Ebola virus infection induces a type I interferon response and the shutdown of key liver functions in human iPSC-derived hepatocytes


 Liver damage and an exacerbated inflammatory response are hallmarks of Ebola virus (EBOV) infection. Little is known about the intrinsic response to infection in human hepatocytes and their contribution to the observed inflammatory response. Here, we present an iPSC-derived hepatocyte platform to define the hepato-intrinsic response to EBOV infection. Transcriptomics analysis revealed a delayed host response with minimal transcriptomic changes at one day post infection (dpi) followed by a general downregulation of genes associated with hepatic functions and upregulation of interferon signaling at two and three dpi. Using RNA-FISH, we showed at single cell resolution that IFNβ and CXCL10 were mainly expressed in bystander cells or cells with weak EBOV mRNA signal intensity. We did not observe an inflammatory signature at any timepoint. In conclusion, iPSC-derived hepatocytes are an immune competent platform to study intrinsic responses to EBOV infection that have not been observed in EBOV-infected hepatocarcinoma cell lines.


Introduction
Ebola virus (EBOV) is a non-segmented negative sense (NNS) RNA virus and belongs to the family Filoviridae (1). EBOV infection causes a severe disease in humans with case fatality rates ranging from 40 to 90% (2). Despite the development of an approved EBOV vaccine and two promising antibody-based therapeutics (3,4), there are currently no treatment options available to mitigate late stage Ebola virus disease (EVD).
Severe liver damage, coagulopathy, and thrombocytopenia are hallmark clinical manifestations of Ebola virus disease (EVD) (5). Histopathological changes in liver autopsies of fatal EBOV cases include hepatocellular necrosis, Councilman bodies, Kupffer cell hyperplasia, and the presence of EBOV inclusions in infected cells (5,6). Circulating monocytes and macrophages have been identi ed as the earliest target cells in the liver of EBOV-infected nonhuman primates (NHPs) (7,8). Infected Kupffer cells were detected as early as 3 days post infection (dpi) and infected hepatocytes by day 4 to 5 (8). Liver damage in EVD patients and experimentally infected NHPs is indicated by elevated serum levels of liver enzymes, including aspartate aminotransferase (AST), alanine aminotransferase (ALT), alkaline phosphatase, and γ-glutamyltransferase, and decreasing levels of albumin and calcium (8)(9)(10)(11)(12).
Decreasing serum protein C levels and increasing brinogen levels indicate dysregulation of clotting and coagulopathy (13)(14)(15). In a low-dosage NHP EBOV infection, no signi cant changes in AST, ALT, or albumin levels were observed among the survivors, whereas the fatal cases exhibited changes in all of these markers, indicating severe liver damage (16). Massive hepatocyte apoptosis has been observed in murine models of EVD (17,18), and two recent studies comparing EBOV and Reston virus pathogenesis in mice engrafted with human hematopoietic cells show that the magnitude of viral replication in the liver correlates with disease severity and fatal outcome (19,20). Taken together, these reports indicate that the liver plays a pivotal role in severe EVD and therefore could be a target for therapeutic intervention.
In addition to organ damage, there is evidence that a dysregulated proin ammatory response is a hallmark of severe EVD (14). Monocytes and macrophages have been shown to be activated upon EBOV infection (7,(21)(22)(23)(24), and this proin ammatory environment may impact liver damage during EBOV infection. While hepatocytes constitute over 80% of the liver volume, their immune response to EBOV has yet to be characterized. Most previous reports that have examined the intrinsic hepatic response to EBOV infection utilized immortalized hepatocarcinoma cell lines (25)(26)(27)(28), which have an impaired antiviral response compared to primary human hepatocytes (PHHs) (29)(30)(31)(32). While PHHs represent a more reliable platform, limited accessibility, donor-to-donor variability, and complex in vitro conditions required for long-term culture have restricted their use (33)(34)(35)(36).
Here, we have employed human induced pluripotent stem cell (iPSC)-derived hepatocytes as a model for EBOV infection. These cells can be terminally differentiated and maintained in a monoculture for approximately 10 days and have previously been shown to effectively model viral infections (37)(38)(39)(40)(41)(42)(43). We demonstrate that iPSC-derived hepatocytes recapitulate aspects of the innate human host response to EBOV infection that cannot be modeled in immortalized cell lines. Our data show that EBOV infection leads to a delayed type I interferon (IFN) response and the downregulation of genes associated with liver function in iPSC-derived hepatocytes, illuminating a mechanism by which hepatocytes participate in EVD severity and liver damage.

Results
iPSC-derived hepatocytes are physiologically more similar to primary hepatocytes than immortalized Huh7 cells To generate human hepatocyte-like cells from iPSCs, we used a previously published step-wise directed differentiation protocol that follows normal hepatic development as a roadmap for iPSC lineage speci cation (Fig. 1A) (39). At day 5 (D5), approximately 80-90% of the cells were CXCR4/cKit doublepositive, indicating robust differentiation into de nitive endoderm (Fig. 1B). By D20 of differentiation, most cells in culture adopted a cuboidal, highly vacuolated morphology, typical of primary hepatocytes (Fig. 1C). A molecular, histological and functional characterization of the iPSC-derived hepatocytes in comparison to either immortalized hepatocarcinoma Huh7 cells or fetal and adult liver tissue showed that the iPSC-derived hepatocytes retained expression of the fetal marker alpha feto protein (AFP), likely re ecting the embryonic-like characteristics of the cells in this culture system (Fig. 1D). As expected, iPSCderived hepatocytes expressed many hepatic markers, including albumin, AAT and the transcription factors FOXA1, FOXA2 and HNF4a. The mRNA levels of these hepatic markers were generally higher in Huh7 cells compared to iPSC-derived hepatocytes. However, immuno uorescence analysis revealed striking differences in the protein levels and localization of the hepatic transcription factors in both cell types (Fig. 1E). While HNF4α, FOXA1, and FOXA2 were translocated into the nuclei of iPSC-derived hepatocytes, they were expressed at low protein levels in Huh7 cells and homogenously distributed throughout the cytoplasm. Albumin expression was observed in both cells types (Fig. 1E), whereas transferrin, a clotting factor and carrier protein produced exclusively by hepatocytes in the liver, was robustly expressed in iPSC-derived hepatocytes but not in Huh7 cells (Fig. 1E), further emphasizing the differences between the two platforms. Functional analysis of iPSC-derived hepatocytes compared to Huh7 cells revealed major de ciencies in the Huh7 cells. One major function of hepatocytes in vivo is to store and metabolize lipoproteins, and an inability to bind or retain lipoproteins indicates a de ciency in hepatic function (44). While the majority of iPSC-derived hepatocytes were positive in the low-density lipoprotein (LDL) uptake assay, LDL uptake was impaired in Huh7 cells (Fig. 1F). Similarly, Periodic acid-Schiff (PAS) staining, used to detect glycogen storage, resulted in the typical magenta staining in the iPSC-derived hepatocytes, whereas the Huh7 cells showed weak and diffuse purple staining (Fig. 1G).
Finally, we tested whether the iPSC-derived hepatocytes have a proper cytochrome P450 function. Hepatocytes produce cytochrome P450 enzymes, which are involved in the metabolism of over 80% of all prescribed therapeutics (45). Immortalized hepatocytes express low levels of these enzymes unless speci cally cultured in con uent conditions (46). To measure the enzymatic activity of the predominant cytochrome P450 enzyme, CY3A4, in iPSC-derived hepatocytes and Huh7 cells, we used a luminescentbased assay. In contrast to Huh7 cells that had no CY3A4 activity, iPSC-derived hepatocytes showed a strong luminescent signal indicative of CY3A4 enzymatic activity (Fig. 1H).
It was previously reported that infection with vesicular stomatitis virus (VSV), a strong inducer of the type I IFN response, leads to type I IFN expression in PHHs but not in Huh7 cells (29). To examine if the iPSCderived hepatocytes generated by us were able to adequately respond to an acute viral infection, iPSCderived hepatocytes and Huh7 cells were infected with VSV and IFNβ expression was analyzed by qRT-PCR. In contrast to Huh7 cells, IFNβ expression was strongly upregulated in infected iPSC-derived hepatocytes (Fig. S1). This indicates that similar to PHHs, iPSC-derived hepatocytes are IFN-competent and able to respond to viral stimuli. It also con rms that hepatocarcinoma cell lines may not represent a suitable platform to analyze antiviral responses to viral infections.
Overall, these results indicate that iPSC-derived hepatocytes recapitulate functional and molecular hepatocyte features more faithfully than Huh7 cells do and therefore provide a better model to study EBOV hepatic infection.
iPSC-derived hepatocytes and primary human hepatocytes are susceptible to EBOV infection Next, we determined if iPSC-derived hepatocytes were permissive to EBOV infection. As a comparison, we also infected primary human hepatocytes using a previously established platform based on cryopreserved adult primary hepatocytes collected from donors, allowing for the maintenance of primary hepatocytes in culture for up to several weeks (33,34). We initially used micropatterned cocultures (MPCCs) that consist of PHH islands supported by J2-3T3 murine embryonic broblasts. The MPCC system was successfully used to for hepatitis C and hepatitis B virus infection studies (34). However, despite robust EBOV infection of the PHHs in this coculture system ( Fig. S2A and B), the MPCCs were less suitable for EBOV because the mouse broblasts were infected as well in addition to the PHHs, albeit at low infection rates ( Fig. S2A and B). Since the goal of this study was to dissect the host response of human hepatocytes to EBOV infection, infected mouse broblasts could potentially confound the results. Therefore, we took advantage of a controlled apoptosis system to remove the 3T3-J2s broblasts from the MPCCs (47) prior to infection.
iPSC-derived hepatocytes from three different donors (BU1-3), PHHs from 2 donors in broblast-depleted MPCCs, and Huh7 cells were infected with EBOV over a time course of 7 days and analyzed by immuno uorescence staining. At 1 dpi, about 50% of iPSC-derived hepatocytes and PHHs were infected and about 90% of the Huh7 cells ( Fig. 2A). Viral inclusion formation, a hallmark of EBOV infection, was observed in all three cell types (Fig. 2B). In parallel, we collected supernatants from the infected cells to determine viral titers over the course of infection. Viral titers peaked at 3 dpi, with the highest titers produced in Huh7 cells (5x10 5 ) followed by iPSC-derived hepatocytes (2x10 4 ) (Fig. 2C). The increase in viral titers correlated with an increase in the percentage of infected cells in culture, which consistently showed 80-90% infection by 3 dpi (Fig. 2A). Infected PHHs showed a modest increase in viral production and PHHs over time with overall lower titers (1x10 3 ) compared to the iPSC-derived hepatocytes. In summary, these data show that iPSC-derived hepatocytes and PHHs are permissive to EBOV infection and that iPSC-derived hepatocytes support EBOV replication more e ciently than PPHs do.
Since the iPSC-derived hepatocytes showed robust EBOV infection, we focused on this cell type for a more phenotypical characterization. To visualize spatial and temporal viral RNA production in infected cells at early and late time points, we performed single-cell, single-molecule RNA in situ hybridization using RNAscope technology (48). The used probes are directed against EBOV VP35 positive sense RNA which includes VP35 mRNA and antigenomic RNA. The positive sense antigenomic RNA is a complementary copy of the viral genome and serves as a replication intermediate (49). Robust viral mRNA production was observed at 8 hours post infection (hpi) with a punctate staining pattern throughout the cytoplasm, each small dot likely representing a single mRNA molecule, whereas larger bright puncta may represent RNA aggregates (48) (Fig. 3A). At 2 dpi, the viral RNA was homogenously distributed throughout the cytoplasm and aggregated in clusters that colocalize with viral inclusions (50).
Electron microscopy (EM) analysis of EBOV-infected iPSC-derived hepatocytes con rmed the formation of EBOV nucleocapsids in the cytoplasm of the infected cells and the release of viral particles (Fig. 3B). Newly synthesized EBOV nucleocapsids accumulated in viral inclusions and were arranged in parallel tubes in a hexagonal pattern in which each nucleocapsid is surrounded by 6 neighboring nucleocapsids, in line with previously reported observations ( Fig. 3C) (51). The accumulation of viral inclusions and the release of viral particles was observed at 1 and 2 dpi for all 3 donors (Fig. 3D-F).
To follow the course of EBOV infection in iPSC-derived hepatocytes by live cell imaging, we used a recombinant EBOV clone expressing GFP from an additional transcription unit (52). Although the cells were infected with a high MOI of 10, we observed peak infection only at 3 dpi (Fig. S3). At 7 dpi, the infected cells showed a pronounced cytopathic effect (CPE), leading to reduced GFP expression. Overall, these data indicate that iPSC-derived hepatocytes can be productively infected with EBOV and that the infection leads to cell damage at late time points.

EBOV infection induces a delayed type I IFN response in human hepatocytes
To analyze the transcriptional pro le induced by EBOV infection across the three infection platforms, iPSC-derived hepatocytes, PHHs, and Huh7 cells were infected with puri ed Ebola virions and the cellular RNA was harvested 24 hpi for bulk RNA sequencing analysis. The chosen time point was based on the transcriptional response to EBOV infection in human monocyte-derived macrophages, which display a strong transcriptional signature at 24 hpi (53). We calculated the log fold change (logFC) in expression for each aligned transcript between the EBOV-infected sample and mock control to determine which transcripts were differentially expressed for each condition. However, our results indicated that there was no signi cant differential gene expression in infected compared to noninfected cells in either cell type (Fig. S4).
It has been shown that delayed interferon responses to viral infections may contribute to disease severity (54,55). In addition, we observed peak infection of EBOV-infected iPSC-derived hepatocytes as late as 3 dpi (Fig. S3). This inspired us to explore the transcriptional response to EBOV infection in liver cells at later time points. iPSC-derived hepatocytes were infected with EBOV and cellular RNA was isolated at 1, 2, 3, and 7 dpi for bulk transcriptomics analysis. Principle component analysis (PCA) of mock-infected compared to EBOV-infected samples at all time points showed clear clustering of the replicates, separating the non-infected from the infected samples at each time point, particularly at 2, 3 and 7 dpi (Fig. 4A). We observed minimal gene expression changes at 1 dpi, con rming our previous RNA sequencing data. When we ranked the expressed genes on normalized log counts per million (lcpm), there were only 14 upregulated genes and 18 downregulated genes with a p-value ≤ 0.01 at 1 dpi (Table S1). We then applied the same raw-normalized log counts per million (lcpm) analysis for the 250 most variable genes across all time points in mock-infected and EBOV-infected iPSC-derived hepatocytes over time and plotted the results as a heatmap (Fig. 4B). Clear gene expression signatures related to EBOV infection emerge at 2 and 3 dpi, converging into a different signature by 7 dpi. Starting 2 dpi, we observed a pronounced type I and III IFN signature, speci cally upregulation of IFNβ, IFNλ1-3, and a set of IFNstimulated genes (ISGs). There was a strong overlap of the top 30 DEGs at 2 and 3 dpi (20/30), including 13 ISGs (Fig. 4B, C and Fig. S5A). To further assess the induction of a type I/III IFN response in EBOVinfected cells, we plotted the fold change expression of a panel of type I/III IFNs, and 50 ISGs for each time point, con rming the upregulation of ISGs at 2 and even more pronounced, 3 dpi, including IFNλ1-3, IFNβ, chemokine ligand 5 (CCL5), 2'-5'-oligoadenylate synthetase like (OASL), OAS1, interferon-induced protein with tetratricopeptide repeats 1 (IFIT1), IFIT2, and the poorly characterized interferon-induced protein 44 (IFI44) that was shown to have antiviral activity (56-58) (Fig. 4C). By 7 dpi, the IFN response signature was no longer present (Fig. 4C).
A striking feature of the host response signature in the infected iPSC-derived hepatocytes was a general downregulation of genes at later time points, many of them associated with characteristic liver functions. For instance, several of the downregulated genes were associated with aminotransferases and ion transport (9/30) (Fig. S5B). AGXT, cytochrome P450 enzyme CYP3A7, the bile acid transporter nuclear receptor subfamily 1 group H member 4 (NRIH4), SERPINA7, ITIH1, and ITIH3 were all signi cantly downregulated at 7 dpi (Fig. S5B). Genes involved in plasma formation and clotting, including ALB, insulin-like growth factor 2 (IGF2), podocan (PODN), and n-acetyltransferase 16 (NAT16), were also downregulated at this time point (4/30) (Fig. S5B). We further con rmed the signi cant downregulation of liver associated genes in EBOV-infected iPSC-derived hepatocytes 3 dpi by using a curated list of liver speci c genes for comparison (39), including albumin (ALB), serpina genes (SERPINA 1 and SERPINA 3), brinogen genes (FGA, FGB, FGG), apolipoprotein genes (APOA1, APOA2 and APOC1) and angiopoetin (ANG) among others (Fig. 4D). This indicates that EBOV infection leads to severe liver cell damage associated with loss of hepatic functions.
To identify pathways which were perturbed during EBOV infection, we performed Gene Set Enrichment Analysis (GSEA) on the DEGs in EBOV-infected cells compared to mock-infected controls using WebGestalt (59). Gene ontology (GO) analysis showed moderate upregulation of IFN response pathways in EBOV-infected cells at 2 and 3 dpi (Fig. 5). Hallmark pathway analysis con rmed these results, highlighting that the IFNα and IFNγ response pathways are upregulated at 2 and 3 dpi (Table 1).
Importantly, there was no upregulation of in ammatory response pathways, suggesting that EBOV infection does not trigger a proin ammatory response in primary-like hepatocytes. While there were only few moderately upregulated pathways in EBOV-infected cells, gene downregulation was more pronounced over the course of infection. This includes pathways related to metabolic processing, protein synthesis, protein transport, protein membrane localization, and extracellular structure organization, indicating cellular shutdown (Fig. 5). By Hallmark pathway analysis there were no upregulated pathways at 7 dpi (Table 1). By GO analysis, response to type I IFN and response to virus were slightly upregulated at 7 dpi, driven by the expression of IFNβ and IFNλ at this time point (Fig. 5). Hallmark pathway analysis suggested a downregulation of coagulation and xenobiotic metabolism pathways, indicating that EBOVinfected hepatocytes may have a reduced capacity to produce clotting factors or metabolize drugs, which are suggested implications of liver damage during EBOV infection of NHPs (Table 1) (8,13). We utilized WebGestalt to perform a transcription factor-based network analysis to determine the ciselements that may be highly expressed and dominating the transcriptomic changes observed in our data. The major transcription factor targets based on the transcriptomic signature predicts that at 2 dpi the transcriptional response is directed by transcription factor binding to interferon-sensitive response element 1 (ISRE1) and an unknown gene set (Table S2). The unknown gene set is listed on the GSEA molecular signatures database (MSigDB) as 31 genes containing a highly conserved KRCTCNNNNMANAGC motif within 4 kb of the transcription start site. At 3 dpi, the network analysis predicts that the gene signature is dominated by transcription factor binding to ISRE1 and interferon regulatory factor 2 (IRF2), indicating an upregulation of the IFN response (Table S2). Finally, this network analysis of the DEGs at 7 dpi highlights that the gene expression pro le is dominated by the activation of HNF1A, a hepatic-speci c transcription factor that is critical for regulating cellular function in the adult liver (Table S2).

EBOV-infected iPSC-derived hepatocytes express IFNβ
Bulk RNA analysis only provides the average gene expression levels of a diverse cell population and does not allow to discriminate between the host response signatures in infected and non-infected cells. Given the clear but moderate type I/III IFN response in EBOV-infected iPSC-derived hepatocytes, we hypothesized that there might be signi cant heterogeneity regarding the IFN response in this cell population. To assess the IFN and in ammatory response at single cell level, we performed uorescent in situ RNA hybridization (RNA-FISH) analysis with infected iPSC-derived hepatocytes using probes targeting IFNβ, CXCL10, and IL6 mRNA in combination with EBOV-speci c probes targeting VP35 mRNA.
The number of the cells with detectable VP35 mRNA expression increased from about 20% at 1 dpi to 60-70% at 2 and 3 dpi ( Fig. 6A and B). Consistent with the RNAseq data, less than 2% of the cells expressed CXCL10 (1.6%) or IFNβ (0.2%) at 1 dpi (Fig. 6A, C, and D). We observed an increase in both CXCL10 and IFNβ expressing cells at 2 and 3 dpi with up to 2% of IFNβ and 6% of CXCL10 expressing cells. Of note, CXCL10 and IFNβ were mainly expressed in uninfected bystander cells. The ratio of uninfected to infected IFNβ-expressing cells increased from 50% at day 1 to 92% at day 3 (Fig. S6A). Cells expressing IFNβ -particularly at 3 dpi -tended to be found in regions with a lower infection rate compared to the average infection rate of the whole slide. In addition, the signal intensity of EBOV mRNA was weak in cells expressing IFNβ or CXCL10, suggesting that these cells represent early stages of infection with low expression rates of VP35 and VP24, the known EBOV IFN antagonists (60). There was no detectable IL6 mRNA expression in infected iPSC-derived hepatocytes at any time point post infection, con rming the absence of an in ammatory response in these cells (Fig. S6B). In contrast to the iPSCderived hepatocytes, we did not observe IFNβ or CXCL10 expression in Huh7 cells (Fig. S6C). Overall, the RNA-FISH results corroborate the ndings from the transcriptomic analysis and suggest that EBOV infection of iPSC-derived hepatocytes induces a delayed and moderate type I IFN response that is not observed in infected Huh7 cells.
EBOV can be transmitted from infected macrophages to iPSC-derived hepatocytes. Since there is evidence that migrating monocytes or tissue-resident macrophages are early target cells in the liver (7,8), we explored whether EBOV can be transmitted from infected macrophages to iPSC-derived hepatocytes in a co-culture model. Human primary monocyte-derived macrophages (MDMs) were infected with recombinant EBOV expressing GFP at a multiplicity of infection (MOI) of 10. One dpi, the cells were detached, rigorously washed, and seeded onto iPSC-derived hepatocytes. Spread of infection was visualized by GFP expression (Fig. 7). While only single green cells were observed at 2 dpi, likely representing infected MDMs, the infection spread throughout the entire iPSC-derived hepatocyte monolayer over the time course of 7 days, indicating that EBOV infection of hepatocytes can be initiated by infected macrophages.

Discussion
Although the liver plays a crucial role in EBOV disease, the intrinsic response of human hepatocytes to EBOV infection is not well de ned. In vitro hepatocyte platforms are limited in their capacity to recapitulate human immune responses. Immortalized hepatocytes like Huh7 cells have impaired IRF3 signaling (29), which our study highlights as an important signaling cascade in the EBOV-induced hepatic host response. Additionally, investigations have shown that immortalized cell lines can vary signi cantly from lot to lot, preventing an accurate comparison of results across studies, even within the same cell line (61). PHHs are di cult to culture and require intricately engineered culture systems to support in vitro survival (33,34). In the co-culture system presented in this study, PHHs destabilize 2 to 3 days after the removal of the broblast stroma (47) which challenges the use of this PHH culture system for longer infection times. For these reasons, we used iPSC-derived hepatocytes as an alternative to traditional immortalized and primary culture models. We have demonstrated that iPSC-derived hepatocytes are terminally differentiated hepatocytes and primary-like in their functional capacity. iPSC-derived hepatocytes are also differentiated as a monoculture that can be sustained in a mature state for 10 days, allowing for longer infection studies than PHH culture systems. iPSC-based platforms also offer additional advantages, such as the ability to differentiate into multiple cell types from one donor to create organ-speci c co-culture systems within one genetic background. This advantage is applicable to EBOV pathogenesis studies since EBOV infects many different cell types and organs, which may contribute to disease progression.
Using the iPSC-based platform, we began to de ne the intrinsic hepatic response to EBOV infection. The most prominent transcriptional changes are de ned by downregulation of pathways involved in critical cellular processes, such as protein production and transport, and extracellular structure organization. This general shutdown of cell function is contrasted by a moderate upregulation of type I and type III IFN responses dominated by the expression of IFNβ, IFNλ1, and select ISGs beginning 2 dpi and increasing at 3 dpi. At single cell level, IFNβ and CXCL10 expression was observed in uninfected cells and cells with low expression levels of EBOV mRNA. This is in line with RNA-FISH analysis on liver samples obtained from EBOV-infected NHPs, showing expression in uninfected bystander cells (7). Although EBOV generally suppresses type I IFN induction in the infected cells (62), there is a strong proin ammatory and IFN response in EVD patients and infected animals associated with specialized immune cells, including macrophages and monocytes (7,14,24,53,63,64). In contrast to monocytes and macrophages, there was no induction of a proin ammatory response in EBOV-infected iPSC-derived human hepatocytes, indicating that, indeed, immune cells are the main drivers of EBOV-induced in ammation (23,24).
A crucial question that still needs to be addressed from our observations is that hepatocytes are consistently silent 1 dpi, Increasing levels of brin in whole blood samples is associated with fatality in EBOV patients, and in a study of 5 patients, brinogen D-dimer products in whole blood decreased in patients who survived infection (14,15). Tissue factor (F3) is also upregulated in EBOV infection, and inhibition of F3 was protective against EBOV infection in NHPs (5,6,8,13,67,68). In our study, iPSC-derived hepatocytes downregulate the coagulation cascade 2-7 dpi. We also do not observe an increase in acute phase reactants that has been observed in NHPs (69). iPSC-derived hepatocytes downregulate the expression of critical clotting factors and plasma proteins, including ALB, brinogen chains alpha, beta, and gamma (FGA, FGB, and FGG), protein c (PROC), and F3 as early as 2 dpi. This discrepancy could be evidence of a compensatory feedback loop designed to prevent coagulation dysfunction or consumptive coagulopathy.
Fibrinogen levels are lower than healthy controls in EBOV patients and increase in patients who survive, and low levels of serum ALB is a hallmark indicator of EBOV-induced liver damage in NHPs (15,70). This is emphasized by the network analysis of the 7 dpi DEG that highlighted HNF1A as a critical regulator of the gene expression pro le. In EBOV-infected NHPs, another hepatic nuclear factor homeobox gene (HNF4A) was signi cantly downregulated in the liver, and the authors attributed the downregulation of this transcription factor with the associated downregulation of genes involved in cholesterol synthesis, coagulation, and hepatic metabolism (71). HNF1A is known to regulate the expression of HNF4A and can bind to its promoter, which may highlight overlap in the transcriptional signatures between our EBOVinfected iPSC-derived hepatocytes and bulk NHP liver tissue (71,72 Primary Hepatocyte Cell Culture The PHH co-culture with murine embryonic broblasts 3T3-J2s were generated by seeding cryopreserved primary human hepatocytes purchased from BioIVT on collagen-patterned 96-well plates as described previously (33,34 (52) were propagated in Vero E6 cells as described before (53). Cell supernatants were centrifuged at 1,000xg for 10 minutes at 4°C to remove cellular debris. To remove soluble factors, including cytokines and chemokines, clari ed supernatants were puri ed over a 20% sucrose cushion by centrifugation at 25,000 rpm for 2 hours at 4°C and the pelleted viral particles were resuspended on PBS. Viral titers were determined by tissue culture infectious dose 50 (TCID 50 ) assays and focus forming unit (FFU) assays as previously described (53 CYP3A4 P450-Glo Assay The quanti cation of CYP3A4 in 1x10 6 iPSCs, iPSC-derived hepatocytes, and Huh7 cells was performed using the P450-Glo Assay kit (Promega) according to manufacturer's instructions. One day prior to the assay, Huh7 cells were treated with cell culture media containing 100 nM Dexamethasone. One the day of the assay, the Luciferin Glo substrate was added to the appropriate cell culture medium for each cell type. Cells were incubated with luciferin reagents at 37°C for 4 hours. Supernatant was clari ed at 300xg for 5 minutes at room temperature. Clari ed supernatant and detection reagent were incubated at room temperature for 20 minutes. Luminescence was measured on a LUMIStar Omega Luminometer (BMG Labtech).
Immuno uorescence Analysis 1x10 6 or 4x10 5 cells were xed with either 4% PFA in DMEM for 10 minutes at room temperature or 10% formalin for at least 6 hours at 4°C. Cells were washed in PBS, permeabilized with a solution of acetonemethanol in a 1:1 volume ratio for 5 minutes at -20°C, and incubated in blocking buffer for at least 20 minutes at room temperature. Due to cross-reactivity with the BSA in the blocking buffer, the cells were not incubated in blocking buffer when stained for albumin. Cells were incubated with primary antibody overnight at 4°C, and primary antibody names and dilutions are listed in Table 2. Cells were incubated with secondary antibody and 4′,6-diamidino-2-phenylindole (DAPI) for one hour at room temperature (dilutions listed in Table 2). VSV-GFP or VSV-mCherry at an MOI of 10. 1 dpi, cells were imaged for GFP expression and cellular RNA was puri ed for qRT-PCR analysis.
LPS stimulation for NFκB Activation Samples were examined in a FEI Tecnai Spirit Twin transmission electron microscope, operating at 80 kV.
SmartSeq2 Sequencing 1x10 6 iPSC-derived hepatocytes or Huh7 cells were seeded in 6-well tissue culture plates, and 1x10 4 PHHs were seeded in 96-well glass-bottom plates. The cells were infected with EBOV at an MOI of 10 or left uninfected and lysed 1 dpi. Cellular RNA was isolated using TRIzol reagent (ThermoFisher) according to the manufacturer's instructions. 50 ng of puri ed RNA was used for sequencing by diluting to a nal concentration of 5 ng/µL in a nal volume of 10 µL. Samples were pipetted into a twin.tec® PCR 96-well plate (Eppendorf) and sent out for sequencing on dry ice. Sequencing was performed at the Broad Institute Genomics Services Core. The quality of the raw data was assessed using FastQC v.0.11.7. The sequence reads were aligned to a combination of the human genome reference (GRCh38) and the Ebola

Macrophage Differentiation From Peripheral Mononuclear Blood Cells
MDMs were generated from leukopaks (NY Biologics Inc.) or apheresed peripheral blood mononuclear cells (PBMCs) using Ficoll separation (GE Healthcare). Whole blood was diluted to a nal volume of 30 mL with PBS. 15 mL of Ficoll was added to a 50 mL conical tube, and the diluted whole blood was overlayed onto the Ficoll. Blood was centrifuged for 30 minutes at 450xg with the brake turned off of the centrifuge. The blood separated into three layers: the off-white plasma layer, a thin white buffy coat containing leukocytes, and the pelleted erythrocytes. The plasma layer was removed, and the buffy coat was isolated into a fresh 50 mL conical tube. The buffy coat was diluted to a nal volume of 50 mL using 37ºC PBS. Cells were centrifuged at 350xg for 10 minutes with the centrifuge brake on. The cell pellet was then washed three times with 50 mL of warm PBS and centrifuged at 350xg for 10 minutes. After the nal wash, cells were resuspended in 1 mL of warm PBS and counted. 1x10 7 cells were plated in a T75 ask with RPMI medium for one hour at 37ºC and 5% CO 2 , non-adhered cells were removed, and 10 ml RPMI medium with 10% FBS, 1% L-glutamine, 10 mM HEPES, 10ng/mL GM-CSF, and 50 U/mL penicillin with 50 mg/mL streptomycin (MDM medium) was added to the cells. Medium was changed every 2 to 3 days for a minimum of 6 days before differentiated macrophages were split into 6-well TC plates for infection experiments.

Co-Culture of iPSC-Derived Hepatocytes and Macrophages
On the day prior to infection, MDMs seeded in a T75 ask were split by incubating cells with 0.1% Trypsin-EDTA solution for 15 minutes at 37°C and 5%CO 2 . Cells were centrifuged at 300xg for 5 minutes at 4°C and resuspended in 1 mL of MDM medium. Cells were counted and 1x10 6 cells were seeded into 6well low-attachment plates (Corning). On the day of infection, macrophages were infected with EBOV-GFP at an MOI of 10. At 1 dpi, MDMs were incubated with 1 mL of Gentle Cell Dissociation buffer (GCD; Gibco) for 2 minutes at 37°C and 5%CO 2 . Cells were centrifuged at 300xg for 5 minutes at 4°C. GCD was removed, and cell pellets were washed three times in 1 mL of PBS. After the third wash, cells were resuspended in 1 mL of the appropriate medium and seeded in 6-or 12-well tissue culture plates alone or onto D20 iPSC-derived hepatocytes at a 1:1 ratio.        Transmission of EBOV-GFP from monocyte-derived macrophages (MDM) to iPSC-derived hepatocytes.
Top, infection scheme and work ow. Live uorescent imaging of MDM and iPSC-derived hepatocyte cocultures at various time points p.i. 10x magni cation. Images taken from one well over the period of 1-7 dpi. Scale bar is 20 µm. Images are representative of 2 independent experiments.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. SupplementaryInformation.pdf