Advanced iCCA poses a remarkable challenge as there are no curative treatments available, except for early surgical resection1. However, the standard diagnostic approaches for iCCA, which include CT, MRI, PET scans, and pathological examination, have limitations, including issues related to accuracy and potential complications10. In this context, CA19-9 has gained widespread attention as a biomarker for iCCA, but it exhibits only modest sensitivity and specificity7. A meta-analysis of 31 studies, each with varying control groups and cut-off values, revealed that CA19-9 has an overall pooled sensitivity of 72% (ranging from 38 to 100%) and a specificity of 84% (ranging from 31.35 to 100%), with an overall AUC of 83% for diagnosing CCA7. Our findings align well with these ranges, particularly when considering studies with healthy control groups similar to our study. Although CA19-9 has utility, its sensitivity could be improved, especially for early iCCA detection, where high sensitivity is critical for preventing false negatives and improving treatment outcomes. This highlights the continued necessity for the development of more reliable non-invasive diagnostic biomarkers for iCCA screening.Our research has uncovered a promising array of autoantibody biomarker candidates, revealing 19 potential signatures with remarkable AUC values exceeding 95%. On average, these signatures demonstrated a sensitivity of 92.9%, a specificity of 87.3%, and an AUC of 95.9%, surpassing the reported values for CA19-9 alone7. These findings highlighted the potential of these signatures for iCCA diagnosis within our Thai cohort. However, it is worth noting the observed similarity in performance among these signatures, signifying their equal discriminatory power between iCCA and CTR. This likely stems from iCCA’s inherent heterogeneity, making it challenging to pinpoint one diagnostic signature that can fully capture iCCA’s diverse nature. iCCA’s heterogeneity may result from factors like distinct molecular subtypes17 and/or varying disease progression among patients. Consequently, no specific set of biomarkers consistently outperforms others, as what works for one subgroup may not work as effectively for another due to their differences. Nonetheless, our focus is on effectively distinguishing iCCA from non-cancer individuals rather than classifying subtypes or states of the disease, which falls outside the scope of this research.We also observed that out of the 66 differentially expressed autoantibodies we identified, the majority (89.4%) target intracellular antigens. Specifically, 59.1% target cytoplasmic proteins, 30.3% target nuclear proteins, and the remaining 10.6% target secreted and cell membrane proteins. These intracellular antigens, particularly nuclear proteins, are typically shielded from immune surveillance and are not involved in the negative selection process of B-cell maturation37. We speculate that their presence may result from the release of intracellular contents due to apoptotic or necrotic cells in cancer, potentially leading to an augmented immune response characterized by increased production of autoantibodies against them38. The analysis combining ORA with the MCL technique has revealed that our autoantibody-targeted antigens are linked to general cancer pathways shared across diverse cancer types. For example, metabolic reprogramming is a shared feature in many cancers, providing the energy and building blocks essential for cancer cell growth39. Cellular response mechanisms likely involve signaling cascades that contribute to cancer cell survival, proliferation, and evasion of the immune system40. However, the most prominent pathways, as confirmed by GSEA, are related to changes in cell structure (e.g., intermediate filament and cytoskeleton fiber), suggesting potential adaptations in cellular organization linked to cell motility, which is involved in cancer metastasis41. Thus, targeting these pathways for the development of biomarkers and therapies could be a promising approach to enhance diagnostic accuracy and potentially impede disease progression, thereby improving treatment effectiveness and clinical outcomes.Two key structural proteins, KRT8 (CK8/K8) and KRT19 (CK19/K19), though not part of our final biomarker signature, show a significant increase in autoantibodies, hinting at potential clinical relevance for iCCA. As intermediate filament proteins, these two have already been extensively studied and proposed as potential therapeutic or biomarker candidates in various reports42,43,44,45. KRT8 has shown promise as a pan-cancer early biomarker in a comprehensive study involving over 17,000 samples, exhibiting significant overexpression in various cancers46. Meanwhile, KRT19, commonly utilized in immunohistochemical analysis, aids in distinguishing iCCA (K19 +) from hepatocellular carcinoma or HCC (K19-), reflecting its prevalence in iCCA47. Reports have also shown its association with poor prognosis in both iCCA and HCC45,48. Despite their roles in maintaining the structural integrity of epithelial cells49, the increased autoantibody levels against these keratins suggest possible abnormalities in their expression or modification in iCCA cells, which in turn may impact the structural stability of these cells. The continuous phosphorylation of keratin can prompt the reorganization of the keratin network in cells, resulting in the degradation of keratin structure, a characteristic of the Epithelial-Mesenchymal Transition (EMT) that has been associated with the aggressiveness, invasion, and metastasis of tumors49. Consequently, a plausible hypothesis emerges: in iCCA, EMT initiation might be triggered by the hyperphosphorylation of keratins. Thus, during apoptosis in cancer, these hyperphosphorylated keratins could stimulate an amplified immune response, evident in increased autoantibody production against them. With this knowledge, modulating keratin phosphorylation and reorganization represents a potential novel approach to control EMT and metastasis of iCCA. Notably, using kinase inhibitors tailored to keratins like KRT8, KRT19, and KRT15 may reduce their phosphorylation, thereby suppressing EMT and improving clinical outcomes49.Two key candidates from our final autoantibody signature, VIM and NDE1, are also associated with EMT and cell migration processes50,51. Especially, VIM (vimentin), an intermediate filament protein, is known to induce EMT, thereby promoting metastasis50. It emerges as a potential therapeutic or biomarker candidate in various cancer types50. Abnormal VIM expression in iCCA is linked to poor prognosis and decreased overall survival rates, signifying a more aggressive tumor phenotype52. For NDE1, existing research predominantly focuses on its role in microtubule organization, mitosis, cell migration, and neuronal development51. Despite limited attention in the cancer field, NDE1 does hold potential relevance to cancer mechanisms. NDE1 regulates cell division and the cell cycle, interacting with proteins like LIS1 and dynein to control microtubule dynamics and ensure proper mitotic spindle formation51,53. Therefore, we hypothesize that aberrant NDE1 may contribute to mitotic defects, potentially resulting in chromosome aneuploidy54, and genomic instability55—characteristics frequently observed in cancer cells56. Furthermore, since both NDE1 and VIM are associated with cell migration and invasion50,51, their elevated autoantibodies might suggest a potential role in promoting EMT, contributing to the metastatic behavior of iCCA cells. Importantly, the novel status of NDE1 in iCCA, unexplored as a cancer biomarker or therapeutic target, holds transformative potential for future intervention and diagnosis.The final candidate in our selected signature, PYCR1, exhibits increased expression across diverse malignancies and correlates with poor clinical outcomes57. Moreover, it is overexpressed in cancer-associated fibroblasts (CAFs), non-tumor cells implicated in regulating the tumor microenvironment by supporting collagen production for the extracellular matrix (ECM), which fuels cancer cell growth and metastatic dissemination58. It is known that iCCA tissue is recognized for its abundance of desmoplastic stroma, including CAFs, and the demonstrated crosstalk between them plays a pivotal role in tumor growth and development59. In our study, the elevated PYCR1 autoantibody level might suggest potential aberration in PYCR1 expression or modification in iCCA or CAF cells, contributing to tumor growth and metastatic spread. Despite unclear links between PYCR1 and iCCA progression, its role in remodeling the tumor microenvironment proposes a plausible association.Finally, FADD, despite not being included in our final biomarker signature, is still regarded as one of our final candidates and warrants discussion. Prior research has shown that FADD, serving as a ubiquitous adaptor protein, actively participates in and modulates various signaling complexes such as necrosomes, endosomes, and inflammasomes60. Therefore, FADD holds a crucial role in apoptosis, inflammation, innate immunity, and carcinogenesis60. Reports have hinted at FADD’s upregulation linked to tumor progression and poor prognoses in specific cancer cases61,62,63. A pan-cancer analysis indicated that FADD was highly expressed in CCA, showing exceptional diagnostic performance with an AUC of 94%64. The increased presence of FADD autoantibodies in our study suggests a potential abnormality in FADD expression, indicating its possible involvement in iCCA development. We speculate that this aberration could disrupt apoptotic signaling, allowing abnormal cells to evade cell death mechanisms. Additionally, FADD’s role in inflammation might further contribute to iCCA progression by influencing the tumor microenvironment and promoting inflammation-related mechanisms.These promising autoantibodies not only capture the diverse molecular landscape of iCCA but also pave the way for personalized therapeutic strategies. However, while elevated levels of autoantibodies might suggest their involvement, the precise mechanisms linking them to iCCA remain unclear. Autoantibody responses can be complex and may not always directly reflect the functions of the targeted proteins. Additionally, our study arrays exclusively focused on protein antigens, which cannot detect autoantibodies targeting non-protein antigens like CA19-9, a carbohydrate antigen. Although the literature does not provide significant evidence of autoantibodies specifically targeting CA19-9 in iCCA, the immune system could produce autoantibodies against such non-protein molecules. This suggests that our study may not capture the complete autoantibody landscape in iCCA. Future studies should incorporate platforms detecting autoantibodies against a broader range of antigens, including non-protein molecules, to provide a more comprehensive understanding of the autoantibody profile in iCCA.Importantly, our study is also constrained by a small sample size with a specific focus on the Thai population. To address these limitations and validate the clinical applications of our findings, larger studies involving iCCA from diverse populations are essential. Moreover, while our study utilized mRNA expression data to validate candidate autoantigens, we acknowledge that protein expression analysis would provide a more direct assessment of antigen levels in tumor versus non-tumor tissues. Future research should incorporate protein expression studies to validate and extend our findings, thereby providing a more comprehensive understanding of the candidate autoantigens’ roles in disease. These future investigations will elucidate the roles of autoantibodies in iCCA, ensuring their efficacy across a broader spectrum of cases and facilitating a comprehensive assessment of diagnostic sensitivity, specificity, and reliability.In summary, our study emphasizes the significance of autoantibody biomarkers as valuable complements to the existing tools in the diagnosis of iCCA. The combination of an autoantibody signature panel, comprising NDE1, PYCR1, and VIM, along with the conventional biomarker CA19-9, holds the potential to improve diagnostic accuracy for iCCA.
