Recapitulating dengue virus infection with human pluripotent stem cell-derived liver organoids for antiviral screening

Categories: Disease & Virus

August 29, 2025

Generation and characterization of hPSC-derived liver organoids

We used human induced pluripotent stem cell (hiPSC) to generate human liver organoid (hPLO), by mimicking embryonic liver development through a stepwise differentiation process (Fig. 1a), modified based on previous protocols^38,39,40,41. These cells sequentially differentiated to produce definitive endoderm (DE) and hepatic specification (HS) lineages under monolayer cultures (Supplementary Fig. 1a–b). Subsequently, HS cells were embedded in Matrigel to generate organoids through the three-dimensional (3D) cultures within 10 days (Fig. 1a and Supplementary Fig. 1a). At this stage, expandable hepatic organoids could be passaged for further large-scale cultivation and cryopreservation, or alternatively, continued differentiation for another 15 days to form more mature hepatic organoids (Fig. 1a). The morphology of the 3D cultures evolved from an initial hollow spherical structure, measuring 20-30 μm on the second day of culture, to 600–800 μm after 8–10 days of rapid proliferation (Fig. 1b). Then, the hepatic organoids gradually transitioned from a hollow spherical structure to a solid aggregate within 15 days of maturation (Fig. 1c and Supplementary Fig. 1c). These hPLOs could be maintained in culture for at least 100 days (Fig. 1c).

**Fig. 1: Generation and characterization of hPLOs.**

In day 25 hPLOs, hepatic markers, such as AFP, ALB, HNF4A and A1AT, were detected by immunofluorescence (Fig. 1d and Supplementary Fig. 1d). The presence of KI67 indicated cellular proliferation (Fig. 1e). Additionally, cholangiocyte-related markers (KRT19 and KRT7) were found in the hPLOs, suggesting the presence of various hepatic cell types beyond hepatocytes (Fig. 1e). In day 50 hPLOs, ALB, HNF4A, and other mature hepatocyte markers like TF were observed (Fig. 1f and Supplementary Fig. 1e).

RNA-seq analysis was performed to reveal the gene expression profiles during the formation of hPLOs (Fig. 1g–k). We observed robust expression of stem cell markers (OCT4, SOX2, NANOG) at the pluripotent stage (day 0) (Fig. 1g). As differentiation progressed, definitive endoderm markers such as SOX17, CXCR4, hepatic endoderm marker FOXA2 and hepatic progenitor gene TBX3 were expressed sequentially, but their levels decreased in hPLOs (Fig. 1g and Supplementary Fig. 1f–g). In hPLOs, hepatic makers like HNF4A, ALB, AFP, A1AT, TTR, along with liver function-related genes including ALDOB, APOA1, APOC3, NPC1L1, BMP1, AKR1C1, CYP3A4, CYP3A5, CYP3A7, CYP2C9, COMT, UGT1A1, UGT2B15, UGT2B7, UGT3A1, UGT2A3, PROX1, MAOB, were all expressed (Figs. 1g–k and Supplementary Fig. 1h –1k). Interestingly, we observed the expression of the mesodermal gene MESP1 expressed on Day 10 (Fig. 1i). Furthermore, expression of the mesenchyme-associated gene PDGFRB during the 3D culture stage suggested the spontaneous emergence of hepatic stellate cells, potentially arising from the MESP1-positive population⁴² (Figs. 1i, l).

To further validate the differentiation efficiency of our protocol, two different wild-type human embryonic stem cell lineages (hESCs, H1 and H9) were differentiated using the same methods. These hESC-derived liver organoids displayed similar morphology and marker expression profiles to the hiPSC-derived liver organoids (Fig. 1m and Supplementary Fig. 1l).

To further investigate the cell-type composition, we performed 10x Genomics single-cell RNA sequencing (scRNA-seq) on hPLOs. Uniform Manifold Approximation and Projection (UMAP) analysis identified 4 major cell clusters after the stringent exclusion of low-quality cells (Fig. 1n). These cells include hepatocyte-like cells which expressed hepatic marker SERPINA1, ALB, APOA2, TTR, AMBP, APOB, RBP4, HNF4A, AFP; They also include proliferating hepatocyte-like cells, which expressed MKI67, PCNA, TOP2A, ONECUT2, FOXA2 along with hepatic markers (Fig. 1o and Supplementary Fig. 2a). These hepatic marker-positive cells exhibit significant proliferative activity and are consistent with previous findings^39,43; Additionally, cholangiocyte-like cells are present in the hPLOs, which expressed KRT19, KRT7, KRT18, KRT8, KRT17 (Fig. 1o and Supplementary Fig. 2a), and stellate-like cells were characterized by DCN, COL3A1, SPARC, COL21A1, MFAP4, PDGFRβ (Fig. 1o and Supplementary Fig. 2a). Stem cell and definitive endoderm markers, including NANOG, POU5F1, SOX2, and SOX17, were largely absent or weakly expressed in hPLOs (Supplementary Fig. 2b).

Functionality characterization of hPLOs

Subsequently, we sought to determine whether the prepared hPLO is functional. Using immunofluorescence staining and transmission electron microscopy (TEM), we identified tight junctions (TJ) between adjacent cholangiocytes near the apical membrane, along with abundant microvilli (MVs) on the apical surface, both of which are indicative of intrahepatic bile duct formation (Fig. 2a, b). Cholangiocytes, that form the intrahepatic bile ducts in the liver, are also involved in regulating bile composition via the excretion of ions, solutes, and water^44,45. To assess secretory and transport functions in Day 25 hPLOs, 2’,7’-dichlorofluorescin diacetate (CDFDA) staining was performed to visualize bile canaliculi formation, indicative of MRP-2 transporter activity (Fig. 2c)^39,46. Additionally, we further evaluated the active transport capability of Day 25 hPLOs using Rhodamine 123, a dye commonly used to assess multidrug resistance protein 1 (MDR1) transporter activity in normal cholangiocytes. The dye was transported into the luminal space, and its absence in the lumen when treated with Verapamil (an MDR1 inhibitor) confirmed the active transport capability of our hPLOs (Fig. 2d).

The functionality of hPLOs was further characterized by the uptake and release of indocyanine green (ICG), a dynamic assessment of liver function based on clearance parameters. As shown in Fig. 2e, ICG was completely released at 7 h and 14 h. The hPLOs also demonstrated glycogen storage capacity, as confirmed by periodic acid-Schiff (PAS) staining (Fig. 2f). Finally, we assessed the secretory and metabolic functions of the organoids and compared them with those of functional human liver organoids and primary human hepatocytes (PHHs) reported in previous studies^39,47, and found that the hPLOs exhibited comparable functions. Specifically, we observed that both Day 25 and Day 50 hPLOs secreted human albumin (ALB), human α1-antitrypsin (AAT) and produced urea (Fig. 2g–i). These two stages of hPLOs exhibited metabolic activity, as indicated by the activity of phase I enzyme (CYP3A4, CYP2C9) (Fig. 2j–k). Overall, our results clearly demonstrate that the established hPLOs exhibit numerous characteristics of liver function. Therefore, we have successfully established a platform for efficient production of hPLOs.

hPLOs support productive DENV-2 infection

To leverage hPLOs for DENV research, Day 25 or Day 50 hPLOs were inoculated with DENV-2 (Fig. 3a). Remarkably, the hPLOs (Day 25) were highly susceptible to DENV-2 infection. DENV-2 infection led to dramatic morphological changes since day 2 post infection, and healthy hPLOs in hollow and transparent appearance transformed into small, solid spheres (Fig. 3b, c and Supplementary Fig. 3a, b), and over 90% of the hPLOs lost their healthy appearance at day 6 post infection (Fig. 3c). As expected, MOCK treatment with inactivated DENV-2 showed no obvious impact on the appearance and status of hPLOs (Fig. 3b, c). Live cell imaging documented this entire morphological transformation, showing the gradual shrinking of DENV-2 infected hPLOs (Day 25) from hollow, vesicle-like structures to solid spheres (Supplementary Movie 1). Live/Dead staining with calcein-AM (AM) and ethidium homodimer-1 (EthD-1) showed that DENV-2 infection led to cell death (red) at day 4 post infection, while all organoids in MOCK group remain viable (green) (Fig. 3d and Supplementary Fig. 3b).

Cell death was observed in DENV-2 infected Day 50 hPLOs as well, but to a lesser extent, with no significant morphological changes detected after infection (Supplementary Fig. 3c). This could be attributed to the denser structure of Day 50 hPLOs, which may limit the extent of morphological alterations. Additionally, results from hPLOs derived from H1 and H9 pluripotent stem cell lines confirmed that DENV-2 infection caused dramatic morphological transitions in Day 25 hPLOs (Supplementary Fig. 3d, e) but not in Day 50 hPLOs (Supplementary Fig. 3f, g). Therefore, Day 25 hPLOs were used for subsequent virological assays.

To clarify the replication kinetics of DENV-2 in Day 25 hPLOs, culture supernatants were collected at the indicated time points and subjected to RT-qPCR and plaque forming assays. As expected, rapid increase of DENV-2 genomic RNA and infectious particles was detected from day 2 to 8 post infection (Fig. 3e, f). More importantly, DENV-2 antigens were widely distributed in the infected organoids at day 6 post infection (Fig. 3g). Additionally, DENV-2 particles were visualized using TEM (Supplementary Fig. 4a). Together, all these results demonstrate that DENV-2 could efficiently infect and replicate in hPLOs and lead to cell death and morphological changes.

Furthermore, DENV-2 antigens were detected in association with markers of distinct cell types, including HNF4A-positive hepatocyte-like cells (Fig. 3h and Supplementary Fig. 4b), KRT7-positive cholangiocyte-like cells (Fig. 3i), and proliferative cells marked by KI67 (Fig. 3j and Supplementary Fig. 4b). Collectively, these data suggest that hPLOs support the complete life cycle of DENV-2 infection.

Single-cell transcriptome profiles of DENV-2 infected hPLOs

To investigate DENV-2 infection in hPLOs at the single cell level, 10x Genomics scRNA-seq was performed on MOCK and DENV-2 infected hPLOs (Day 25) at day 1, 2 and 8 post infection. Clustering of 59,029 high-quality cells revealed four major cell clusters annotated using well-established cell-type-specific markers (Fig. 4a and Supplementary Fig. 5a, b). UMAP visualization revealed marked differences in cell distribution between the MOCK and DENV-2 infected groups (Fig. 4b and Supplementary Fig. 5c). The cell density heatmap showed that alterations in cell distribution began to emerge by day 2 following DENV-2 infection and became more pronounced by day 8 (Fig. 4c and Supplementary Fig. 5d).

We next analyzed the relative proportions of the four major cell types at all three time points (D1, D2, D8) in both MOCK and DENV-2 infected hPLOs (Fig. 4d–g and Supplementary Fig. 5e). The relative proportions of stellate-like and cholangiocyte-like cells increased, while those of hepatocyte-like and proliferating hepatocyte-like cells decreased in DENV-2 infected hPLOs compared with the MOCK group (Fig. 4d). Specifically, at day 1 post DENV-2 infection, the relative proportions of each cell type exhibited minor changes ( <2% variation) compared to the MOCK group (Fig. 4e and Supplementary Fig. 5e). This aligns with earlier findings showing minimal phenotypic changes in hPLOs at day 1 post DENV-2 infection (Fig. 3c and Supplementary Fig. 3b). As DENV-2 infection progressed, the relative proportion of proliferating hepatocyte-like cells decreased by ~6% on day 2, up to 10% on day 8 (Fig. 4f, g and Supplementary Fig. 5e). Interestingly, hepatocyte-like cells showed almost no change during the first two days but exhibited an ~5% decline on day 8 (Fig. 4f–g and Supplementary Fig. 5e). In contrast, the proportions of cholangiocyte-like and stellate-like cells increased by about 5% and 10%, respectively, on day 8, indicating greater tolerance or resistant to DENV-2 infection (Fig. 4f, g and Supplementary Fig. 5e).

We further examined the distribution and expression levels of DENV-2 across different cell types. DENV-2 could be detected in all four cell compositions (Fig. 4h-i and Supplementary Fig. 5f), with expression levels increasing as the infection progressed (Fig. 4i and Supplementary Fig. 5g–h). Quantification of DENV-2 positive rates across cell types revealed that the relative proportions of hepatocyte-like cells and proliferating hepatocyte-like cells positive for DENV-2 were ~1% on Day 1, both increasing to ~45% by Day 8. Similarly, the proportions of DENV-2 positive cholangiocyte-like cells and stellate-like cells were around 0.5% on Day 1 and rose to over 35% by Day 8 (Fig. 4j). DENV-2 positive cells were also detected among different hepatic lineage cell types by flow cytometric analysis (Supplementary Fig. 5i, j). These findings suggest that DENV-2 infection directly drives dynamic changes in the proportions of four major cell clusters, with proliferating hepatocyte-like cells being more affected by DENV-2 infection. In contrast, cholangiocyte-like and stellate-like cells are less affected by DENV-2 infection.

Exploring the pathogenesis in hPLOs during DENV-2 infection

To further investigate the pathogenesis of DENV-2 infection in hPLOs, we firstly focused on the differentially expressed genes (DEGs) in DENV-2 infected hepatocyte-like cells across different time points, given the high DENV-2 expression levels (Fig. 4i). A comparison of DEGs between DENV-2 D8 vs D2 and DENV-2 D2 vs D1 revealed that transcriptomic changes were more pronounced by day 8 post-infection (Fig. 5a). Genes that were upregulated in both comparisons were classified as “Both-UP,” indicating a consistent increase in expression from day 1 to day 8 post DENV-2 infection. Conversely, genes downregulated in both comparisons were designated as “Both-DOWN”, reflecting a continuous decrease over the same period. H19, associated with oxidative stress and pathological conditions^48,49, was identified as a Both-UP gene. Conversely, C1QBP and GPX1 associated with mitochondrial function^50,51 and significant antioxidant function^52,53, were identified as Both-DOWN genes, signifying their downregulation during the same period (Fig. 5a). We also found that several mitochondrial genes, including MT-CO1, MT-CO2, MT-CO3, MT-CYB, MT-ND3, and MT-ATP6, as well as hepatic genes such as APOA2, ALB, AFP, and TTR, showed decreased expression on day 8 post infection compared with day 2 (Fig. 5a). These findings indicated that mitochondria dysfunction may start as early as the second day of DENV-2 infection in hepatocyte-like cells, and by day 8, massive transcriptome changes were observed.

**Fig. 5: DENV-2 pathogenesis in hPLOs.**

To further explore the pathogenesis in other cell types after DENV-2 infection, we analyzed the DEGs at day 8, as day 8 presented the highest level of DENV-2 and the most significant changes in cell type compositions (Fig. 4e–i). DENV-2 RNA was found in all cell clusters as upregulated DEGs (Fig. 5b, c and Supplementary Fig. 6a). In proliferating hepatocyte-like cells, the upregulated genes were enriched in virus infection-related, apoptotic process, response to oxidative stress and defense response signaling pathways (Fig. 5d). For example, PLIN2, one of the top 10 upregulated genes (Fig. 5b), has been reported to play a role in host defense against viral infection⁵⁴, and NEAT1 (Fig. 5b), a long noncoding RNA, has been implicated in liver-related disorders such as acute liver failure, liver fibrosis, and liver carcinoma^55,56,57,58. NEAT1 is also known to be related to sense mito-stressors, which indicated that DENV-2 may cause mitochondrial dysfunction⁵⁹. Indeed, the downregulated genes in proliferating hepatocyte-like cells were mainly enriched in mitochondrial-related signaling pathways (Fig. 5d), suggesting mitochondria dysfunction, which could inhibit proliferation or trigger apoptosis. In additional, of the ten most downregulated genes, four are mitochondrial genes (MT-CO2, MT-CO3, MT-CYB, MT-ND3) (Fig. 5b), which are crucial for the mitochondrial respiratory chain⁶⁰. Similarly, in DENV-2 infected hepatocyte-like cells, the relatively significantly downregulated genes also included mitochondrial genes, and enriched in pathways related to viral infection, apoptotic processes, defense responses, and mitochondrial signaling (Fig. 5a and Supplementary Fig. 6b). In contrast, DENV-2 infected cholangiocyte-like cells and stellate-like cells showed downregulation of genes related to liver function, such as APOA2, CRABP1 (Fig. 5c and Supplementary Fig. 6a). The upregulated genes were enriched in pathways related to antiviral defense and immune responses (Supplementary Fig. 6c, d). The most upregulated gene, IFI6, ISG15 have been found to exhibit antiviral activity against virus replication^61,62,63 (Fig. 5c and Supplementary Fig. 6a).

We conducted an unbiased Gene Set Variation Analysis (GSVA) of MOCK and DENV-2 infected samples and found that immune response pathways were more active in cholangiocyte-like cells and stellate-like cells on day 8 post infection (Supplementary Fig. 6e). Furthermore, the DEGs networks of the four cell type compositions were visualized (Fig. 5e-f and Supplementary Fig. 7a-e). In cholangiocyte-like cells group, we identified a subgroup of upregulated DEGs network as Module 1 (Fig. 5e), which was related to the immune response, including genes such as MX1⁶⁴, ISG15⁶³, and STAT1^65,66, which are associated with antiviral responses. Notably, these genes were absent in proliferating hepatocyte-like cells (Fig. 5f).

We then analyzed the mitochondrial function-related and immune response-related scores across four distinct cell types. The scores for inflammatory response, interferon-alpha response, and interferon-gamma response were significantly higher in cholangiocyte-like and stellate-like cells compared to proliferating hepatocyte-like and hepatocyte-like cells on day 8 post DENV-2 infection (Fig. 5g–i). However, the scores for mitochondrial fragmentation involved in apoptotic process, apoptotic mitochondrial changes, were higher in proliferating hepatocyte-like and hepatocyte-like cells compared to cholangiocyte-like and stellate-like cells on day 8 post DENV-2 infection (Fig. 5j-k). In contrast, the scores for mitochondrial electron transport cytochrome c to oxygen were lowest in proliferating hepatocyte-like and hepatocyte-like cells on the same day (Fig. 5l). At the same time, we observed elongated and swollen mitochondria by TEM following viral infection (Supplementary Fig. 8a). MitoTracker staining showed a diffuse, smeared signal pattern in DENV-2 infected hPLOs (Supplementary Fig. 8b). Together, these phenomena may indicate altered mitochondrial function.

Pseudotime analysis was performed on MOCK and DENV-2 infected samples, and the results also confirmed that differences emerged on day 2 post DENV-2 infection and became more pronounced by day 8 as time progressed (Supplementary Fig. 9a). DENV-2 RNA was also detected in all four cell types as demonstrated by pseudotime analysis (Supplementary Fig. 9b). In addition, within the DEGs networks, mitochondrial-related genes ATP5F1E and MT-CO1, as well as immune response-related genes IFITM3 and HLA-C, exhibited a decreasing trend in proliferating hepatocyte-like and hepatocyte-like cells compared to cholangiocyte-like and stellate-like cells (Supplementary Fig. 9c–f).

Collectively, our results indicate that mitochondrial dysfunction in proliferating hepatocyte-like cells and hepatocyte-like cells contributes to the observed cell death and morphological transitions in hPLOs upon DENV-2 infection. In contrast, antiviral responses in cholangiocyte-like cells and stellate-like cells rendered these cells more resistant to DENV-2 infection.

Screening of anti-DENV drugs using hPLOs

The hPLOs are suitable for large scale production and exhibit pronounced phenotypic changes upon DENV-2 infection, making hPLOs an ideal platform for anti-DENV drugs screening. To validate its utility, we tested a panel of chemical compounds with known anti-DENV activities, including the broad-spectrum RNA virus inhibitor 7-deaza-2’-C-methyladenosine (7DMA)^20,67, and pre-identified anti-DENV drugs such as JNJ1802³, JNJ-A07²⁰, ML-SA1⁶⁸, NITD-688⁶⁹, were subjected to hPLOs evaluation (Fig. 6a and Supplementary Fig. 10a). Among these, 7DMA and JNJ1802 (Supplementary Fig. 10b, c) were the most effective in rescuing the phenotypes of DENV-2 infected hPLOs (Fig. 6b, c) and significantly reducing viral loads (Supplementary Fig. 10d–e). However, some drugs like NITD688 showed hepatotoxicity (Supplementary Fig. 10a).

**Fig. 6: Screening and evaluating the efficacy of anti-DENV drugs using hPLOs.**

Given that scRNA-seq data suggested oxidative stress and mitochondrial damage as key contributors to cell death, we further tested two mitochondrial-protective small molecules, including oxyresveratrol (Oresveratrol, ORES)⁷⁰ and Resveratrol^71,72. ORES effectively rescued DENV-2 infected hPLOs phenotypes (Fig. 6d–f), similar to 7DMA and JNJ1802. ORES also reduced the viral loads significantly (Fig. 6g, h). Previous studies have suggested that ORES can alleviate oxidative stress by activating the Nrf2 pathway, thereby protecting hepatocytes from oxidative stress and mitochondrial dysfunction⁷⁰. Moreover, DENV targets Nrf2 for degradation, disrupting Nrf2 antioxidant signaling and impairing the host’s antiviral immune response⁷³. Thus, we hypothesized that the anti-DENV effect of ORES may result from its activation of the Nrf2 signaling pathway. To confirm this hypothesis, we treated DENV-2 infected hPLOs with omaveloxolone (RTA 408), an FDA-approved Nrf2 activator⁷⁴ (Fig. 6i). RTA 408 rescued the phenotypes (Fig. 6j-k) and reduced intracellular viral RNA levels in infected hPLOs (Fig. 6l, m), suggesting that Nrf2 activation could be a promising therapeutic pathway for DENV-2 treatment. Furthermore, downstream genes in the Nrf2 pathway were also restored in the ORES and RTA 408 treatment groups (Fig. 6n).

To assess whether the anti-DENV-2 drugs could restore liver function, we first tested glycogen storage capacity. DENV-2 infected hPLOs displayed reduced storage capacity, but ORES and RTA 408 restored this function, as determined by PAS staining (Fig. 6o). Treatment groups also restored some metabolic activities, as measured by phase I enzyme (CYP3A4) activities, lactate dehydrogenase (LDH) activities, compare to DENV-2 groups (Fig. 6p, q). Notably, the elevated LDH release observed in the DENV-2 infected groups served as an indicator of apoptosis and reduced cell viability (Fig. 6q). Additionally, human albumin and urea secretion were restored in the treatment groups, with levels close to those in the MOCK group (Fig. 6r, s).

To mimic clinical treatment, we tested delayed drug administration, treating DENV-2 infected hPLOs with 7DMA, JNJ1802, ORES, and RTA 408 on day 2 post infection, when phenotypic changes were evident (Supplementary Fig. 10f). Delayed treatment could inhibit DENV-2 replication as well (Supplementary Fig. 10g-j). Collectively, these data showed that ORES and RTA 408 efficiently inhibited DENV-2 infection and restore the function in hPLOs by targeting the NRF2 signaling.

Validation of anti-DENV activity in cell and mouse model

The anti-DENV-2 efficacy of ORES and RTA 408 was further validated in Huh7 cells and AG6 mice, as they are frequently used models for DENV studies. DENV-2 infected Huh7 cells were treated with ORES and RTA 408, and DENV-2 RNA levels were quantified on day 2 and exhibited significant inhibition with these two drugs (Fig. 7a, b). Next, the prophylactic efficacy of ORES and RTA 408 against DENV-2 was tested in 4-6 weeks old AG6 mice, which are interferon alpha and gamma receptor-deficient (IFNα/γR1-/-)⁷⁵ (Fig. 7c). Oral administration was performed twice daily (b.i.d.) with 60 mg/kg ORES and once daily with 15 mg/kg RTA 408 for 5 consecutive days (Fig. 7c). Viral challenge with 1 × 10⁵ PFU of DENV-2 was performed after two days of pretreatment. Tissues and serum were collected three days post-infection to quantify viral RNA levels. Significantly lower viral RNA levels were observed in both the liver and serum under both drug treatments (Fig. 7d-e). Additionally, to verify whether the antiviral effects of these two drugs are related to NRF2, we knocked down NRF2 in the Huh7 cell line and found that the antiviral effect was attenuated in the knockdown group (Supplementary Fig. 10k, l). Taken together, these results extend the findings from hPLOs, demonstrating that both ORES and RTA 408 exhibit anti-DENV-2 activity in vitro and in vivo.