Functional Characterization of Organoids Derived From Irreversibly Damaged Liver of Patients With NASH

Abstract

BACKGROUND AND AIMS:

NASH will soon become the leading cause of liver transplantation in the United States and is also associated with increased COVID-19 mortality. Currently, there are no Food and Drug Administration–approved drugs available that slow NASH progression or address NASH liver involvement in COVID-19. Because animal models cannot fully recapitulate human NASH, we hypothesized that stem cells isolated directly from end-stage liver from patients with NASH may address current knowledge gaps in human NASH pathology.

APPROACH AND RESULTS:

We devised methods that allow the derivation, proliferation, hepatic differentiation, and extensive characterization of bipotent ductal organoids from irreversibly damaged liver from patients with NASH. The transcriptomes of organoids derived from NASH liver, but not healthy liver, show significant up-regulation of proinflammatory and cytochrome p450–related pathways, as well as of known liver fibrosis and tumor markers, with the degree of up-regulation being patient-specific. Functionally, NASH liver organoids exhibit reduced passaging/growth capacity and hallmarks of NASH liver, including decreased albumin production, increased free fatty acid–induced lipid accumulation, increased sensitivity to apoptotic stimuli, and increased cytochrome P450 metabolism. After hepatic differentiation, NASH liver organoids exhibit reduced ability to dedifferentiate back to the biliary state, consistent with the known reduced regenerative ability of NASH livers. Intriguingly, NASH liver organoids also show strongly increased permissiveness to severe acute respiratory syndrome–coronavirus 2 (SARS-CoV-2) vesicular stomatitis pseudovirus as well as up-regulation of ubiquitin D, a known inhibitor of the antiviral interferon host response.

CONCLUSION:

Expansion of primary liver stem cells/organoids derived directly from irreversibly damaged liver from patients with NASH opens up experimental avenues for personalized disease modeling and drug development that has the potential to slow human NASH progression and to counteract NASH-related SARS-CoV-2 effects.

NAFLD is the most prevalent chronic liver disease worldwide and is closely associated with type 2 diabetes mellitus.^(1,2) NAFLD, defined by hepatic fat accumulation exceeding 5% of the liver weight, can be asymptomatic and reversible. However, a subset of patients with NAFLD will develop NASH, characterized by inflammation that often progresses to cirrhosis and HCC. NASH is rapidly becoming the leading cause of end-stage liver disease and transplantation,^(3,4) with the prevalence of diagnosis projected to reach 18 million in the United States, Japan, and the EU 5 (England, France, Germany, Italy, and Spain) by 2027.⁽⁴⁾ Preexisting NASH is also strongly associated with severe COVID-19 outcome.⁽⁵⁾ No Food and Drug Administration–approved drugs are currently available to treat NASH or NASH-related COVID-19 complications, making their development critically urgent.

While mouse and rat models have been invaluable for defining liver disease mechanisms, it is becoming clear that many aspects are unique to primates and should be studied in primate or human tissue models. For example, diacylglyceroltransferase 2 inhibitors failed to reduce plasma triglyceride levels in rhesus monkeys, despite promising results in murine models.⁽⁶⁾ Further, comparison of liver gene expression profiles from different NASH mouse models and patients with NASH showed little overlap at the gene level.⁽⁷⁾

Hepatocyte-like cells (iHeps) derived from human induced pluripotent stem cells (iPSCs) have been used extensively for liver disease modeling but require time-consuming differentiation and quality control steps⁽⁸⁾ and, by definition, lack the precise epigenetic signature of normal or diseased human liver. Recent studies revealed that DNA methylation patterns in liver from patients with NASH are strikingly distinct from those of healthy livers. Further, it is well documented that derivation and long-term culture of iPSCs introduces spontaneous mutations at a relatively high rate compared to primary cells.⁽⁹⁾ Taken together, these factors may compromise the interpretation of functional data from normal donor or patient iPSC-derived iHeps.

In 2015, the Clevers group presented a method for the direct isolation of proliferating bipotent ductal organoids from transplant donor livers. These organoids, derived from cholangiocyte-like “oval” cells, are genetically an order of magnitude more stable than iPSCs and are amenable to both genetic manipulation and clonal expansion.^(9,10) Soon after exchanging expansion with differentiation medium, three-dimensional (3D) organoid growth slows, and the cells within the organoids start to acquire many properties of adult human hepatocytes and can engraft in fumarylacetoacetate hydrolase double knockout mouse livers after splenic injection.⁽⁹⁾

While the epigenome of healthy liver–derived organoids may reflect that of healthy liver tissue, it likely cannot faithfully reflect NASH liver. It follows that patient liver is an attractive source of organoids that more accurately represent the disease than any alternative; however the capacity of diseased liver tissue to yield viable organoids has not been established. Here, we present data showing that irreversibly damaged liver from patients with NASH can consistently give rise to long-term expandable bipotent ductal organoids that readily undergo hepatic differentiation and functionally recapitulate important aspects of patient-specific NASH pathology.

Materials and Methods

ORGANOID ISOLATION, SPLITTING, AND PASSAGING

Wedge biopsies received from the Penn Transplant Institute at the University of Pennsylvania from both donor and explant livers were taken from the peripheral liver edge to assure comparable location within the liver across samples. Patient consent was obtained (Penn institutional review board protocol #828085). For isolation of organoids a modified version of a previously published protocol⁽¹⁰⁾ was used. Details are described in the Supporting Information. For splitting and passaging, organoid culture droplets were scraped from the surface of the well and transferred to a 15-mL tube with cold basal medium. The mixture was incubated on ice for 5–10 minutes, followed by pipetting up and down 20 times with a p10 tip attached to a p1000 tip. Cold basal medium (was added, 2 mL per culture well. Samples were mixed and spun for 5 minutes at 1,300 rpm. Basal medium was removed, and the pellet was resuspended in fresh basement matrix and seeded onto suspension plates. Plates were incubated for 10–20 minutes at 37°C and then overlaid with expansion medium. For additional experimental details, required equipment, composition of all solutions, and media see Supporting Information.

HEPATIC DIFFERENTIATION, DUCTAL DEDIFFERENTIATION, AND HEPATIC REDIFFERENTIATION OF ORGANOIDS

Expanding organoids were differentiated by supplementing the expansion medium with bone morphogenetic protein 7 for approximately 5 days before switching to complete differentiation medium. Hepatic organoids were considered differentiated and ready for assessment after 11–14 days in differentiation medium. For ductal dedifferentiation, hepatic organoids were split into fresh basement matrix, overlaid with expansion medium containing Y-27632, and cultured for at least 1 week before routine passaging. For redifferentiation, dedifferentiated expanded organoid cultures were subjected to the same differentiation procedure that was used for the first differentiation. For detailed protocols including for doubling time determination, Ki67 assay, media composition, and required equipment, see Supporting Information.

RNA ISOLATION FROM ORGANOIDS, EXPLANT LIVER, HUMAN PRIMARY HEPATOCYES, iPSC-iHeps

Organoid cultures from 6 patients with NASH and 3 healthy donors were cultured as described for each experiment and were either in the biliary state or the hepatic state. Individual wells were harvested in TRIzol and stored at −80°C. For further details, see Supporting Information.

PREPARATION AND ANALYSIS OF RNA-SEQUENCING LIBRARIES

Individual complementary DNA libraries were synthesized from 50 ng of purified RNA as described in the Supporting Information. For each experimental condition, three parallel cultures and libraries were generated for sound statistical data. (1) Global transcriptome analysis. Upon completion, all libraries were subjected to multidimensional scaling to assess comparability. (2) Heatmap. For comparison of liver organoids, human primary hepatocytes and iPSC-iHeps, a heatmap consisting of genes characteristic for either hepatocytes or cholangiocytes was generated as described in the Supporting Information. Differential gene expression analysis algorithm EdgeR (EdgeR) algorithmic analysis provided outputs such as log2 fold change values for differential gene expression analysis and log10 P values for significance assessment. (3) Ingenuity Pathway Analysis (IPA). Genes above the ±1.5 log2 fold change and 0.05% false discovery rate (FDR) value thresholds were input into the online IPA portal to identify pathological pathways related to the differentially expressed gene profile. Gene expression data have been deposited at the NCBI Gene Expression Omnibus (GEO) database (GEO accession # GSE180882). (4) Establishment of the manually procured gene list. See Supporting Information. (5) Database for Annotation, Visualization, and Integrated Discovery (DAVID) functional annotation. See Supporting Information. For protocol modifications, instrument specifications, and further details of data analysis, see Supporting Information.

ALBUMIN SECRETION, AMMONIA DEGRADATION, CYTOCHROME P450 3A4 ACTIVITY, AND FORSKOLIN SWELLING ASSAY

See Supporting Information.

FREE FATTY ACID–INDUCED LIPID ACCUMULATION AND APOPTOSIS

Expanding organoids were split and seeded into 96 wells, and differentiation medium was added immediately to induce flat cell growth. After either 7 days or 12 days of culture, apoptosis and lipid accumulation were induced by the indicated amounts of palmitic acid and oleic acid for 24 hours, respectively. For simultaneous measurement of apoptosis and lipid accumulation, a mix of the caspase 3/7 detecting green fluorescent reagents Cell Event (1:1,000 dilution), red LipidTox (1:1,000 dilution), and nuclear stain Hoechst (1 μM) was added. Imaging was done on a Nikon TE2000 inverted microscope, and digital images were analyzed by Nikon Elements software. For further details, see Supporting Information.

VIRAL INFECTION

Severe acute respiratory syndrome–coronavirus 2–vesicular stomatitis virus expressing red fluorescent protein (SARS-CoV-2-VSV-RFP) pseudovirus and VSV-RFP was generated according to Lederer et al.⁽¹¹⁾ and as described in the Supporting Information. Logarithmically growing biliary organoids were mechanically disrupted as described in the splitting procedure, and virus was added at a multiplicity of infection of 0.5 (SARS-CoV-2-VSV-RFP pseudovirus) or 1.25 (VSV-RFP) and subjected to spinning in a tabletop culture plate centrifuge at 600 rpm for 1 hour at 37°C, followed by incubation of 3 hours in the CO₂ incubator. Organoids were seeded onto 24-well plates suspended in basement membrane extract, taking care not to disturb the transduced organoids. After 24-hour culture, images were taken with a Nikon TE2000 inverted fluorescent microscope and analyzed by Nikon Elements software. For additional details, see Supporting Information.

Results

BIPOTENT DUCTAL ORGANOIDS CAN BE DERIVED FROM SMALL AMOUNTS OF EXPLANT LIVER FROM PATIENTS WITH NASH

We successfully derived bipotent ductal organoids from 300–500 mg of the peripheral liver edge of patient liver explants (Table 1 and Fig. 1), otherwise discarded after transplantation, by implementing key modifications to a previously published protocol.⁽¹⁰⁾ To our surprise, no explant sample failed to yield stem cells capable of long-term expansion (12 samples as of September 2019; see Fig. 1). The observed organoid yield for explants was on average ~5-fold lower than for healthy livers, likely due to patient-specific characteristics, as storage medium was identical and time from liver retrieval until stem-cell isolation was between 17 and 20 hours in all cases (Table1; Supporting Fig. S1). We also showed that 100 mg liver from patients with NASH is still sufficient with a proportional reduction in initial yield but no impact on morphology and growth properties (Supporting Fig. S1A). Transplant donor organoids were all derived from small healthy tissue samples (50–100 mg) (Table1), which are harder to obtain, in part due to the requirement for patient or familial consent; therefore, all control liver data in this study are based on the analysis of 3 transplant donors with healthy livers, 2 living and 1 deceased; and they are referred to as “healthy organoids,” while organoids from diseased livers are referred to as either as “NASH organoids” or more broadly as “cirrhotic organoids.”

TABLE 1.

Characteristics of healthy donors and NASH patients and their corresponding liver derived organoids

ID#	Ethnicity	Gender	Donor type or Transplant Recipient	History of fatty liver	Age first diagnosed with fatty liver	Age first diagnosed with NASH, AC or CF	Age first diagnosed with liver cirrhosis	Age at Transplantation	Presence of type 2 diabetes	BMI	Alcohol consump-tion (standard drinks per week)	Smoker status 0 = never smoker 1 = former smoker 2 = current smoker	If smoker cigarettes/day	MELD score	ALT	AST	Albumin	Presence of cirrhosis based on general appearance of biopsy/explant (see also Fig. 1)	Fibrosis stage of biopsy/explant (Masson’s trichrome stain)	Steatosis of biopsy/explant (reticulin stain)	Intrahepato-cellular PAS positive A1AT globules	Intrahepato-cellular hemosiderin (iron stain)	Kupffer cell hyperplasia (PAS stain)	Number and size of visible foci of HCC or CC of explant liver	Other notable features	Time from harvest till organoid derivation
FDD	NA	Female	Deceased Donor	no	N/A	N/A	N/A	62	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	normal	no fibrosis	none	none	none	none	none	none	~20 hours
FLD	non hispanic white	Female	Living Donor	no	N/A	N/A	N/A	29	no	26.59	0	2	1–5 c/d	N/A	N/A	N/A	N/A	normal	no fibrosis	none	none	none	none	none	none	~20 hours
MLD	non hispanic white	Male	Living Donor	no	N/A	N/A	N/A	28	no	26.08	10	0	N/A	N/A	N/A	N/A	N/A	normal	no fibrosis	none	none	none	none	none	none	~2 hours
NASH1	non hispanic white	Male	Recipient	yes	43	43	43	49	yes	43.76	0	0	N/A	31	56	61	3.2	very cirrhotic,	stage 4	<5%	yes, presence of M1Z alpha antitrypsin phenotype	none	yes, mild	1, 2.1cm CC	portal vein thrombosis, von Meyenburg complex, glycogenated hepatocellular nuclei	~20 hours
NASH2	non hispanic white	Male	Recipient	yes	50	50	50	53	yes	46.28	0	2	1–5 c/d	34	22	50	2.8	cirrhotic	stage 4	<5%	none	none	none	1, 1.1cm HCC	hepatic encephalopathy, ascites, stage III chronic kindney disease, portal hypertension	~20 hours
NASH3	non hispanic white	Male	Recipient	yes	66	67	67	67	yes	30.12	2	0	N/A	28	67	68	4.1	very cirrhotic	stage 4	10% macrove-sicular steatosis	none	none	small Kuffer cell aggregates	1, 1.1cm HCC		~20 hours
NASH4	non hispanic white	Male	Recipient	yes	N/A	64	67	67	yes	27.5	0	0	N/A	32	85	73	2.6	cirrhotic	stage 4	<5%	none	none	none	2, 1.0cm and 1,1cm HCC		~20 hours
NASH5	non hispanic white	Female	Recipient	yes	N/A	49	55	55	yes	27.41	0	1	1–5 c/d	28	45	85	2.3	cirrhotic	stage 4	<5%	yes, presence of M1Z alpha anti trypsin phenotype	none	yes, mild	none	jaundice, hyperparathyroidism, GI bleeding, portal hypertension	~20 hours
NASH6	non hispanic white	Male	Recipient	yes	N/A	60	61	63	yes	28.91	0	1	>11 c/d	30	107	187	2.6	cirrhotic	stage 4	<5%	none	yes	yes, mild	2, 1.3 cm and 2.6cm HCC		~20 hours
AC (alcohol-associated cirrhosis)	non hispanic white	Female	Recipient	yes	N/A	N/A	N/A	55	no	20	heavy drinker 2006–2018, quit 1 year before transplant	1	1 pack/day for 37years quit 1 year before transplant	18	18	39	2.7	cirrhotic	stage 4	<5%	in scattered periportal hepatocytes, alpha anti trypsin phenotype N/D	none	none	none	portal hypertension, ascites, hepatic hydrothorax	~17 hours
CF (cystic fibrosis ΔF508)	non hispanic white	Female	Recipient	yes	N/A	as child	N/A	26	yes, cystic fibrosis related	17.92	0	0	N/A	10	25	22	4	cirrhotic	stage 4	<5%	none	none	none	none	presesence of cystic fibrosis homozygous Δ F508 allele, bile duct paucity, chronic upper respiratory infections, ascites, esophageal varices	~20 hours

FIG. 1.

Derivation of long-term proliferating bipotent ductal organoids from irreversibly damaged liver from patients with NASH. (A) Schematic of bipotent ductal organoid derivation, propagation, hepatic differentiation, and growth characteristics. (B) Efficiency of derivation is lower for organoids from cirrhotic NASH, AC, or CF liver compared to healthy liver (left panel, individual yields; right panel, average yields). (C) Average doubling times of NASH, AC, and CF liver organoids are significantly longer compared to those of healthy liver organoids (D) Long-term culture capacity of NASH, AC, and CF liver organoids is significantly reduced compared to healthy liver organoids. (E) Bipotent ductal organoids were isolated from 50–100 mg of healthy donor liver and from 300–500 mg or 100 mg of NASH, AC, or CF patient liver explants. All organoids were cultured for 5–6 passages and hepatically differentiated for ~12 days. Representative images of their development throughout long-term culture and differentiation are shown. In contrast to all other NASH organoids, a significant percentage of NASH2 organoids show irregular morphology that persists throughout passaging but not differentiation.

DUCTAL ORGANOIDS ISOLATED FROM HUMAN NASH EXPLANT LIVER EXHIBIT DEVELOPMENTAL DELAY, REDUCED MAXIMAL PASSAGE NUMBERS, AND INCREASED DOUBLING TIMES

Liver organoids from patients with NASH exhibited a significant developmental delay after isolation (Supporting Fig. S1A), taking up to 2 weeks longer than donor organoids to reach similar sizes and densities. This could not be explained by differences in the amount of starting material, variations in isolation procedure, or variation in media preparations as all patient-derived organoids continued to exhibit a reduced growth rate throughout long-term culture, with NASH 3 and NASH 4 being the most severely affected (Fig. 1). A high percentage of NASH 2 organoids exhibited irregular shapes, while all others consistently displayed a spherical morphology (Fig. 1E). All liver organoid cultures continued to expand without changes to their characteristic growth speed and shape until about passage 6–12, after which NASH organoid lines slowed down at various rates and stopped dividing after passage 8–16 (Fig. 1D). In striking contrast, all healthy liver organoids continued to expand normally until at least passage 17, slowed down expansion only slightly thereafter, and grew to date until passage 19 (with longer passaging not attempted), consistent with earlier reports^(9,10) (Fig. 1D).

HUMAN HEPATICALLY DIFFERENTIATED, LIVER-DERIVED ORGANOIDS EXHIBIT A HYBRID BILIARY HEPATIC TRANSCRIPTOME SIMILAR TO iPSCs-iHeps

Transcriptome analysis reveals that biliary organoids exhibit higher cholangiocyte marker expression and much higher cell cycle gene expression compared to their hepatically differentiated, growth-arrested counterparts, although they also express some hepatocyte markers (Supporting Fig. S1B). The marker genes were selected based on their ability to distinguish between all liver cell types as determined by recent single-cell RNA-sequencing (RNA-seq) data of whole human liver.⁽¹²⁾ Our adult liver–derived, hepatically differentiated organoids and iPSCs-iHeps express both hepatocyte markers (albumin, apolipoprotein E, argininosuccinate synthetase 1, apolipoprotein B, cytochrome P450 3A4 [CYP3A4], LDL receptor) and cholangiocyte markers (epithelial cell adhesion molecule and keratins 7, 18, and 19) (Fig. 2A). They are also similar to human fetal hepatobiliary hybrid progenitor cells recently identified by singe-cell RNA-seq,⁽¹³⁾ and distinct from primary human primary hepatocytes that do not express cholangiocyte markers (Fig. 2A).

FIG. 2.

All human liver–derived organoids express markers of hepatocytes and cholangiocytes, but NASH liver–derived organoids exhibit patient-specific dysregulation of NASH-related pathological pathways. The three organoid lines from healthy donors (FDD, FLD, and MLD), the 6 NASH, as well as the AC and CF liver lines were expanded and differentiated for 12 days in differentiation medium in triplicate cultures. RNA was isolated and subjected to RNA-seq. The transcriptomes of the three healthy liver organoids were averaged and used as a baseline to evaluate log2 fold changes of the individual cirrhotic organoid lines, using the EdgeR algorithm. (A) Heatmap showing hepatocytes and cholangiocyte marker expression of liver organoids compared to iPSCs-iHeps and primary human hepatocytes. All data are from RNA-seq libraries. Selection of liver cell type-specific markers was based on single-cell RNA-seq of human liver.⁽¹²⁾ (B,C) IPA disease pathways up-regulated or down-regulated in six combined NASH organoids compared to three combined healthy liver organoids, using all Pubmed data (B) or only human liver–relevant data (C). Activation z scores of ±2 or higher are considered of biological significance. IPA indicates oncogenic, proapoptotic, and proinflammatory but at the same time strongly growth-arresting gene expression signatures in the six combined NASH liver organoids. If only human liver–specific literature is taken into account (C), several pathways that attract immune cells to the liver emerge in the NASH organoids. Down-regulated IPA pathways are almost all cell cycle–related (B,C). (D) IPA-defined apoptosis-related gene expression pattern for the six combined NASH liver organoids versus the three combined healthy liver organoids. The graph shows the directionality and extent of expression change for IPA apoptosis genes. Genes regulated in a proapoptotic direction are green, and those in an antiapoptotic fashion are red. Most genes are changed toward the apoptosis pathway. While these data are obtained from 3D organoids, similar results are seen also when the same organoids are grown under 2D conditions (see Supporting Fig. S5). Abbreviations: EC, endothelial cell; LSEC, liver sinudoidal endothelial cell. All abbreviations for gene names are according to the online human gene database GeneCards.

HUMAN NASH LIVER–DERIVED ORGANOID TRANSCRIPTOMES SHOW UP-REGULATION OF MANY PATHOGENIC PATHWAYS

We probed the transcriptomes of hepatically differentiated organoids from 3 healthy and 8 cirrhotic explant livers, including six NASH livers, one cystic fibrosis (CF) liver, and one alcohol-associated cirrhosis (AC) liver for comparison, each cultured in triplicate and passaged 5–6 times prior to differentiation (Table 1 and Fig. 1). Multidimensional scaling of the corresponding RNA-seq libraries revealed that individual healthy liver transcriptome clusters grouped together, while the transcriptomes of individual patients with NASH were diverse and formed no observable pattern (Supporting Fig. S2A). Volcano plots indicated highly significant differential gene regulation between healthy and NASH organoids (see Supporting Fig. S2B,C).

To assess the functional roles of genes in combined NASH organoids compared to combined donor organoids, we used IPA. Up-regulated IPA disease and function pathways included organismal death, frequency of tumor, development of digestive organ tumor, apoptosis, glucuronidation of hormone, transport of ion, metabolism of vitamin, metabolism of retinoid, and inflammatory response (Fig. 2B–D). In particular, oncogenic pathways, cell death pathways, and attraction of lymphocytes (Fig. 2B–D) were prominent features of NASH organoids (Fig. 2B,C). Down-regulated pathways were almost all cell cycle–related and growth-related, confirming the reduced growth capabilities of NASH organoids.

We next assessed the regulation of individual genes to gain insight into the molecular basis of NASH pathology. We probed the differential expression of the genes from our manually procured gene list (Supporting Table S1) in the combined NASH organoids compared to combined donor organoids (Supporting Fig. S3), thus revealing potential disease pathways. While several of these pathways were also predicted by the IPA, others were new, including fibrosis and bile acid synthesis pathways (Supporting Fig. S3).

Further we assessed differential regulation of individual cirrhotic organoids using our manually procured gene list. When using as baseline the average of the combined healthy organoids for each gene, relative up-regulation of specific genes among individual patients was generally much stronger than up-regulation of these genes for the combined patients. Further, while cell cycle genes/pathways are consistently down-regulated among all individual patients with NASH, log2 fold change values vary strongly in a patient-specific manner for pathways related to lipids, fibrosis, inflammation, tumor formation, and xenobiotic metabolism (Figs. 3 and 4). For alternative analysis of patient-specific differential gene expression, we used the DAVID Functional Annotation Online Portal (Supporting Fig. S10). This analysis did in part confirm our IPA data but indicated a higher number of proinflammatory pathways including those caused by bacterial and viral infection (see Supporting Fig. S10). The following paragraphs summarize pathological pathways that are induced in NASH organoids based on our transcriptome data (Fig. 4):

FIG. 3.

Down-regulation of cell cycle genes is characteristic for all cirrhotic liver–derived organoids, while lipid metabolism changes are specific for individual cirrhotic organoid lines. The transcriptomes of the three healthy liver organoids from Fig. 2A were averaged and used as a baseline to evaluate gene expression changes of the individual cirrhotic organoid lines (from Fig. 2A) for each of the genes in the manually procured gene list (Supporting Table S1). The panels show patient-specific log2 fold gene expression changes for each of the six organoids from patients with NASH directly compared to the combined healthy donor organoids based on the EdgeR algorithm. For comparison, gene expression patterns for CF and AC organoid lines are shown. All cirrhotic liver organoid lines show strong down-regulation of cell cycle pathways but variable gene expression changes related to lipid-related pathways. EdgeR FDR * < 0.01, ** < 0.001, and *** < 0.001. Abbreviations: DNL, de novo lipogenesis; FFA, free fatty acid; PUFA, polyunsaturated fatty acid; TG, triglycerides. All abbreviations for gene names are according to the online human gene database GeneCards.

FIG. 4.

End-stage NASH hepatic organoids show patient-specific up-regulation of fibrosis, tumor, inflammation, and p450-related pathways. Log2 fold gene expression changes were determined as described in Figs. 2 and 3. Log2 fold changes for each of the six NASH, CF, and AC patient organoids versus the combined healthy organoids based on the EdgeR algorithm are shown. Fibrosis and tumor markers are up-regulated in all cirrhotic organoids. Inflammatory markers and xenobiotic markers are up-regulated in a patient-specific way. EdgeR FDR value * < 0.01, ** < 0.001, and *** < 0.001. All abbreviations for gene names are according to the online human gene database GeneCards.

Fibrosis Markers

Some, but not all, known human liver fibrosis markers are up-regulated in end-stage NASH organoids, including aldo-keto reductase family 1 member B10 (AKR1B10), which has been used as a clinical blood marker of advanced liver fibrosis.^(14,15) Additionally, the known human fibrosis-associated proteins ubiquitin D (UBD), cyclin D2 (CCND2), collagen type VI alpha 2 chain (COL6A2),^(15,16) dickkopf-related protein 3 (DDK3),⁽¹⁷⁾ and peroxidasin (PXDN)⁽¹⁸⁾ are up-regulated in at least two of our six NASH organoid populations, while secreted phosphoprotein 1 (SPP1) is found mostly down-regulated.

Tumor Markers

NASH organoids 1, 2, and 5, but not 3 and 4, overexpress the clinical Hepatocellular Carcinoma (HCC) marker alpha-fetoprotein (AFP).^(19,20) Interestingly, unlike patients 1 and 2, patient 5 does not show overt HCC lesions (Table 1), consistent with the fact that AFP overexpression is also found in some patients with isolated liver cirrhosis.⁽²¹⁾ Likewise, mucin 16 (MUC16), a well-known tumor marker,⁽²²⁾ is overexpressed in some, but not all, patient organoids. It seems highly unlikely that our NASH organoids contain HCC tumor cells because HCC organoids have a highly irregular, compact, clump-like morphology distinct from the organoids of patients with NASH.⁽²³⁾ Thus, AFP and MUC16 expression in NASH organoids may indicate merely a tumorigenic micro-environment. Importantly, mouse HCC models⁽²⁴⁾ and clinical data⁽²⁵⁾ indicate an intersection between HCC, fibrotic, and inflammatory pathways, which is consistent with the simultaneous up-regulation of these pathways in our NASH organoids.

Immune System

Strong and patient-specific up-regulation of innate and adaptive immune system components occurs despite the absence of any proinflammatory challenge in the organoid medium, although individual organoids vary considerably in their expression for many of the genes shown in Fig. 4. Up-regulated innate immune response genes in NASH organoids include PXDN, which directly binds to and kills gram-negative bacteria,⁽²⁶⁾ and serum amyloids A1 and A2 (SAA1, SAA2), which are induced in many inflammatory conditions and can act as chemokines for neutrophils and monocytes,⁽²⁷⁾ two cell types often found at elevated levels in NASH livers. Key components of the adaptive immune system are also up-regulated in the hepatically differentiated organoids of select patients, most strikingly the major histocompatibility complex (MHC) class II subunits human leukocyte antigens DRB1, DQB1, and DRB5. Up-regulated MHC class II proteins have been found specifically in the hepatocytes from NASH and alcohol-associated cirrhosis livers,⁽²⁸⁾ and it has been suggested that hepatocyte degradation, a cardinal feature of NASH, may occur as T cells are recruited and activated in the liver by MHC class II–expressing hepatocytes exposed to the blood through sinusoidal cell fenestrations.⁽²⁹⁾

The Cytochrome p450 (p450–Uridine 5′-Diphospho-Glucuronosyltransferase (UGT) System

The p450–Uridine 5′-Diphospho-Glucuronosyltrans ferase (p450-UGT) system involved in metabolism of xenobiotics, steroid hormone synthesis, retinol metabolism, and bile acid synthesis was strongly up-regulated in NASH organoids in the absence of any stimulus, in a highly patient-specific way, and at levels comparable to the corresponding livers (Supporting Fig. S6), consistent with an epigenetic memory. It is unclear if up-regulation is due to therapeutic drugs patients took before liver transplantation or to native molecules.

HUMAN NASH LIVER–DERIVED ORGANOIDS SHOW REDUCED ALBUMIN SECRETION, INCREASED CYP3A4 ACTIVITY, AND VARIABLE AMMONIA ELIMINATION

Both healthy and cirrhotic organoids show significantly higher CYP43A4 activity than the human liver cancer cell line Huh7 but with NASH1, NASH2, and CF organoids at the highest level (Fig. 5), confirming CYP3A4 mRNA expression patterns (Fig. 4). Because all organoids were grown in parallel without specific stimulus, these observed differences may be due either to endogenous NASH metabolites or to drugs the patients took before liver harvest. Albumin secretion appears to be reduced in all NASH organoids (Fig. 5), consistent with the reduced albumin production of NASH livers (Table 1). In contrast, ammonia degradation appears highly variable especially among cirrhotic organoids (Fig. 5).

FIG. 5.

NASH liver–derived organoids show reduced LDL uptake activity as well as reduced albumin production and highly patient specific ammonia elimination and CYP3A4 activity. All indicated organoid lines were hepatically differentiated for 12 days in parallel in duplicate cultures and subjected to analysis as follows: (A) LDL uptake was measured by fluorescence microscopy on flat cells emerging from 3D organoids. (B,C) Organoid cultures were double-stained for 4 hours with fluorescent LDL (green) and LipidTOX (red), and the intensity of the LDL stain per cell area (as measured by overexposing the LipidTOX channel) was deterimined using NIS elements software. Values are averages of at least eight independent images per organoid line. Relative LDL uptake is lower in NASH organoids compared to healthy organoids. *P > 0.03, **P > 0.005. (D) Ammonia elimination was done by measuring ammonia levels 24 hours after adding the ammonium elimination medium to the 3D cultures. (E) Albumin secretion was measured by ELISA of culture medium added to the organoids 24 hours before. (F) CYP3A4 activity was measured 3.5 hours after addition of detection medium, by luminescence produced by mono-oxygenation of the CYP3A4 substrate luciferin–IPA (Promega), which as a result becomes luminogenic. Abbreviation: N/D, not determined.

HEPATICALLY DIFFERENTIATED NASH ORGANOIDS EXHIBIT INCREASED SENSITIVITY TO APOPTOTIC STIMULI

Because NASH liver is known to exhibit higher apoptosis rates than healthy liver, we investigated the sensitivity of the NASH liver organoids to the potent apoptotic stimulant palmitic acid. Using a two-dimensional (2D) organoid platform we developed to allow efficient microscopy-based measurements (Figs. 5 and 6), we found overall strongly increased sensitivity to apoptosis of NASH organoids compared to healthy organoids, with the degree specific to individual organoid lines (Fig. 6A–C). Intriguingly, sensitivity levels of both healthy and NASH liver organoids were much lower than those for Huh7 cells (Fig. 6A–C), perhaps reflecting tolerance of the liver to the high concentrations of free fatty acids and triglycerides that enter through the portal vein. We are confident that these data are physiologically meaningful because transcriptome analysis of 2D and 3D NASH organoids revealed the presence of a strong apoptosis signature for both 2D and 3D NASH organoids (Supporting Figs. S4 and S5).

FIG. 6.

Hepatic NASH organoids show increased sensitivity to apoptotic stimuli and increased free fatty acid–stimulated accumulation of triglycerides. Biliary organoids at passage 5–6 were hepatically differentiated for 7 days in 2D format on a 96-well plate in the presence of 5% Matrigel in the medium. (A) Palmitic acid–bovine serum albumin (0, 1.2 and 2 mM) and Hoechst dye were added for 6 hours, and organoids were imaged by fluorescence microscopy. Palmitic acid–bovine serum albumin induced caspase 3/7 activity, and the total number of nuclei in hepatic healthy and cirrhotic liver organoids was measured by green fluorescent caspase assay and Hoechst nuclear stain, taking eight images per organoid culture. The caspase reaction product is green fluorescent and binds to nuclear DNA. As control, the liver cell line Huh7 was also analyzed. (B) Average ratios of caspase fluorescent nuclei (green) versus total nuclei by Hoechst stain (blue) for all donor organoids versus all cirrhotic liver organoids were analyzed by NIS elements software. (C) Individual organoids at 0 and 2 mM palmitic acid–bovine serum albumin show highly patient-specific caspase 3/7 activity. (D) The organoids from (A-C) and Huh7 cells were in parallel subjected to 0, 1.2, and 2 mM oleic acid–bovine serum albumin for 24 hours and stained with red LipidTOX and blue Hoechst dye. (E,F) Cell area with a defined intensity range of red LipidTOX fluorescence per total number of nuclei per image was determined. Eight images per organoid culture were taken. Images were analyzed by NIS software, and the average lipid accumulation for both the combined healthy organoids and the cirrhotic organoids was determined. (F) Lipid accumulation for individual organoids at 0 and 2 mM oleic acid–bovine serum albumin. The 1.2 mM values were only taken for the three healthy liver organoid lines and the two NASH liver organoid lines NASH 4 and 5, with the strongest 2 mM lipid accumulation. Abbreviations: OA, oleic acid; PA, palmitic acid.

HEPATICALLY DIFFERENTIATED NASH ORGANOIDS SHOW INCREASED FREE FATTY ACID–INDUCED LIPID ACCUMULATION BUT DECREASED LDL UPTAKE

NASH liver organoids grown in 2D (described in Figs. 5 and 6) show significantly higher oleic acid–induced lipid accumulation compared to healthy liver organoids (Fig. 6D–F). Interestingly 2 mM oleic acid is needed to achieve significant lipid accumulation in organoids but does not cause any apoptosis in the organoids. In contrast, 2 mM oleic acid induces strong apoptosis in Huh7 cells, as expected (Fig. 6D–F). At 2 mM oleic acid, lipid accumulation in the NASH5 organoid is about 4.5-fold higher compared to the average of healthy organoids and ~9-fold higher than in Huh7 cells, perhaps reflecting human NASH liver fat accumulation. This effect can be seen at both day 7 and day 12 of differentiation (compare Fig. 6D–F; Supporting Fig. S7C,D). In contrast, LDL accumulation is about 2-fold reduced in NASH organoids (Fig. 5A), with unclear implications for NASH patient physiology.

CF LIVER–DERIVED BILIARY ORGANOIDS EXHIBIT REDUCED FORSKOLIN-INDUCED SWELLING

Healthy ductal organoids swell within minutes of forskolin addition, consistent with their strong CF transmembrane conductance regulator (CFTR) mRNA expression, while CFTR ΔF508 organoids from CF liver did not show strong swelling (Supporting Fig. S8). This is similar to intestinal organoids and their CFTR ΔF508 counterparts⁽³⁰⁾ and expected as CFTR ΔF508 inactivates the CFTR calcium channel.

NASH LIVER ORGANOIDS SHOW REDUCED ABILITY TO DEDIFFERENTIATE FROM THE HEPATIC STATE BACK TO THE BILIARY STATE

Sophisticated murine lineage tracing indicates that hepatocytes can behave as bipotent stem cells to repopulate the biliary tree after cholangiocyte ablation⁽³¹⁾ and that cholangiocytes can revert to functional hepatocytes after chemical injury to native hepatocytes.⁽³²⁾ Thus, we hypothesized that a similar dedifferentiation might be possible through the addition of expansion medium to differentiated organoids upon splitting, allowing the cells to reenter a proliferative state. Healthy organoids cultured for 12 days in differentiation medium efficiently overcame hepatic growth arrest after being split into fresh basement matrix and expansion medium. As early as 1 week postsplitting, the degree of dedifferentiation observed was such that the growth rate and morphology were comparable to ductal organoids grown exclusively in expansion medium (Fig. 7A,B). In contrast, all NASH organoids demonstrated a significant reduction in their ability to reenter the proliferative state (Fig. 7A). This reduction is unlikely to be caused by increased apoptosis because under these conditions apoptosis is at background levels for all organoids (Supporting Figs. 6A–C and S7C,D), and the number of live cells does not decrease significantly more strongly for NASH compared to healthy liver organoids during differentiation (Fig. 7C).

FIG. 7.

NASH liver–derived organoids show decreased regenerative ability and increased permissiveness to SARS-CoV-2-VSV pseudovirus. (A) Hepatic NASH organoids fail to dedifferentiate back to the initial biliary state. Healthy and NASH liver organoids were differentiated in parallel for 13 days to the hepatic state, then split into expansion medium and further incubated for 12 days. (B) Healthy liver organoids repeatedly cycle between hepatic and biliary states. Expanding healthy organoids were hepatically differentiated for 13 days, split directly into expansion medium, cultured for 12 days, split once more, and differentiated identically to the first procedure. Black arrows indicate the time at which RNA samples were taken for RNA-seq and transcriptome analysis, depicted in Supporting Fig. S9. (C) Live cells per culture decrease only slightly during 12 days of differentiation for both healthy and NASH liver organoids. (D) SARS-CoV-2-VSV-RFP pseudovirus infection of biliary organoids derived from three healthy donor (FDD, FLD, and MLD) livers, six NASH livers, one CF liver, as well as HEK293 cells and NIH3T3 cells. Three fluorescent and phase contrast images per organoid line were taken 16 hours after infection. (E) Left panel: Low ACE2 cell line controls show little SARS-CoV-2-VSV-RFP infection. ACE2-dependent pseudovirus infectivity measured by RFP cell area per total cell area is ~9-fold increased in the diseased versus healthy organoids and only poorly infects mouse NIH3T3 or human HEK293 cells that express none or low levels of ACE2, respectively. Right panel: Regular VSV-RFP infects cirrhotic liver–derived organoids with ~4-fold increased efficiency and infects HEK293 cells efficiently as its cell entry is not dependent on ACE2. (F) Increased SARS-CoV-2-VSV-RFP pseudovirus permissiveness in cirrhotic liver organoid lines correlates more strongly with induced UBD expression and increased sensitivity to apoptotic stimulus than with increased ACE2 and TMPRSS2 expression, especially for NASH 5 and NASH 6 organoids. Apoptosis assays were those from Fig. 6, and UBD expression was taken from the RNA-seq data shown in Fig. 4. Abbreviation: PA, palmitic acid

Further, we successfully redifferentiated the two dedifferentiated healthy organoid cultures one more time through a second and identical differentiation program (Fig. 7B). RNA was isolated at the end of both the first and second differentiations for transcriptome analysis, which showed overall few changes in expression within a panel of 90 hepatically expressed genes (Supporting Fig. S9 and annotated gene panel in Table S1), indicating that normal donor organoids can repeatedly and efficiently cycle between the biliary and hepatic states (Fig. 7B). Our finding that normal hepatic organoids can revert back to a proliferative biliary-like state is consistent not only with murine lineage tracing studies^(31,32) but also with in vivo and in vitro studies using bipotent ductal mouse organoids.^(33,34)

NASH LIVER BILIARY ORGANOIDS SHOW STRONGLY INCREASED PERMISSIVENESS TO VSV AND TO SARS-CoV-2-VSV PSEUDOVIRUS

Prompted by our discovery of robust induction of the liver fibrosis marker UBD (also called FAT10)^(15,16) in NASH liver organoids and by the recent discovery that UBD is also an inhibitor of RNA virus–induced interferon signaling,^(35,36) we assessed whether NASH organoids would exhibit increased viral permissiveness. We infected healthy and NASH organoids with RFP-tagged VSV or SARS-CoV-2-VSV pseudovirus that expresses the SARS-CoV-2 spike protein and makes entry of the VSV single-stranded RNA genome fully dependent on angiotensin 1 converting enzyme 2 (ACE2) and transmembrane serine protease 2 (TMPRSS2), the two essential membrane proteins for SARS-CoV-2 viral entry. NASH liver biliary organoids are ~9-fold more permissive to SARS-CoV-2 pseudovirus and ~4-fold more to VSV compared to healthy liver. organoids (Fig. 7D,E). Further, permissiveness levels in individual organoid lines correlate better with UBD mRNA levels than with ACE2 or TMPRSS2 levels (Fig. 7F), consistent with a role for increased UBD in increased viral permissiveness of NASH liver organoids.

Discussion

We established a procedure that allows reliable derivation, propagation, and hepatic differentiation of bipotent ductal organoids from ~100 mg of irreversibly damaged liver from patients with NASH, an amount that is only ~2-fold higher than that yielded by a dual-pass 25 mm 16-gauge needle biopsy (2 × 28 mg = 56 mg), a routine scenario for liver biopsies.⁽³⁷⁾ This suggests that a relatively modest improvement to our organoid isolation procedure may be sufficient to allow use of routine liver biopsies for patient-specific organoid production. Although patients are generally reluctant to consent to this procedure, attitudes could change if functional data from organoids revealed credible routes of discovery of anti-NASH drugs. The low efficiency of biliary organoid derivation from cirrhotic NASH livers is remarkable, considering that replication of biliary cells in response to injury (ductular reaction) is a hallmark of NASH. Perhaps the inflammatory, fibrotic, and apoptotic environment in cirrhotic NASH livers weakens these cells, resulting in low survival after isolation.

Our observation that NASH patient organoids display a marked reduction in regenerative capacity as well as a universal down-regulation of major cell cycle pathways is consistent with the fact that cirrhotic NASH strongly reduces the ability of a patient liver to regenerate.⁽³⁸⁾ Remarkably, the NASH organoids exhibited also significantly increased sensitivity to apoptosis, increased free fatty acid–induced lipid accumulation, and reduced albumin secretion, all of which are associated with NASH liver. Our transcriptome analysis is consistent with these functional changes and additionally predicts inflammation, fibrosis, and tumor formation. One example of a consistently up-regulated gene in our eight cirrhotic organoid lines is the multifunctional inducible ubiquitin UBD, a known clinical fibrosis marker,^(15,16) inducer of apoptosis,⁽³⁹⁾ and component of human NASH-associated Mallory-Denk bodies.⁽⁴⁰⁾ It will be interesting to elucidate by methylome analysis if these profound gene expression changes are due to a strong epigenetic memory in the NASH organoids that originated in the corresponding NASH liver. In addition, functional verification of the inflammatory, fibrotic, xenobiotic, and tumorigenic pathways by coculture of NASH organoids with HSCs, T cells, macrophages, and/or Kupffer cells would enhance our data. In contrast, the recently described iPSC-derived organoids that contain several types of liver cells⁽⁴¹⁾ may be less capable of faithfully reflecting the environmentally determined epigenetics in the liver from patients with NASH.

Of most urgent interest, cirrhotic NASH liver organoids are more permissible to RNA viruses, including SARS-CoV-2-VSV pseudovirus and VSV, providing a possible explanation for the severe COVID-19 outcome of patients with cirrhosis and NASH.⁽⁵⁾ While SARS-CoV-2 has been shown to infect healthy biliary organoids,⁽⁴²⁾ we are currently studying if cirrhotic organoids show increased permissiveness to SARS-CoV-2. We also found that UBD levels correlate much better with viral permissiveness than ACE2 levels. Because UBD is a known inhibitor of the antiviral interferon response through direct interaction with the double-stranded RNA sensor retinoic acid–induced gene I (RIG-I),^(35,36) it will be interesting to see if therapeutic down-regulation of UBD may help inhibit proinflammatory SARS-CoV-2 effects on NASH liver. Intriguingly, the SARS-CoV-2 N and M proteins were recently shown to inhibit RIG-I-mediated interferon production,^(43,44) consistent with SARS-CoV-2 being at least in part sensed by RIG-I, as shown for other coronaviruses.⁽⁴⁵⁾

A limitation of this study is the relatively small number of donors and patients. In addition, after the start of these studies, the Clevers groups described organoids derived directly from the hepatocytes of healthy adult human liver that seem to be functionally somewhat closer to hepatocytes than hepatically differentiated bipotent ductal organoids. Because passage times for these organoids (~8 days)⁽⁴⁶⁾ are about twice as long as those for bipotent ductal organoids (~4 days) and increase even further after extended passaging⁽⁴⁶⁾ and because it may be difficult to isolate viable hepatocytes from cirrhotic liver from patients with NASH, we decided to continue with isolation of ductal organoids. Further, early–NASH stage organoids are needed to allow functional elucidation of progressive NASH pathology, which in turn may suggest specific early therapeutic interventions. In addition, combined analysis of methylomics, metabolomics, and transcriptomics across NASH stages should yield important clues regarding the mechanism(s) that triggers fibrosis, the most crucial medical aspect of NASH. Finally, human liver chimeric mouse or rat models and nonhuman primate models would enhance the functional annotation of our findings.

In summary, we show that it is possible to derive, culture, and functionally analyze organoids from irreversibly damaged human NASH liver. Our findings open up a direct pathway toward the personalized functional study of liver from patients with NASH-specific differences. Indeed, NASH patient–derived organoids, perhaps in combination with Hepatic Stellate Cells (HSCs), Liver Sinuosidal Endothelial Cells (LSECs), Kupffer cells, and/or blood derived macrophages, may allow direct screening for different classes of anti-NASH drugs for general and personalized treatments. NASH organoids show increased permissiveness to SARS-CoV-2-VSV pseudovirus, which correlates with increased levels of UBD, a known inhibitor of antiviral interferon induction. Future organoid and animal studies should reveal if there is a causal relationship between high UBD levels in preexisting NASH liver and severe COVID-19 outcome.

Supplementary Material

Additional Supporting Information may be found at onlinelibrary.wiley.com/doi/10.1002/hep.31857/suppinfo.

Abbreviations:

AC

alcohol-associated cirrhosis

ACE2

angiotensin 1 converting enzyme 2

AFP

alpha-fetoprotein

BME

basement membrane extract

CC

Cholangiocarcinoma

CF

cystic fibrosis

CFTR

CF transmembrane conductance regulator

CYP3A4

cytochrome P450 3A4

2D/3D

2-dimensional/3-dimensional

EdgeR

differential gene expression analysis algorithm EdgeR

FDR

false discovery rate

HCC

Hepatocellular Carcinoma

HSC

Hepatic Stellate Cells

iHeps

induced hepatocyte-like cells

IPA

Ingenuity Pathway Analysis

iPSC

induced pluripotent stem cell

LSEC

Liver Sinusoidal Endothelial Cells

MHC

major histocompatibility locus

p450

cytochrome p450

RIG-I

retinoic acid–induced gene I

RNA-seq

RNA sequencing

SARS-CoV-2

severe acute respiratory syndrome–coronavirus type 2

TMPRSS2

transmembrane serine protease 2

UBD

ubiquitin

UGT

Uridine 5’-Diphospho-Glucuronosyltransferase

VSV-RFP

vesicular stomatitis virus expressing red fluorescent protein