Research ArticleBioengineering

An extracorporeal bioartificial liver embedded with 3D-layered human liver progenitor-like cells relieves acute liver failure in pigs

See allHide authors and affiliations

Science Translational Medicine  08 Jul 2020:
Vol. 12, Issue 551, eaba5146
DOI: 10.1126/scitranslmed.aba5146

Beneficial bioartificial livers

Bioartificial livers are an attractive option as a bridge to transplant or to promote liver regeneration in cases of acute liver failure. Here, Li et al. tested an extracorporeal bioartificial liver system composed of human liver progenitor-like cells cultured on macroporous scaffolds in a bioreactor that provides alternating air-liquid exposure. Three hours of treatment improved survival, reducing inflammation and promoting native liver regeneration in pigs with drug-induced acute liver failure. Results suggest extracorporeal cell-based bioartificial livers may be a promising treatment for acute liver failure.


Clinical advancement of the bioartificial liver is hampered by the lack of expandable human hepatocytes and appropriate bioreactors and carriers to encourage hepatic cells to function during extracorporeal circulation. We have recently developed an efficient approach for derivation of expandable liver progenitor-like cells from human primary hepatocytes (HepLPCs). Here, we generated immortalized and functionally enhanced HepLPCs by introducing FOXA3, a hepatocyte nuclear factor that enables potentially complete hepatic function. When cultured on macroporous carriers in an air-liquid interactive bioartificial liver (Ali-BAL) support device, the integrated cells were alternately exposed to aeration and nutrition and grew to form high-density three-dimensional constructs. This led to highly efficient mass transfer and supported liver functions such as albumin biosynthesis and ammonia detoxification via ureagenesis. In a porcine model of drug overdose–induced acute liver failure (ALF), extracorporeal Ali-BAL treatment for 3 hours prevented hepatic encephalopathy and led to markedly improved survival (83%, n = 6) compared to ALF control (17%, n = 6, P = 0.02) and device-only (no-cell) therapy (0%, n = 6, P = 0.003). The blood ammonia concentrations, as well as the biochemical and coagulation indices, were reduced in Ali-BAL–treated pigs. Ali-BAL treatment attenuated liver damage, ameliorated inflammation, and enhanced liver regeneration in the ALF porcine model and could be considered as a potential therapeutic avenue for patients with ALF.


Acute liver failure (ALF) is a severe clinical syndrome with a mortality rate of up to 80% worldwide (1, 2). Liver transplantation remains the standard therapy for medically refractory ALF (3). However, organ shortages, contraindications, and lifelong immunosuppressive therapy limit the applicability of the procedure (4). Currently, several alternative therapies are under investigation, including hepatocyte transplantation, tissue engineering, and liver support systems. Supportive therapies such as cell-based bioartificial livers (BALs) are designed to detoxify blood by removing waste molecules and provide synthetic organ function, with the auspicious goal of bridging the gaps in liver transplantation or facilitating liver self-regeneration (59). The therapeutic efficacy of BAL relies on the functionality of the applied cells and the bioreactor configuration.

Human hepatocytes (10), hepatoblastoma cell lines (11), human-induced hepatocytes (hiHeps) (6), and porcine hepatocytes (5, 7) have been proposed for use in BAL devices. Human hepatocytes are generally regarded as the gold standard for clinical BAL applications, but obtaining sufficient and functional hepatocytes is often impractical due to limited availability and phenotypic instability. The hepatoblastoma cell line C3A has been used in the extracorporeal liver assist device (ELAD; Vital Therapies) (11, 12). Although cell lines are capable of in vitro expansion and secretion of albumin (ALB), they lack some hepatic functions, such as normal urea cycle enzymes (13), which can contribute to the development of cerebral edema, neurological dysfunction, and hepatic encephalopathy (HE) (14). HiHeps derived from human fetal fibroblasts have the potential for metabolic detoxification. Shi et al. (6) applied hiHeps in a BAL support system (hiHep-BAL) in a porcine model of drug overdose–induced ALF to improve survival. However, the process of preparing adequate and high-quality hiHeps to meet the needs of critical patients is cumbersome and time consuming (15). Porcine hepatocytes are the most commonly used heterogeneous cells because they have synthetic detoxification functions similar to human hepatocytes. Glorioso et al. (5, 7) successfully used porcine hepatocytes in a spheroid reservoir bioartificial liver (SRBAL) to relieve ALF in porcine models. Because porcine hepatocytes pose some ethical concerns due to the potential risk of xenozoonosis and immunological barriers (4), it is imperative to seek alternative human cell sources for BAL applications.

Another challenge in BAL applications is finding an appropriate bioreactor configuration for cell expansion and function. Bioreactors have many design challenges, such as the fabrication of three-dimensional (3D) cell patterns, the balanced delivery of oxygen and nutrients, and the amelioration of mechanical shear forces (16). Many current BALs, including ELAD (5, 12), Academic Medical Center Bioartificial Liver (17), HepatAssist System (18), and modular extracorporeal liver support (19), use hollow fiber bioreactors. In most semipermeable hollow fiber bioreactors, an external site is used for cell attachment, whereas perfusion medium is circulated through the internal lumen. These devices avoid exposing cells to shear stress but have the distinct disadvantage of mass transfer restrictions. The most widely tested device is the ELAD, consisting of a dual pump dialysis system and hollow fiber cartridges. In clinical pilot-controlled trials performed in patients with ALF, improvements were seen in mental status, renal function, and hemodynamic stability, but not in survival (11, 12). Other bioreactors such as flat plate and packed bed systems need to be further modified because of nonuniform perfusion, shear forces, low surface area–to–volume ratios, and reduced hepatic function as 2D monolayers (2022). The SRBAL allows porcine hepatocytes to form spheroid 3D structures and still remain functionally stable in suspension. SRBAL treatments improved survival and reduced hyperammonemia in porcine ALF models (5, 7, 22), highlighting the importance of the 3D configuration for BAL development.

Recently, we have converted mouse and human primary hepatocytes into expandable hepatocyte-derived liver progenitor-like cells (HepLPCs) through chemical reprogramming (23, 24). Upon in vivo transplantation in Fah−/−mice, the cumulative liver repopulation reached as high as 78.52%, and no tumorigenesis was observed (23). HepLPCs became mature hepatocytes after hepatic differentiation and were therefore competent as a prospective cell resource for BAL devices. Here, we developed an air-liquid interactive bioartificial liver (Ali-BAL) support system embedded with 3D-layered human immortalized HepLPCs (iHepLPCs). The safety and efficacy of the Ali-BAL were evaluated in a randomized experiment using a porcine model with ALF induced by d-galactosamine (d-gal).


Optimization of iHepLPCs

We previously identified small-molecule–based culture conditions that allowed for the conversion of mouse and human hepatocytes into HepLPCs in vitro (23, 24). Because most human HepLPCs can only proliferate for 10 passages or fewer, we immortalized human HepLPCs (iHepLPCs) with HPV E6/E7 genes. The mean doubling time of iHepLPCs was about 23 hours, and this growth speed could be maintained through at least 40 passages without growth arrest (25). However, the metabolic profile and secretion ability were decreased in iHepLPCs compared with the parental cells. Hepatic transcription factors including HNF1A, PROX1, FOXA3, ATF5, HNF6, CEBPA, HNF4A, and FOXA2 are known to be crucial to the determination of the fate of hepatic cells (2629); therefore, we overexpressed them in iHepLPCs by transient transfection and analyzed the changes in expressions of ALB, cytochrome P450 family 3 subfamily A member 4 (CYP3A4), carbamoyl-phosphate synthase 1 (CPS1), and other hepatic genes as surrogate maturation markers. Although most transcription factors exerted little effect on the expression of mature hepatocyte markers, the overexpression of FOXA3 in iHepLPCs significantly increased hepatic function (P < 0.05 compared to the vector control; Fig. 1A). In agreement with this observation, the expression of FOXA3 was much lower than other transcription factors in iHepLPCs as compared to fresh primary hepatocytes (F-PHH) (P < 0.0005; Fig. 1B). After overexpression of FOXA3, iHepLPCs-FOXA3 displayed more mature hepatic cell morphology (Fig. 1C). Cell proliferation was slightly slower, with doubling times of 23.20 ± 1.27 hours and 25.09 ± 1.34 hours in iHepLPCs and iHepLPCs-FOXA3, respectively (Fig. 1D). The expressions of hepatic genes were increased in iHepLPCs-FOXA3 cells compared with the parental cells (Fig. 1E). Western blot analyses revealed up-regulation of CYP3A4, a-1-antitrypsin (AAT), ALB, arginase 1 (ARG1), argininosuccinate synthase 1 (ASS1), CPS1, ornithine carbamoyltransferase (OTC), and glutamine synthetase (GS) (Fig. 1F). iHepLPCs-FOXA3 cells acquired the mature functions of ALB and AAT secretion (Fig. 1, G and H) and increased urea production, ammonia elimination, and glutamine secretion (Fig. 1, I to K). Collectively, these results indicate that the functional properties of iHepLPCs improved after overexpressing FOXA3.

Fig. 1 Characterization of optimized iHepLPCs.

(A) Heatmap of hepatic functional gene expression after transcription factor transfection (n = 3). (B) Expressions of hepatic transcription factors between iHepLPCs and freshly isolated primary human hepatocytes (F-PHH). The relative gene expression was normalized to HepG2/C3A (n = 3). (C) Bright-field microscopy showing morphology of iHepLPCs-FOXA3 observed by an inverted microscope. Scale bar, 50 μm. (D) Cell proliferation curve and doubling time of iHepLPCs and iHepLPCs-FOXA3 (n = 3). (E) qPCR analyses of hepatic gene expression between iHepLPCs and iHepLPCs-FOXA3 (n = 4). (F) Western blot analysis of the expression of CYP3A4, AAT, ALB, ARG1, ASS1, CPS1, OTC, and glutamine synthetase (GS). GAPDH was used for normalization. (G and H) Quantitative analysis of ALB and AAT secretion in supernatants (n = 3). (I to K) Urea production, ammonia elimination, and glutamine secretion. Data are means ± SD. n.s., nonsignificant. *P < 0.05, **P < 0.005, ***P < 0.0005 by unpaired-tailed Student’s t test.

Large-scale 3D expansion of iHepLPCs-FOXA3 and establishment of the Ali-BAL

Another technical challenge for BAL application is the production of iHepLPCs-FOXA3 cells on a large scale of up to 109 to 1010 cells. To this end, we developed an Ali-BAL capable of supporting 3D expansion of iHepLPCs-FOXA3 cells on polyethylene terephthalate–based macroporous carriers (Fig. 2A and movie S1). iHepLPCs-FOXA3 cells were expanded from 5 × 108 to 3 × 109 (weighing about 27 g) in the Ali-BAL over 14 days (Fig. 2, A and B) with a stable pH and glucose culture environment (fig. S1, A to C). The cells attached to the fibers of the carrier, filling the spaces and forming cluster-like 3D constructs (Fig. 2, C to E). As compared to the monolayer culture, cells grown on the carriers displayed increased expressions of hepatic genes such as CPS1 and CYP3A4 (Fig. 2, F and G), improved synthetic functions, such as ALB and AAT secretion (Fig. 2, H and I), elevated urea production, ammonia elimination, and glutamine secretion (Fig. 2, J to L). Among all the different batches of iHepLPCs-FOXA3 prepared, hepatic genes were expressed at comparable quantities and consistent numbers of cells were harvested, indicating that the expansion procedure was robust and reproducible. The Ali-BAL thus was evaluated in a porcine ALF model.

Fig. 2 Large-scale expansion in Ali-BAL.

(A) Schematic diagram of the large-scale expansion of iHepLPCs-FOXA3 procedure. Cell numbers are indicated below schematic. (B) Large-scale expansion of iHepLPCs-FOXA3 cultures from 5 × 108 to about 3 × 109 cells in 14 days in the Ali-BAL bioreactor (n = 3). (C) Photograph of the microporous carrier: length, 10 mm; width, 3 mm; thickness, 0.3 mm; inclination angle, 45°; culture area, 2400 cm2/g. Scale bar, 10 mm. (D) Immunofluorescence morphology of cells on carriers. Scale bar, 100 μm. (E) Scanning electron microscopy (SEM) image of the carrier. Left, empty carrier; right, carrier with cells growing in the gap. White arrows indicate cells grown in interspaces between fibers. Scale bar, 400 μm. (F) qPCR analyses of the expressions of hepatic genes in monolayer iHepLPCs-FOXA3 cells and cells grown in carriers (n = 4). (G) Western blot analysis of the expression of CYP3A4, AAT, ALB, ARG1, ASS1, CPS1, OTC, and GS. GAPDH was used for normalization. (H and I) Quantitative analysis of ALB and AAT secretion in supernatants (n = 3). (J to L) Ammonia elimination, urea production, and glutamine secretion (n = 3). Data are shown as means ± SD. *P < 0.05, **P < 0.005, ***P < 0.0005 by unpaired-tailed Student’s t test.

Establishment of drug-induced porcine ALF model

To test the therapeutic effect of Ali-BAL treatment, we induced ALF in Bama miniature pigs by intravenous injection of d-gal (5, 6, 30) at a dosage of 0.5 g/kg body weight, a model that reproduces the features of ALF in humans. All 18 animals demonstrated biochemical profiles consistent with ALF 24 hours after administration of d-gal (Table 1). Each of the three treatment groups had comparable concentrations of alanine aminotransferase (ALT; P = 0.45), aspartate aminotransferase (AST; P = 0.91), ammonia (NH3; P = 0.43), total bilirubin (TBIL; P = 0.35), γ-glutamyl transpeptidase (γ-GT; P = 0.57), alkaline phosphatase (ALP; P = 0.29), creatinine (Cr; P = 0.16), international normalized ratio (INR; P = 0.31), and fibrinogen (Fib; P = 0.88) at the initiation of therapy. All pigs with ALF manifested loss of appetite, unsteady gait, anorexia, slight restlessness, and yellowish urine at 24 hours.

Table 1 Lab values in pigs with ALF before extracorporeal therapy (n = 6, mean ± SD), t = 24 hours.

Alanine aminotransferase, ALT; aspartate aminotransferase, AST; ammonia, NH3; total bilirubin, TBIL; γ-glutamyl transpeptidase, γ-GT; alkaline phosphatase, ALP; creatinine, Cr international normalized ratio, INR; fibrinogen, Fib.

View this table:

Prevention of ALF progression by 3-hour Ali-BAL treatment

The Ali-BAL–based extracorporeal circulation system included three micropumps, a heparin pump, two plasma filters, and an Ali-BAL embedded with macroporous fiber carriers (Fig. 3A). Figure 3B represents the outline of experimental design. Blood exits the pig via the right jugular vein catheter and is flowed through the first plasma filter, entering the plasma circuit. After passing through the bioreactor, the filtrate goes through a second plasma filter before re-entering the heated blood circuit and ultimately re-enters the pig via the right femoral vein catheter.

Fig. 3 Experimental design.

(A) Schematic diagram depicts the apparatus of the Ali-BAL support system. The red line represents the blood circuit; the yellow line indicates the plasma circuit. Pressures and temperature were detected by individual sensors. Flow rates were determined by pumps. The semipermeable membrane of plasma filter 1 was used to separate plasma from blood to the plasma circuit. The bioreactor contained hepatocytes with fluid entering at the top and exiting from the bottom. Plasma filter 2 served as an immunoprotective barrier. (B) Experimental time line showing the sequence of events and BAL treatment of ALF pigs. HE, hepatic encephalopathy.

Animals treated with Ali-BAL displayed markedly improved survival (5 of 6, 83% surviving to the 168-hour study end point) compared to ALF control (1 of 6, 16.7%, P = 0.013) or the application of the Ali-BAL apparatus alone without cells (no-cell) (0 of 6, 0%, P = 0.003) groups (Fig. 4A and Table 2). Biochemical measurements showed that Ali-BAL treatment reduced ammonia concentrations at 48 hours after extracorporeal circulation treatment (ECT) compared with ALF control and no-cell groups (Fig. 4B). Pigs in the ALF control and no-cell groups showed obvious symptoms of HE 48 hours after d-gal administration, including fatigue, unresponsiveness, and apparent drowsiness (movies S2 to S4). Grade IV HE (defined by the onset of coma with no reaction to pain stimulation) was observed in the ALF control and no-cell groups at 58.67 ± 13.09 hours and 57.33 ± 21.81 hours, respectively. It is of note that one of the animals in the control group that eventually survived also showed grade IV HE at 90 hours. On the contrary, none of the animals in Ali-BAL group presented with the characteristic symptoms of HE, although one of them died at 96 hours, possibly due to the less-optimized anticoagulation treatment.

Fig. 4 Evaluation of Ali-BAL treatment in pigs.

(A) Kaplan-Meier survival curve of ALF control, no-cell, and Ali-BAL–treated ALF pigs (n = 6 per group). **P < 0.005, log-rank test. (B) Plasma ammonia in each group (n = 6). (C) Dynamic changes in the biochemical parameters in the three groups. Plasma ALT, AST, DBIL, TBIL, ALP, γ-GT, albumin, and Cr (n = 6 per group). (D) Changes in coagulation function parameters. PT, TT, INR, APTT, and Fib (n = 6 per group). (E) Blood gas analysis in each group. pH, HCO3, PO2, Na+, K+, and glucose (n = 6 per group). Data are means ± SD. *P < 0.05, **P < 0.005, ***P < 0.0005 by unpaired Student’s t test.

Table 2 Ali-BAL therapy of ALF pigs.

View this table:

A decrease in ALT and AST in all groups was evident starting 48 hours after d-gal administration, whereas the Ali-BAL group displayed the greatest improvement. In addition, the concentrations of total bilirubin (TBIL) and direct bilirubin (DBIL) were steadily increased in the ALF control and no-cell groups, whereas those of Ali-BAL group were decreased (Fig. 4C). Liver recovery was observed in the Ali-BAL group, and the bilirubin and liver enzymes were improved. The concentrations of Cr in pigs were sharply increased before death, accompanied by hepatorenal syndrome and multiple organ failure (Fig. 4C). The prothrombin time (PT), INR, activated partial thromboplastin time (APTT), and thrombin time (TT) were gradually shortened in the Ali-BAL group, but these parameters became worse in the ALF control and no-cell groups. Consistently, Fib was only found to be increased in Ali-BAL group. These data indicate that Ali-BAL had a marked therapeutic effect (Fig. 4D). According to blood gas analysis, the internal environment of pigs in the Ali-BAL group was relatively stable but that of the other two groups gradually worsened as evidenced by liver failure symptoms (Fig. 4E). Together, Ali-BAL treatment reduced blood ammonia, improved coagulation, and stabilized the internal environment in the d-gal–induced ALF porcine model.

Reduction of inflammatory reaction after Ali-BAL treatment

ALF shares marked similarities with septic shock in regard to the features of systemic inflammation, progression to multiple organ dysfunction, and immunoparesis (31). Plasma samples were collected to quantify endotoxin, inflammatory cytokines [interleukin-2 (IL-2), IL-6, IL-10, and tumor necrosis factor–α (TNF-α)], hepatocyte growth factor (HGF), and alpha-fetoprotein (AFP) every 24 hours after d-gal injection. Starting 48 hours after d-gal injection, HGF and AFP concentrations were much higher in the Ali-BAL group (Fig. 5A), whereas endotoxin concentrations were lower than in the control and no-cell groups (Fig. 5B). Strong inflammatory responses were marked by significant increases in the concentrations of TNF-α and other proinflammatory cytokines including IL-6 and IL-2 (P < 0.05). This suggests intense activation of inflammatory cells during the development of ALF. These plasma inflammatory cytokines decreased after Ali-BAL treatment. IL-10, an anti-inflammatory cytokine, increased in ALF control and no-cell groups, suggesting that the survival outcome may not depend on the establishment of counter-regulatory homeostasis to prevent inflammation (Fig. 5C). In addition, the plasma proinflammatory cytokines were ascertained further using cytokine arrays. After Ali-BAL treatment, proinflammatory cytokines were notably reduced and returned to normal by 168 hours (Fig. 5D). From the heatmap and scanning images, it was found that IL-10, IL-18, IL-21, and transforming growth factor–β1 (TGF-β1) were evidently altered (Fig. 5, D to F, and table S1). Thus, Ali-BAL treatment alleviated inflammatory reactions in the ALF model.

Fig. 5 Inflammatory reaction after Ali-BAL therapy.

(A) HGF and AFP concentrations in each group (n = 6 per group). (B) Endotoxin concentrations in each group (n = 6 per group). (C) Inflammatory parameters of endotoxin, IL-2, IL-6, IL-10, and TNF-α (n = 6 per group). *P < 0.05, **P < 0.005, ***P < 0.0005 by unpaired-tailed Student’s t test. (D) A porcine cytokine antibody array was used to analyze the amounts of proinflammatory cytokines of each group using plasma samples. A heatmap of 48 inflammatory cytokines was established for the ALF control, no-cell, and Ali-BAL groups. (E) Scanning images showing cytokine expression. Red boxes represent the locations and expression values of four differently expressed cytokines (IL-10, IL-18, IL-21, and TGF-β1). (F) Quantification of mean amounts of IL-10, IL-18, IL-21, and TGF-β1 (n = 2). AU, arbitrary units.

Promotion of liver regeneration by Ali-BAL treatment

Dynamic progression of ALF was observed after examining histology and immunohistochemistry. Hepatocyte damage was confirmed by liver histopathology at 24 hours after d-gal injection. Large areas of hemorrhage, typical hepatocyte vacuolar degeneration, and apoptotic bodies were found around the central vein lobule at that time point (Fig. 6A). Extensive hepatocyte necrosis and collapse of the hepatic lobular structure became progressively worse in the liver tissue specimens of the two control groups until death (Fig. 6A). By contrast, liver injury was markedly alleviated in the Ali-BAL group (Fig. 6, A and B), whereas apparently increased numbers of hepatic parenchymal cells were found at 72 and 96 hours, indicating that compensatory hepatocyte regeneration occurred (Fig. 6A). The mean degree of liver injury was highest in the no-cell group and lowest in the Ali-BAL group (P < 0.0005; Fig. 6B). A significant negative correlation was observed between liver injury score and survival time (P < 0.0001; Fig. 6C).

Fig. 6 Liver histology and immunohistochemistry.

(A) H&E staining of ALF liver histology in ALF control, no-cell, and Ali-BAL groups. Scale bars, 300 μm. Insets shown at higher magnification below. (B) Liver injury score in ALF control (n = 4), no-cell (n = 3), and Ali-BAL (n = 6) groups. **P < 0.005, ***P < 0.0005 by unpaired-tailed Student’s t test or one-way ANOVA. Scores were assigned in a blinded fashion. (C) Correlation between survival time and liver injury score. (D) Staining for regenerative marker Ki-67 in all groups every 24 hours. Scale bars, 200 μm. Insets shown at higher magnification below. (E) Regenerative index (Ki-67) in all groups every 24 hours. Regenerative index is quantified by Ki-67+ nuclei per 40× high-power field; a minimum of five microscopic fields per animal were counted (n = 2). **P < 0.005, ***P < 0.0005 by unpaired-tailed Student’s t test.

Ki-67 staining was used to examine hepatocyte proliferative activity. Some Ki-67+ hepatocytes appeared at 24 hours, particularly around the portal area. Remarkable liver regeneration was observed after Ali-BAL treatment at 48 hours after d-gal infusion. The highest degree of hepatocyte regeneration was observed in the Ali-BAL group at 96 hours and lasted until 168 hours. The liver specimens in the ALF control and no-cell groups also displayed increasing Ki-67+ regeneration indices; however, the number of regenerated hepatocytes was significantly lower than that of Ali-BAL group at the same time points (P < 0.001; Fig. 6, D and E). These results indicate that Ali-BAL could alleviate liver injury and promote regeneration.

Ali-BAL performance and mechanism

Next, we explored the therapeutic mechanism of Ali-BAL on ALF. We measured several toxins and inflammatory molecules in the plasma before and after it perfused the bioreactor. Although the concentrations of toxins such as bilirubin and bile acids were not significantly decreased after plasma perfused the bioreactors (P > 0.05; Fig. 7A), the accumulation of urea in the bioreactor suggests that the ammonia was removed by Ali-BAL via ureagenesis (P < 0.0001; Fig. 7B). As lactate is metabolized primarily by the liver, the reduction in lactate concentrations after bioreactor perfusion further confirmed a hepatic metabolic function of Ali-BAL (P < 0.05; Fig. 7B). On the other hand, the porcine inflammatory cytokines such as TNF-α, IL-2, IL-6, and IL-10 remained unchanged during the perfusion (Fig. 7C). However, the elevated concentrations of human HGF and TGF-α, which are strong mitogens for hepatocytes in vitro and in vivo (32), as well as ALB and AAT, were detected in pig plasma after perfusion (P < 0.05; Fig. 7, D and E). This suggests that cytokines, growth factors, and other substances synthesized and secreted by the cells within the bioreactor could be infused into ALF pigs. These data collectively demonstrate that Ali-BAL treatment could metabolize and detoxify ammonia and lactate via biotransformation and that the BAL secreted hepatotrophic factors such as HGF and ALB to promote hepatocyte regeneration and liver recovery. Further, the reduction in concentrations of ALT and AST after Ali-BAL perfusion with respect to the no-cell group indicates that the progression of liver injury was partially halted during the treatment (P < 0.05; Fig. 7F). In addition, the oxygen partial pressure (PO2) and pH values remained stable, suggesting that the Ali-BAL can provide a controlled environment required for homeostatic cell function (Fig. 7G). During extracorporeal circulation, the vital signs of the animals were stable, and no complications were observed during the operation (fig. S2). These data demonstrate that the Ali-BAL was safe and effective for ALF treatment in pigs.

Fig. 7 Ali-BAL performance and mechanism.

(A) Toxic parameters of bilirubin and bile acids in the pig plasma before and after it perfused the bioreactors in the no-cell and Ali-BAL groups during treatment (n = 3). (B) Urea production and lactate concentrations during the treatment (n = 3). (C) Porcine inflammatory cytokines: TNF-α, IL-6, IL-2, and IL-10 (n = 3). (D and E) Hepatotrophic factors of human origin: HGF, TGF-α, AAT, and ALB (n = 3). U.D., undetected. (F) Hepatic damage parameters of ALT and AST (n = 3). (G) pH value and PO2 in the Ali-BAL environment during treatment (n = 3). *P < 0.05, **P < 0.005, ***P < 0.0005 by unpaired-tailed Student’s t test.


The major challenges of bioartificial liver support therapy are the limited availability and efficacy of cell sources. Recently, we developed a highly efficient approach for derivation of HepLPCs through the delivery of developmentally relevant cues (24). By immortalizing HepLPCs with HPV E6/E7, we enabled efficient cell expansion without growth arrest. These iHepLPCs could easily differentiate into metabolically functional hepatocytes in vitro, although the activity of some essential hepatocytic functions was lower than those of primary human hepatocytes. We hypothesized that iHepLPCs lost the expression of some transcription factors that are imperative to hepatic maturation (2629). After a parallel screening of hepatic transcription factors, FOXA3 was identified as a strong promoter of hepatic differentiation and was then overexpressed in the cells. iHepLPC-FOXA3 cells displayed increased expression of hepatic genes, improved metabolic detoxification, and elevated hepatic protein synthesis, confirming that FOXA3 is an indispensable transcription factor for reprogramming cell maturation (33). Under 3D culture conditions, iHepLPC-FOXA3 cells formed hepatic spheroids with enhanced liver functions that were comparable to those of primary human hepatocytes, including urea synthesis and albumin production (25); these functions are limited in other tumor-derived, immortalized, or reprogrammed hepatocyte cell lines used in bioartificial liver support systems (13).

Another technical challenge for BAL development is to scale the 3D expansion of iHepLPCs-FOXA3 to therapeutic numbers (109 to 1010 cells). Unlike continuous perfusion bioreactor designs such as hollow fiber, flat plate, spheroid, and packed bed reactors, we used an air-liquid interactive bioreactor configuration in which cells grew into 3D structures on macroporous carriers and were alternately exposed to aeration and nutrition via the decompression and compression of bellows holding the culture medium. The gentle vertical oscillation of the culture medium created a dynamic interface between the air and culture medium on the surface of the cells, providing a low shear stress and high aeration 3D culture environment, enabling rapid 3D expansion of iHepLPCs-FOXA3 to obtain 3 × 109 cells in about 2 weeks. In addition to the potential for scale-up to therapeutic numbers, a clinically effective BAL device should accommodate adequate bidirectional mass transport and provide an in vivo–like environment to maintain the viability and functionality of large numbers of iHepLPCs-FOXA3 cells (25). As the cells growing on the carriers form porous 3D structures, the air-liquid interactive configuration offers markedly improved mass transfer by allowing direct contact of cells on or inside macroporous carriers with the perfusing media. Furthermore, the gentle upward and downward tide motion provided cells with ample oxygen without the use of an oxygenator, regardless of the bioreactor scale. Together, the bioreactor design of Ali-BAL supported a large-scale and high-density 3D expansion of iHepLPCs-FOXA3 cells in multiple macroporous carriers and integrated them as packed, bioengineered liver tissue with homogeneous oxygen supply and efficient mass transport.

In the drug-induced porcine ALF model, Ali-BAL treatment improved survival rates. Brain edema is commonly associated with HE and is the major cause of death in ALF models (5). Although HE was observed in all animals in the control and no-cell groups, the neurological signs were considerably improved or nearly absent after Ali-BAL treatment. Consistent with this observation, the concentrations of blood ammonia were much lower in Ali-BAL–treated animals than the control groups, providing evidence of the ammonia removal capacity of the Ali-BAL. In support of this notion, urea concentrations were steadily elevated in the bioreactor during the treatment, confirming that iHepLPC-FOXA3 cells were able to detoxify ammonia via ureagenesis. In conjunction with the reduction in the lactate concentrations after bioreactor perfusion, these data collectively demonstrate that the Ali-BAL was able to metabolize and detoxify ammonia and lactate via biotransformation. The lack of increase in ammonia after completing the extracorporeal perfusion further indicates the effect of Ali-BAL on native cell function rather than an ongoing effect of BAL.

Consistent with this notion, Ali-BAL treatment showed a potential benefit by stimulating native hepatocyte regeneration. We noticed a marked increase in Ki-67+ hepatocytes in the animal treated with Ali-BAL that died at around 96 hours after d-gal administration. We next designed a separate study with a similar setup in which animals treated with Ali-BAL were euthanized at different time points after drug administration so the tissues could be obtained and compared for treatment effects. Immunohistochemistry analysis confirmed that Ali-BAL therapy promoted liver regeneration as evidenced by the quantification of Ki-67+ cells. In addition, several liver regeneration markers including plasma concentrations of porcine HGF and AFP, which were also noted in patients with ALF with favorable prognosis (34), started to be up-regulated from the controls on day 2 after initiating Ali-BAL treatment, suggesting the importance of early and sustained liver regeneration to survival. It is well known that liver injury can lead to rapid regenerative responses. After peak hepatotoxicity, liver regeneration factors and markers including cytokines (TNF-α and IL-6), growth factors (HGF and TGF-α), and plasma markers (AFP and AAT) have been implicated in the promotion of hepatocyte regeneration (34, 35). In the present study, we demonstrated that the cytokine and growth factor milieu was altered after Ali-BAL perfusion. Multiple human growth factors including HGF and TGF-α were secreted into the perfused porcine plasma, suggesting that, in conjunction with the reduction of detrimental toxins (ammonia and lactate) after the perfusion, exogenous human growth factors and cytokines may tip the balance toward the resolution of toxic reactions and promote early liver recovery, which could be the primary cause for improved survival noted in this model.

Another factor contributing to the improvement of animal survival was the amelioration of inflammatory reactions (36). Animal models for drug-induced liver injury are often exposed to lethal quantities of chemical stress, with changes in proinflammatory cytokines expression, such as TNF-α and IL-6 (37). Inflammatory cytokines may perpetuate liver damage and extend the inflammatory cascade (38), thereby aggravating liver injury. In addition, cytokines like TNF-α may reduce numbers of branch chain amino acids and increase peripheral type benzodiazepine receptors in the cortex and striatum (36, 39), thus further worsening encephalopathy (14). After Ali-BAL intervention, porcine inflammatory cytokines including TNF-α, IL-2, IL-6, and IL-10 were reduced compared with control groups. Noting that these inflammatory porcine cytokines remained unchanged during perfusion, this effect appears to be secondary to reduced toxic reactions and, more importantly, enhanced liver regeneration. The inhibition of such aberrant immune activation may further limit the extent of parenchymal damage and simultaneously promote liver recovery (31).

A few limitations of this study should be acknowledged. Inflammation, HE, and liver regeneration are extraordinarily complex processes and are influenced by multiple factors. The underlying mechanisms for the therapeutic effects of Ali-BAL treatment are still not fully understood. Examining the effects in rodent models of liver failure using encapsulated iHepLPC-FOXA3 cells as an implantable bioreactor, for example, could help to clarify the mechanism by which the Ali-BAL enhanced survival in this study (31, 40). Patients with ALF or acute-on-chronic liver failure can exhibit much more heterogeneous etiologies and clinical presentations than animal models, with conditions that are not as readily reversible as toxic injury (41). It is, therefore, necessary to demonstrate the survival benefits of Ali-BAL treatment in ALF models of other complications such as post hepatectomy liver failure (7). A dose-response relationship study should also be performed to determine the safety and potency of Ali-BAL treatment in these models. On this basis, clinical studies in humans are warranted to examine the true effectiveness of the bioreactor device.

In conclusion, we have developed an extracorporeal liver support device that exhibited a remarkable capacity to support liver function by detoxification of ammonia, promotion of native liver regeneration, and suppression of inflammation, leading to marked recovery and survival of animals with ALF. Further randomized clinical trials of Ali-BAL therapy are necessary to determine whether this approach could provide a therapeutic benefit to patients with ALF.


Study design

In this study, we have developed a 3D Ali-BAL using iHepLPCs and evaluated the safety and efficacy of the device in a porcine model of drug-overdose ALF. HepLPCs were immortalized with HPV E6/E7 genes and then functionally enhanced through a selective screening of hepatic transcription factors. The cells were expanded in a 3D microfiber network on a large scale in the Ali-BAL. Ali-BAL design was based on the tide motion principle, where the compression and decompression of bellows allows the intermittent exposure of cells to nutrients and air. Animals received infusions of hepatotoxin d-gal to induce ALF and were treated extracorporeally with the Ali-BAL. Blood ammonia, liver function, coagulation index, internal environment, and cytokines were measured every 24 hours. Cytokine arrays were used to determine the inflammatory cytokines. Hematoxylin and eosin (H&E) staining and immunohistochemistry were used to examine the liver injury and regeneration. This parallel, randomized study was performed at the Animal Experimental Center, Renji Hospital, Shanghai Jiaotong University School of Medicine, Shanghai, China, with the approval of the Animal Care Ethics Committee of Renji Hospital (approval no. RJ-2019004-2). All the experiments were designed to use the smallest number of miniature pigs that can allow us to perform appropriate statistical analyses. Pigs were randomly assigned to three intervention groups. Outlier values were not excluded in the study design. Blinding was not possible during treatment due to the visual symptoms of the ALF but was used whenever possible during the quantification and assessment of the sample material. Histological sections were assigned a randomized blinded code before quantification by a separate researcher, and the randomization was decoded at the time of final data analysis. All in vivo experiments were performed at least twice, and all in vitro experiments were performed at least three times. Primary data are reported in data file S1.

Cell culture and hepatocytes optimization

Human hepatocytes were plated on a Matrigel-coated (Corning) culture dish (NEST Biotechnology) at 0.5 to 2 × 104 cells/cm2 and cultured in transition and expansion medium (TEM) as previously described (23, 24). HepLPCs were immortalized with HPV E6/E7 overexpression to achieve expansion without growth arrest in vitro (25). The functionally enforced iHepLPCs were derived from selected immortalized hepatocytes by transduction with FOXA3. Cells were cryopreserved in liquid nitrogen using cell freezing medium (Cryowise Medical Technology Inc.).

Large-scale expansion

Large-scale expansion was performed in an Ali-BAL. The Ali-BAL consists of a column stent and is embedded with macroporous carriers (see Fig. 2C and the “BAL system” section below). iHepLPC-FOXA3 cells (5 × 108) were seeded onto microcarriers and expanded in the Ali-BAL for 2 weeks to obtain 3 × 109 cells. Glucose, lactate dehydrogenase, and pH values of the medium were measured. Cell count was calculated by the Crystal Violet Dye (CVD) Nuclei Count Kit (ESCO) every day.

Functional hepatocytes studies

To determine the secretion of human ALB and AAT, supernatants were collected after 24-hour culture and analyzed using a human ALB and AAT enzyme-linked immunosorbent assay (ELISA) kit (both from Bethyl Laboratories). To determine the ammonia elimination and urea synthesis abilities, cells were incubated in TEM supplemented with 3 mM NH4Cl. The supernatant was collected at 24 hours after NH4Cl induction. NH4+ concentrations were measured using the enzymatic colorimetric assays (Megazyme International). Urea concentrations were measured using a QuantiChrom Urea Assay Kit (BioAssay Systems).

RNA extraction quantitative polymerase chain reaction

Total RNA of cells was extracted by TRIzol (Beyotime Biotechnology). A total of 1 μg of RNA was reverse-transcribed into complementary DNA with M-MLV Reverse Transcriptase (Promega) according to the manufacturer’s instructions. Quantitative polymerase chain reaction (qPCR) was performed with SYBR Green qPCR Mix (2×, High ROX) (Beyotime Biotechnology) on an ABI 7300 Plus Real-Time PCR platform (Life Technologies). Gene transcription was evaluated by the ∆∆Ct method normalized to the housekeeping gene 18S ribosomal RNA (rRNA). Primers sequences were respectively listed in table S2. All qPCR data were repeated three times.

Western blot

The cells were homogenized in radioimmunoprecipitation assay lysis and extraction buffer, and protein concentrations were measured by the bicinchoninic acid (BCA) Protein Assay Kit (both from Beyotime Biotechnology). Proteins were loaded onto 8 to 10% SDS–polyacrylamide gel electrophoresis and transferred to polyvinylidene fluoride membranes (Merck). After blocking with skim milk (Yeasen), the membranes were incubated at 4°C overnight with the following primary antibodies: rabbit anti-CYP3A4 (1:2000; ABclonal), rabbit anti-AAT (1:5000; Abcam), rabbit anti-ALB (1:1000; ProteinTech), rabbit anti-ARG1 (1:1000; ProteinTech), rabbit anti-ASS1 (1:1000; ProteinTech), rabbit anti-CPS1 (1:2000; Abcam), rabbit anti-OTC (1:1000; ProteinTech), rabbit anti-GS (1:1000; ProteinTech), and rabbit anti-GAPDH (1:1000; Beyotime Biotechnology). The membranes were washed with tris-buffered saline [150 mM NaCl (Sangon Biotech) and 0.01 M tris-HCl (pH 7.5)] containing 0.1% Tween 20 (Sangon Biotech) and incubated with horseradish peroxidase–conjugated secondary antibodies specific to the species of the primary antibodies (1:1000; Beyotime Biotechnology) for 1 hour at room temperature. Last, the immune complexes were detected by a Chemiluminescence Imaging System (CLINX) using the ECL Prime Western blotting detection reagent (Beyotime Biotechnology).

Scanning electron microscopy

Cell morphology on carriers was viewed under the scanning electron microscope (SEM) (Hitachi S3400N, Hitachi). Briefly, carriers were fixed with 2.5% glutaraldehyde at 4°C overnight and then suffered from dehydration in a series of ethanol solutions (50, 75, 90, 95, and 100%). Afterward, the resulting samples were dried and sputter-coated with gold, followed by SEM observation at a working voltage of 15 kV.


Each animal was housed in a singular standard air-conditioned cage (20° to 25°C), with a 12-hour light/12-hour dark cycle. Standard laboratory water and diet were provided ad libitum. All experimental animals received humanitarian care. The pigs acclimated to the experimental environment for at least 1 week before being used in this study.


Female healthy Bama experimental miniature pigs (range, 20 to 30 kg; Pig Breeding Center) were randomized into three intervention groups: (i) ALF-control group (n = 6), in which animals received only intensive care; (ii) no-cell group (n = 6), which involved sham Ali-BAL treatment without hepatocytes; and (iii) Ali-BAL group (n = 6), in which animals received Ali-BAL treatment implanted with functional hepatocytes.

BAL system

The BAL system consists of a blood circuit and a plasma circuit (Fig. 3A). The components of the BAL system include three micropumps, a heparin pump, two plasma filters (Fresenius), and an Ali-BAL embedded with macroporous fiber carriers. The Ali-BAL bioreactor design comprises two compartments: upper and lower chambers. The upper chamber is a 550-ml column stent with an inside diameter of 6.5 cm and a height of 34.5 cm to accommodate 11-g carriers. The compressible lower chamber containing the medium could be compressed and released by a vertically movable stage. The macroporous carrier is made of polyethylene terephthalate which size is 10-mm long, 3-mm wide, and 0.3-mm thick, with 45° inclination angle and 2400 cm2/g culture area. The Ali-BAL reactor is a product of Shanghai Celliver Biotechnology Inc. Ltd. with independent intellectual property rights. Aliver is the registered trademark of Ali-BAL bioreactor. The blood circuit is indicated in red in Fig. 3A, whereas the plasma circuit is shown in yellow. Blood pressure, temperature, and oxygen pressure were detected by individual sensors. Flow rates of pumps were determined by pig blood velocity. About 3 × 109 iHepLPCs-FOXA3 were grown on the macroporous fiber carriers to play a role in detoxification and synthesis. Plasma filter 2 served as an immunoprotective barrier.

Standard surgery of ALF pigs

Figure 3B presents the outline of experimental events. Animals in all groups underwent general anesthesia [inhalation of 2 to 3% isoflurane, followed by intramuscular injection with Zoletil 50 (25 mg/kg)] for venous catheterization. The right jugular and femoral veins were cannulated with a 16-gauge 20-cm single-lumen catheter (Sungwon Medical Co. Ltd.) guided by a portable ultrasound (MyLabOne MylabTwice 2010, Esaote) for blood sampling, drug administration, and extracorporeal dialysis. ALF was induced with d-gal (0.5 g/kg; Sigma-Aldrich). Before d-gal injection, a 10-ml baseline blood sample was collected through the femoral vein. d-gal was dissolved in 5% glucose solution (0.5 g/ml), and the pH value was adjusted to 6.8 using 1 M NaOH. It was infused through the femoral vein catheter immediately. Both jugular and femoral vein catheterizations were completed, and then pigs were returned to cages after awakening from anesthesia.

Monitoring steps included blood draws every 24 hours and clinical observations such as food and water intake, gait, anorexia, and so on, every 4 to 6 hours. Chemical analysis (liver and renal function, coagulation, and blood ammonia) was performed in the Renji Hospital Clinical Chemistry Laboratory. Blood gas analysis was detected by the GEM Premier 3000 system (Instrumentation Laboratory). All blood glucose values were less than 4 mM, and all pigs were treated with 10 ml of 50% dextrose. The animals were clinically observed, and blood was sampled until the animals were euthanized at 168 hours after d-gal injection, which was the end point of the study. Dying pigs received humane care. The pigs that survived the critical point of ALF were euthanized at 168 hours. Liver tissue was collected from all animals at the end for H&E staining and immunohistochemistry examination.

Extracorporeal circulation treatment

ECT groups were initiated in the no-cell and Ali-BAL device groups at 24 hours after d-gal infusion. Light sedation during BAL treatment was limited to propofol (0.1 mg/kg per min; Fresenius). The heart rate, blood pressure, and tongue oximetry of animals were continuously monitored. Activated clotting time (ACT) was measured during ECT, and heparin was administered to maintain ACT above 200 s. The bioreactor was washed six times with 3 liters of warm normal saline (ns) before use, and then 450 ml of ns was added as a supplement to ensure that there was a large enough volume for detoxification. Cell viability was greater than 95%. The pigs were connected to the extracorporeal device via the right jugular and femoral venous catheters (Fig. 3A). Before initiation of treatment, all animals received 250 ml of hydroxyethyl and 250 ml of ns, followed by fluid energy support with 250 ml of 5% glucose solution to ensure hydration during treatment. The flow rates of three pumps in the BAL system were set as follows: 30 to 50 ml/min in whole blood cycle (pump 1) and 10 to 15 ml/min in plasma separation and bioreactor cycle (pumps 2 and 3). Ali-BAL was placed in the incubator at 37°C. Parameters of Ali-BAL were adjusted to ensure that the cells performed detoxification and secretion functions effectively.

We tried to maintain stable body temperatures during ECT. The operating room was kept at a constant temperature of 25°C, and pigs were covered with a blanket to avoid temperature loss. Also, we used a 37°C heater with the external circulation system to warm the blood before it flowed back to the body. According to hepatocyte viability and function, we treated ALF pigs for 3 hours. The plasma concentrations of toxins (bilirubin, bile acids, urea, and lactate), porcine inflammatory cytokines (TNF-α, IL-2, IL-6, and IL-10), and human growth factors (HGF and TGF-α), as well as human AAT and ALB, were detected before and after the plasma perfused the bioreactor. The environment in the Ali-BAL bioreactor was also monitored. After the BAL treatment, all pigs were awoken and placed in the cages and then received standard care for 7 days.

Measurement of porcine cytokines and endotoxin in plasma

Plasma concentrations of IL-2, IL-6, IL-10, TNF-α, HGF, and AFP were measured using the porcine-specific ELISA Quantitation Kit (Abcam), according to the manufacturer’s instructions. Plasma endotoxin was measured using an End-point Chromogenic Tachypleus Amebocyte Lysate kit (Xiamen Bioendo Technology Co. Ltd.).

Cytokine arrays for measuring plasma cytokines

To identify cytokines in pigs with ALF in the three treatment groups, plasma samples which were harvested from randomly chosen pigs before the pigs died or were euthanized were analyzed using porcine cytokine arrays (RayBiotech Inc.) containing antibodies against 48 cytokines according to the manufacturer’s instructions. All cytokine array assays contained positive, negative, blank, and internal controls. The quantities of each cytokine were measured in duplicates.

Tissue histology and immunohistochemistry examination

Liver specimen was obtained from each pig immediately after death and was 4% paraformaldehyde-fixed and paraffin-embedded for H&E staining as well as immunohistochemistry with anti–Ki-67 antibody. Liver injury score (5) and Ki-67+ cells (Ki-67 index) (30) were counted by two observers blinded to the specimen groups, and the means of the results were used for the analysis.

Statistical analysis

Data were presented as means ± SD. An unpaired two-tailed Student’s t test and analysis of variance (ANOVA) were used to calculate statistical significance. For survival time, the Mantel-Cox log-rank test was performed. P < 0.05 was considered statistically significant. Statistical calculation was performed using GraphPad Prism 7.


Fig. S1. Ali-BAL environmental monitoring.

Fig. S2. Vital signs during ECT.

Table S1. Cytokine array normalized data.

Table S2. Primer list.

Movie S1. Large-scale expansion in Ali-BAL.

Movie S2. Pig from ALF control group 48 hours after d-gal injection.

Movie S3. Pig from no-cell group 48 hours after d-gal injection.

Movie S4. Pig from Ali-BAL group 48 hours after d-gal injection.

Data file S1. Primary data.


Acknowledgments: W. Wang provided us the ACT-plus equipment for ACT monitoring. Y.-G. Hu provided technical and manufacturing support of the ECT.We thank X. Zhou for expert care of ALF animals. We are grateful to Y. Chen, D. Tang, and Y. Peng for assistance of animals’ experiments. We acknowledge B. Xiao for technical assistance in Ali-BAL equipment. Funding: This work was supported by National Key R&D Program (2018YFA0108200 and 2016YFC1101402), the National Natural Science Foundation of China (31872823), Shanghai Academic/Medical Research Leader Program (2018BR14), Shanghai Municipal Education Commission-Gaofeng Clinical Medicine Grant Support (20181710), and Shanghai Celliver Biotechnology Co. Ltd. Author contributions: H.-.X.Y. conceptualized the study, led the experimental design and development of the protocol with input from all authors, and wrote the manuscript. W.-.J.L. led the experimental design and development of the protocol and performed experiments. X.-.J.Z. assisted in animal experiments’ design and writing the manuscript. T.-.J.Y. assisted in conducting experiments and collecting data. Z.-.Y.W. assisted in cell culture, cell functional evaluation, and collecting data. Z.-.Q.B. assisted in animals’ experiments. H.-.S.J. assisted in conducting experiments, large-scale expansion, and collecting data. X.S. assisted in venipuncture catheterization. C.-.Y.C. and Y.-.P.S assisted in extracorporeal therapy. G.-.B.F. assisted in study design and performed the statistical analysis of study data. W.-.J.H. performed the blinded examination of liver pathology. Q.L. assisted in liver function, coagulation, and blood ammonia tests. M.Z. assisted in animal experiments and care of ALF animals. H.-.D.Z. and H.-.P.W. assisted in data collecting and analysis. W.-.F.Y. and B.Z. supervised experiments, collected data, and assisted in writing the manuscript. Competing interests: Shanghai Celliver Biotechnology Inc. Ltd. hold patents related to this research including ZL2019104363685 (Biological reaction device and biological reaction system), CN110438157A (Liver precursor-like cell line, construction method, and application in the field of bioartificial liver), and CN2019106234369 (Construction method of liver progenitor-like cell bank, cell lines thereof and application thereof). X.-.J.Z., H.-.D.Z., and M.Z. are full-time employees of Shanghai Celliver Biotechnology Co. Ltd. Shanghai, China. H-.X.Y. and B.Z. are co-founders of Shanghai Celliver Biotechnology Co. Ltd., Shanghai, China and have equity interest in Celliver Biotechnology Inc. All other authors declare that they have no competing interests. Data and materials availability: All data associated with this study are present in the paper or the Supplementary Materials.

Stay Connected to Science Translational Medicine

Navigate This Article