Introduction
Alpha-1 antitrypsin deficiency (AATD) is a rare genetic disorder with a high prevalence in northwestern Europe and North America.1 AATD is caused by a single mutation in the SERPINA1 gene, resulting in misfolding and polymerizing of the Z mutant form of alpha-1 antitrypsin (AAT) in hepatocytes, rather than its secretion.2 AAT is the most abundant serine protease inhibitor in the plasma. While intrahepatic accumulation of misfolded AAT can lead to liver disease, the resultant low levels of AAT in the circulating plasma can cause lung inflammation in AATD patients.3 In adult AATD individuals, liver disease is often underdiagnosed until the final stage of cirrhosis, when liver transplantation is the only treatment option.4 In AATD children, some may present with jaundice or cirrhosis, also requiring liver transplantation.5
Liver fibrosis and cirrhosis are major causes of worldwide morbidity and mortality, with limited therapeutic options, often leading to hepatocellular carcinoma (HCC). While cirrhosis is irreversible, new evidence indicates that liver fibrosis is dynamic and reversible.6 Liver fibrosis in AATD patients is primarily caused by hepatotoxicity due to the abnormal accumulation of AAT in hepatocytes.7 It is crucial to assess the extent of the hepatic AAT burden and identify individuals in the early stages of the disease to implement curative treatments before cirrhosis develops. Currently, liver biopsy is the gold standard for diagnosing AATD-mediated liver disease and assessing the stage of fibrosis.7 However, the biopsy is an invasive procedure associated with various complications. Additionally, the limited size of liver samples and the subjective nature of pathological assessment limit the accuracy and reproducibility of histologic diagnoses.8 Given these challenges, alternative approaches are needed to overcome the limitations of liver biopsy in diagnosing and assessing liver disease in AATD patients.
Extracellular vesicles (EVs), with diameters ranging from 30 to 200 nm, are released by all cells in the body.9 EVs selectively carry signaling biomolecules such as microRNAs (miRNAs), messenger RNAs, proteins, and lipids. Most biofluids, including plasma and serum, contain a significant number of EVs, which coordinate multiple physiological and pathological processes through cargo transfer between cells and organs.3,9 In the liver, almost all cells, including hepatocytes, release and receive EVs. EVs play an important role in the initiation and progression of liver diseases, including inflammatory disorders and fibrosis.9 The study of EV contents has emerged as a promising area for discovering new biomarkers in liver disease.9
MiRNAs are commonly associated with EVs,10 and several EV-associated miRNAs have been found to be altered and linked to different types of liver diseases.11,12 These miRNAs have been shown to be associated with the clinical stage of various liver diseases, such as metabolic chronic liver disease and viral hepatitis.11 AATD-mediated liver disease is characterized by a toxic “gain-of-function” mechanism involving ER and oxidative stress, mitochondrial dysfunction, and lipotoxicity.7 However, there has been no systematic screening of plasma EV-associated miRNAs for hepatotoxic features of AATD. Additionally, there is no consensus on the potential use of circulating miRNAs as diagnostic indicators for detecting AATD-associated liver disease.
In this study, we employed an innovative approach to compare miRNA profiles of plasma-derived EVs from AATD individuals with and without liver disease, as compared with non-AATD healthy controls. We chose to focus specifically on EV-associated miRNAs rather than free-circulating miRNAs in plasma for several reasons. While free miRNAs may originate from non-specific sources, such as general cell death or injury, EV-associated miRNAs have higher stability, a more targeted role in cell-to-cell communication, are more disease-specific, and have reduced background noise in plasma, thereby increasing the likelihood of identifying disease-specific biomarkers.13 The current gold standard for diagnosing AATD involves measuring serum AAT levels, followed by genotyping to identify mutations in the SERPINA1 gene. It is important to note that our objective was not to identify miRNA biomarkers to distinguish AATD individuals from non-AATD subjects. Instead, our integrative approach involved analyzing both AATD individuals and non-AATD healthy controls, followed by filtering steps to identify miRNAs that could provide insights into AATD liver disease status and aid in detecting AATD individuals at risk of developing pathological liver disease. Although they are not yet part of the standard diagnostic process, EV miRNAs could serve as molecular markers for early detection, monitoring liver disease progression, or assessing treatment response in AATD patients. However, further validation and large-scale studies are needed before EV miRNAs can be integrated into routine clinical practice.
Methods
Human subjects
The study protocol was approved by the Clinical Research Ethics Committee of the University of Florida (IRB202101148), and written informed consent was obtained from all subjects in accordance with the Declaration of Helsinki Principles.14 AATD individuals (n = 62; ages 35–70 years; mean 59 years; male:female ratio 9:10) were recruited by pulmonologists specialized in AATD at the University of Florida Shands Hospital (n = 17 in the testing cohort and n = 45 in the validation cohort). Diagnosis of AATD was established by genotyping, isoelectric focusing of serum proteins, and measurement of AAT serum levels by nephelometry (Behring Diagnostics, Marburg, Germany) using in-house standards and controls.15 Percutaneous liver biopsy was performed on a well-characterized cross-section of adults with AATD, as previously reported.7 A group of age- and gender-matched non-AATD control individuals (n = 20; ages 37–79 years; mean 52 years; male: female 2: 3) from the Alpha-1 Foundation DNA and Tissue Bank were also recruited. Detailed clinical data of subjects enrolled in this study are shown in Table 1.
Table 1Demographics of study subjects
| MM (n = 20) | ZZ PASD 0–1 (n = 24) | ZZ PASD 2–3 (n = 38) |
|---|
| Age (years) | 52 (37–79) | 55 (25–71) | 58 (40–70) |
| Sex (Female%) | 60 | 71 | 47 |
| Augmentation (%) | N/A | 58 | 60 |
| FEV1 (%) | N/A | 67.3 | 61.2 |
| ALT (U/L) | N/A | 22.7 | 24.0 |
| AST (IU/L) | N/A | 29.7 | 29.1 |
Sample collection and processing
Blood samples were collected from participants following guidelines set by the International Society for Extracellular Vesicles.16 The collected blood was centrifuged at 2,500xg for 10 m at room temperature, and plasma was separated, aliquoted, and stored at −80°C. Plasma aliquots were thawed on ice, quickly swirled in a 37°C water bath until clear, and then centrifuged for 5 m at 18,000xg at 4°C. The supernatant was removed and passed through a 0.8 mm filter prior to EV extraction. Four milliliters of plasma were used per sample.
EV isolation
EVs were isolated from cryopreserved plasma using differential centrifugation, as described previously.3 Briefly, plasma samples were ultracentrifuged in a Beckman Coulter Optima XE90 Ultracentrifuge with an SW-40Ti swinging rotor at 110,000xg for 2 h at 4°C. EV pellets were resuspended in PBS, filtered through a 0.22 µm filter (Millipore, Billerica, MA), and centrifuged at 110,000xg for 70 m at 4°C. The EVs were then resuspended in 200 µL of PBS and stored for further analysis.
RNase and proteinase treatment of isolated plasma EVs
Free RNA and proteins were removed from the isolated EV fractions by treatment with 100 µg/mL of Proteinase K (Qiagen, Hilden, Germany) for 30 m at 37°C, followed by 10 µg/mL of RNase A (Qiagen, Hilden, Germany) for 15 m at 37°C.
EV protein quantification
Protein concentrations of EV fractions were quantified using the Pierce BCA Protein Assay Kit (Thermo Scientific, Waltham, MA) according to the manufacturer’s instructions. Four microliters of each standard and lysed EV sample were mixed into a 96-well plate with 200 µL of reaction buffer. The plate was incubated at 37°C for 30 m, and absorbance at 562 nm was measured using a spectrophotometer (SpectraMax, Molecular Devices). A standard curve was used to determine the protein concentration.
EV characterization
Purified EVs were characterized for size, concentration, presence of specific EV markers, and morphology. To assess morphology and size, isolated EVs were subjected to transmission electron microscopy and nanoparticle tracking analysis (NTA), as previously described.3 EVs were also analyzed for the presence of CD81, CD63, and TSG101 as specific EV markers and for the absence of the ER marker Calnexin by western blot analysis.
RNA isolation and library preparation
RNA from EV fractions was isolated using the exoRNeasy Maxi Kit (Qiagen, Germantown, MD), and the quantity and quality of RNA and small RNA were determined using the Agilent RNA 6000 Pico Kit and Small RNA Kit on the Agilent Bioanalyzer, following the manufacturer’s instructions. Sequencing libraries were constructed using approximately 6 ng of total EV-derived RNA with the Small RNA Library Prep Set for Illumina (New England BioLabs), according to the manufacturer’s protocol. Each adapter-ligated RNA was mixed with ProtoScript II reverse transcriptase, RNase Inhibitor, and first-strand synthesis reaction buffer (NEB, catalog number E7560S) and incubated for 60 m at 50°C. Index codes were added to each reaction product for sample differentiation. Size selection was performed using a 6% polyacrylamide gel. For miRNA, bands corresponding to ∼150 bp were isolated. Library quality was assessed with the Agilent Bioanalyzer 2100 and Qubit, and libraries were sequenced using the Illumina MiSeq 1 × 150 cycles V3 kit at the University of Florida Interdisciplinary Center for Biotechnology Research.
Mice and treatments
The study protocol was approved by the ethics committee of the University of Florida (IACUC 202300000056). Wildtype, C57/B6 mice, aged 12 weeks, were housed under controlled environmental conditions (12-h light/dark cycle; temperature 22–24°C) and provided food and water ad libitum in the Animal Care Facility. Mice were divided into three groups: Group 1 (n = 6) consisted of control animals injected intraperitoneally with CCl4 (Sigma-Aldrich, St. Louis, MO, USA) (0.3% CCl4, 10 µL/g of body weight, twice weekly) for four weeks to induce liver fibrosis. Groups 2 and 3 (n = 6/group) were injected intraperitoneally with CCl4 using the same schedule and dosing as Group 1, along with intravenous injections of EVs derived from either normal or AATD plasma (1 × 109 EVs, pooled from three subjects, twice weekly). All animals were euthanized 48 h after the last dose of CCl4, and their livers were collected for analysis.
Small RNA-sequencing and bioinformatics
After removing adapters using cutadapt software,17 the small RNA sequencing reads were aligned to the Genome Reference Consortium Human Build 38 (hereinafter referred to as GRCh38) using bowtie2 software18 with “–very-sensitive-local” option for local alignment, which does not require reads to align end-to-end. miRNA expression count data were extracted using HTSeq software19,20 with the miRBase Sequence Database (release 22)21 as the gene annotation reference. The resulting miRNA expression data were further normalized using the count per million reads method, followed by log2(x+1) transformation, where x is the expression value and 1 is the pseudo count. Low-quality sequences were trimmed, and poor-quality reads were removed using Trimmomatic.22 Star Aligner was used to map high-quality paired-end reads to the human genome (GRCh38).23 Gene expression was obtained using RSEM.24
Functional pathway and network analyses
We performed pathway enrichment analysis using Ingenuity Pathway Analysis (IPA) (Ingenuity Systems, Inc., Redwood, CA). IPA core analyses are based on previous knowledge of associations within causal networks and their downstream target genes archived in the Ingenuity knowledge base.25 An overlap p-value, based on the significant overlap between networks and their targets, and an activation z-score were computed. The activation z-score is used to infer likely activation or inhibition states of causal networks. Networks were considered significantly activated or inhibited if the overlap p-value was ≤0.05 and the z-score was ≤2.0 (or ≥−2.0).
Bibliographic search and miRNA selection
A literature search of PubMed and the Liver Disease MicroRNA Signature26 databases was conducted to identify dysregulated miRNAs in various liver diseases. We combined this bibliographic information with our results from next-generation sequencing (NGS) of plasma-derived EVs from AATD and healthy control samples. A group of highly expressed miRNAs (each producing over 5,000 reads in the NGS assay and belonging to the most abundant miRNAs identified by NGS) were further considered for biomarker studies as molecular identifiers for the severity of AATD-associated liver disease.
Real-time qPCR validation of selected miRNAs
Reverse transcription reactions were performed using TaqMan Advanced miRNA Assays (Applied Biosystems Inc., CA) with 2 µL of cell-free RNA extracted from plasma-derived EVs to prepare cDNA. Real-time PCR reactions were performed in duplicate on an Applied BioSystems 7500 Fast thermocycler, programmed as follows: 95°C for 20 s, followed by 40 cycles of 95°C for 1 s and 60°C for 20 s. We used hsa-miR-21-5p and hsa-miR-26a-5p, two of the most stable miRNAs in terms of read counts, as endogenous controls. All fold-change data were obtained using the 2−ΔΔCt method.27
Liver histology
Percutaneous liver biopsy on AATD individuals was performed with a 16-gauge BioPince™ core biopsy needle. The samples were formalin-fixed and processed for further examination. Hematoxylin & Eosin (H&E), Masson’s Trichrome, and Periodic Acid-Schiff positive, Diastase-resistant (PASD) stains were performed on the liver sections. Two pathologists scored each biopsy for fibrosis using METAVIR stages F0-F4 based on Masson’s Trichrome staining. Clinically significant fibrosis was defined as stage ≥ F2. PASD staining was also used to identify AAT accumulation and assess AATD-mediated liver disease severity. PASD globules present in a high-power field were scored from 0–3 as follows: 0 - None, 1 - Rare (<5 hepatocytes with globules), 2 - Few (5–20 hepatocytes with globules), 3 - Numerous (≥20 hepatocytes with globules).7 Mouse liver tissues were also formalin-fixed and stained for collagen deposition (Picrosirius Red) or α-smooth muscle actin as an indication of hepatic stellate cell activation.
Statistical analysis
After removing non-expressed miRNAs, differential analysis was performed using the edgeR package,28 which employs a negative binomial distribution to model count data. For RNA-Seq data, we performed differential expression analysis using edgeR.27,28 Thresholds were set at a false discovery rate (q-value) ≤ 0.05 and a log2 fold change of >1 or <−1 (linear fold change of >=< 2). Each miRNA’s predictive power was determined by a receiver operating characteristic (ROC) curve. To ensure accurate and reliable ROC analysis, miRNA data were normalized by count per million reads, log2-transformed, and standardized. In the ROC curve, PASD status (no liver disease [PASD 0–1] vs. liver disease [PASD 2–3]) was the outcome variable, and the expression level of each miRNA was the predictor. By varying miRNA expression criteria in the prediction of PASD status, we obtained a series of prediction sensitivities and specificities. The optimum sensitivity and specificity were chosen such that the Youden index was maximized, where the Youden index = sensitivity + specificity + 1. To assess over-predictive power, we calculated the area under the curve (AUC). An AUC of one indicates perfect prediction, while an AUC of 0.5 is equivalent to random guessing. To explore the combined predictive power of miRNA panels (e.g., sets of three or four miRNAs), we constructed composite scores using logistic regression and assessed their performance using ROC analysis. To ensure validity and avoid overfitting, we tested these composite scores on an independent cohort of 45 AATD cases.
Results
Isolation and characterization of EV
EVs were isolated from the plasma of 20 non-AATD control subjects (MM) and 62 AATD individuals (ZZ) with different stages of liver disease using a combination of filtration and ultracentrifugation. EV fractions were fully characterized by analyses of particle size, distribution, and concentration.3 NTA was performed to assess size distribution and concentration of EVs in both AATD patients and control groups. Figure 1A presents the average particle size profiles for each group, with no significant differences in size distribution between samples. EV isolated from the plasma of both MM and ZZ groups were within the normal range for EV size (below 150 nm), and the EV recovery in ZZ individuals was the same as in the MM control group (Fig. 1A). No AUC analysis was conducted, as NTA is designed to measure particle size and concentration, rather than providing data suitable for AUC comparison. The cup-shaped morphology of the EVs was confirmed in both MM (top) and ZZ individuals (bottom) by transmission electron microscopy (Fig. 1B). Results from western blot analysis confirmed the expression of CD63, CD81, and TSG101 EV markers in all EV fractions, while Calnexin (EV negative control marker) was not detected (Fig. 1C). Coomassie blue staining showed that Proteinase K treatment degraded most proteins in both plasma and the isolated EV fraction (Fig. 1D). After Proteinase K and RNase treatment, we observed that EV isolated from the plasma of the MM group had slightly higher concentrations (p-value < 0.05) of total RNA compared to EV isolated from the plasma of the ZZ group (Fig. 1E).
Plasma EV-derived miRNA profile of AATD and control individuals and validation of miRNAs by RT-qPCR
To obtain the profile of plasma EV-derived miRNA in AATD individuals and non-AATD controls, we first analyzed 37 plasma EV samples (17 AATD, 20 non-AATD controls) by small RNA sequencing. Based on raw data, an average of 6,116,653 raw reads per sample were obtained for non-AATD control samples and 11,846,649 reads per sample for the AATD group (Supplementary Table 1), reflecting a significant difference between the two groups (p-value < 0.001). A total of 682 mature known miRNAs were identified among all samples. More than 80% of these miRNAs (574 miRNAs) were already included in the vesicular database Vesiclepedia, and 36% (246 miRNAs) were included in EVpedia as related to EVs from human samples (Fig. 2A). The complete list of identified miRNAs has been submitted to the EVpedia database under the title of this manuscript. In differential expression analysis (|log2(F.C.)| > 1, q-value < 0.05), 178 miRNAs were found to be differentially expressed between AATD individuals (regardless of whether they had liver disease) and non-AATD healthy controls. Out of these 178 differentially expressed miRNAs (DEMs), 92 miRNAs were upregulated, and 86 miRNAs were downregulated in AATD individuals compared to healthy controls (Fig. 2B and C). Our small RNA-seq data indicated that miR-25-3p, miR-99b-3p, and miR-30a-5p were highly upregulated, while miR-451a, miR-374a-5p, miR-146a-5p, miR-16-5p, miR-26b-5p, miR-96-5p, and miR-142-5p were highly downregulated in the plasma EVs of AATD individuals compared to non-AATD healthy controls (Supplementary Fig. 1A). To validate these small RNA sequencing results, 20 miRNAs were randomly chosen from the 178 DEMs to ensure an unbiased validation of our findings. This approach allowed us to confirm the reproducibility and reliability of the sequencing results across a diverse set of miRNAs. Relative expressions of these miRNAs were determined by RT-qPCR measurements using hsa-miR-26a-5p as a reference.29 Expression of hsa-miR-125a, hsa-miR-128-3p, hsa-miR-99a-5p, hsa-miR-335-5p, hsa-miR-361-3p, hsa-miR-150-5p, hsa-miR-30e-5p, hsa-miR-193a-5p, and hsa-miR-23a-3p were significantly upregulated in AATD individuals (Supplementary Fig. 1B) compared to controls (Mann–Whitney p-values < 0.001). Expression of hsa-miR-16-5p and hsa-miR-142-3p was significantly downregulated in AATD individuals (Supplementary Fig. 1C) compared with controls (Mann–Whitney p-values < 0.001). We also observed that with a 95% confidence interval, the fold change values obtained from both methods overlapped in 10 cases (Fig. 2D). Moreover, fold change values obtained from these two different methods were highly correlated (Pearson’s correlation 0.86, p-value 0.0013).
Correlation of plasma EV-miRNA with PASD and fibrosis scores in AATD individuals
As we have previously shown,7 AATD-mediated liver disease is characterized by hepatic accumulation of misfolded AAT, forming PASD globules in the hepatocytes (PASD scores 0–3), and subsequent liver fibrosis (METAVIR scores 0–3), as measured by Masson’s Trichrome staining (Fig. 3A). In this study, we conducted two separate analyses to detect correlations between EV-associated DEMs with (A) PASD, and (B) fibrosis scores in 17 AATD individuals, considering each individual’s PASD and fibrosis scores, respectively. We identified 89 miRNAs whose expression increases with increasing PASD scores, and 68 miRNAs whose expression was associated with increasing fibrosis scores in these 17 AATD individuals. Notably, there was an overlap of 60 DEMs where expression was associated with both PASD and fibrosis scores (Fig. 3B). IPA target analysis revealed that these 60 DEMs, associated with both PASD and fibrosis scores, predominantly target genes such as Bcl2, Col1A1, Col4a, EGFR, FN1, and Pdgfr, and are involved in liver cell proliferation, activation, and fibrogenesis. The cord plot from ingenuity target analysis in Supplementary Figure 1D indicates target genes that are involved in hepatic stellate cell activation, proliferation, and fibrogenesis. The 18 miRNAs displayed in the cord plot show the most relevant interactions with genes specifically linked to liver fibrogenesis. This result is consistent with our previous demonstration that the severity of fibrosis correlates with PASD scores in the liver of AATD individuals.7 To further investigate the impact of AATD circulating EVs and their miRNA content on hepatic stellate cell activation in vivo, we employed the established CCl4 liver fibrosis model and administered either control or AATD plasma-derived EVs. After four weeks, mice treated with CCl4 and AATD EVs exhibited a significant increase in α-smooth muscle actin–positive cells (activated hepatic stellate cells) within fibrotic septa and enhanced collagen deposition compared to those receiving CCl4 with control EVs or CCl4 alone (Fig. 3C).
Exploring potential plasma EV-associated miRNA biomarker candidates to assess the severity of liver disease in AATD
We first divided the experimental set of 17 AATD individuals into two subgroups: those with no liver disease (PASD scores 0–1) and those with liver disease (PASD scores 2–3). Our DEM analysis (|log2(F.C.)| > 1, p < 0.05) unveiled 39 differentially expressed miRNAs (Fig. 4A). We then individually screened these 39 DEMs for their performance as potential biomarkers of AATD-mediated liver disease by executing a ROC and AUC analysis and calculating their specificity and sensitivity. Most candidate DEMs (24 miRNAs) displayed a specificity of 0.7–1.0 and a sensitivity of 0.8–1.0 (Table 2). From these DEMs, a total of 6 miRNAs were selected based on significant differential expression, as well as theoretical evidence for clinical application in liver disease based on previous studies (Table 3).30–42 To validate the specificity of these six plasma EV-derived miRNAs for AATD individuals in the testing set, ROC analysis was performed, and AUC was calculated for each of these miRNAs in both the testing (n = 17) and validation (n = 45) cohorts of AATD individuals (total of 62 AATD patients) (Fig. 4B).
Table 2Top 24 miRNAs ranked by sensitivity specificity (miRNA-seq data)
| miRNA_ID | log2FC | p-value | Sensitivity | Specificity | Sen × Spe |
|---|
| hsa-miR-30a-5p | 1.73 | 5.05E-06 | 1 | 0.9 | 0.9 |
| hsa-miR-10b-5p | 2.42 | 0.00034 | 1 | 0.9 | 0.9 |
| hsa-miR-361-3p | 1.17 | 0.0077 | 0.88 | 1 | 0.88 |
| hsa-miR-10a-5p | 2.16 | 3.12E-05 | 1 | 0.85 | 0.85 |
| hsa-miR-186-5p | −2.19 | 0.000117 | 1 | 0.85 | 0.85 |
| hsa-let-7c-5p | 1.79 | 0.000356 | 1 | 0.85 | 0.85 |
| hsa-miR-451a | −3.94 | 2.08E-07 | 0.94 | 0.9 | 0.85 |
| hsa-miR-223-3p | 1.89 | 3.48E-05 | 0.94 | 0.9 | 0.85 |
| hsa-miR-3615 | 2.48 | 5.99E-05 | 0.94 | 0.9 | 0.85 |
| hsa-miR-20a-5p | −2.28 | 6.69E-05 | 0.94 | 0.9 | 0.85 |
| hsa-miR-146a-5p | −1.92 | 1.57E-06 | 1 | 0.8 | 0.8 |
| hsa-miR-96-5p | −4.03 | 2.50E-06 | 0.94 | 0.85 | 0.8 |
| hsa-miR-30a-3p | 2.04 | 1.42E-05 | 0.94 | 0.85 | 0.8 |
| hsa-miR-28-3p | 1.61 | 0.000104 | 1 | 0.8 | 0.8 |
| hsa-miR-125a-5p | 1.99 | 0.006954 | 0.94 | 0.85 | 0.8 |
| hsa-miR-99b-5p | 1.43 | 0.000576 | 0.82 | 0.95 | 0.78 |
| hsa-miR-99a-5p | 1.77 | 0.000865 | 0.94 | 0.8 | 0.75 |
| hsa-miR-1244 | 1.45 | 0.00574 | 0.94 | 0.8 | 0.75 |
| hsa-miR-16-5p | −3.75 | 2.07E-06 | 0.88 | 0.85 | 0.75 |
| hsa-miR-194-5p | −2.49 | 4.83E-05 | 0.88 | 0.85 | 0.75 |
| hsa-miR-370-3p | 1.88 | 0.000773 | 0.88 | 0.85 | 0.75 |
| hsa-miR-150-5p | 2.04 | 0.00156 | 0.88 | 0.85 | 0.75 |
| hsa-miR-139-5p | 1.58 | 0.00182 | 1 | 0.75 | 0.75 |
| hsa-miR-93-5p | −1.59 | 0.00613 | 1 | 0.75 | 0.75 |
Table 3Circulating plasma miRNAs related to liver diseases
| miRNA_ID | Reported in patients’ plasma (Ref) | Reported in liver disease (Ref) |
|---|
| hsa-let-7a-5p | Chronic Hepatitis C31 | Hepatic Fibrosis32 |
| hsa-let-7f-5p | Hepatocellular Carcinoma33 | Hepatocellular Carcinoma34 |
| hsa-miR-15-5p | Hepatocellular Carcinoma35 | Liver Cancer36 |
| hsa-miR-233-3p | Acute-on-chronic liver failure37 | Hepatocellular Carcinoma38 |
| hsa-miR-23a-3p | NAFLD39 | NAFLD40 |
| hsa-miR-374a-3p | HBV-Liver fibrosis41 | Hepatitis B-related acute-on-chronic liver failure42 |
Diagnostic efficacy comparison of miRNA panel candidates in AATD-associated liver disease
To further verify miRNA combination panels for AATD-associated liver disease diagnosis, we performed a diagnostic performance analysis in our independent validation cohort containing 45 AATD individuals. To develop a classification method, we used Fisher’s linear discriminant analysis and designed comprehensive classifiers consisting of three and four miRNAs in the training set (Supplementary Tables 2 and 3). The most significant results of the three-miRNA combinations are shown in Figure 5A (AUC = 0.73), and the results from four-miRNA combinations are demonstrated in Figure 5B (AUC = 0.75). The performance of miRNA panels in the testing cohort (AUC = 0.95–1) is higher for AATD-associated liver disease diagnosis than in the validation cohort.
Discussion
Here, we conducted a comprehensive analysis of plasma EV-associated miRNAs in AATD individuals, comparing them to non-AATD controls.43 We also focused on comparing miRNA profiles from AATD individuals with and without liver disease. We identified 6 miRNAs as potential molecular identifiers for the detection of AATD liver disease. We suggest that panels consisting of three and four miRNAs could offer potential biomarkers for assessing the severity of AATD liver disease.
Our results revealed that among the DEMs associated with AATD, miR-25-3p, miR-96-5p, and miR-30a-5p have been previously reported as HCC biomarkers.44–51 Consistent with previous literature, our results also suggest that AATD liver disease is a risk factor for HCC.7 This suggests that plasma EV-associated miRNAs can serve as molecular signatures of HCC in AATD individuals. Studies have shown high extracellular levels of miR-99-3p,52 and low levels of miR-1653 are linked to hepatic fibrosis. Given that liver fibrosis is found in 35% of AATD individuals,7 changes observed in AATD plasma EVs may represent a liver fibrosis diagnostic option in these patients. Furthermore, our in vivo experiments showed that plasma EVs derived from AATD patients enhanced CCl4-induced liver fibrogenesis compared to control EVs, supporting the idea that AATD EVs may contribute to the promotion of fibrosis. This suggests that the miRNAs carried within these EVs likely play a role in liver disease progression by influencing fibrogenesis.
Network analysis of plasma EV-associated DEMs in AATD suggested an increase in the activity of different causal networks, including FOXP3, which is essential for immune cell function. Activated hepatic stellate cells in the fibrotic liver induce naive T cells to differentiate into Foxp3(+) cells.54 While activation of hepatic stellate cells is the hallmark of liver fibrosis,55 this finding suggests that FOXP3 activation may be downstream of hepatic stellate cell activation in the AATD liver. IPA also predicted inhibition of integrin β4 (ITGB4) in the plasma EVs of AATD individuals (Supplementary Fig. 2A). While ITGB4 can promote liver fibrogenesis,56 inhibition of ITGB4 may be a compensatory mechanism against AATD liver fibrosis.
AATD patients with positive liver PASD scores and liver fibrosis show an important pathological overlap, hampering correct diagnosis and management.7 The overlapping miRNA signatures in our dataset suggest that PASD and fibrosis phenotypes share the same miRNA profiles, which points to a common injury process within the AATD liver. Our data also suggest common molecular pathways that may be relevant to other chronic liver diseases. This expands the potential clinical application of these biomarkers beyond AATD, opening possibilities for their use in other liver conditions. Among these overlapping miRNAs, we observed that the hsa-let-7 family represents 48% of miRNAs identified (Fig. 2B). While let-7a-5p was predicted to negatively correlate with collagen expression, previous studies have shown the inhibitory role of let-7a in liver fibrogenesis.57 This suggests that the upregulation of let-7a-5p may be an anti-fibrotic defense mechanism in the AATD liver. In terms of translational research, the identified miRNAs provide important mechanistic insights into AATD-related liver disease. For example, miR-23a-3p, associated with endoplasmic reticulum stress,58 mitochondrial dysfunction,59 and lipid accumulation,60 offers new avenues for understanding the pathophysiology of AATD liver disease. These mechanistic insights could be harnessed to develop targeted therapies aimed at addressing the underlying causes of liver injury in AATD.
Conclusions
Here, we show unique properties of six plasma EV-associated miRNAs, and their combinations offer a valuable tool for early detection and risk stratification. With high diagnostic accuracy (AUC > 0.93 in training batches and AUC > 0.75 in validation batches), our designed panels may enable clinicians to detect liver disease in AATD patients early and potentially intervene before severe complications, such as cirrhosis or HCC. The use of these miRNAs as non-invasive disease signatures is especially advantageous for patients, as it reduces the need for invasive liver biopsies. This facilitates regular monitoring of liver health in AATD individuals, particularly those who may not yet show clinical symptoms of liver disease. Our results have significant clinical and translational implications, particularly in improving the detection and management of liver disease in AATD individuals. Moreover, the study suggests that these miRNA panels could be used to monitor therapeutic responses to anti-fibrotic treatments. This could accelerate the development of new therapies by providing a quicker and less invasive way to assess treatment efficacy in clinical trials.
Finally, this study contributes to the growing field of precision medicine by identifying miRNA signatures that can help predict disease severity and progression in AATD patients. These disease molecular signatures could facilitate personalized treatment plans based on individual risk profiles, improving patient outcomes and reducing unnecessary interventions for those with a lower likelihood of developing advanced liver disease.
While our findings point to the potential use of plasma EV miRNAs as molecular identifiers for liver disease in AATD individuals, it’s important to note that further research on plasma-derived free miRNAs is needed to confirm the effectiveness of plasma EV-associated miRNAs. We acknowledge that our study has several limitations, such as the small number of individuals included, which may impact the robustness of our model. However, it is important to emphasize that AATD is a rare condition, with a frequency of approximately one in 1,800 individuals, presenting significant challenges in terms of participant recruitment. Many AATD individuals remain undiagnosed or misdiagnosed, as the condition can mimic more common respiratory diseases like COPD, reducing the number of available patients for study. Despite this, we believe that the insights gained from our study contribute meaningfully to the understanding of this under-researched disease. We also acknowledge that out of the 20 randomly chosen miRNAs to be validated by qPCR, only 11 were successfully validated due to technical constraints, including issues with qPCR primer design and the limited expression levels of some miRNAs. Furthermore, our study uses cross-sectional data, which restricts our ability to evaluate changes in miRNA expression over time. Longitudinal studies would be valuable in addressing this gap, allowing for the assessment of temporal changes in miRNA expression. Future research incorporating longitudinal data is needed to understand temporal dynamics and their implications for disease management. Further studies in other chronic liver diseases can also be performed to evaluate the specificity of miRNAs for liver injuries.
Supporting information
Supplementary Table 1
Raw reads table.
(XLSX)
Supplementary Table 2
ROC cobmine3.
(XLSX)
Supplementary Table 3
ROC cobmine4.
(XLSX)
Supplementary Fig. 1
Top 10 differentially expressed miRNAs in plasma EVs from AATD individuals.
(A) Bar graph indicating upregulated and downregulated differentially expressed miRNAs from our plasma EV fractions miRNA dataset. (B) qPCR results from upregulated and (C) downregulated differentially expressed miRNAs from our plasma EV-derived miRNA dataset. (D) Cord plot of ingenuity target analysis indicating target genes involved in hepatic stellate cells’ proliferation and fibrogenesis. Red indicates upregulated miRNAs, and green indicates downregulated miRNAs from our dataset. Orange indicates upregulation, and blue indicates downregulation of target genes, as predicted by IPA software. EV, extracellular vesicle; AATD, alpha-1 antitrypsin deficiency; IPA, Ingenuity Pathway Analysis.
(TIF)
Supplementary Fig. 2
Pathway identification and functional enrichment analysis of DEMs.
(A) Causal networks revealing causal relationships for the 178 DEMs identified in our plasma EV fractions of AATD individuals compared to healthy controls, as predicted by IPA. (B) Percent abundance of the DEM families in our plasma EV-derived miRNA dataset. EV, extracellular vesicle; AATD, alpha-1 antitrypsin deficiency; IPA, Ingenuity Pathway Analysis; DEMs, differentially expressed miRNAs.
(TIF)
Declarations
Acknowledgement
We would like to thank Ashish Sharma, Shannon Holiday, and Asha Rani for their advice and suggestions on this work, as well as members of our laboratories for their contributions to various aspects of our research. The ultrastructural studies were conducted at the University of Florida College of Medicine Electron Microscopy Core Facility under the kind supervision of Dr. Jill Verlander, with the assistance of Dr. Sharon Matthews. Small RNA sequencing and data analysis were performed at the Interdisciplinary Center for Biotechnology Research in the University of Florida’s premier core research facility under the supervision of Drs. Yanping Zhang and Tongjun Gu.
Ethical statement
The study protocol was approved by the Clinical Research Ethics Committee of the University of Florida (IRB202101148), and written informed consent was obtained from all subjects in accordance with the Declaration of Helsinki Principles.
Data sharing statement
Data and code used to generate the miRNA-seq associated plots are hosted at the following repository: https://github.com/zfg2013/EV-miRNA-and-AATD/tree/update. All correspondence regarding the data & code used should be sent to Nazli Khodayari ([email protected]) or Zachary Greenberg ([email protected]).
Funding
This work was supported by a grant from the Alpha One Foundation (AGR00019116). Dr. Zhou is the recipient of a Research Career Scientist Award from the Department of Veterans Affairs (IK6BX004477).
Conflict of interest
The authors have no conflict of interests related to this publication.
Authors’ contributions
Acquisition of data, technical support, review & editing (RO), data curation, formal analysis, review & editing (ZH), formal analysis, review & editing (ZG), formal analysis, review & editing (SD), acquisition of data (VC), formal analysis, review & editing (HZ), acquisition of data (JW), writing, review & editing (MH), conceptualization, supervision (MB), conceptualization, funding acquisition, formal analysis, data curation, supervision, writing the original draft, and review & editing (NK). All authors have approved the final version and publication of the manuscript.