cytokine signature with the highest levels of IFN-γ, TNF-α, IL-9, and IL-17, aligning with AIH data from others [
70,
71]. Additionally, our results identified IL-4 as a significant predictor of biochemical response to treatment in multivariate analysis. IL-4, a multifunctional cytokine produced by various immune cells, exerts its effects by binding to the IL-4 receptor and prompting the differentiation of naive T cells into Th2 cells. IL-4 possesses anti-inflammatory properties by inhibiting the production of pro-inflammatory cytokines and promoting the activity of Tregs, crucial for maintaining immune homeostasis [
72]. Elevated levels of IL-4 and Th2 cells collectively suppress inflammation in patients, potentially facilitating biochemical remission, though further investigation is warranted.
We developed a nomogram to visually represent and validate the predictive capability of the combined model based on the aforementioned indicators. Despite the study's relatively modest sample size, the findings are promising. The multi-omics-based combined model demonstrated excellent predictive efficacy, exhibiting an AUC of 0.986 in the primary cohort and 0.940 in the validation cohort for anticipating complete biochemical response. This suggests that genes, metabolites, and cytokines may offer insights into pathological changes associated with the disease process earlier than conventional laboratory markers. Previous investigations predominantly concentrated on assessing biochemical response in PBC/AIH VS patients through clinical symptoms, pathological staging, and biochemical indicators [
12]. To our knowledge, our study represents the first attempt to utilize multi-omic features for predicting insufficient biochemical response in PBC/AIH VS patients. Our findings hold significant generalizability and application potential. The selection of pretreatment indicators can aid in identifying high-risk patients with a poor response, enabling the implementation of early interventions.
This study is subject to several limitations. Firstly, the retrospective design introduces significant selection bias, potentially hindering the accurate representation of real clinical conditions. Secondly, the limited sample size and challenges in sampling led to most patients being assessed for metabolism and cytokines only prior to treatment, with a lack of gene expression testing. Future investigations should prospectively involve a larger sample size to identify more robust indicators for predicting biochemical responses in patients with PBC/AIH VS. Subsequent large-scale multicenter studies are essential to validate our model. Lastly, the absence of suitable animal models prevented further exploration of the identified indicators' role in disease development. Further validation of the underlying mechanisms is warranted in future studies.
In conclusion, our integrated analysis of whole transcriptomics, metabolomics, and cytokineomics revealed substantial alterations in lipid metabolism and immune responses, particularly in Th cells and their associated factors among patients with PBC/AIH VS. We identified ACAA1 and CACAN1H genes that likely play regulatory roles in these processes. The amalgamation of these features allowed us to construct a predictive model, suggesting an insufficient biochemical response in PBC/AIH patients post-treatment. Moreover, a nomogram incorporating potential risk factors emerges as a valuable tool for clinicians, enabling the early identification of patients with insufficient response and facilitating timely interventions.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.