Introduction
Liver cancer is the sixth most common malignant neoplasm and the third leading cause of cancer-related fatalities worldwide, with an estimated 865,000 new cases and 758,000 deaths in 2022, as reported by the International Agency for Research on Cancer.1,2 Hepatocellular carcinoma (HCC) accounts for 75%–85% of all primary liver cancers.1,2 Primary risk factors include chronic infection with hepatitis B virus (HBV), hepatitis C virus (HCV), nonalcoholic fatty liver disease (NAFLD), and excessive alcohol consumption.3–5 Despite substantial progress in imaging, chemotherapy, interventional radiology, and surgical approaches, HCC patients continue to face high recurrence rates and poor overall survival. Early-stage HCC typically presents with few or no symptoms, leading to missed opportunities for effective treatment; the overall five-year survival rate for advanced HCC remains below 20%.6,7 In contrast, early detection can improve the five-year survival rate substantially, from 60% to 70%.8
Current clinical guidelines recommend imaging techniques—such as computed tomography (CT), magnetic resonance imaging (MRI), and ultrasound screening—along with serum alpha-fetoprotein (AFP) testing for early HCC diagnosis.9 However, these methods have notable limitations in diagnostic accuracy and sensitivity. A meta-analysis indicated that the sensitivity of ultrasound for early-stage HCC screening was below 50%, increasing to only 63% when combined with AFP.10 CT and MRI, the primary detection modalities for HCC with approximately 84% sensitivity,10 demonstrate relatively low accuracy in diagnosing early-stage HCC or tumors smaller than 2 cm with atypical imaging features.11 Histological biopsy, the gold standard for HCC diagnosis, is unsuitable for broad screening owing to its invasive nature. These limitations collectively underscore an urgent clinical need for novel, minimally invasive diagnostic approaches. In this context, extracellular vesicles (EVs) have emerged as highly promising candidates for the early detection of HCC.
EVs are membrane-derived nanoparticles that encapsulate proteins, nucleic acids, lipids, metabolites, and even organelles; they are broadly categorized as exosomes (typically <200 nm in diameter) and ectosomes/microvesicles (>200 nm in diameter), reflecting their distinct biogenetic origins in the endosomal system and plasma membrane, respectively.12 EVs feature a lipid bilayer membrane and contain various bioactive molecules, including proteins, messenger RNAs (mRNAs), microRNAs (miRNAs), circular RNAs (circRNAs), and long non-coding RNAs (lncRNAs).13 Given the lack of consensus on subtype-specific biomarkers, this review adopts the term ‘EVs’ as recommended by the International Society for Extracellular Vesicles (ISEV). Most studies referring to ‘exosomes’ or ‘ectosomes/microvesicles’ in the existing literature describe heterogeneous EV populations rather than vesicles defined by specific biogenesis pathways.12 EVs play a crucial role in HCC development and progression by facilitating intercellular transfer of diverse molecular cargoes. Their unique lipid-raft structure protects EV contents from degradation by ribonucleases and proteases in the circulation. Studies have demonstrated significantly elevated EV levels in the blood of patients with HCC compared with healthy individuals,14 making EVs a promising component of liquid biopsies for noninvasive early HCC screening (Fig. 1). This review aims to provide a comprehensive overview of current EV-derived molecular cargoes as potential biomarkers for early HCC diagnosis,15 while also addressing existing technical challenges in EV isolation and characterization and critical gaps in the literature that must be resolved to realize the full clinical application potential of EV-based diagnostics.
The diagnostic performance of EVs in HCC
EV-derived miRNAs
miRNAs are a class of small noncoding RNAs, typically 18–25 nucleotides in length, that regulate gene expression post-transcriptionally by targeting the 3′-untranslated regions of mRNA transcripts, leading to translational repression or mRNA degradation. This regulatory capacity confers critical roles in cell growth, differentiation, and metabolism.22 EV-derived miRNAs have demonstrated considerable promise as biomarkers for the early detection of HCC. A meta-analysis encompassing 5,125 HCC patients and 6,561 controls reported pooled sensitivity and specificity of 0.80 and 0.73, respectively, for miRNAs in differentiating HCC from non-HCC cases.23 Importantly, miRNAs encapsulated within EVs are significantly more stable and abundant than free miRNAs in serum or plasma,24 supporting EV-derived miRNAs as a novel approach for early HCC liquid biopsy screening (Table 1).25,26–38
Table 1The diagnostic efficacy of EV-derived miRNA for HCC patients
| Author | Country | Year | Sample size | Categories | Diagnostic efficacy | Reference controls |
|---|
| Wang et al.26 | China | 2020 | 50 NC
50 CLD
50 HCC | miR-21
miR-96
miR-122
miRNA panels | 0.912
0.802
0.853
0.996 | U6 |
| Ghosh et al.27 | India | 2020 | 35 NC
98 CLD
38 HCC | miR-10b
miR-221
miR-21
miR-223
miRNA panels
albumin | 0.65
0.69
0.78
0.63
0.80
0.75 | miR-39 |
| CHO et al.28 | Korea | 2020 | 28 NC
60 CLD
90 HCC | miR-10b
miR-18a
miR-215
miR-940
AFP | 0.968
0.842
0.936
0.827
0.707 | miR-1228 |
| Zhang et al.29 | China | 2019 | 70 NC
153 CLD
78 HCC | miR-212
AFP
CA125 | 0.793
0.84
0.777 | U6 |
| Asmaa et al.30 | Egypt | 2018 | 18 NC
42 CLD
60 HCC | miR-1262
RNA panels | 0.92
0.883 | U6 |
| Xue et al.31 | China | 2018 | 23 NC
85 HCC | miR-93 | 0.825 | miR-16 |
| Hao et al.32 | China | 2020 | 50 NC
55 CLD
104 HCC | miR-320a | 0.86 | miR-39 |
| Wang et al.33 | China | 2018 | 50 NC
80 CLD
50 HCC | miR-122
miR-148a
miR-1246
AFP
Combination | 0.816
0.891
0.785
0.712
0.931 | cel-miR-39 |
| Xue et al.34 | China | 2019 | 30 NC
80 HCC | miR-122
miR-125b
miR-145
miR-192
miR-194
miR-29a
miR-17
miR-106a | 0.746
0.65
0.535
0.752
0.738
0.703
0.85
0.704 | miR-16 |
| Yang et al.35 | China | 2022 | 50 NC
50 CLD
50 HCC | miR-26a
miR-29c
miR-199a
Panels | 0.890
0.917
0.850
0.994 | U6 |
| Cui et al.36 | China | 2019 | 50 NC
89 HCC | miR-224 | 0.91 | let-7 |
| Fründt et al.37 | Germany | 2021 | 20 NC
51 CLD
86 HCC | miR-146a
miR-16
miR-221
miR-192
Panels | 0.8
0.63
0.69
0.59
0.78 | cel-miR-39
miR-484 |
| Cho et al.38 | Korea | 2020 | 22 NC
50 CLD
72 HCC | AFP/miR-4661
AFP
miR-4661/miR-4746 | 0.921
0.704
0.942 | U6 |
miR-21 is a liver-enriched miRNA. It plays a core role in the proliferation, invasion, and apoptosis of HCC cells by targeting multiple genes, including SMAD7,39 kruppel-like factor 6 (KLF6),40 and suppressor of cytokine signaling 6 (SOCS6).41 Based on a meta-analysis of 1,589 subjects from 14 publications, Zhang et al.42 reported that the pooled sensitivity, specificity, and area under the curve (AUC) of miR-21 were 0.83, 0.80, and 0.88, respectively, suggesting that miR-21 is a valuable biomarker for early HCC diagnosis. Nevertheless, several publications have indicated that miR-21 is mainly enriched in EVs, and its expression level is significantly higher than that in EV-depleted serum or plasma.43 Wang et al. assessed the potential of EV-derived miR-21 as a diagnostic biomarker, and the results showed a powerful ability to diagnose early HCC with an AUC value of 0.912, which was significantly higher than those of AFP and serum-derived miR-21.26 The sensitivity, specificity, and AUC of EV-derived miR-21 for the early diagnosis of HBV- or HCV-induced HCC in India were 0.74, 0.77, and 0.77, respectively.27 Interestingly, many researchers have established new, more rapid, and sensitive detection methods for EV-derived miR-21 compared to quantitative reverse transcription PCR (RT-qPCR), thereby accelerating its clinical application.44 It is worth noting that serum EV-derived miR-10b-5p had the best diagnostic performance (AUC of 0.968)28 for differentiating HCC from healthy controls (HCs) among all reported EV-derived miRNAs, which was remarkably higher than that of miR-18a-5p (AUC = 0.842), miR-212 (AUC = 0.79), miR-1262 (AUC = 0.92), miR-93 (AUC = 0.825), miR-320a (AUC = 0.86), and AFP (AUC = 0.84).27,29–32 However, Ghosh et al. reported that single miR-10b-5p may be difficult to demonstrate good sensitivity in distinguishing HCC from non-HCC.27 Therefore, further large-scale prospective studies are required to assess their role in HCC. miR-122 is considered a key regulator of innate immunity, EMT, apoptosis, and proliferation of HCC and has powerful potential in the early diagnosis of HCC.45 Recently, the diagnostic potential of EV-derived miR-122 was evaluated for early HCC screening in several different studies,26,33,34 and the highest AUC value of 0.9 (sensitivity, 100%; specificity, 92%) was obtained. Other studies further indicated that miR-122 in EVs showed better accuracy than that in serum; thus, EV-derived miR-122 could be used as a new diagnostic biomarker for early HCC.
Chronic liver disease (CLD) is the primary cause of HCC, particularly due to conditions like liver cirrhosis (LC), hepatitis from HBV or HCV, alcoholic hepatitis, and NAFLD. The effectiveness of EV-derived miRNAs as diagnostic tools can differ among various types of CLD. Yang et al.35 conducted a diagnostic evaluation of miR-26a, miR-29c, and miR-199a using 50 HCC patients, 50 LC patients, and 50 healthy subjects and found that plasma EV-derived miR-29c had the highest diagnostic value with an AUC of 0.917 for identifying early HCC, while miR-26a and miR-199a had AUC values of 0.890 and 0.850, respectively. Moreover, the EV-derived miRNA panel significantly differentiated patients with HCC from other CLD or HCs. Previous research also indicated that EV-derived miR-320a was significantly reduced in HCC, with sensitivity, specificity, and AUC values of 76.1%, 81.8%, and 0.829, respectively, for detecting HCC in patients with alcoholic hepatitis.32 EV-derived miRNAs may serve as more effective biomarkers than AFP for monitoring malignant lesions in patients with LC. For example, EV-derived miR-224 levels were linked to larger tumors (>3 cm), achieving an AUC of 0.910 for early HCC diagnosis.36 Conversely, low levels of EV-derived miR-638 in HCC indicate more advanced stages (III/IV).46 In Wang’s study, the AUCs of serum EV-derived miR-122, miR-148a, and miR-1246 in detecting early HCC from LC were 0.816, 0.891, and 0.785, respectively, which were significantly higher than that of AFP (0.712).33 This advantage was further supported by the validation group results (AUCs of 0.795, 0.86, and 0.861 compared to 0.665 for AFP). Additionally, Ghosh et al. found that EV-derived miR-21 was a more effective biomarker for distinguishing HCC from chronic hepatitis than HCs, with specificity and AUC values of 83% versus 68% and 0.83 versus 0.71, respectively.27 The expression of EV-derived miR-215-5p increases with liver disease progression and is associated with tumor stage advancement and vascular invasion. Further studies indicated that serum EV miR-215-5p could be a promising biomarker for early HCC detection in healthy individuals, with an AUC of 0.936, thus outperforming AFP levels. However, its diagnostic performance (AUC of 0.7) in distinguishing early HCC at stages I or II from non-tumor patients (including those with LC and chronic hepatitis) was less impressive.28
Combining multiple markers can appreciably improve the specificity and accuracy of early HCC screening compared with a single indicator. For instance, the AUC values of EV-derived miR-21, miR-96, and miR-122 in distinguishing early HCC from HC were 0.912, 0.802, and 0.853, respectively, which were better than those of AFP and serum miRNAs. Their combination exhibited a better diagnostic performance with a sensitivity, specificity, and AUC of 96%, 98%, and 0.996, respectively. This study represents a preliminary investigation with a limited sample size, and the relevant conclusions need to be further verified by subsequent large-sample studies.26 Further research found that the diagnostic AUC value of the EV-derived miRNA panel for tracking the progression of LC to HCC increased to 0.924.26 In another study, EV-derived miR-221 showed low specificity (63%) and AUC (0.75) in HCC screening, whereas the combination of four miRNAs (miR-21, miR-10b, miR-221, and miR-223) led to a remarkable improvement, with 86% specificity and 0.86 diagnostic accuracy in differentiating early HCC from chronic hepatitis.27 Fründt et al. reported that the specificity of plasma EV-derived miR-16, miR-221, and miR-96 for early diagnosis was 7.6%, 27%, and 10%, respectively, whereas the diagnostic efficiency of miRNA panels improved to 81% for sensitivity, 86% for specificity, and 0.78 for AUC.37 Current evidence indicates that AFP alone is no longer recommended for the early diagnosis of HCC, owing to its limited sensitivity and specificity. Cho et al. reported that AFP (cut-off value 20 ng/mL) showed a low diagnostic value in detecting HCC in LC and HC patients, with a sensitivity of 15.9%, specificity of 71.4%, and AUC of 0.541, whereas combined AFP and serum EV-derived miR-4661-5p represented a promising diagnostic strategy with a sensitivity, specificity, and AUC of 95.4%, 72.4%, and 0.921, respectively. This study is among the first to report on the potential clinical significance of serum EV-derived miR-4661-5p in HCC.38 Further analysis demonstrated the superiority of combined AFP and EV-derived miR-4661-5p (AUC = 0.91) over AFP alone (AUC = 0.604) for detecting early HCC. In another study, the diagnostic performance of HCC in early-stage CLD patients was significantly improved using a combination of serum EV-derived miR-122, miR-148a, and AFP, with a sensitivity of 87%, specificity of 90%, and AUC of 0.947.33 Thus, EV-derived miRNA panels combined with AFP may be a more valuable assay for increasing the diagnostic performance of HCC.
High-throughput sequencing has identified numerous EV-derived miRNAs that are significantly differentially expressed between HCC patients and HCs, providing a rich repertoire of candidate biomarkers for HCC.47 Multi-marker EV-derived miRNA panels, when combined with traditional biomarkers such as AFP, substantially enhance the efficiency and accuracy of early-stage HCC diagnosis. Nevertheless, several important gaps remain. The expression profiles of EV-derived miRNAs in HCC have shown inter-study inconsistencies and variable diagnostic performance across different ethnic groups and HCC etiologies; most existing studies have focused on Asian populations with HBV-related HCC, highlighting the need for larger, diverse prospective cohorts. Additionally, EV-derived miRNA expression correlates with Tumor–Node–Metastasis (TNM) stage and tumor size,38 and differs between virus-related and CLD-induced HCC.43 Systematic identification of the most discriminating miRNA categories or panels for different HCC etiologies remains an unmet priority for the field.
EV-derived lncRNAs
lncRNAs are a class of noncoding RNA transcripts exceeding 200 nucleotides in length, once regarded as transcriptional ‘noise’ without coding potential.48 Advances in molecular biology have since revealed their indispensable roles in maintaining cellular homeostasis. lncRNAs participate in diverse biological processes, including transcriptional and post-transcriptional regulation (e.g., mRNA degradation, translational inhibition, competing endogenous RNA (ceRNA) networks, and m6A methylation).49–51 By acting as ceRNAs, lncRNAs function as miRNA sponges and thereby modulate downstream gene expression.52 Recent evidence has demonstrated that EV-mediated transport of lncRNAs facilitates intercellular communication between tumor cells, promoting HCC proliferation, immune evasion, and drug resistance.53 The dysregulation of EV-derived lncRNAs during HCC development and progression correlates with tumor stage and size, establishing EV-derived lncRNAs as promising novel biomarkers for early HCC screening (Table 2).36,54,55,56–62
Table 2The diagnostic efficacy of EV-derived lncRNAs for HCC patients
| Author | Country | Year | Sample size | Categories | Diagnostic efficacy | Reference controls |
|---|
| Huang et al.54 | China | 2020 | 52 NC
43 CLD
112 HCC | lnc85 | 0.873 | GAPDH |
| Yao et al.55 | China | 2020 | 45 NC
90 CLD
45 HCC | lnc-EPC1-4
lnc-FAM72D-3
lnc-ZEB2-19
lnc-GPR89B-15 | 0.521
0.629
0.772
0.508 | GAPDH |
| Wang et al.56 | China | 2021 | 100 NC
166 HCC | lnc-CRNDE | 0.839 | cel-miR-39 |
| Tan et al.57 | China | 2019 | 14 NC/CLD
31 HCC | lnc-RN7SL1S | 0.93 | MIR1228 |
| Kim et al.58 | Korea | 2020 | 29 NC
63 CLD
90 HCC | LINC00853
AFP | 0.965
0.604 | HMBS |
| Sun et al.59 | China | 2018 | 76 NC
76 HCC | LINC00161 | 0.794 | cel-miR-39 |
| Abd El et al.30 | Egypt | 2018 | 18 NC
42 CLD
60 HCC | lnc-RP11-513I15.6
miR-1262
RAB11A
Combination | 0.964
0.868
0.754
0.877 | Beta-actin
U6 |
| Yao et al.60 | China | 2022 | 148 NC
101 HCC | H19-204
THEMIS2-211
PRKACA-202
AFP
PRKACA-202+THEMIS2-211
H19-204+THEMIS2-211
THEMIS2-211+PRKACA-202+ AFP | 0.67
0.816
0.812
0.824
0.826
0.825
0.947 | U6 |
| Xu et al.61 | China | 2018 | 120 NC
241 CLD
115 HCC | ENSG00000258332.1
LINC00635
AFP
Combination | 0.719
0.750
0.666
0.894 | GAPDH |
| Lu et al.62 | China | 2019 | 200 NC
200 CLD
200 HCC | ENSG00000248932.1
ENST00000440688.1
ENST00000457302.2
AFP
Panels
Combination | 0.530
0.632
0.883
0.400
0.818
0.833 | Cel-miR-39 |
Based on RNA sequencing, 8,572 plasma EV-derived lncRNAs were found to be differentially expressed between patients with HCC and HCs.54 EV-derived lnc85, which showed the most significant expression, was further validated in HCC cell EVs, and its knockdown significantly affected the proliferation and migration of HCC cells by sponging miR-324-5p. Moreover, the AUC, sensitivity, and specificity of lnc85 for distinguishing HCC patients from HCs and LC patients were 0.873, 80.0%, and 74.5% (cut-off = 1.563), respectively.54 Therefore, EV-derived lnc85 may be a promising biomarker for HCC diagnosis. Another study identified over 40,000 lncRNAs dysregulated in HCC and hepatitis patients compared to the control group, providing numerous candidate targets for HCC detection.55 Additionally, the expression of serum EV-derived lncRNAs FAM72D-3 and EPC1-4 was found to be closely linked to the proliferation and apoptosis of HCC cells, showing potential diagnostic efficacy for HCC, with AUC values of 0.584 and 0.852, respectively.55 The lncRNA CRNDE, recognized as a significant oncogene in cancer, is notably elevated in HCC tissues and can enhance the proliferation, migration, and invasion of HCC cells through various pathways involving the miR-217/MAPK1,63 miR-33a-5p/CDK6,64 and miR-136-5p/ILF2 axis.65 A recent study indicated that high levels of serum EV-derived lncRNA CRNDE were associated with larger tumor sizes and more advanced stages of HCC, and diagnostic assessments revealed an AUC of 0.839 with 69.3% sensitivity and 85.0% specificity for EV-derived lncRNA CRNDE in early HCC screening.56 Furthermore, lncRNA RN7SL1S was found to be highly expressed in both HCC serum and EVs, where it inhibited p53 translation by targeting and regulating TP53 expression.57,66 Tan et al. reported that the diagnostic performance of the lncRNA RN7SL1S fragment in plasma (AUC = 0.87) was significantly higher than that of AFP (AUC = 0.712) and EV-derived RN7SL1S (AUC = 0.75), indicating that plasma-derived lncRNA RN7SL1S may be much more suitable for the screening of early stages of HCC.57 Currently, serum EV-derived LINC00853 has the best diagnostic value among reported EV-derived lncRNAs, with an AUC of 0.934.58 More importantly, EV-derived LINC00853 was found to have a sensitivity of 93.75% and specificity of 89.77% for the diagnosis of early-stage HCC (stage I), which was much better than that of AFP (9.38% sensitivity and 72.73% specificity). In particular, EV-derived LINC00853 showed excellent discriminatory ability in AFP-negative and AFP-positive early HCC, yielding positivity rates of 97% and 67%, respectively.58 Thus, EV-derived LINC00853 detection is expected to be a new breakthrough in the early noninvasive screening of HCC. At present, the specific function of EV-derived LINC00853 in the progression of HCC remains unclear and deserves further investigation. LINC00161 has been identified as a promising and independent biomarker for HCC diagnosis and prognosis, and its increased expression promotes tumor migration and invasion in HCC, leading to poor survival.67 Sun et al. found that serum LINC00161 expression was closely related to AFP concentration and TNM stage in HCC, and its AUC, sensitivity, and specificity were 0.794, 75%, and 73.2%, respectively, in early HCC detection.59 Further analysis showed that EV-derived LINC00161 expression in patients with HCC was remarkably higher than that in HCs68; however, no statistical significance was detected between serum EV-free samples and control serum samples, demonstrating that LINC00161 is mainly present in EVs.59 EV-carried LINC00161 can promote HCC tumorigenesis and metastasis by inhibiting the miR-590-3p/ROCK2 signaling pathway.68
Serum EV-derived lncRNA RP11-513I15.6 was significantly increased in the malignant group compared to the chronic HCV and HC groups, while EV-derived miR-1262 was downregulated.30 Further studies showed that lncRNA RP11-513I15.6 could reduce the repression of RAB11A mRNA by competing with miR-1262, and that the sensitivity and specificity of lncRNA RP11-513I15.6 were 96.7% and 95%, respectively, in detecting HCC in chronic HCV patients and HC; however, when combined with EV-derived miR-1262 and AFP, the specificity dropped to 75%. Similarly, in early-stage HCC screening, lncRNA RP11-513I15.6 alone had a higher diagnostic accuracy (AUC = 0.964) than their combination (AUC = 0.877).30 Therefore, the early diagnostic efficiency of EV-derived lncRNA RP11-513I15.6, combined with AFP, in HCC is still worth evaluating. Conversely, EV-derived lncRNA panels have been reported as a better choice for HCC diagnosis. Plasmatic EV-derived H19-204, THEMIS2-211, and PRKACA-202 were increased in patients with HCC, and their AUC values in HCC screening were 0.67, 0.816, and 0.812, respectively, which were inferior to that of AFP (AUC = 0.824).60 In contrast, PRKACA-202 or H19-204 + THEMIS2-211 dramatically elevated the diagnostic efficiency (0.826 and 0.825); the combination of the above three markers had a better AUC value (0.887), and THEMIS2-211 + PRKACA-202 + AFP showed the highest diagnostic performance (AUC = 0.947) in discriminating HCC patients from normal controls. Crucially, Yao et al. found that H19-204, THEMIS2-211, and PRKACA-202 in plasma EVs were highly stable because of the strong protective effect of the bilayer membranes through treatment with RNase or low-temperature storage.60 Hence, EV-derived H19-204, THEMIS2-211, and PRKACA-202 have been viewed as novel biomarkers with high stability for early-stage HCC diagnosis. Xu et al. reported that EV-derived ENSG00000258332.1 and LINC00635 expression was significantly upregulated in serum samples of patients with HCC, and their high levels were strongly associated with portal vein tumor emboli, lymph node metastasis, and TNM stage of HCC patients.61 The AUC of ENSG00000258332.1 (cut-off = 1.345), LINC00635 (cut-off = 1.690), and AFP (cut-off = 20 ng/mL) were 0.719, 0.75, and 0.666, respectively, in distinguishing HCC patients from CLD controls, whereas their combination substantially improved the diagnostic efficiency with an AUC of 0.894, sensitivity of 83.6%, and specificity of 87.7%.61 In addition, a recent study reported that upregulated plasma EV-derived ENSG00000248932.1, ENST00000440688.1, and ENST00000457302.2 played a crucial role in the development and progression of HCC, and the diagnostic AUC of three lncRNAs alone, AFP, their combination, and lncRNA panel combined with AFP were 0.794, 0.571, 0.538, 0.510, 0.838, and 0.905.62 Accumulating evidence indicates that EV-derived lncRNAs can serve as key regulators of tumorigenesis and metastasis in HCC.69 More importantly, EV-related lncRNAs are sufficiently stable and suffer minimal effects during repeated freezing and thawing.60,62 Thus, EV-derived lncRNA panels combined with AFP represent a powerful diagnostic strategy for HCC screening.
EV-derived circRNAs
circRNAs are a distinct class of noncoding RNAs characterized by a covalently closed circular structure, which differentiates them from linear RNA species and confers exceptional resistance to exonuclease degradation and RNase activity.70 This structural stability is a key advantage, as circRNA levels often exceed those of their corresponding linear mRNAs, making them highly attractive as diagnostic biomarkers. Functionally, circRNAs harbor multiple miRNA-binding sites and act as competitive ceRNAs, attenuating the suppressive effect of miRNAs on target mRNA expression.71,72 Diverse ceRNA networks involving circRNAs, miRNAs, and mRNAs have been characterized in HCC.73 Next-generation sequencing has revealed widespread dysregulation of circRNAs in blood, urine, saliva, and tumor tissues of HCC patients. Notably, EV-derived circRNA expression levels correlate closely with TNM staging and tumor size, reflecting their tissue specificity and potential as stage-specific diagnostic markers.74
circ-0072088, also known as circ-103809 and circZFR, is highly expressed in EVs isolated from HCC cell lines and the plasma of HCC patients, and its increase suggests tumor node metastasis and unfavorable prognosis of patients.75,76 circ-0072088 is an important regulator of the proliferation, migration, invasion, and TMT of HCC cells by sponging miRNAs to affect downstream gene expression, including miR-1270/PLAG1-like zinc finger 2 (PLAGL2),76 miR-377-3p/fibroblast growth factor receptor 1 (FGFR1),77 and miR-620.78 A recent study showed that upregulated EV-derived circ-0072088 secreted by HCC cells could promote cell invasion and migration by mediating the degradation of miR-375 to regulate MMP-16.75 Plasma EV-derived circ-0072088 showed promising performance for HCC diagnosis, with an AUC of 0.899 (Table 3).75,79–85 Similarly, Huang et al. reported that circANTXR1 was highly and stably expressed in serum EVs of HCC patients and could facilitate the proliferation and migration of HCC cells by being transported by EVs between cells to induce the miR-532-5p/X-ray repair cross-complementing 5 (XRCC5) axis. In clinical diagnostic evaluation, ROC curve analysis demonstrated that serum EV-derived circANTXR1 achieved an AUC of 0.76 for HCC diagnosis.79 In contrast, circ-0051443 was dramatically decreased in the plasma EVs of patients with HCC, and its overexpression in normal cells via EVs significantly suppressed the proliferation of tumor cells by regulating the miR-331-3p/BAK1 axis.80 More importantly, the AUC value for EV-derived circ-0051443 to differentiate patients with HCC from controls was 0.8089. In another study, EV-derived circ-0004277 was remarkably increased in the plasma of patients with HCC, and its AUC, sensitivity, and specificity for diagnosing HCC were 0.816, 58.3%, and 96.7%, respectively.81 Further studies found that circ-0004277 could be transmitted from HCC cells to surrounding normal cells by EVs to induce its overexpression, thereby stimulating EMT progression and inhibiting ZO-1 expression.81 These results demonstrated that EV-derived circ-0072088, circANTXR1, circ-0051443, and circ-0004277 have great potential as novel and independent diagnostic markers and therapeutic targets for HCC.
Table 3The diagnostic efficacy of EV-derived circRNAs for HCC patients
| Author | Country | Year | Sample size | Categories | Diagnostic efficacy | Reference controls |
|---|
| Lin et al.75 | China | 2021 | 50 NC
50 HCC | circ-0072088 | 0.899 | GAPDH |
| Huang et al.79 | China | 2021 | 50 NC
70 HCC | circANTXR1 | 0.76 | GAPDH |
| Chen et al.80 | China | 2020 | 60 NC
60 HCC | circ-0051443 | 0.809 | GAPDH |
| | | | | | |
| Zhu et al.81 | China | 2020 | 60 NC
60 HCC | circ-0004277 | 0.816 | GAPDH |
| Sun et al.82 | China | 2020 | 40 NC
71 HCC | circ-0004001
circ-0004123
circ-0075792
Panels | 0.79
0.73
0.76
0.89 | GAPDH |
| Lyu et al.83 | China | 2021 | 54 NC
108 CLD
111 HCC | circ-0070396
AFP
Combination | 0.857
0.781
0.938 | 18s rRNA |
| Guo et al.84 | China | 2021 | 30 NC
87 HCC | circ-0006602
AFP
CEA
circ-0006602
+AFP | 0.907
0.694
0.589
0.942 | GAPDH |
| Wang et al.85 | China | 2021 | 56 CLD
104 HCC | circ-0028861
AFP
Combination | 0.82
0.76
0.86 | GAPDH |
circRNA signatures combined with AFP tend to show higher diagnostic performance in distinguishing early-stage HCC from controls. Sun et al. reported that the accuracies of circ-0004001, circ-0004123, and circ-0075792 in HCC screening were 0.79, 0.73, and 0.76, respectively, while their circRNA signature contributed to increased sensitivity (90.5%), specificity (78.1%), and AUC values (0.89) for diagnosis.82 circ-0070396 was highly overexpressed in the plasma of HCC patients, whereas its free expression level in HCC plasma samples was not significantly different from that in non-cancerous controls, indicating that circ-0070396 is mainly packaged into EVs.83 The efficiency of EV-derived circ-0070396 (AUC = 0.857) in detecting HCC from HC was substantially higher than that of AFP (AUC = 0.781) and free circ-0070396 (AUC = 0.587), whereas the AUC, sensitivity, and specificity of the combination of EV-derived circ-0070396 and AFP were 0.938, 82%, and 100%, respectively. In addition, the combination had optimal diagnostic performance (AUC of 0.85 and 0.75) for discriminating HCC from LC and chronic hepatitis B.83 In another study, 143 upregulated EV-derived circRNAs were identified by full transcriptome sequencing, in which EV-derived circ-0006602 showed the highest fold-change. Biological cellular experiments have shown that silencing EV-derived circ-0006602 expression dramatically suppresses the proliferation, invasion, and migration of HCC cells.84 In clinical diagnosis, the AUC of plasma EV-derived circ-0006602 was 0.907, especially 0.956 for early-stage HCC, superior to that of serum tumor markers AFP (AUC = 0.694) and carcinoembryonic antigen (CEA) (AUC = 0.598), while it improved to 0.942 when combined with AFP.84 Likewise, Wang et al. reported that serum EV-derived circ-0028861 exhibited an AUC of 0.79 for screening all HCC patients from HBV and LC patients, similar to AFP (AUC = 0.76); whereas the combination of EV-derived circ-0028861 and AFP had a better accuracy (0.86).85
Dysregulated EV-derived circRNAs play crucial roles in HCC development and malignant progression and show considerable promise as biomarkers for early-stage HCC owing to their distinctive structural properties. However, several important challenges hinder their clinical translation. For instance, EV-derived circRNAs including circ-0074854,86 circAKT3,87 and circFBLIM188 are all highly expressed in HCC serum and associated with poor survival and elevated recurrence risk, yet sufficient clinical data to rigorously assess their diagnostic capabilities remain unavailable. Moreover, the lack of standardized EV isolation and circRNA detection protocols across studies makes it difficult to directly compare diagnostic performance metrics. Large-scale, multicenter prospective studies enrolling diverse HCC patient cohorts are urgently needed to identify stable, high-specificity EV-derived circRNA signatures. Improvements in circRNA extraction, amplification, and quantification technologies will further facilitate rigorous diagnostic evaluation.
EV-derived mRNAs
Unlike noncoding RNAs, which are transferred to recipient cells via EVs to modulate downstream gene expression, EV-derived mRNAs can be translated in recipient cells to directly regulate molecular signaling pathways, thereby promoting tumor growth and migration in HCC and surrounding stromal cells.89 Furthermore, key mRNAs encoding critical epigenetic information about HCC cells can be packaged into EVs and secreted into peripheral blood and ascites, exhibiting high heterogeneity and tissue specificity that may be exploited for diagnostic purposes.89 Recent transcriptomic studies have begun to define the EV-derived mRNA landscape in HCC: Huang et al. identified 9,440 dysregulated mRNAs in HCC plasma EVs by RNA sequencing, although the majority showed relatively low tumor specificity.54 Wu et al. further identified 159 differentially expressed mRNAs using an EV-related online database, suggesting that elevated SMARCA5, CDC42, and UBC mRNA levels in HCC patient EVs represent promising candidate biomarkers.90 Detection of EV-derived mRNA thus offers new avenues for minimally invasive HCC diagnosis.
Cancer-testis antigen lactate dehydrogenase C (LDHC), a specific class of the lactate dehydrogenase (LDH) isoenzyme family, was first reported in human spermatozoa and spermatogenic cells.91 LDHC, which has limited expression in normal tissues, is primarily involved in the maintenance of glycolysis and ATP homeostasis throughout spermatogenesis.92 High-throughput sequencing results showed that LDHC is significantly expressed in some malignancies and participates in regulating tumor cell invasion and metastasis, such as lung adenocarcinoma93 and breast cancer.94 Interestingly, Cui et al. found that the expression level of LDHC mRNA was remarkably increased in HCC serum and serum EVs, with 68% and 60% positivity rates, respectively. ROC analysis showed that the AUC of LDHC mRNA in serum EV (0.945) was significantly higher than that in free serum (0.838), suggesting that EV-derived LDHC mRNA is a suitable peripheral screening tool for HCC.95 Iron metabolism is a vital pathway in the promotion and progression of HCC. Abnormal iron accumulation is mainly enriched in hepatocytes, which in turn promotes mitochondrial DNA damage, lipid peroxidation, and protein denaturation.96 HEPCIDIN plays a crucial role in iron metabolism by affecting intestinal absorption pathways. Sasaki et al. found that the HEPCIDIN mRNA variant was more abundant in HCC serum EVs and hepatoma-derived cell lines than in controls; meanwhile, the copy number of the EV-derived HEPCIDIN mRNA variant could be used to discriminate HCC patients from CLD controls, with a sensitivity of 52.2% and specificity of 100%.97 Similarly, heterogeneous nuclear ribonucleoprotein H1 (hnRNPH1) gene was listed for the first time as a new and promising biomarker for early-stage HCC diagnosis. The hnRNPH1 gene is an important RNA-binding protein involved in all key events in cellular RNA biogenesis, and its expression is upregulated in multiple cancer cells.98 Quantitative analysis of serum EV-derived contents in Xu’s study revealed that increased EV-derived hnRNPH1 mRNA indicated advanced Child-Pugh classification and TNM stage.99 In further diagnostic evaluation, serum EV-derived hnRNPH1 mRNA displayed satisfactory efficiency in differentiating HCC patients from CHB patients (AUC 0.865, sensitivity 85.2%, specificity 76.5%), while serum AFP had an AUC of 0.785. However, both markers yielded similar AUC values in diagnosing HCC from LC patients (0.647 for hnRNPH1 mRNA vs. 0.674 for AFP). Notably, EV-derived hnRNPH1 mRNA in combination with AFP increased the diagnostic accuracy with 0.891 AUC, 87.5% sensitivity, and 84.8% specificity.99
RNA expression profiling is widely used to construct predictive signatures for cancer diagnosis and progression. For instance, myosin light chain 6 B (MYL6B) is a class of light chains of the myosin family, which is mainly responsible for material transport and cell adhesion processes.100 A previous study showed that MYL6B could induce ubiquitination and degradation of the p53 protein to suppress HCC cell apoptosis.101 The THO complex subunit 2 gene (THOC2) is located on chromosome Xq25 and encodes a subunit of the Transcription-Export (TREX) complex.102 THOC2 plays an essential role in nuclear mRNA export and maintaining highly conserved properties. Dysregulated THOC2 or mutations result in abnormal retention of mRNA in the nucleus, which can cause neurodevelopmental disorders and tumorigenesis.103 Zhu et al. found that EV-derived MYL6B and THOC2 expression were closely related to clinical features of HCC patients and HCC cell migration. They established a highly effective model for diagnosing early-stage HCC: P = −23.0370 + (2.2566 × expression level of EV-derived MYL6B) + (3.8247 × expression level of EV-derived THOC2). This model had an AUC of 0.879, sensitivity of 76.0%, and specificity of 82.0%.104 Sun et al. constructed a diagnostic model with 10 specific plasma-EV mRNAs for early detection of HCC, with an AUC value of 0.93 (sensitivity = 94.4%, specificity = 88.5%).105 However, these gene signatures require further clinical data to verify their practical application.
EV-derived proteins and lipids
Although traditional serum protein markers remain the primary minimally invasive approach for HCC screening, their clinical utility is limited, particularly for early-stage disease or for monitoring disease progression.106 Growing evidence indicates that EV-derived proteins and lipids play dual roles: they serve as essential structural and functional components for EV biogenesis and cargo transport, while simultaneously carrying substantial molecular information that reflects the physiological and pathological state of the originating cells.107–109 Compared with conventional serum protein markers, which are subject to dilution in blood and interference from complex protein interactions, EV-derived proteins and lipids are highly concentrated and can be enriched through EV purification procedures.108 Multiple studies have now established the diagnostic and prognostic utility of EV-derived proteins and lipids across various cancer types, including gastric cancer,110 HCC,111 pancreatic cancer,112 and colorectal cancer.113
Data-independent acquisition (DIA) is a robust and precise proteomic quantification technology based on mass spectrometry that can be used to provide highly reproducible measurements of complex protein samples.114 Recently, DIA was developed for protein biomarker discovery and validation through bottom-up proteomics.115 By performing DIA analysis, Zhao et al. clustered a spectral library containing 1,251 proteins from the serum EV-derived proteomic profiling of HCC patients and identified ten EV-derived protein panels as candidate biomarkers, such as hemopexin (HPX).111 HPX glycan has been verified as a promising marker for distinguishing HCC patients from negative controls, with a sensitivity and specificity of 79% and 93%, respectively.116 Chaperonin containing TCP1 subunit 8 (CCT8) is widely known to play an essential role in maintaining the correct folding of various oncoproteins and tumor suppressors. Studies have found that CCT8 is significantly overexpressed in HCC cells and blood samples. It correlates directly with the histological grade and malignancy of HCC.117,118 A study found that CCT8 attenuates the migration and invasion of HCC cells via glucose-regulated protein 94 (GRP94)/c-Jun/EMT signaling.119 Likewise, cofilin 1 (CFL1) is increased in human HCC, and its high expression is associated with a low overall survival rate and poor progression in patients with HCC.120 Both in vivo and in vitro experiments have revealed that silencing CFL1 remarkably suppresses the growth and metastasis of HCC cells.121 This evidence indicates that CCT8 and CFL1 may serve as potential targets for the diagnosis and treatment of HCC. Interestingly, a recent study reported that CCT8 and CFL1 proteins in EVs isolated from HCC cells and serum had higher expression levels than in controls, and ROC analysis of the validation cohort showed that the AUC of serum EV-derived CCT8 protein (cutoff > 31 ng/mL) and CFL1 protein (cutoff > 1,040 pg/mL) were 0.698 and 0.677, respectively, to detect HCC, while the AUC of serum AFP (cutoff > 20 ng/mL) was 0.628.122 However, the combination with AFP had the highest diagnostic performance, with an AUC of 0.838.122 Hence, EV-derived CCT8 and CFL1 proteins represent promising novel biomarkers for HCC screening (Table 4).95,97,99,104,105,122 Further research involving evaluation of their diagnostic efficiency in free serum will contribute to the selection of a better strategy.
Table 4The diagnostic efficacy of EV-derived mRNAs and proteins for HCC patients
| Author | Country | Year | Sample size | Categories | Diagnostic efficacy | Reference controls |
|---|
| Cui et al.95 | China | 2020 | 100 NC
112 HCC | LDHC | 0.945 | GAPDH |
| Sasaki et al.97 | Japan | 2019 | 30 NC
17 CLD
20 HCC | HEPCIDIN | 0.60 | NA |
| Xu et al.99 | China | 2017 | 68 NC
135 CLD
88 HCC | hnRNPH1
AFP
Combination | 0.865
0.785
0.891 | GAPDH |
| Zhu et al.104 | China | 2021 | 202 NC
243 HCC | MYL6B+THOC2 | 0.907 | NA |
| Sun et al.105 | China | 2020 | 23 NC
51 CLD
46 HCC | AFP/GPC3/ALB/APOH
/FABP1/FGB/FGG/AHSG
/RBP4/TF
AFP | 0.93
0.65 | NA |
| Cho et al.122 | Korea | 2021 | 34 NC
58 CLD
132 HCC | CCT8
CFL1
AFP
Combination
AFP+CCT8
AFP+CFL1
CCT8+CFL1 | 0.698
0.677
0.628
0.838
0.728
0.773
0.829 | HMBS |
EVs are rich in lipids, which not only protect their contents from degradation but also play a crucial role in the entire process of EV formation, including the creation of multivesicular bodies, endosomal sorting, secretion, and interactions with target cells.123 The types and distribution of lipids in EV membranes vary depending on cell type. Additionally, studies have shown that different biological states of cells, including cancerous conditions, can significantly influence the composition and quantity of lipids in EVs.124 Therefore, EV-derived lipids carry valuable information about the physiological state of the cells that produce them, making them potential biomarkers.109 For instance, a recent study using ultra-high-resolution mass spectrometry identified 2,864 distinct EV-derived lipids across 52 classes that were differentially expressed in HCC patients compared with non-HCC patients.107 Certain lipid species, such as dilysocardiolipins, lysophosphatidylserines, cardiolipins, and sphingosines, show specific expression levels in early-stage HCC, indicating their diagnostic potential.109 However, it is important to note that EV lipidomics is a relatively new area of research, and a limited number of studies have explored it. According to statistical data, the majority of current research on EV biology (over 95%) has focused on genomics and proteomics.123 Nevertheless, advancements in technologies for isolating, purifying, and quantifying EVs have led to growing interest in the biological functions of EV-derived lipids. EV-derived lipids are being explored as new therapeutic and diagnostic markers for cancer.124 Currently, there are limited reports on the detection capabilities of EV-derived lipids in HCC patients, but substantial clinical data are being gathered to identify potential EV-derived lipid biomarkers that could enhance the positive detection rate in HCC patients.
The endogenous controls for EVs
Reference genes—commonly referred to as housekeeping genes—are stably expressed across diverse tissues and cell types and are routinely used as endogenous controls to normalize target gene expression and minimize experimental variability.138 Selecting an appropriate endogenous control is particularly critical for accurate quantitative analysis of EV-derived nucleic acids, as the EV isolation process introduces additional sources of variability not encountered in conventional cell-based assays. Discrepancies in reference gene selection across studies contribute substantially to inter-study inconsistencies in HCC diagnostic accuracy and further impede meaningful cross-study comparisons. Researchers have advocated reference gene validation specific to cell type, HCC subtype, and sample source (including EVs).139,140
Currently, the snRNAs U6 and miR-16 are commonly used to normalize miRNAs in HCC EVs and plasma/serum.141 However, recent findings indicate that U6 and miR-16 are highly unstable for normalizing EV-derived miRNAs, owing to significant expression variability within groups.142 Tang et al. found that U6 was the least stable endogenous control when assessing miRNA levels in serum samples from patients with HCC, CLD, HBV, and healthy individuals.143 Similarly, Jacobsen et al. noted a significant bias when using U6 and miR-211 between HBV-replicating cells and liver cells, highlighting their inadequacy as reference genes.144 Additionally, some studies have raised concerns that U6 and miR-16 from EVs might be affected by contamination, particularly because U6 is primarily found in the nucleus, which contradicts its EV origin.145 Consequently, new candidate reference genes have been proposed to enhance miRNA quantification accuracy, such as miR-106a143 and miR-1280146 for most HCC, miR-4644 for HBV/LC-HCC,147 and miR-24-3p and miR-151a-3p for HBV-HCC.144 Furthermore, cel-miR-39 has gained popularity as a normalization choice in many laboratories owing to its stable expression. Li et al. identified miR-221, miR-191, let-7a, miR-181a, and miR-26a as the optimal endogenous controls for EV-derived miRNAs in patients with HCC and HBV.142 Notably, another independent study found that using miR-221, miR-181a, and miR-26a for normalization improved the quantitative accuracy of EV-derived miRNAs in HCC patients.148
Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) and β-actin (ACTB) are the housekeeping genes most commonly used in cancer research. GAPDH is a key enzyme in glycolysis, whereas ACTB is an important component of muscle filaments and cytoskeletal microfilaments.149 Our analysis shows that these two genes are frequently used as internal reference genes in studies assessing the diagnostic value of EVs. However, only Bahri et al.150 reported the precise standardization of suitable reference genes in EVs derived from cancer cells. Controversially, the three traditional genes ACTB, TBP, and HPRT1 were nearly absent in the EVs of liver cancer cells. This may be limited by the use of different algorithms (Delta Ct, NormFinder, RefFinder). However, subsequent validation confirmed the stability of the reference GAPDH. In subsequent research, YWHAZ and UBC could also be considered for data normalization in the study of liver cancer EVs. Some studies have indicated that the same reference genes may not be applicable to different cell types or culture environments. Therefore, it is recommended to validate the internal reference genes before evaluation, select appropriate reference genes, and achieve higher accuracy for data normalization.
Future perspectives
Extracellular vesicles have emerged as a transformative platform for early HCC liquid biopsy, offering superior biomarker stability, specificity, and sensitivity compared with traditional serum markers. Unlike most existing reviews that catalogue biomarkers in isolation, this work integrates mechanistic evidence, quantitative diagnostic performance data, methodological standardization considerations, and a critical appraisal of clinical trial quality to provide a clear, evidence-grounded translational roadmap for EV-based HCC diagnostics.
A central conclusion of this review is that EV-derived biomarkers often outperform AFP in many studies for detecting early-stage and AFP-negative HCC. Among the most clinically validated candidates, miRNA panels (particularly EV-miR-21, miR-10b-5p, miR-122, and miR-148a), lncRNA LINC00853, circRNA signatures (circ-0006602, circ-0070396), and combined EV protein markers (CCT8 and CFL1 paired with AFP) achieve AUC values consistently above 0.90 across independent cohorts (Fig. 3), approaching the performance expected of clinical-grade diagnostic tools. Despite these compelling results, widespread clinical implementation remains constrained by inconsistent EV isolation protocols, the absence of large-scale multicenter validation, and a lack of universally accepted diagnostic cutoff values. To translate EV-based liquid biopsy into clinical practice, a prioritized implementation strategy is proposed. In terms of target populations, EV-based screening should first be deployed in well-defined high-risk groups—specifically individuals with HBV or HCV infection, LC, advanced NAFLD-related fibrosis, or AFP-negative early HCC—where unmet diagnostic needs are greatest and the incremental benefit over the current standard of care is most pronounced. Regarding specific biomarker targets, priority should be assigned to EV-miRNA panels combined with AFP (achieving AUC ≥ 0.94), lncRNA LINC00853 (sensitivity 93.75%, specificity 89.77% for stage I HCC), and combined EV protein signatures, given their demonstrated accuracy, analytical feasibility, and clear biological relevance to HCC initiation and immune evasion. Concerning the translational pathway, adoption of MISEV2023-compliant standardized protocols for EV isolation and characterization is the first indispensable step, followed by prospective multicenter validation studies enrolling diverse etiologies and ethnicities to establish robust, universally applicable diagnostic thresholds. From a therapeutic standpoint, engineered EVs also hold substantial promise as targeted drug delivery vehicles (Fig. 4), leveraging their natural hepatotropism, low immunogenicity, and superior membrane permeability; sorafenib- or doxorubicin-loaded EV formulations have already shown enhanced efficacy in preclinical HCC models.20 Ultimately, integrating EV biomarkers with imaging data, conventional liquid biopsy panels, and clinical risk scores will generate clinically actionable precision models for HCC prevention, early diagnosis, and targeted treatment.