We interrogated the 154 BA-associated SNPs against regulatory annotations from both ENCODE and HaploReg databases. SNPs overlapping histone marks, DNase I hypersensitive sites, protein binding regions, or transcription factor motifs were predicted to reside within cis-regulatory elements (CREs). We performed systematic screening of the SEdb 3.0 database to identify super enhancers that co-localize with experimentally validated enhancer-active CREs.

Human H69 cholangiocytes were cultured at 37 °C with 5% CO₂ in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% fetal bovine serum (Gibco, California, United States) and 100 IU/mL penicillin (Gibco, California, United States).

DNA fragments flanking the 28 candidate SNPs were cloned into the pGL4.23 luciferase vector (Promega, United States). H69 cells in 24-well plates (Corning, United States) were co-transfected with either the pGL4.23 constructs or the empty pGL4.23 vector, along with the Renilla luciferase control vector, using Lipofectamine 8000 (Beyotime, Shanghai, China). Twenty-four hours post-transfection, luminescence was measured with the Dual-Luciferase Reporter Assay System (Yeasen, Shanghai, China) on a Synergy™ H1 Hybrid Multimode Microplate Reader (BioTek, United States). Firefly luciferase activity was normalized to Renilla luciferase activity for data analysis.

All zebrafish experiments were performed using wild-type AB-strain reared at Xinhua Hospital. Procedures were conducted in accordance with institutional guidelines and approved by the Xinhua Hospital Animal Care and Use Committee (XHEC-WSJSW-2018-029).

Total RNA was extracted from 5 h post-fertilization (hpf) embryos using TRIzol reagent (Invitrogen, United States) in accordance with the manufacturer’s instructions. RT-PCR was performed using SYBR Green Master Mix with fluorescent labeling (Applied Biosystems, Cat#A25742, United States) on a QuantStudio Dx Real-Time PCR Instrument (Applied Biosystems, CA, United States). The 18S ribosomal RNA (18S rRNA) gene served as the normalization control. All assays were conducted in triplicate. Relative gene expression levels were calculated using the ΔΔ^CT method, expressed as RQ values (RQ = 2^−ΔΔCT). Primer sequences used for RT-PCR are listed in (Supplementary Table S1).

First-strand cDNA was then synthesized from 1 μg total RNA using PrimeScript RT Master Mix (Takara, Japan). PCR primers were designed to generate a 456-bp antisense riboprobe targeting zebrafish add3a (Supplementary Table S2). The PCR products were subsequently cloned into the pGEM-T Easy Vector (Promega, WI, United States). Digoxigenin-labeled antisense and sense RNA probes were then synthesized using the linearized plasmid templates (Roche Applied Science, Penzberg, Germany).

At 24 hpf, larvae were maintained in 0.003% 1-phenyl-2-thiourea (PTU; Sigma-Aldrich) to inhibit melanin formation. For fixation, zebrafish embryos were incubated overnight at 4 °C in 4% paraformaldehyde (Sangon, Shanghai, China), followed by dehydration and storage in 100% methanol (MeOH) at −20 °C for ≥2 h prior to processing. The chorions of larvae were manually removed under a stereomicroscope using fine forceps. Embryos >1 dpf were permeabilized with proteinase K (10 μg/mL) for 15–30 min (RT) to facilitate probe penetration.

Following prehybridization (2–4 h, 60 °C), embryos were hybridized overnight with 600 ng antisense probe at 60 °C. After blocking (3 h, RT), samples were incubated overnight at 4 °C with anti-DIG-AP antibody (1:5000; Roche). Signal detection was performed using BM Purple AP substrate (Roche) with subsequent imaging on a Nikon SMZ25 stereomicroscope (Chiyoda, Japan).

Morpholino antisense oligonucleotides (MOs) targeting add3a were designed and synthesized by GeneTools, LLC. A standard MO was used as the negative control. Translation-blocking (TMO) and splice-blocking (SMO) were designed to inhibit the 5′translational start site and a splice acceptor site respectively (Supplementary Table S3). Knockdown of add3a were mediated by injecting SMO and TMO into the yolk of 1 to 4-cell stage embryos. MOs were diluted to 0.25 mM in sterile double distilled water and approximate 2 nL/embryo MO was injected. To validate the efficiency of add3a splice-blocking morpholino (SMO), zebrafish embryos at 3 dpf were collected after microinjection. Total RNA was extracted, reverse-transcribed into cDNA, and analyzed by RT-PCR followed by gel electrophoresis to detect aberrant splicing products indicative of successful knockdown. To validate TMO efficacy, we PCR-amplified a 118-bp fragment encompassing the TMO target sequence from zebrafish cDNA and subcloned it into the pEGFP-N1 plasmid. Fluorescence attenuation following co-injection of TMO with this reporter construct confirmed successful translational inhibition. Primers are listed in Supplementary Table S4.

Using cDNA from 5 dpf zebrafish eggs as the template, the full-length coding sequence of add3a was amplified by RT-PCR with the high-fidelity enzyme. Primers are listed in Supplementary Table S5. The full-length add3a fragment was purified by gel extraction. Homologous sequences corresponding to the linearized vector ends and restriction enzyme sites were added to the 5′ends of the primers. The pCS2+ empty vector was linearized by digestion with EcoR I. The purified PCR product was then cloned into the linearized pCS2+ vector in the SP6 forward orientation using the Hieff Clone Plus One Step Cloning Kit (YEASEN, China). The plasmid containing add3a was linearized with KpnI and used as a template for in vitro transcription to generate full-length mRNA. Capped mRNA was transcribed using the mMESSAGE mMACHINE Kit with SP6 polymerase (Thermo, United States). Embryonic microinjection 100 pg add3a mRNA was performed into one-to four-cell embryos.

PED6 is a fluorogenic substrate for phospholipase A2. It is metabolized in the liver and excreted into bile, accumulating in the gallbladder. To directly visualize the gallbladder, 5 dpf zebrafish embryos were incubated with 0.1 mg/mL PED-6 (Invitrogen, United States) for over 2 hours. Images from the PED-6 assay were acquired using a stereo fluorescence microscope (Leica, German) (So et al., 2018). Embryos in each group were categorized as “normal”, “faint”, or “absent” based on gallbladder size and the intensity of PED-6 fluorescence (Cui et al., 2013).

Zebrafish larvae at 5 dpf were fixed in methanol:DMSO (4:1) for 2 h at room temperature, followed by post-fixation in 100% ice-cold methanol for long-term storage. After rehydration through a graded methanol series in PBSX (PBS containing 0.5% Triton X-100), the larval epidermis was carefully removed to enhance antibody penetration. Samples were blocked for 1 h in 10% bovine serum albumin (BSA) with 0.5% Triton X-100 at room temperature.

Primary immunostaining was performed using mouse anti-Keratin 18 monoclonal antibody (Ks18.04; PROGEN, Cat#61028; 1:100 dilution in blocking buffer) overnight at 4 °C with gentle agitation. After extensive washing with PBSX, specimens were incubated with Alexa Fluor 488-conjugated goat anti-mouse IgG secondary antibody (YEASEN, Cat#33206ES60; 1:500 dilution) overnight at 4 °C.

Immunostained larvae were mounted in glycerol and imaged using a Leica TCS SP8 confocal microscope (Leica Microsystems, Germany). Images were processed using Adobe Photoshop CS6, and quantitative morphometric analyses (intrahepatic duct length and gallbladder area) were performed using ImageJ software (NIH).

Statistical analyses were performed using GraphPad Prism software. The data presented as mean ± SEM. Normality was assessed using the Shapiro-Wilk or Kolmogorov-Smirnov test (as appropriate), followed by Levene’s test for homogeneity of variance when parametric assumptions were met. Normally distributed data were analyzed using either a two-tailed Student’s t-test (for two-group comparisons) or one-way ANOVA (for multiple groups). The categorical data were analyzed using the χ² test. A two-sided P-value < 0.05 was considered statistically significant for all analyses.

In a GWAS of BA susceptibility in Chinese individuals, we identified 154 SNPs reaching genome-wide significance (P < 5 × 10⁻⁸), including rs17095355 as the lead variant. All of these SNPs lie in the 5′upstream intergenic region or introns of ADD3 and are in high LD with rs17095355 (r² > 0.9). eQTL analysis revealed that the risk allele of rs17095355 correlated with increased ADD3 expression. To identify causal variants among the 154 BA-associated SNPs, we screened for SNPs located within cis-regulatory elements (CREs) by querying data from ENCODE and HaploReg. This analysis revealed 28 SNPs across 10 distinct CREs (Figure 1), with spatially clustered SNPs co-localizing within the same CREs (Figure 1A). Among the 28 associated SNPs, 20 were located in regions containing histone marks characteristic for enhancers, 19 resided in DNaseI hypersensitive sites, seven overlapped transcription factor binding sites, and 24 were predicted to alter transcription factor motifs (Supplementary Tables S6, S7).

We conducted in vitro functional tests of enhancer activity for the 10 CREs harboring 28 susceptibility SNPs. We cloned DNA fragments centered on the variants into the PGL4.23 luciferase vector and transfected them into Human H69 cholangiocytes (Table 1; Figure 1A). Totally, 10 elements were tested (Figure 1B; Table 1). Our results revealed that eight elements (E1, E2, E3, E4, E7, E8, E9 and E10) demonstrated significant enhancer activity (P < 0.0001 and > 2-fold increase in reporter activity compared to promoter-only control vector) and four elements (E2, E4, E9 and E10) showed allele-specific reporter activity differences (Figure 1C). Among the latter, 75% of risk haplotypes (E2, E4 and E10) exhibited higher enhancer activity than their non-risk counterparts (Figure 1D). Although the E1 region containing lead SNP rs17095355 showed strong enhancer activity in our assays, the similar activity levels between risk and non-risk haplotypes (P = 0.13) imply that this SNP might not be the functional variant responsible for the observed genetic association.

Super-enhancers (SEs) are clusters composed of multiple transcriptional enhancers that coordinately regulate gene expression (Wang et al., 2022). In this region, eight elements exhibiting enhancer activity were identified, suggesting that the locus may function as a SE. To explore this further, we queried SEdb 3.0. The results indicated this region acts as a SE across a broad range of tissues and cell types, including the liver tissue (Supplementary Table S8) (Wang et al., 2022).

At E2, the risk haplotype CdeletionT (rs9630101-rs35533450-rs9630102) showed 6.1-fold increased luciferase activity compared to the control (basal promoter), and 3.1-fold higher activity compared to the non-risk haplotype (P = 1.9 × 10⁻⁵). At E4, the risk haplotype TA (rs12240333-rs61614979) showed 8.2-fold increased luciferase activity compared to the control (basal promoter) and 2.7-fold higher activity compared to the non-risk haplotype (P = 1.4 × 10⁻⁶). At E10, the risk haplotype TTTT (rs2122517-rs7904096-rs7079713-rs7083619) showed 8.7-fold increased luciferase activity compared to the control (basal promoter) and 2.4-fold higher activity compared to the non-risk haplotype (P = 2.3 × 10⁻⁵). Our previous study found that the risk alleles were associated with increased expression of ADD3, which is overexpressed in BA cholangiocytes and hepatocytes (Cui et al., 2023). Taken together with the present results, these findings suggest that BA-associated risk alleles contributed to BA pathogenesis by upregulating ADD3 expression.

To explore the role of ADD3 dysregulation in hepatobiliary development, we employed zebrafish as a model system. We firstly used RT-PCR and WISH experiments to instigate whether add3a, the ortholog of human ADD3 in zebrafish, was expressed in the developing liver of zebrafish. The qPCR results showed that the expression of add3a was present at 4 hpf, gradually increases until it peaks at 48 hpf, after which its expression slightly decreases and stabilizes (Figure 2A). The WISH results showed that add3a was mainly expressed in the head at 24-96 hpf, and was observed to be expressed in the liver at 72 hpf and 96 hpf (Figure 2B). These observed expression patterns strongly indicate the potential involvement of add3a in the development of hepatobiliary system.

Following our add3a expression analysis, we overexpressed add3a in zebrafish to model the upregulation effects observed in BA livers on hepatobiliary function and structure. Previous study reported that knockdown of add3a produced intrahepatic defects and impaired biliary function; therefore, we also performed a knockdown analysis for comparison (Tang et al., 2016).

We overexpressed add3a by injecting its mRNA into zebrafish embryos to assess its effect on hepatobiliary development. RT-PCR showed a marked upregulation of add3a expression (Figure 3A). PED6 accumulation in the gallbladder was significantly reduced in add3a mRNA–injected larvae relative to controls (Figures 4A,B), implying impaired hepatobiliary function (Cui et al., 2013). We examined the liver and gallbladder by cytokeratin 18 immunofluorescence staining. Biliary abnormalities were evident in the add3a-overexpressing larvae compared with control larvae (Figures 4C,D). Relative to control MO larvae, add3a-overexpressing larvae exhibited reduced intrahepatic duct density and a decreased number of interconnecting ducts and terminal ductules (Figure 4C). Furthermore, the gallbladders were smaller, with significantly fewer cells and a smaller average cell area in the add3a-overexpressing larvae than in controls (Figure 4D).

We performed MO–mediated knockdown to reduce add3a expression, using a SBMO and a TBMO. The SBMO targets the exon 1 splice acceptor site of add3a, while the TBMO targets the translation start region in the first exon (Figure 3B). We confirmed MO efficacy: SBMO injection reduced band brightness on RT-PCR (Figure 3C), and TBMO injection abolished PEGFP-N1 plasmid fluorescence (Figure 3D). Co-injection of SBMO with overexpression constructs yielded higher band intensity than SBMO alone (Figure 3B). Rescue experiments with co-injection of full-length add3a overexpression mRNA alongside SBMO or TBMO partially restored the MO-induced phenotypes (Figure 4B), supporting MO specificity. SBMO- and TBMO-mediated knockdown reduced PED6 accumulation in the gallbladder and impaired intrahepatic ductal development as well as gallbladder morphology (Figures 4A–D).

Overall, our findings indicate that add3a is essential for normal hepatobiliary development in zebrafish. Hepatobiliary development is highly sensitive to add3a levels, as both overexpression and knockdown cause disruption, indicating precise dosage control is essential. Collectively, these in vivo data support the notion that BA-associated risk alleles contribute to BA pathogenesis by upregulating ADD3 expression.