Reprogramming and redifferentiation of mouse hepatocytes

The purified PHs were seeded on collagen I- or Matrigel-coated six-well or 12-well plates (1% collagen I in PBS, 2% Matrigel in DMEM/F12) at 2 × 10⁴ cells/cm² and were cultured in HREM. HREM was based on basal medium DMEM/F12 containing 1% fetal bovine serum, insulin-transferrin-selenium-ethanolamine solution (ITS-X, Gibco), 20 ng/ml HGF, and 20 ng/ml EGF (Peprotech), 10 μm Y-27632 (ROCK kinase inhibitor), 1 μm A-8301 (TGF-β signaling inhibitor), 3 μm CHIR99021 (Wnt signaling activator) (MedChem Express), 1 μm sphingosine-1-phosphate, and 5 μm lysophosphatidic acid (yes-associated protein signaling agonists) (Santa Cruz). We named the last five small molecules supplemented in DMEM/F12 as “5 Additives” (5A). About 5–7 days after seeding, cells were dissociated with Accutase (Gibco) and were passaged at a 1:2–1:3 ratio. Approximately after five serial passages of 1:2–1:3 subculture, cells that entered into stable growth and exponential proliferation state were called LPCs (wild-type [WT]-LPCs and knockout [KO]-LPCs). The doubling time was calculated by using the online tool http://www.doubling-time.com/compute.php. ^[11]

For cell redifferentiation, LPCs (passage 15) were cultured in HREM for 3 days to reach 80% confluence. Then, the medium was changed to hepatocyte maturation medium (HMM) as described previously.^[10] HMM contained DMEM/F12, ITS-X, 20 ng/ml oncostatin M (Peprotech), 10 μm N-[N-(3,5-difluorophenacetyl)-l-alanyl]-S-phenylglycine t-butyl ester and SB431542 (MedChem Express), and 10 μm dexamethasone (Dex) (Sigma-Aldrich). The medium was changed every other day for 3 weeks, and the redifferentiated cells (LPC-Heps) were used in subsequent experiments.

Statistical analysis

Results were presented as mean ± SD or median ± interquartile range (IQR). All data were subjected to the normality test with SPSS 23.0 software (IBM). Comparisons were evaluated by unpaired two-tailed Student t test (2 groups) or one-way ANOVA followed by Student-Newman-Keuls post-test (≥3 groups) with Prism version 7.00 (GraphPad Software). p values < 0.05 were considered statistically significant. Ratios of ATP7B, CD45, sirius red, and alpha-smooth muscle actin (α-SMA) positive areas in liver sections were analyzed using ImageJ version 1.52 (National Institutes of Health) by taking representative pictures from different mice, and 3–5 areas of each mouse were counted.

For experiments involving animals, we assured that all animals received humane care according to the criteria outlined in the NIH Guide for the Care and Use of Laboratory Animals. And all animal studies complied with the guidelines of the Ethics Committee of Shanghai Tong-Ren Hospital and conformed to the Animal Research: Reporting of In Vivo Experiments (ARRIVE) guidelines.

Further methodology may be found in the Supporting Materials.

PH-LPCs exhibit high proliferative capability

First, we isolated PHs from livers of WT mice or ATP7B^−/− mice (KO). PHs were both α-antitrypsin (AAT) and albumin (ALB) positive at 4 h after isolation (Figure S1A). Next, we optimized the hepatocyte expansion system by adding combinations of growth factors and 5A (Figure 1A). To explore the optimal plate-coating conditions, cell proliferation on collagen-coated plates was compared with Matrigel-coated plates. We found that PHs exhibited a more rapid expansion on Matrigel than on collagen at 3 and 7 days in HREM (Figure S1B). With Matrigel coating and HREM culture, WT-PHs and KO-PHs could be reprogrammed and serially passaged over 40 times without obvious morphological changes (Figure S1C). Karyotyping of two independently established KO-LPC lines (passage 17) showed that donor 1–derived LPCs exhibited generally normal diploid karyotypes except for aneuploidy of the no. 19 chromosome (dotted box). Donor 2–derived LPCs showed diploid, triploid, and tetraploid changes of karyotypes (Figure S1D). These results were in line with previous reports that aneuploidy was one of the characteristics of hepatocyte proliferation.^[15]

We further evaluated the expression of proliferation-related marker Ki-67 and apoptosis-related marker caspase-3 in KO-LPCs (passage 15). We found that 79.8% ± 7.1% of LPCs were Ki-67⁺, whereas caspase-3⁺ cells were not detected, suggesting the extensive proliferation capacity of LPCs (Figure 1B). Meanwhile, we also found that WT-LPCs grew faster than KO-LPCs (passage 15) and KO-LPCs had longer doubling times than WT-LPCs (12.40 h vs. 8.59 h) (Figure 1C). We postulated that this was due to abnormal copper metabolism in KO-LPCs due to loss of ATP7B function. Once withdrawing the 5A, KO-LPCs became senescent and lost their proliferative capability as shown by SA-β-gal staining (Figure 1D). Expression of the cell cycle–related gene CDK2 was reduced, and CDK4 showed a declining tendency, whereas senescence genes like P21 and P53 were up-regulated after the withdrawal of 5A (Figure 1E). These results also implied that the proliferative cells did not undergo oncogenic transformation.

To further clarify the characteristics of KO-LPCs, we assessed the expression of hepatic markers in KO-LPCs. We found that from passage 0 to 14, percentages of ALB and AAT positive hepatocytes gradually decreased, whereas an increased rate of cytokeratin 19 (CK19) positive duct-like liver progenitors was observed (Figure 2A). Consistent with these findings, mRNA analysis indicated that mature hepatocyte-associated genes (ALB, AAT, glucose-6-phosphatase, catalytic subunit [G6PC], and hepatocyte nuclear factor [HNF] 4α) exhibited dramatic down-regulation whereas progenitors-associated genes (CK7, CK19, SRY [sex-determining region Y]-box 9 [SOX9], and epithelial cell adhesion molecule [EpCAM]) were elevated, even in late passages (Figure 2B). Collectively, these data suggested that KO-LPCs acquired stable progenitor features and displayed powerful proliferation potentials when maintained in HREM.

KO-LPCs exhibit the bi-potency of hepatic and biliary differentiation

First, we tested whether KO-LPCs could differentiate toward hepatic lineages (KO-LPC-Heps). In line with a previous report,^[10] the morphology of KO-LPCs gradually changed from pebbled to polygonal shapes after 3 weeks of differentiation, and ATP7B was not expressed in KO-LPC-Heps (Figure 3A). In addition, increased expression of hepatocyte markers (ALB⁺, 86.2% ± 4.0% and AAT⁺, 82.1% ± 8.4%) and decreased expression of progenitor markers (CK19⁺, 5.0% ± 1.1%) were present in KO-LPC-Heps. It was noted that during the redifferentiation process, LPCs (CK19⁺), biphenotypic cells (ALB⁺/CK19⁺), and LPC-Heps (ALB⁺) could be observed in the same field (Figure 3B). Consistently, gene expression levels of ALB, AAT, cytochrome P450 (CYP) 1A2, HNF4α, and G6PC were up-regulated whereas CK19, SOX9, EpCAM, HNF1β, and AFP levels were down-regulated after differentiation (Figure 3C). Compared with KO-LPCs, KO-LPC-Heps exhibited typically hepatic glycogen synthesis and 5(6)-carboxy-2’,7’-dichlorofluorescein diacetate (CDCFDA) secretion to bile canaliculi, which was comparable to PHs (Figure 3D). Furthermore, drug metabolism activities of CYP1A2, CYP3A4, and CYP2C9; ALB secretion; and urea synthesis were also elevated in KO-LPC-Heps (Figure 3E). Our results supported the view that relatively mature hepatocytes and liver progenitors could be mutually converted in vitro.^{[6, 12]} Next, we detected the biliary differentiation potential of LPCs. We found the cyst, duct, and sphere formation in three-dimensional cultures (Figure S2A), and cholangiocyte marker CK19 and tight junction marker ZO-1 were coexpressed in the cyst structure (Figure S2B). In addition, rhodamine 123 dye was transported to the luminal space, and verapamil (multidrug resistance protein 1 inhibitor) blocked the efflux process of dye, suggesting that the KO-LPC-Chols were functional (Figure S2C).

Recovery of ATP7B subcellular localization and copper-export function in LPC-ATP7B-Heps

ATP7B gene was overexpressed in KO-LPCs through an LV delivery system. Because the large size of ATP7B genome (5.2 kb) led to unsatisfactorily low-yield virus production, a miniATP7B was constructed based on previous research.^[16] To maximize the function of miniATP7B, we kept the species consistency between donor cells and the ATP7B genome. After homology comparison of amino acids between mouse and human ATP7B, we constructed mouse miniATP7B cDNA by deleting the DNA sequences corresponding to amino acids 74–488, and then we packaged this into LV-miniATP7B-green fluorescent protein (GFP; Figure S3A–C). Virus-transfected KO-LPCs expressed green fluorescence, and we estimated that ATP7B mRNA level was 200 times higher than in LPCs with only LV-GFP transfection. After 21 days of differentiation, ATP7B expression in LV-ATP7B-Heps was nearly 600 times higher than in LV-GFP-Heps. Flow cytometry showed that the transduction efficiency was 62.8% ± 0.8% before cell sorting (Figure 4A). Further immunofluorescence suggested that LV-GFP-transfected KO-LPCs expressed GFP but not ATP7B. In contrast, LV-ATP7B-GFP-transfected KO-LPCs expressed both GFP and ATP7B. Moreover, ATP7B was also expressed in redifferentiated binucleated LPC-ATP7B-Heps (Figure 4B). Besides the expression of ATP7B, correct subcellular localization is also a key precondition for copper export. It is known that ATP7B localizes in the Golgi apparatus under conditions with normal intracellular copper. Under high copper conditions, ATP7B will migrate to the lysosomes, exporting excess copper into bile ducts.^[17] In KO-LPC-Heps, due to the lack of LV-ATP7B-GFP transfection, no ATP7B signal was found in the Golgi (as detected by Golgin-97⁺) under conditions with no added CuCl₂ and was also not found in lysosomes (as detected by LAMP-1⁺) under conditions with added 50 μm CuCl₂ for 2 h (Figure 4C). By contrast, in LPC-ATP7B-Heps, ATP7B almost completely colocalized with Golgi instead of lysosome under with no CuCl₂. Once cells were treated with 50 μm CuCl₂ for 2 h, some of the ATP7B migrated to the apical membrane and colocalized with lysosomes (arrows). When cells were treated with the copper chelation bathocuproine sulphonate (BCS), 50 μm, for 1 h, ATP7B was no longer localized in lysosomes but returned to the Golgi. These data suggested that the overexpressed ATP7B was capable of proper subcellular localization in response to changes of cellular copper levels (Figure 4D). Finally, copper levels in the supernatants of LPC-ATP7B-Heps were 3.74, 3.56, and 3.41 times higher than levels in supernatants of LPC-GFP-Heps at 3, 6, and 9 h after copper withdrawal, indicating functional copper export of LPC-ATP7B-Heps (Figure 4E).

RNA sequencing reveals the advantages of KO-LPC-ATP7B-Heps under copper stress

With the increased copper level from 0 to 600 μm in medium, cell viabilities declined much slower in the KO-ATP7B group than those in KO-GFP and WT-GFP groups, suggesting the capacity of copper resistance after ATP7B transduction (Figure 5A). To gain an in-depth understanding of gene expressions in cells under high copper stimulation, transcriptome analyses were performed. There were 1709 differentially expressed genes (DEGs; fold change >2) when comparing KO-ATP7B_Cu with KO-GFP_Cu cells, among which 1100 genes were also differentially expressed in WT-GFP_Cu cells. The heatmap illustrated that KO-GFP_Cu clustered closer to WT-GFP_Cu cells than KO-ATP7B_Cu cells (Figure 5A). The process and pivotal proteins involved in intrahepatic copper trafficking were displayed in Figure 5B. We found that in KO-GFP_Cu and WT-GFP_Cu cells, due to their absent or insufficient expression of ATP7B, non-ATP7B mediated copper metabolic pathways were mobilized, leading to compensatory elevations of metallothionein 3/4, superoxide dismutase (SOD)1, copper metabolism domain containing 1 (COMMD1), X-linked inhibitor of apoptosis protein (XIAP), cytochrome coxidase 17, antioxidant, and copper chaperone for superoxide dismutase (CCS). When ATP7B was restored in KO-ATP7B_Cu cells, all these genes above were down-regulated, accompanied with up-regulated expressions of ATP7B, ceruloplasmin (CP), or even biliary epithelial marker aquaporin 1 (Figure 5B). Furthermore, KO-ATP7B_Cu cells displayed enrichment in genes involved in copper ion binding and the positive regulation of cell proliferation. By contrast, genes involved in the apoptotic process were enriched in KO-GFP_Cu cells (Figure 5C). Further quantitative PCR (qPCR) verified consistent expression of several key genes in pathways of copper metabolism (ATP7B, copper metabolism domain containing 1 [COMMD1], and XIAP), proliferation process (TGFβ2, platelet-derived growth factor α [PDGFα], insulin-like growth factor 1 [IGF1], HGF, FGF18, and VEGFα, which represented common growth factors), and apoptotic process (B cell lymphoma 2 [BCL2], BCL2–associated X protein [BAX], and Caspase-3, which were star molecules in this pathway) (Figure 5D). To further explore if there were unexpected other gene expression profile differences (fold change > 2, p value < 0.05), the Gene Ontology analysis was performed. There were eight DEGs in the copper ion binding pathway (Figure S4A), 56 DEGs in the apoptotic process pathway (Figure S4B), and 62 DEGs in the cell proliferation pathway (Figure S4C).

Oxidative stress was one of the most important events that occurred after cellular copper accumulation.^[18] We found that compared with KO-ATP7B_Cu cells, pathways of glutathione (GSH) metabolism and oxidative stress were more enriched in KO-GFP_Cu cells (Figure S5A). And expressions of key genes such as SOD3, ribonucleotide-diphosphate reductase subunit M2 B (RRM2B), glutathione peroxidase (GPX) 1/3, and peroxidasin (PXDN) were also decreased (Figure S5B). Furthermore, lower reactive oxygen species (ROS) levels and higher mitochondrial membrane potentials (5,5,6,6′-tetrachloro-1,1′,3,3′ tetraethylbenzimi-dazoylcarbocyanine iodide red polymers/green monomers) were observed in KO-ATP7B_Cu cells (Figure S5C,D). Besides, both the GSH level and GPX activity were higher in KO-ATP7B_Cu cells than in KO-GFP_Cu cells (Figure S5E), suggesting the pivotal role of ATP7B in relieving cellular oxidative stress.

Next, we compared the transcriptome of KO-ATP7B_Cu versus WT-GFP_Cu under copper stimulation. We found that KO-ATP7B_Cu cells displayed enrichment in genes involved in copper ion binding, zinc ion binding, and cellular defense response, whereas apoptotic process-related genes were enriched in WT-GFP_Cu cells (Figure S6A,B). The qPCR analyses confirmed the expressional differences in pathways of copper ion binding (ATP7B and CP), oxidative stress (SOD2/3, RRM2B, GPX1/3/8, and PXDN), apoptotic process (BCL2, BAX, and their ratio), and cell proliferation (TGFβ2, PDGFα, IGF1, VEGFα, FGF18, EGF, and HGF) (Figure S6C).

LPC-ATP7B-Heps transplantation reduces copper deposition and improves liver functions

Liver sections from 8-week-old ATP7B^−/− mice showed inflammation compared with WT mice, and no ATP7B positive cells were found in ATP7B^−/− mice (Figure S8B). Based on the data above, LPC-ATP7B-Heps were then ready to be transplanted into ATP7B^−/− mice (Figure 6A). Cell engraftment of LPC-ATP7B-Heps in mice reached an average of 31.5% after 4 months following cell transplantation, which was comparable to mouse healthy PHs transplantation (29.4%). Instead, no ATP7B⁺ cells were observed in the group of LPC-GFP-Heps (Figure 6B). Consistent with this result, copper staining and liver copper was increased in the LPC-GFP-Heps group compared with the other two groups (Figure 6C). Serum free copper (non-CP bound copper, NCC) was significantly reduced in PHs and LPC-ATP7B-Heps mice, although total copper remained unchanged (possibly due to the offset of NCC). Serum CP activities and biochemical parameters like alanine aminotransferase (ALT) and aspartate aminotransferase (AST) were all improved in LPC-ATP7B-Heps mice compared with LPC-GFP-Heps mice (Figure 6D). Moreover, pathological phenotypes of liver inflammation and fibrosis were also improved in PHs and LPC-ATP7B-Heps groups as evaluated by the presence of CD45⁺ leukocytes, sirius red⁺ collagen fibers, and α-SMA⁺ activated hepatic stellate cells (Figure S7).