To model region-specific gastrointestinal infection, we generated human iPSC-derived intestinal (HIO) and colonic (HCO) organoids using a modified protocol [17], yielding proximal (small intestine–like) and distal (colon-like) epithelium, respectively, in the absence of mesenchyme (Fig 1A–1E). Single-cell RNA sequencing confirmed robust expression of intestinal markers including CDX2 and Villin1, with multicellular complexity reflective of native human gut epithelium [19].

We employed this model to assess the susceptibility of HIOs and HCOs to EBOV infection, using recombinant EBOV expressing ZsGreen as a reporter gene [21]. HIOs and HCOs were derived from two genetically distinct iPSC donors, with three or more independent differentiations per donor line. HIOs and HCOs were infected with EBOV-ZsGreen at a multiplicity of infection (MOI) of 10 on day 35 of culture. At 1 day post infection (dpi), robust EBOV-ZsGreen expression was observed in all organoids, characterized by numerous ZsGreen-positive columnar basal cells distributed throughout the epithelial monolayer (Fig 2). In contrast, mock-infected organoids showed no ZsGreen-positive cells under identical fluorescent exposure conditions. Viral spread became evident by 3 dpi, with a noticeable increase in the number of infected cells across the organoid cultures (Fig 2). By 7 dpi, organoids exhibited significant cytopathic effects (CPE), leading to the eventual collapse of the organoid structures (Fig 2). Due to the extent of virus-induced damage, organoids at 7 dpi were not included in downstream analyses. These results establish that both HIOs and HCOs are permissive to EBOV infection, supporting viral replication and virus-induced cell damage in a regionally defined human gut model.

Building on our work with the fluorescent EBOV-ZsGreen virus model and following the determination of optimal infection timepoints, we next investigated the susceptibility of HCOs to wild-type EBOV (Mayinga isolate) and MARV (Musoke isolate) at an MOI of 10. Confocal microscopy was employed to visualize caudal type homeobox 2 (CDX2)-GFP labeling, marking intestinal identity in green, and red fluorescence corresponding to immunofluorescence staining of filoviral nucleoprotein (NP) or nucleocapsid (NC), which confirmed viral infection. At 1 dpi, minimal viral presence was detected, characterized by small, punctate fluorescence signals for both EBOV and MARV (Fig 3A and 3E). By 3 dpi, viral expression was widespread throughout the organoid culture, indicating active viral replication and dissemination (Fig 3B and 3F). Higher magnification imaging revealed the presence of inclusion bodies, a hallmark of filovirus infection, localized within the perinuclear region of infected cells (Fig 3C and 3G). These findings confirm that iPSC-derived distal HCOs are permissive to both EBOV and MARV infection. Additionally, comparable infection efficiencies were observed in iPSC-derived HCOs from a genetically distinct donor, further supporting the robustness of this model (S1 Fig). The BU310-Cre2 iPSC line, used in this experiment, lacks the CDX2-GFP knock-in promoter. Consequently, HCOs were sorted based on CD26 (dipeptidyl peptidase 4 (DPP-4)) positivity, a marker of intestine identity and a potential biomarker for intestinal health [22]. To further confirm intestinal identity, the HCOs were stained for Villin1, an apical brush border protein specific to the intestinal epithelium (S1 Fig).

Flow cytometry analysis, utilizing an antibody against EBOV NP, further quantified the infection rate, revealing that approximately 42% of cells were infected with EBOV by 3 dpi (Fig 3D). Attempts to quantify MARV infection using flow cytometry were unsuccessful, likely due to antibody incompatibility and the stringent viral inactivation protocols required for BSL-4 pathogen handling. Therefore, immunohistochemistry (IHC) was employed to quantify both EBOV and MARV infections, providing spatially resolved and reproducible measures of viral spread within the organoids. These findings were corroborated by IHC, where the localization and expression patterns of EBOV matrix protein VP40 and MARV glycoprotein (GP) mirrored those observed by immunofluorescence assay. Infection progression was characterized by minimal viral presence at 1 dpi, which increased significantly by 3 dpi, in contrast to mock-infected controls (Fig 3H and 3J). This time-dependent progression in viral replication and spread was supported by quantifying viral antigen positive tissue areas, with similar trends observed for both EBOV and MARV (Fig 3I and 3K). These results confirm that HCOs are permissive to both EBOV and MARV infections, facilitating efficient viral replication, dissemination, and the induction of cytopathic effects. Together, the increase in antigen-positive cells, broader spatial distribution of viral RNA, and rising viral transcript levels support intra-organoid viral spread, although direct measurement of infectious virus release into the supernatant was not performed.

Having established a human colonic epithelial model for EBOV and MARV infection using iPSC-derived HCOs, we next aimed to delineate the global, time-dependent transcriptomic responses of these organoids to filoviral infection. To investigate the host epithelial responses to filovirus infection, we performed bulk RNA sequencing (RNA-seq) on HCOs infected with EBOV or MARV (Fig 4A). Organoids were infected at day 35 of differentiation and harvested at 1 and 3 dpi. Transcriptomic profiles were compared to mock-infected controls (n = 3 biological replicates per condition). Principal component analysis (PCA) revealed time-dependent transcriptional changes, with the largest variation attributed to post-infection time (PC1, 19.3%) and infection status (PC2, 16.9%) (Fig 4B and 4C). By 3 dpi, infected organoids exhibited distinct separation from controls, reflecting progressive viral replication and host response. LOESS regression confirmed increasing viral RNA expression in EBOV- and MARV-infected samples over time, with no viral transcripts detected in mock-infected controls (S2 Fig). Read counts of viral transcripts confirmed productive viral replication in both infection models (Fig 4D and 4E).

To characterize virus-specific transcriptional signatures, we identified differentially expressed genes (DEGs) relative to mock-infected controls (LogFC > 2 or <–2; p < 0.05). Venn analysis revealed both shared and virus-specific DEGs (Fig 4F). At 1 dpi, 160 upregulated genes were common to both EBOV and MARV, while no significantly downregulated genes were shared. By 3 dpi, 106 upregulated and 19 downregulated genes overlapped between the two infections, highlighting distinct and evolving host responses.

At 1 dpi, both viruses induced significant enrichment of pathways associated with hypoxia, EMT, TNF-α signaling, inflammatory responses, glycolysis, bile acid metabolism, and key epithelial programs, including apical surface and tight junction integrity (Fig 4G). These pathways are essential for maintaining intestinal epithelial polarity, selective permeability, and absorptive function. Their dysregulation suggests an early disruption of epithelial architecture and barrier function, which may underlie clinical features such as malabsorption and diarrhea commonly observed in filovirus-infected patients. Simultaneous downregulation of DNA repair mechanisms further indicates compromised epithelial cell homeostasis and increased vulnerability. By 3 dpi, many of these transcriptional signatures persisted; however, a key divergence emerged in interferon signaling (Fig 4H). MARV-infected HCOs exhibited sustained and significant upregulation of interferon response pathways, whereas EBOV-infected HCOs demonstrated suppression of these same pathways. This differential regulation highlights a virus-specific modulation of the host innate immune response during the later stages of infection.

Unsupervised hierarchical clustering of the top 250 DEGs revealed virus- and time-specific transcriptional profiles (Fig 5A). As expected, viral transcripts were among the most highly upregulated genes, confirming active replication. In addition to viral gene expression, infection induced significant transcriptional changes in host gene sets associated with critical intestinal epithelial functions. These included markers of hypoxic stress, glycolytic metabolism, EMT, apical membrane organization, tight and gap junction integrity, and core regulators of absorptive and barrier function. By 3 dpi, MARV-infected HCOs showed robust induction of ISGs, including OASL, MX1, IFIT1, IFIT2, IFI6, and CXCL10 (Fig 5B), consistent with a strong innate immune activation. In contrast, EBOV-infected organoids did not exhibit a delayed ISG response.

Transcriptomic data also indicated metabolic and structural disruption. Genes associated with epithelial barrier integrity, including solute carriers (SLC10A2, SLC9A3, SLC26A3, SLC5A1) and epithelial adhesion (e.g., EPCAM), were significantly downregulated in both infections by 1 dpi (Fig 5C). Additionally, transcripts implicated in diarrheal pathogenesis (e.g., TTC37, TTC7A, GUCY2C) were dysregulated. These findings suggest compromised absorptive function and epithelial barrier integrity early in infection, which may underlie gastrointestinal symptoms commonly observed in filovirus disease.

Hallmark EMT-related genes were significantly upregulated by 1 dpi, reflecting the early onset of epithelial remodeling. EMT is characterized by the loss of epithelial polarity and intercellular adhesion, accompanied by a transition to a mesenchymal state with enhanced motility and invasive capacity. This phenotypic shift compromises epithelial barrier integrity, a critical feature of intestinal homeostasis, and may facilitate viral dissemination within the tissue microenvironment. The resulting disruption of barrier function likely contributes to clinical manifestations such as diarrhea and malabsorption, which are common in filovirus-infected individuals. Moreover, EMT activation is associated with the induction of pro-inflammatory signaling cascades and has been implicated in viral immune evasion through suppression of antiviral responses [23].

In parallel, we observed significant dysregulation of oxidative stress response pathways and a pronounced downregulation of oxidative phosphorylation (OXPHOS), particularly at 3 dpi and most prominently in MARV-infected HCOs (Fig 4G). OXPHOS is essential for ATP generation and mitochondrial function in intestinal epithelial cells; its impairment is associated with increased production of reactive oxygen species (ROS), mitochondrial dysfunction, and cellular injury. These changes reflect a virus-induced metabolic shift from OXPHOS to glycolysis, a well-documented strategy exploited by many viruses to support replication by increasing the supply of metabolic intermediates and energy [24,25].

Concurrently, genes associated with hypoxic stress, including PDK1, BNIP3L, VEGF, and LOX, were significantly upregulated in both EBOV- and MARV-infected HCOs, consistent with prior reports of hypoxia-related signaling during viral infection [26,27]. This hypoxia gene signature overlapped with EMT activation, suggesting a coordinated cellular stress response involving inflammation, metabolic reprogramming, and epithelial remodeling [28,29] (Fig 5D). Together, these findings suggest that filovirus infections perturb epithelial structure and function, as supported by both transcriptional signatures and immunofluorescence evidence of disrupted epithelial organization (Figs 2 and 3).

To further investigate filovirus infection in distinct regional gut areas, we performed parallel infections using iPSC-derived HIOs (small intestine-like) with EBOV and MARV at an MOI of 10. Confocal microscopy confirmed the intestinal epithelial identity of the HIOs via expression of CDX2-GFP (green), while filoviral nucleoprotein (NP) was detected in infected cells (red), and nuclei were visualized by DNA staining (blue). At 1 dpi, NP signal for both viruses appeared as sparse, punctate foci, indicating limited early infection (Fig 6A and 6D). By 3 dpi, viral expression became widespread, indicative of active replication and dissemination (Fig 6B and 6E), with high-resolution imaging revealing viral inclusion bodies in the cytoplasm of the infected cells (Fig 6C and 6F).

Building on previous findings from HCOs, where virus-specific transcriptional responses were observed at later stages of infection, we performed bulk RNA sequencing of HIOs at 3 dpi in triplicate to characterize the transcriptional landscape and identify viral-specific differences (Fig 7A). Our data show that both EBOV and MARV infect HIOs and produce infectious virions, similar to HCOs. PCA revealed significant, time-dependent shifts in the transcriptional profiles of the infected organoids, with the primary sources of variation attributed to time post-infection (PC1, 35.6%) and viral infection status (PC2, 20.0%) (Fig 7B). Read counts of viral RNA confirmed productive infection in both EBOV- and MARV-infected samples (Fig 7C and 7D).

Transcriptomic profiling of infected HIOs at 3 days post-infection (dpi) revealed significant upregulation of immune-related gene programs. Both EBOV- and MARV-infected HIOs showed enrichment of type I and III interferon signaling pathways, inflammatory response genes, apoptotic pathways, and components of the JAK/STAT signaling cascade (Fig 7E). Notably, the interferon-related gene expression signature was consistent with patterns previously observed in non-infected bystander cells in EBOV infection [30–32].

Combined In situ hybridization (ISH) and IHC at 3 dpi revealed virus- and region-specific epithelial immune responses and viral antigen respectively, which partially aligned with transcriptomic profiles. Consistent with RNA-seq data, MARV infection elicited significantly greater epithelial expression of MX1, IFNb, and CXCL10 mRNA in distal HCOs compared to EBOV infection, demonstrating robust ISG induction in MARV-infected HCOs (S3A–S3C Fig). Overall, we observed little overlap between MX1, IFNb and CXCL10 mRNA staining (yellow) and the MARV staining (red), suggesting that mainly non-infected bystander cells were activated by MARV virus infection. In contrast, EBOV-infected HCOs exhibited minimal expression of these markers, indicating limited immune activation. In proximal HIOs, transcriptomic analysis showed upregulation of interferon signaling pathways for both viruses; however, the pronounced MARV-associated ISG induction observed in HCOs was not evident. Nonetheless, ISH revealed increased MX1, IFNb, and CXCL10 mRNA levels in MARV-infected HIOs compared to EBOV, indicating comparatively stronger immune activation (S3A–S3C Fig). Notably, MX1 gene expression was consistently greater in distal HCOs than proximal HIOs across infections, suggesting regional differences in epithelial antiviral responsiveness (S3D–S3F Fig). Collectively, these findings highlight virus- and region-specific innate immune activation, with MARV eliciting a more robust epithelial antiviral response, particularly in non-infected cells in the distal gut.

To evaluate the functional consequences of filovirus infection on epithelial ion transport, forskolin-induced swelling assays were conducted in iPSC-derived HIOs and HCOs. This assay provides a quantitative measure of cystic fibrosis transmembrane conductance regulator (CFTR)-mediated fluid transport, which is highly active in the gut and is based on forskolin-driven cAMP signaling and activation of epithelial chloride channels [33].

The forskolin-induced swelling assay is utilized to assess cystic fibrosis transmembrane conductance regulator (CFTR) function by leveraging forskolin’s mechanism of action. Forskolin activates adenylate cyclase, which increases intracellular cAMP levels. Elevated cAMP activates protein kinase A (PKA), resulting in phosphorylation of CFTR and subsequent opening of the CFTR chloride channel. This enables chloride and sodium ions to cross the membrane, establishing an osmotic gradient that promotes water uptake and causes organoid swelling. Thus, this assay provides a functional readout of CFTR activity and serves as a tool to assess the impact of CFTR mutations on ion transport and epithelial function.

Treatment with 5 µM forskolin led to significant organoid swelling in mock-infected, EBOV-infected, EBOV-ZsGreen-infected, and MARV-infected HIOs. Swelling was detectable at 24 hours post-treatment and continued through 48 hours, indicating preservation of functional cAMP-dependent ion and fluid transport in the proximal intestinal epithelium (Fig 8A). No swelling was observed in untreated control organoids over the same time course. EBOV-ZsGreen-infected HIOs showed an inverse correlation between viral load and swelling, organoids with lower fluorescence intensity—indicative of lower infection—exhibited greater forskolin-induced swelling, whereas those with higher viral burden displayed attenuated swelling (Fig 8A). MARV-infected HIOs also retained the ability to swell in response to forskolin stimulation over 48 hours, consistent with functional CFTR activity.

Unexpectedly, EBOV- and MARV-infected HCOs exhibited a complete absence of forskolin-induced swelling, with organoid size remaining comparable to untreated controls throughout the 48-hour assay period (Fig 8B). In contrast, mock-infected HCOs responded robustly to forskolin stimulation, demonstrating a significant increase in size. Quantitative assessment of cross-sectional area (CSA) confirmed this observation (Fig 8C). Morphologically, mock-infected organoids across both models exhibited thin epithelial walls and a defined central lumen, which expanded in response to forskolin. This response was absent in infected HCOs, despite the presence of similar baseline architecture.

These findings demonstrate that filovirus infection impairs CFTR-dependent fluid secretion in a region-specific manner. Infected HIOs retained partial responsiveness to forskolin stimulation, while infected HCOs exhibited a complete loss of swelling capacity, indicating a pronounced impairment in cAMP-dependent epithelial function in the distal colon. Under mock-infected conditions, morphological imaging revealed changes in organoid architecture: both HIOs and HCOs displayed thin epithelial walls and a central lumen, which expanded following forskolin treatment. In both, EBOV- and MARV-infected organoids, HIOs exhibited a significantly more pronounced swelling response compared to HCOs. While HIOs model the proximal intestine with architecture specialized for nutrient absorption and ion transport, HCOs represent the distal colon and are characterized by a denser population of goblet cells and a more complex mucosal barrier [19,34,35] (Fig 8D). These regional distinctions in morphology and function likely contribute to the differential impact of EBOV and MARV infection on fluid homeostasis along the intestinal tract.

To benchmark our findings against primary cell-derived colonic epithelial organoids (PCOs), we utilized organoids derived from two healthy adult male donors, following standard culturing, expansion, and maintenance protocols [36] (S4A Fig). PCOs were derived from two genetically distinct donors, with three or more independent experimental replicates per donor to ensure biological reproducibility. Similar to iPSC-derived HIOs and HCOs, PCOs exhibited typical multicellularity, comprising both secretory and absorptive epithelial cell types, mirroring human intestinal physiology. IHC analysis using hematoxylin (blue) and 3,3’-diaminobenzidine-based antibody staining (brown) revealed the presence of intestinal epithelial markers. Staining for chromogranin A (CHGA, enteroendocrine marker), lysozyme (LYZ, Paneth cell marker), mucin 2 (MUC2, goblet cell marker), and villin (VIL, brush border component) confirmed a fully differentiated intestinal epithelium (S4B Fig). Magnified images demonstrated CHGA localized near the basement membrane, LYZ in cytoplasmic granules, MUC2 at the apical surface, and villin distributed across the brush border, consistent with the characterization of iPSC-derived HCOs [36–38].

To evaluate the permissiveness of PCOs to EBOV infection, organoids were infected with EBOV-ZsGreen at different MOIs, and infection progression was monitored using live-cell fluorescence microscopy over 6 days. As the MOI increased, the number of EBOV-ZsGreen-positive cells rose, indicating viral replication and spread. At 2 dpi, fluorescent signals were minimal and localized to small punctate regions (S4C Fig). Robust EBOV-ZsGreen infection was observed at high MOIs, as evidenced by numerous ZsGreen-positive columnar basal cells throughout the epithelial monolayer. In contrast, mock-infected PCOs exhibited no ZsGreen-positive cells under the same fluorescence exposure (25 ms). By 6 dpi, viral spread was evident, and infected cells exhibited significant CPE, including structural changes leading to the collapse of PCO spheres. Dose- and time-dependent increases in infection efficiency were observed across MOIs ranging from 0.1 to 100, with MOI 100 reaching saturation of EBOV ZsGreen expression by 6 dpi (S4D Fig). Viral spread in PCOs was slower and less efficient compared to iPSC-derived HIOs and HCOs, with variability in infection levels across donors, suggesting donor-dependent differences in susceptibility.

To identify the intestinal epithelial cell types permissive to EBOV infection, PCOs were infected with EBOV at an MOI of 10 and analyzed at 3 dpi. We examined colocalization of EBOV protein VP35 (yellow) with epithelial markers including MUC2 goblet cell, VIL brush border component, CHGA enteroendocrine cell, and LYZ Paneth cell (magenta), along with DAPI (grey) for nuclei (S5A-S5D Fig). Colocalization was observed exclusively between EBOV VP35 and VIL, indicating that EBOV preferentially infects enterocytes, the predominant absorptive cell type. The colocalization occurred predominantly at the apical brush border of enterocytes (S5B Fig). Minimal colocalization with goblet cells, enteroendocrine cells, and Paneth cells likely reflects their lower abundance and the limitations of 2D sectioning in detecting colocalization in other cell types or regions.

To assess the functional capacity of PCOs for disease modeling and therapeutic applications, we performed the forskolin swelling assay. PCOs were treated with various forskolin concentrations, and swelling was monitored via microscopy. PCOs exhibited a robust, dose-dependent swelling response upon forskolin stimulation at 24 and 48 hours, demonstrating functional CFTR activity and confirming their physiological competence in ion transport regulation. (S6A Fig). PCOs were infected with EBOV or MARV at varying MOIs (0.1, 10, and 100) and treated with forskolin to assess CFTR-mediated ion transport and epithelial function. Mock-infected PCOs showed a clear response to forskolin treatment, with noticeable swelling. At an MOI of 0.1, swelling was pronounced in both EBOV- and MARV-infected PCOs, while at an MOI of 10, swelling was less evident, and no significant swelling was observed at an MOI of 100 for either virus. These results suggest that higher infection rates impair the organoids’ ability to respond to forskolin, indicating that severe infection disrupts the functional integrity of the organoids and their capacity to maintain cAMP signaling-dependent responses (S6B and S6C Fig), similarly to what we observed in infected iPSC-derived HCOs. These results demonstrate that increasing viral load progressively impairs CFTR-mediated swelling responses in PCOs, reflecting a loss of epithelial ion transport function during severe filoviral infection.

Despite its advantages, the model has several limitations. It lacks immune, stromal, vascular, and microbial components, which are critical for capturing the full complexity of host-pathogen interactions. Additionally, the organoids represent a fetal-like state of epithelial maturation, which may not fully recapitulate adult intestinal physiology. Our model is optimized for acute infection and does not capture long-term sequelae such as viral persistence, chronic inflammation, or post-infectious remodeling—hallmarks of filoviral disease progression [57].

We also focused on single isolates of EBOV (Mayinga) and MARV (Musoke); while well-characterized, responses may vary with other viral variants, warranting further investigation. A practical limitation is the absence of infectious viral titer measurements in supernatants. Due to the 3D structure of organoids and biosafety constraints in BSL-4 settings, viral release into the media is often restricted. Future work incorporating optimized titration and sampling methods will be needed to directly assess extracellular viral load.

Another inherent limitation involves the number of donor lines. While our study used two genetically distinct iPSC lines and validated findings in primary tissue-derived organoids from independent donors, expanding donor diversity would improve generalizability by accounting for inter-individual variation. However, the complexity of long-term differentiation protocols and BSL-4 restrictions limited our ability to include more lines. Importantly, our original design included additional iPSC donor lines (including a female donor), but technical limitations precluded successful differentiation into mature 3D gut organoids for some lines—an outcome consistent with well-documented variability in iPSC differentiation potential [69]. To mitigate this, we performed multiple independent differentiations per line and applied orthogonal validation techniques (e.g., immunofluorescence, IHC, ISH, RNA-seq) to ensure biological replication and highlight robust phenotypes unlikely to be donor-specific artifacts [70]. Nonetheless, future studies incorporating more diverse donor lines will be critical for further validation.

Despite these limitations, our iPSC-derived gut organoid platform represents a significant advance for modeling human-specific filovirus pathogenesis. Its scalability, physiological relevance, and adaptability lay a strong foundation for future co-culture models incorporating additional cellular components to better replicate in vivo complexity. As such, it holds substantial promise for elucidating antiviral defense mechanisms, guiding therapeutic development, and improving our preparedness against emerging high-consequence viral pathogens.

All differentiation experiments were performed using de-identified iPSC lines with approval by Boston University’s Institutional Review Board.

All work with wildtype and recombinant EBOV and MARV was performed in the BSL-4 facility of Boston University’s National Emerging Infectious Diseases Laboratories (NEIDL) following standard operating procedures approved by the Boston University Institutional Biosafety Committee.

HIOs and HCOs were derived from human induced pluripotent stem cells (iPSCs). Only previously reprogrammed human iPSC lines were used for this work. Parental iPSC lines, including the BU1 CG iPSC line with a CDX2/GFP reporter (SPC2-ST-B2 clone), were previously published by our group. The BU1 CG line, derived from a healthy male donor, has a normal karyotype (46XY). iPSCs were maintained in feeder-free conditions on growth factor-reduced Matrigel (Corning cat. no. 354277), using mTESR1 (StemCell Technologies), and passaged on hESC Matrigel (Corning) with ReLeSR (Corning) reagent as per manufacturer instructions. Pluripotent stem cell lines used in this study, along with maintenance standard operating procedures and directed differentiation protocols, are available from the CReM iPSC Repository at Boston University and Boston Medical Center and can be found at (https://crem.bu.edu/cores-protocols/).

At >90% confluency, iPSCs were differentiated into HIOs using an established protocol [17]. iPSC colonies were dissociated with Gentle Cell Dissociation Reagent (StemCell Technologies), re-plated at 2 × 10⁶ cells per well on Matrigel-coated six-well plates in mTeSR1 with Y27632 (5 μM) and differentiated into definitive endoderm with the StemDiff Definitive Endoderm Kit. Cells were analyzed by flow cytometry using anti-CXCR4-PE and anti-c-kit-APC antibodies. On day 3, cells were split 1:3 into new hESC Matrigel-coated plates and treated with DS/SB43 (Dorsomorphin and SB431542) and Y27632, then with DS/SB without Y27632. On day 6, cells were split again and cultured in CB/RA medium with CHIR99021, rhBMP4, and retinoic acid. Both DS/SB and CB/RA media were based on complete serum-free differentiation medium (cSFDM). A comprehensive list of reagents and catalog numbers, media recipes, and antibodies can be found in the previously published [17]. For all infection experiments, HIOs were derived from two genetically distinct iPSC lines, each subjected to three or more independent differentiations per line to ensure biological replication and assess reproducibility. A detailed protocol of HCO and HIO differentiation was deposited in protocols.io [71].

PCOs were established from adult human colonic tissue. The primary cell-derived colonic organoids used in this study were kindly provided by Dr. Ryan B. Corcoran, Massachusetts General Hospital, Boston MA. Two lines derived from de-identified healthy adult male donors were utilized.

Organoid Thawing and Plating: Organoids were thawed by removing the cryovial containing the organoids from the freezer and immediately placing it into a 37°C water bath. The cryovial gently swirled for 1–2 minutes until fully thawed. Once thawed, the cryovial was carefully removed from the water bath. Using a sterile pipette, the contents of the cryovial were transferred to a sterile 15 mL conical tube containing pre-warmed selection medium. The suspension was then centrifuged to remove the cryoprotectant. After the aspiration of the supernatant, the organoids were resuspended in fresh selection medium and plated for further culture and maintenance.

Basal Growth Medium (BGM) Preparation: BGM was prepared by combining WNT medium, R-spondin medium, DF20, and necessary supplements as outlined in the Basal Media Growth factors (S1 Table), with DF20 added to increase FBS content.

Organoid Feeding: Medium was replaced every 3–4 days or more frequently if it turned yellow. Media changes were most frequent before passaging and least frequent afterward. Old media was aspirated, and fresh media was added without disturbing the Matrigel dome. Volumes used were 1 mL for 12-well plates, and 500 µL for 24-well plates.

Organoid Passaging: Organoid passaging was performed to dissociate organoids into small cell clusters and reseed them in fresh 3D Matrigel domes, renewing the stem cell population and maintaining pluripotency. Plates were placed on ice, and the BGM was aspirated. Cell Recovery Solution (CRS, Corning; 1 mL for 12-well plates) was added to each well and incubated for 15 minutes. The 3D Matrigel dome was detached by pipetting the solution around its edges and gently breaking it into smaller pieces. The mixture was transferred to a 15 mL conical tube and incubated on ice for 1 hour to aid Matrigel dissociation. After centrifugation at 200 × g for 5 minutes, the supernatant was aspirated, and the cell pellet was resuspended in 3 mL TrypLE and incubated at 37°C for 5 minutes. The suspension was then passed through a 20G needle several times to further dissociate organoid clusters. To neutralize TrypLE, 10 mL of Splitting Media was added, and the sample was centrifuged again. The supernatant was aspirated, and the pellet was resuspended in 3D Matrigel (50 µL for 12-well plates). The plates were incubated for 30 minutes at 37°C to allow for Matrigel solidification and organoid formation, after which fresh cell culture medium was added to the wells. For all infection experiments, PCOs were derived from two independent adult donors. Organoids from each donor were expanded and used in three or more independent experimental replicates to ensure biological replication and assess reproducibility. A detailed protocol of PCO differentiation was deposited in protocols.io [71].

African green monkey kidney cells (Vero E6; ATCC CRL-1586) were used for virus propagation. The cells were maintained in Dulbecco’s modified Eagle medium (DMEM) supplemented with 200 mM L-glutamine and 10% fetal bovine serum (FBS). Cell culture medium was supplemented with 100 µg/mL Primocin. Cells were maintained at 37°C and 5% CO₂.

EBOV isolate Mayinga (GenBank accession number NC_002549) and MARV isolates Musoke (GenBank accession number NC_001608) were kindly provided by H. Feldmann, NIH NIAID Rocky Mountain Laboratories, Hamilton, MT. Recombinant EBOV expressing ZsGreen as a reporter protein (rEBOV-ZsGreen-2PA-VP40; based on Mayinga isolate) was generated in the Mühlberger lab [21]. Filoviruses were propagated in Vero E6 cells as described before [72]. Cell supernatants were centrifuged at 5,250 × g for 10 min at 4°C to remove cellular debris. Clarified supernatants were purified over a 20% sucrose cushion by centrifugation at 80,000 x g for 2 hours at 4°C and viral pellets were resuspended in PBS. Viral titers were determined by tissue culture infectious dose 50 (TCID₅₀) assays using the Spearman Karber algorithm [73].

Live-cell imaging of organoids infected with EBOV-ZsGreen was performed in a BSL-4 setting using the EVOS M5000 Imaging System (Thermo Fisher Scientific). Organoids were embedded in Matrigel and maintained in CKDCI or IMCK medium throughout the course of infection. Exposure time, illumination intensity, and all acquisition settings were standardized and held constant across all conditions and time points to ensure data comparability. Mock-infected organoids, maintained under identical culture and imaging conditions, were included in all experiments as negative controls to account for background fluorescence and baseline morphological variation. ZsGreen fluorescence served as a direct reporter of EBOV infection and was evaluated qualitatively and, where applicable, quantitatively using Fiji. All imaging experiments were independently repeated using multiple biological replicates.

Organoids were harvested from 3D Matrigel cultures using CRS (1 ml/well) for 30 minutes to 1 hour at 4°C. After digestion, organoids were washed with PBS to remove residual CRS. For larger organoids, settling was allowed for 10 minutes without centrifugation, while smaller organoids were briefly centrifuged for 2 minutes at <200 x g and 4°C. Non-infected BSL-2 samples were fixed in 4% paraformaldehyde (PFA) for 30 minutes at room temperature, followed by washing with PBS. Infected and non-infected BSL-4 samples were fixed for at least 6 hours with PFA following approved inactivation SOPs before removal from the BSL-4 lab. After fixation, organoids were transferred to individual Eppendorf tubes and briefly centrifuged for 2 minutes at <200 x g at room temperature after each step to pellet, ensuring minimal damage. For permeabilization, organoids were incubated in PBS-Tx (PBS + 0.5% Triton X-100) for 15 minutes at room temperature. Blocking was performed by incubating the organoids for 60 minutes at room temperature in 500 µL of blocking buffer (PBS-Tx + 4% goat serum) while gently rocking. Primary antibody incubation was performed overnight at 4°C using the respective primary antibodies (S2 Table) diluted in blocking buffer. After the primary antibody incubation, organoids were washed at least three times in PBS-Tx for 30 minutes per wash on a rotating shaker. Secondary antibodies (S2 Table) were diluted in PBS and incubated for one hour at room temperature. Following the incubation, organoids were washed at least three times in PBS-Tx, with each wash lasting 30 minutes on a rotating shaker.

Nuclear staining was carried out by incubating organoids with Hoechst 33342 (1:2,000 dilution) for 10 minutes at room temperature. Hoechst was preferred over DAPI to avoid edge effects on organoid imaging. After the final wash in PBS, the organoids were resuspended in up to 50 µL of ProLong Diamond Antifade Mountant (without DAPI) and applied to the center of a cavity slide. A coverslip was placed on top, ensuring the slide was flipped onto the coverslip, allowing the organoids to sink toward it and preventing them from settling at the bottom of the cavity, which would impair imaging. Organoids were imaged using a Zeiss LSM 710-Live Duo Confocal microscope with two-photon capability at 10x, 20x and 63x for high resolution visualization.

The NIR Live/Dead staining and FACS analysis were performed as follows: NIR dye was reconstituted in DMSO, wrapped in aluminum foil to protect from light, and stored at -20°C. A working solution of NIR Live/Dead stain was prepared by diluting the stock in PBS. FACS buffer was prepared by adding 0.5% PBS and 2 mM EDTA. Organoids were dissociated by adding 1 mL of CRS to each well containing the 3D Matrigel pellet, followed by a 30-minute incubation at 4°C. The Matrigel droplets were then transferred to 1.5 mL tubes using a P1000 pipette with the tip cut off to minimize dissociation. The wells were washed with 500 µL PBS to collect remaining cells, and the suspension was centrifuged at 200 × g for 5 minutes. The supernatant was aspirated, and the pellet was resuspended in 1 mL of the appropriate medium (CKDCI for distal HIOs, CKDI for distal-cAMP HIOs, or IMCK for proximal HIOs). The organoid suspension was then stained with 1:800 diluted NIR Live/Dead stain in 250 µL PBS, mixed thoroughly, and incubated for 30 minutes at room temperature, protected from light. After incubation, the cells were washed with FACS buffer and resuspended in 4% paraformaldehyde (PFA) for a minimum of 6 hours, following the formalin/aldehyde BSL4 inactivation protocol. After inactivation, the samples were transferred to a BSL2 setting for further processing. In the BSL2, the cells were washed with FACS buffer, permeabilized with 0.5% Triton X-100 in PBS for 15 minutes, and blocked with 250 µL of 5% goat serum diluted 1:100 in PBS for 15 minutes. After blocking, the cells were washed and incubated with the primary antibody, diluted in blocking buffer, for 1 hour at room temperature in the dark. Following primary antibody incubation, the cells were washed with FACS buffer and incubated with the secondary antibody diluted in PBS for 1 hour at room temperature. The used antibodies are listed in S2 Table. The cells were washed three times with FACS buffer and resuspended in 350 µL FACS buffer. The samples were analyzed by FACS on a Stratedigm machine, with gating performed first on live/dead cells and then on infection status. Appropriate mock controls were included for all experimental conditions, including mock unstained, virus unstained, mock + NIR, virus + NIR, mock + virus secondary antibody only, virus + virus secondary antibody only, and mock or infected cells with virus antibody at a 1:200 dilution.

Organoids were fixed in 10% neutral-buffered formalin for 72 hours in compliance with institutional standard operating procedures. Organoids were placed in HistoGel (Epredia, Kalamazoo, Michigan, USA) and processed routinely as formalin fixed paraffin embedded (FFPE) blocks. 5 µm sections were cut for subsequent staining. IHC was performed on a Ventana Discovery Ultra autostainer (Roche Diagnostics, Indianapolis, IN, USA). Pretreatment was performed with Benchmark Ultra CC1, a Tris-based antigen retrieval buffer, at 95°C for 32 minutes. Pre-diluted secondary HRP polymer antibodies were used for developing all primary antibodies (MP-7451 or MP-7452, Vector Laboratories, Newark, CA, USA) for 20 min at 37°C following a protein blocking step with Akoya Opal Diluent/Block (ARD1001EA, Akoya Biosciences, Marlborough, MA, USA). Slides were counterstained with hematoxylin or DAPI and coverslipped with Micromount mounting media (Leica, Wetzlar, Germany) or Prolong Gold Antifade Mountant (Invitrogen, Waltham, MA, USA). Information about the used antibodies and parameters for ISH and IHC assays are provided in S3 Table. Whole slide-images were acquired with a PhenoImager HT Automated Quantitative Pathology Imaging System, which included onboard unmixing to minimize background signal (Akoya Biosciences). Exposures for all Opal dyes on the Vectra were set based upon regions of interest with strong signal intensities to minimize exposure times and maximize the specificity of signal detected.

Fluorescent ISH-IHC images were analyzed using HALO software (v4.05107.357; Indica Labs, Inc., Corrales, NM). View settings were optimized to enhance the visibility of immunomarkers and reduce background noise by adjusting threshold gates for minimum signal intensity. Organoid regions were annotated using the MiniNet classifier in HALO AI (Indica Labs). The classifier was trained on several organoids via manual annotations, and after it was run across all fluorescently labeled slides, its automated annotations were reviewed in detail. Any inaccurate annotations were manually corrected or removed. To quantify signal from MX1, IFNb, CXCL10 probes, or viral immunoreactivity, we used the HALO Area Quantification (AQ) module (v2.4.9). The algorithm was customized to detect positive fluorescent signal based on specific color and intensity parameters for each target, and it was run across annotations made by the MiniNet classifier. The module output the percentage of the annotated slide area that displayed positive signal. The resulting data were exported as a.CSV file and analyzed in GraphPad Prism (v10.2.0; Dotmatics, San Diego, CA).

For brightfield IHC with DAB chromogen, whole-slide images were likewise analyzed in HALO using the brightfield Area Quantification (AQ) module (v2.4.9). The algorithm was configured to separate DAB (positive stain) from hematoxylin (counterstain) using color deconvolution, and a positivity threshold was established based on representative control tissues. Organoid regions were annotated with the flood-selection tool, with any inaccurate annotations were manually corrected or removed, and only annotated areas were included in analysis. Within each annotation, the module calculated the proportion of tissue area positive for DAB signal. These values were exported as summary statistics (percent positive tissue area) and used for downstream comparisons in GraphPad Prism. For all analyses, including those shown in S3 Fig, multiple images were quantified per biological replicate. Specifically, at least five representative fields of view per sample (n = 3 biological replicates) from two independent infections were analyzed to ensure robust and reproducible sampling and to avoid potential bias from relying on single representative images.

Organoids infected with EBOV or MARV were harvested for RNA analysis at 1 and 3days dpi. To harvest organoids, culture medium was carefully aspirated from each well while maintaining the integrity of the 3D Matrigel domes. A volume of 500 µL chilled CRS was added to each well, and plates were incubated at 4 °C for 30 minutes to facilitate Matrigel dissociation. The resulting suspension was transferred to 1.5 mL tubes. Wells were subsequently rinsed with 500 µL of PBS to recover any residual organoids or cellular debris, which was combined with the corresponding sample.

Samples were centrifuged at 200 × g for 5 minutes at 4 °C, and the supernatant was carefully aspirated. 1 mL of TRIzol reagent (Thermo Fisher Scientific) was added directly to the pellet, followed by a 10-minute incubation at room temperature. Samples were vortexed thoroughly to ensure complete homogenization. Inactivated samples were transferred to the BSL-2 laboratory in accordance with approved institutional inactivation SOPs.

Total RNA was isolated from human iPSC-derived organoids using TRIzol reagent (Thermo Fisher Scientific), following the manufacturer’s protocol. Briefly, 200 µL of chloroform was added per 1 mL of TRIzol, followed by centrifugation at 12,000 × g for 15 minutes at 4 °C to separate the phases. The aqueous phase was collected, and RNA was precipitated with isopropanol, washed with 75% ethanol, and resuspended in nuclease-free water. RNA concentration. Residual genomic DNA was removed by DNase I treatment. RNA-seq libraries were prepared using a poly(A) selection strategy to enrich for mRNA. Libraries were sequenced on an Illumina NextSeq 2000 platform using a P3 flow cell (2 × 50 bp paired-end reads), generating approximately 1.1 billion total reads per run.

The quality of the raw data was assessed using FastQC v.0.11.7 [75]. The sequence reads were aligned to the GRCh38 reference with added sequences of EBOV and MARV using STAR v.2.6.0 [76]. Counts per gene were summarized using the featureCounts function from the subread package v.2.0.3 [77]. The edgeR package v.4.2.0 [78] was used to import, organize, filter and normalize the counts and the matrix of counts per gene per sample was then analyzed using the limma/voom normalization method [79]. Genes were filtered based on the standard edgeR filtration method using the default parameters for the “filterByExpr” function, which excludes genes with low expression across all samples. Specifically, genes with fewer than 10 counts were removed from the dataset prior to differential expression analysis. After exploratory data analysis with Principal Component Analysis (PCA), contrasts for differential expression testing were done for each of the infected samples vs mock-infected controls at each time point. The limma package v.3.60.0 [79] with its voom method, namely, linear modelling and empirical Bayes moderation was used to test differential expression (moderate t-test). P-values were adjusted for multiple testing using Benjamini-Hochberg correction (false discovery rate-adjusted p-value; FDR). Differentially expressed genes for each comparison were visualized using Glimma v.2.14.0 [80], and FDR < 0.05 was set as the threshold for determining significant differential gene expression. Functional predictions were performed using the fgsea v.1.30.0 package [81] for gene set analysis.

The RNA-seq data have been deposited in the Gene Expression Omnibus (GEO) under accession numbers GSE298600 and GSE300073.

HIOs were dissociated as described in the infection section. Following dissociation, the HIOs were washed with PBS and centrifuged at 200 × g for 3 minutes. The resulting cell pellet was resuspended in fresh 3D Matrigel at 37°C, ensuring a low density to prevent significant spheroid overlap and facilitate downstream imaging and analysis. The resuspended HIOs were plated as droplets onto pre-warmed 37°C tissue culture plates, coated with medium, and allowed to recover for 48–72 hours. The CKDCI medium was supplemented with either forskolin or DMSO control. After 24 hours, high-resolution images of the spheroids were captured using the EVOS M50000 Imaging System. The organoids were then replenished with fresh CKDCI or IMCK medium containing either the 5 µM forskolin (Sigma, catalog #F3917) or vehicle control. After 24 hours, images were acquired and analyzed using the same parameters applied to the pre-forskolin (baseline) images. This procedure was repeated at each designated time point 24, 48, 72 hours post treatment. All images were saved in.tif format and analyzed using OrganoSeg, an open-source MATLAB plug-in previously described [82]. To quantify the cross-sectional area (CSA) of all organoids per well (referred to as “whole-well CSA”), images were segmented using the following parameters: intensity threshold of 0.5, window size of 100, and size threshold of 100. Post-segmentation images were manually reviewed for quality control, excluding spheroids that were on the edge of the field, had burst or flattened, or were not present in both pre- and post-treatment images. OrganoSeg provided output metrics including total sphere number and CSA. “Normalized” whole-well CSA was calculated by comparing pre- and post-forskolin images using the equation: [Whole well CSA post/Whole well CSA pre x100]. Control wells, treated with DMSO (vehicle for forskolin) and/or intestinal media, were included in each experiment. The CSA measurements from control wells were subtracted from those of experimental wells. Each independent experiment included at least three replicate wells per treatment condition. Statistical analyses were performed using GraphPad Prism. Comparisons between treatment groups were conducted using two-way ANOVA followed by Tukey’s multiple comparisons test with a 95% confidence interval. Data are presented as mean ± standard error of the mean (SEM), and p-values ≤ 0.05 were considered statistically significant.