Autophagy suppresses the formation of hepatocyte-derived cancer-initiating ductular progenitor cells in the liver

Autophagy represses YAP/TAZ-dependent hepatocyte dedifferentiation into liver progenitor cells to prevent tumorigenesis.


INTRODUCTION
Liver cancer is predicted to be the third leading cause of cancer-related deaths by 2030 (1). Hepatocellular carcinoma (HCC) is the major form of liver cancer and develops in patients with chronic liver conditions, including viral hepatitis, as well as alcoholic and nonalcoholic fatty liver disease (2). Generally, chronic liver injuries lead to inflammation, stromal activation, regeneration, fibrosis, and cirrhosis before progression to HCC (3).
Autophagy (strictly macroautophagy but hereafter referred to simply as autophagy) is a catabolic membrane-trafficking process that serves to deliver cellular constituents including misfolded proteins and damaged organelles to lysosomes for degradation (4). There is now clear evidence that autophagy is important in various diseases including neurodegenerative diseases, chronic liver diseases, and cancer (5)(6)(7). The role of autophagy in cancer, however, is complex and not fully understood, with seemingly opposing roles described in different tumors and at different stages of tumor evolution (8)(9)(10)(11)(12). In the early stages of malignant transformation, autophagy removes damaged mitochondria responsible for the production of reactive oxygen species (ROS) (13) and prevents genomic instability (14), highlighting its role in preventing tumor initiation. Conversely, in established tumors, autophagy not only can adopt a protumorigenic role, for example, by promoting survival under hypoxic conditions (15) and supporting invasion and metastasis (16), but also can have a tumor-suppressive role by preventing the proliferative outgrowth of disseminated tumor cells from dormant states at metastatic sites (17)(18)(19).
In the liver, autophagy has primarily been described as tumor suppressive (11). Liver-specific deletion of the central autophagyrelated protein 5 (ATG5) or ATG7 in mice leads to the formation of liver steatosis, inflammation, ROS production, oval cell formation, fibrosis, hepatomegaly, and the development of HCCs (11,20). In many cases, loss of autophagy causes accumulation of the autophagy adapter protein p62 (Sqstm1), and this can influence antioxidant responses by affecting the axis between Kelch-like ECH-associated protein 1 (KEAP1) and nuclear factor (erythroid-derived 2)-like 2 (NRF2) (21). In autophagy-deficient livers, studies have shown that p62 accumulation activates the NRF2 signaling pathway to induce metabolic reprogramming, hepatomegaly, and tumorigenesis (22,23).
The liver is a plastic organ in which cell fate can change upon injuries to regenerate liver function loss. Hepatocytes and cholangiocytes, epithelial cells that form the liver parenchyma and the bile duct, respectively, can transdifferentiate into one another to reestablish bile duct or liver parenchyma functions (24,25), with hepatocytes being the primary source of liver regeneration upon injury. Following chronic injury, ductular cells develop in the liver parenchyma when hepatocyte or cholangiocyte function is severely impaired, a process called ductular reaction (26). The ductular reaction is a repair mechanism for generating new hepatocytes or cholangiocytes, depending on which liver cells are injured (27). However, the origin of the ductular reaction and its role in liver tumorigenesis are controversial with reports indicating that ductular cells can arise from cholangiocyte expansion (28,29) or through hepatocyte dedifferentiation (30,31) and reports concluding that the ductular reaction is involved in forming HCC (32,33), while other studies report the opposite (34,35). Autophagy-deficient livers undergo a ductular reaction (36), and we considered this as an excellent system in which to explore its origin and the role, this phenomenon plays in tumorigenesis.
In this study, we report that autophagy prevents hepatocyte dedifferentiation into ductular liver progenitor cells (LPCs). This ductular LPC population affects HCC formation in autophagy-deficient livers. Mechanistically, we show that autophagy deletion activates both yes-associated protein (YAP) and transcriptional coactivator with PDZ-binding motif (TAZ) in hepatocytes, which are connected

Autophagy prevents hepatocyte dedifferentiation into ductular LPCs
Following histological examination, we observed an accumulation of atypical ductular structures in the parenchyma of conditional double knockout livers (Fig. 1E), called ductular reaction. Under normal con ditions, the liver has ductular structures, called the bile duct, that are formed out of cholangiocytes (Fig. 1E). The ductular reaction is a regeneration program that occurs in the liver following chronic liver injury that impairs the hepatocyte capacity to regenerate the liver (27). To evaluate whether hepatocytes are injured upon loss of autophagy, we first assessed the expression of enzymes for liver damage in the serum of 4-to 5-week-old livers. All autophagy-deficient livers had a significant increase in alkaline phosphatase (ALP), alanine aminotransferase (ALT), aspartate aminotransferase (AST) and -glutamyl transferase (GGT) levels in comparison to wild-type (WT) (Alb-Cre + ; Atg7 +/+ or Atg5 +/+ ; Pten +/+ ) mice ( Fig. 2A and fig.  S2, A to D). In addition, we determined whether hepatocytes were dying in our model by looking for cells positive for cleaved caspase 3 (CC3), a marker of apoptosis. We noted a significant augmentation of CC3 + hepatocytes in 4-to 5-week-old autophagy-deficient livers when compared to WT livers (Fig. 2, B and C, and fig. S2E), indicating that autophagy prevents hepatocyte cell death. Next, we observed a significant accumulation of the ductular markers sex-determining region Y-box 9 (SOX9), cytokeratin-19 (CK19), and panCK in Alb-Cre + ; Atg7 fl/fl ; Pten fl/fl or Alb-Cre + ; Atg5 fl/fl ; Pten fl/fl livers in comparison to Alb-Cre + ; Atg7 fl/fl ; or Alb-Cre + ; Atg5 fl/fl single knockout counterparts (Fig. 2, B and D to F, and fig. S2, F to H), confirming that the ductular reaction is occurring in our accelerated model.
As the ductular reaction is a regenerative process for the de novo generation of hepatocytes upon chronic liver injury (28)(29)(30)(31), we hypothesized that ductular cells in our model are LPCs forming to repair injured hepatocytes. To test this, we first looked at the expression of liver stem cell markers in Atg-and Pten-deficient livers and found increased levels of epithelial cell adhesion molecule (EpCAM), CD133, and CD44 within ductular cells ( Fig. 3A and fig. S3, A to C) of autophagy-deficient livers. The expression of the stem cell makers was autophagy dependent but PTEN independent ( Fig. 3A and  fig. S3, A to C), although Pten deletion appears to exacerbate the phenotype caused by Atg5 or Atg7 deletion. In addition, we assessed the expression of -fetoprotein (AFP), a fetal marker reexpressed during HCC and liver stem cell regeneration (39). We observed a significant increase in Afp mRNA levels ( Fig. 3B and fig. S3D) and AFP protein level in the serum ( Fig. 3C and fig. S3E) of autophagydeficient mice when compared to WT counterparts.
We were interested to know how the ductular-reactive cells were forming within the liver parenchyma. It has been established that ductular-reactive cells can originate from dedifferentiated hepatocytes in the parenchyma (30,31) or from the activation and the proliferation of hepatic progenitor cells from the canal of Hering to regenerate the liver parenchyma when the regenerative function of hepatocytes is impaired (29). To determine the cell of origin for the ductular-reactive cells in our model, we crossed Alb-Cre − ; Atg7 fl/fl ; Pten fl/fl or Alb-Cre − ; Atg5 fl/fl ; Pten fl/fl mice with the double reporter Rosa26 LoxP-Tomato-LoxP-GFP (Rosa26 mTmG ) and caused Cre-mediated recombination only in hepatocytes using the AAV8-TBG-Cre adenoassociated virus (AAV) (Fig. 3D and fig. S3F), where the Cre recombinase is expressed under the hepatocyte-specific thyroxine binding globulin (TBG) promoter (29). Following recombination, green fluorescent protein (GFP) will only be expressed in hepatocytes at the membrane, while non-recombined cells and unaffected tissues will remain Tomato + . Using this approach, we found that SOX9 + ductular-reactive cells expressed GFP at the membrane 40 days following AAV8-Cre infection in autophagy-deficient livers (Fig. 3E), confirming the hepatocyte origin of the ducts ( fig. S3G). Together, Data are means ± SD of four or five mice per group and were analyzed by one-way ANOVA with Tukey correction for multiple comparison tests (*P < 0.05, **P < 0.01, and ****P < 0.0001). All data points are the mean from five pictures per mouse. FoV, field of vision. Please note that data are the same controls for WT and Alb-Cre + ; P fl/fl mice as shown in fig. S1 (E to G).  Data are means ± SD of five mice per group and were analyzed by one-way ANOVA with Tukey correction for multiple comparison tests (*P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001). All data points are the mean were from five pictures per mouse. Please note that data are the same controls for WT and Alb-Cre + ; P fl/fl mice as shown in fig. S3 (A to C). (B) Quantitative reverse transcription polymerase chain reaction (RT-PCR) analysis of Afp mRNA isolated from 4-to 5-week-old livers. 18S was used as the internal amplification control. Data are means ± SD of three mice per group and were analyzed by one-way ANOVA with Tukey correction for multiple comparison tests (**P < 0.01 and ****P < 0.0001). All data points are the mean from technical triplicates. CT, cycle threshold. (C) Enzyme-linked immunosorbent assay (ELISA) analysis of AFP from the serum of 4-to 5-week-old mice. Data are means ± SD of three mice per group and were analyzed by one-way ANOVA with Dunnett correction for multiple comparison tests (****P < 0.0001). All data points are the mean from technical triplicates. our data establish that autophagy prevents dedifferentiation of hepatocytes into ductular LPCs.

HCCs originate from ductular LPCs in autophagy-deficient livers
ATG7-deficient livers develop HCCs at around 1 year of age (20). Since the ductular reaction is an early event following autophagy inhibition to regenerate the liver and ductular reactive cells express stem cell markers ( Fig. 3A and fig. S3, A to C) found in cancer stem cells from HCC (40), we hypothesized that ductular LPCs form HCCs in autophagy-deficient livers. To test this, we first assessed whether autophagy-deficient HCCs retain the expression of the duct marker SOX9, and we noted the presence of two distinct hepatocyte populations (SOX9 + and SOX9 − ) in the normal region surrounding liver HCCs, with SOX9 + hepatocytes found adjacent to ductular structures (Fig. 4A). We found that hepatocytes forming HCCs preserved the ductular marker SOX9 (Fig. 4A). To further evaluate the role of the ductular reaction in tumorigenesis, we infected Alb-Cre + ; Atg7 fl/fl ; Pten +/fl and WT mice with the AAV8-TBG-GFP adenovirus at 6 weeks of age to label hepatocytes with GFP ( Fig. 4B). At this age, the ductular reaction is occurring in autophagy-deficient livers, which allows us to distinguish and discriminate between resident hepatocytes (GFP + ) and ductular reactive cells (GFP − ) following AAV8-TBG-GFP infection to trace their role in tumorigenesis. First, we confirmed that at 7 days after AAV8-TBG-GFP infection, SOX9 + LPCs were GFP − , while hepatocytes (SOX9 − ) expressed GFP in autophagy-deficient livers (Fig. 4C), confirming that ductular LPCs are not expressing GFP following AAV8-TBG-GFP infection. We then assessed the expression of GFP in autophagy-deficient HCCs 100 days after AAV8-TBG-GFP infection. This revealed that tumors forming in Alb-Cre + ; Atg7 fl/fl ; Pten +/fl livers expressed no GFP in comparison to the surrounding normal hepatocytes, which retained GFP expression (Fig. 4D), highlighting that the ductular cells initiate tumorigenesis in autophagydeficient livers. We also found that high expression of SOX9 correlates with a decreased survival in human HCCs (Fig. 4E). Together, our data establish that ductular LPCs, formed early upon autophagy deficiency, ultimately lead to the generation of HCCs in autophagydeficient livers.

Autophagy loss in the liver enhances a YAP/TAZ signature within ductular LPCs
Blocking the formation of the ductular reaction would be beneficial in preventing human HCC (41). YAP and TAZ are transcriptional coactivators essential in controlling organ size (42), hepatocyte dedifferentiation (31), stemness (43), and liver tumorigenesis (44,45). The Hippo pathway regulates the activation of YAP and TAZ, and phosphorylation of both coactivators primes them for degradation. As our autophagy-deficient liver model develops severe hepatomegaly ( Fig. 1D and fig. S1D), dedifferentiates hepatocytes into ductular LPCs (Figs. 2 and 3 and figs. S2 and S3), and induces tumorigenesis, we next investigated whether YAP and TAZ are active in early-stage autophagy-deficient livers exhibiting ductular reaction. First, we compared the protein levels of the inactive forms of YAP and TAZ (phosphorylated YAP and phosphorylated TAZ), with the levels of total YAP and total TAZ (active forms) in 4-to 5-week-old livers (Fig. 5A). We noticed that the ratio of phosphorylated YAP and phosphorylated TAZ was reduced in autophagy-deficient livers in comparison to WT counterparts (Fig. 5A), highlighting that unphosphorylated YAP and unphosphorylated TAZ accumulate in autophagy-deficient livers undergoing ductular reaction.
To evaluate whether YAP and TAZ are functionally active in autophagy-deficient livers, we tested for the expression of YAP/TAZ transcriptional targets in 4-to 5-week-old livers. We found that mRNA levels of connective tissue growth factor (Ctgf ), amphiregulin (Areg), and cysteine-rich angiogenic inducer 61 (Cyr61), three YAP/TAZ target genes (46,47), were all significantly up-regulated in autophagy-deficient livers ( Fig. 5B and fig. S4A). At the protein level, CTGF was increased in total liver lysates of all autophagydeficient conditions (Fig. 5A). Next, we assessed the localization of YAP and TAZ in 4-to 5-week-old autophagy-deficient livers and observed that both YAP and TAZ strongly accumulated in the ductular cells, whereas YAP and TAZ were found in the bile duct and the canal of Hering of WT counterparts ( Fig. 5C and fig. S4B). Collectively, our data therefore indicate that autophagy loss in hepatocytes triggers a YAP/TAZ signature within the ductular LPC population.

TAZ is not degraded by autophagy
YAP is turned over not only by the proteasome (48,49), but also by autophagy as shown in recent reports (20,50). As TAZ is a YAP homolog, we next wondered whether TAZ accumulation and activation in our autophagy-deficient livers were due to blockage of autophagy-mediated degradation of TAZ. To test more directly whether TAZ is degraded by autophagy, we first deleted ATG7 or ATG5 expression in the liver cancer cell lines HLE and Huh7 using the CRISPR-Cas9-mediated gene disruption system. Next, we treated each cell line with Earle's balanced salt solution (EBSS), to induce starvation-mediated autophagy, in combination with or without 200 nM bafilomycin A1 (Baf) for 2 hours to prevent lysosomal degradation of autophagosomes. We checked for the efficient disruption of ATG7 or ATG5 expression following lenti-  S5, A and B). Unexpectedly, we also observed that not only YAP levels accumulated under EBSS only and EBSS and Baf conditions but also this occurred in ATG7 CRISPR /ATG5 CRISPR cells, indicating that this was an autophagy-independent effect. Together, our data indicate that TAZ and YAP are not directly turned over by autophagy in liver cells and that the accumulation of YAP and TAZ in autophagy-deficient livers is not the result of the inhibition of the autophagy degradation pathway but instead is due to the expansion of ductular cells in vivo, which are known to express YAP and TAZ ( Fig. 5 and fig. S4) (51).

YAP and TAZ deletion blocks ductular cell formation and tumorigenesis in autophagy-deficient livers
Deletion of YAP partially rescued hepatomegaly, fibrosis, and tumorigenesis induced by autophagy blockage in the liver (20). As a YAP homolog, TAZ can compensate YAP activity if the latter is lost (52). Since we observed in our model that YAP and TAZ are activated within the ductular LPC population, we hypothesized that deleting both YAP and TAZ might prevent the early ductular reaction and subsequent HCC formation in autophagy-deficient livers. First, we evaluated whether TAZ has a role in the phenotype of autophagydeficient livers. To test this, we crossed Wwtr1 flox/flox (encoding TAZ) mice (53) with our liver-specific autophagy-deficient model, and we observed that loss of TAZ significantly reduced liver size of 4-to 5-week-old autophagy-deficient livers ( Fig. 6A and fig. S6A). Next, we found that TAZ loss also significantly reduced the accumulation of activated -SMA + hepatic stellate cells and collagen deposition in 4-to 5-week-old autophagy-deficient livers (Fig. 6B  and fig. S6B), indicating that TAZ contributes to hepatic stellate cell activation and fibrosis in our model. In addition, TAZ loss significantly decreased SOX9 + , panCK + , and EpCAM + cells in 4-to 5-week-old autophagy-deficient livers ( Fig. 6B and fig. S6B), highlighting that TAZ loss hinders the formation of ductular LPCs upon autophagy deficiency in the liver. We next compared tumor formation between Alb-Cre + ; Atg7 fl/fl ; Pten +/fl or Alb-Cre + ; Atg5 fl/fl ; Pten +/fl and Alb-Cre + ; Atg7 fl/fl ; Pten +/fl ; Taz fl/fl or Alb-Cre + ; Atg5 fl/fl ; Pten +/fl ; Taz fl/fl in 140-day-old livers and noted that TAZ deletion caused a highly significant decrease in tumorigenesis in autophagy-deficient livers (Fig. 6, C and D, and fig. S6, C and D) that was accompanied by a significant increase in the survival of autophagy-deficient mice ( Fig. 6E and fig. S6E). Last, we evaluated whether TAZ has a role in the proliferation of ductular LPCs. We found that TAZ loss did not impair the number of Ki-67 + proliferative LPCs in 4-to 5-week-old autophagy-deficient livers ( fig. S7).
To evaluate whether there was any redundancy between YAP and TAZ in our model, we crossed Yap1 flox/flox mice (53) to our liverspecific (Alb-Cre) autophagy-and TAZ-deficient model to evaluate the effect of YAP/TAZ double knockout on the ductular reaction and tumorigenesis of autophagy-deficient livers. Unexpectedly, we observed that 40% (9 of 22 mice) of YAP-deficient mice developed jaundice within 6 to 8 weeks regardless of Atg7, Atg5, Pten, or Wwtr1 genotype. This is likely because YAP is highly expressed in the bile duct of WT mice ( Fig. 5C and fig. S4B), and the Albumin promoter driving Cre recombinase expression is expressed in hepatoblasts, the embryonic progenitor cells generating hepatocytes and cholangiocytes (54). YAP deletion in our Albumin-Cre model can therefore impair cholangiocyte function in the bile duct leading to acute jaundice. To overcome this phenotype for long term studies, we used AAV8-TBG-Cre adenovirus to induce Cre recombination more specifically in the hepatocytes of our Atg7 flox/flox ; Pten flox/flox ; Yap flox/flox ; Taz flox/flox model (Fig. 7A). First, we assessed the effect of YAP/TAZ deletion on the hepatomegaly and ductular reaction of autophagydeficient livers 3 weeks following AAV8-TBG-Cre recombination and confirmed the recombination of Atg7, Pten, Yap, and Wwtr1 alleles in AAV8-TBG-Cre-infected livers ( fig. S8). We found that although YAP or TAZ deletion significantly reduced hepatomegaly of autophagy-deficient livers (Fig. 7B), YAP/TAZ double knockout mice significantly restored liver size to that observed in nonrecombined counterparts infected with the AAV8-TBG-null adenovirus (Fig. 7B). In addition, we noted that while the individual deletion of Yap or Taz significantly impaired the formation of SOX9 + cells in autophagy-deficient livers (Fig. 7, C and D), only YAP/TAZ codeletion completely blocked the formation of SOX9 + cells in autophagydeficient livers (Fig. 7, C and D). In this AAV8-TBG-Cre model, Atg7 / ; Pten / mice had to be culled because of hepatomegaly and did not develop tumors at humane end point. To evaluate the role of YAP/TAZ loss in the tumorigenesis of autophagy-deficient livers, we infected Atg7 flox/flox ; Pten +/flox ; Yap flox/flox ; Taz flox/flox with AAV8-TBG-Cre adenovirus and assessed tumor formation 140 days following AAV8 infection (Fig. 7E). We observed that while Yap or Taz deletion significantly impaired tumorigenesis in autophagydeficient livers (Fig. 7, F and G), only YAP/TAZ codeletion completely prevented tumor formation (Fig. 7, F and G). Our data therefore show that deleting YAP and TAZ suppresses the ductular reaction and tumorigenesis of autophagy-deficient livers. However, in this context, we observed functional redundancy between YAP and TAZ, and only the combined deletion of both these genes could revert the effects on tissue overgrowth and tumor development.

DISCUSSION
We report a new model for extensive ductular reaction upon deletion of ATG5 or ATG7 and PTEN in the murine liver. Although Ptendeficient livers develop steatosis and HCC (37), we observed that hepatic Pten deletion alone did not initiate liver damage, inflammation, hepatic stellate cell activation, fibrosis, or a ductular reaction in young livers, but these effects were observed on hepatic deletion of ATG5 or ATG7. ATG5 and ATG7 are two proteins that are essential for the stage of autophagy that involves LC3 conjugation. ATG5 and ATG7 are also important for two other processes that involve the LC3 conjugation machinery: LC3-associated phagocytosis (LAP) (55) and LC3-associated endocytosis (LANDO) (56). We consider, however, that the core observations in our study relating to tumor development and liver injury are connected to autophagy, as previous studies have shown that they can be reversed by concomitant deletion of the autophagy adapter protein p62 (11,22,57), and autophagy adapter proteins are not thought to be involved in LAP or LANDO (58). We cannot fully discount that some of the effects we observe on deletion of ATG5 or ATG7 may be related to LAP or LANDO rather than autophagy or a combination thereof. Future studies to clarify this point using deletion of other factors such as FIP200 or ATG13 that are involved in autophagy, but not LAP and LANDO (59)(60)(61)(62), would certainly be merited to investigate this possibility.
Autophagy is impaired in Pten-deficient mice due to mTORC1 activation; however, autophagy is not blocked in Pten-deficient livers (63). LC3 is still conjugated to phosphatidylethanolamine leading to autophagosome and autolysosome formation when Pten expression is lost (63). This dictates an important role for autophagy in hepatocytes to prevent the microenvironmental remodeling and ductular reaction in healthy livers, with Pten cooperating with the autophagy-specific phenotype. Pten loss induces cellular senescence to protect from tumorigenesis in different models (64,65). However, we noticed the presence of apoptotic hepatocytes following autophagy abrogation and Pten deletion. The extent of injury in hepatocytes determines their fate toward senescence or cancer (66). Acute injury in hepatocytes results in senescence (67), while chronic injury does not activate senescence in hepatocytes, ultimately leading to HCC (66). Autophagy degrades damaged mitochondria, a process named mitophagy, to maintain cellular homeostasis. In hepatocytes, loss of autophagy leads to ROS accumulation, damaged mitochondria, and dysfunction (11,22,68,69). We suggest that the persistence of chronic damage and defects in damaged mitochondria clearance by mitophagy drive apoptosis and tumorigenesis in our autophagy-and Pten-deficient livers.
In our autophagy-and Pten-deficient model, we observed that following liver injury, hepatocytes dedifferentiate into ductular LPCs. This ductular reactive phenotype is not unique to the loss of autophagy as it has previously been observed in animal models subjected to diet modification, e.g., a diet enriched in 3,5-diethoxycarboncyl-1,4-dihydrocollidine (70) or choline-deficient, ethionine-supplemented diet (71). This indicates that the ductular reaction is likely to be a secondary effect of autophagy inhibition due to liver damage caused by autophagy loss. The origin of the ductular reaction in rodents is still controversial, with reports indicating the role of biliary cells (28,29) or hepatocytes (30,31) in forming LPCs with the capacity for generating new hepatocytes upon liver injury. Here, we show in a genetically modified mouse model that ductular reactive cells arise from mature hepatocytes upon injury induced by autophagy deficiency. Data are means ± SD of five mice per group and were analyzed by one-way ANOVA with Tukey correction for multiple comparison tests (*P < 0.05, **P < 0.01, and ****P < 0.0001). All data points are the mean from five pictures per mouse. X fl/fl , AAV8-null infected; X / , AAV8-Cre infected.
The cellular plasticity of human hepatocytes can also generate ductular cells in a transplantation mouse model (30), strengthening the hepatocyte origin of the ductular reaction in human liver diseases. The plastic differentiation program of the ductular reaction for liver regeneration is defined by the origin of the injuries. Following bile duct injury, resident LPCs/biliary cells (26) and hepatocytederived LPCs (72) regenerate biliary cells. When hepatocyte function is impaired, resident LPCs/biliary cells (28,29,73) or hepatocytederived LPCs (30,31,74) generate new hepatocytes. The decision to recruit biliary cells or hepatocytes during the ductular reaction remains elusive, and future studies will be required to shed further light on this mechanism. Autophagy loss has been previously shown to give rise to HCC in mice (20). Our results suggest that the hepatocyte-derived ductular reaction gives rise to HCC in autophagy-deficient livers. While some studies conclude that the ductular reaction is not involved in liver carcinogenesis (34,35,74), other studies do report a role for the ductular reaction in initiating HCCs (32,33). Although all these studies recombine LPCs for lineage tracing, they differ with respect to the timing between the induction of LPC labeling and the start of the injury. Recombination of LPCs for lineage tracing before inducing liver injury (34,35,74) does not label hepatocyte-derived LPCs, excluding them from the lineage tracing of HCCs. In contrast, recombination of LPCs for lineage tracing following liver injury results in LPC-derived HCCs (32,33). In our autophagy-and Pten-deficient model, we report that hepatocyte-derived LPCs generate SOX9 + hepatocytes that give rise to HCC. The ability of LPCs to induce tumorigenesis has been controversial since it is generally accepted that HCC originates from hepatocytes. Here, we reconcile these findings by showing that HCC does originate from hepatocytes, but these hepatocytes, early upon liver injury, dedifferentiate into LPCs to attempt to regenerate liver function, before transforming into HCC.
In human liver diseases, the accumulation of LPCs is observed in nonalcoholic steatohepatitis-induced cirrhosis preceding HCC (75), and the presence of peritumoral ductular reaction is a poor prognostic factor for human HCC after resection (76), indicating the importance of targeting the ductular reaction in human liver diseases. The gene signature of autophagy-deficient mice is similar to the human transcriptomes of nonalcoholic fatty livers (20), and rat livers from rats fed a high-fat diet reduce their autophagy function (77). Restoring autophagy could therefore be a beneficial treatment in injured livers harboring a ductular reaction.
Mechanistically, we report that YAP and TAZ cooperate to drive hepatocyte dedifferentiation and tumorigenesis in autophagy-deficient livers. Unlike a previous study on YAP (20), we uncovered that TAZ also plays a role in promoting hepatomegaly, ductular reaction, stromal activation, fibrosis, and tumorigenesis in autophagy-deficient livers. TAZ deletion alone, similar to YAP deletion alone (20), only impaired carcinogenesis in autophagy-deficient livers. However, TAZ loss did not impair the proliferative outgrowth of the ductular LPC population. Here, we speculate that TAZ is involved in the differentiation switch in our model as its homolog YAP can directly drive hepatocyte dedifferentiation (31), and, more recently, YAP/TAZ have been described as regulators of stemness and cell plasticity in glioblastoma (78). We found that YAP and TAZ are not directly turned over by autophagy and that their accumulation in the absence of autophagy in vivo is associated with the increased presence of ductular cells, which are known to express YAP and TAZ (79).
YAP and TAZ are mechanosensors and mechanotransducers (80), and their activation is linked to the stiffness of the extracellular matrix (81). As we noted a significant increase in extracellular matrix remodeling and fibrosis ( Fig. 1E and fig. S1, F and G) correlating with a significant increase in YAP + /TAZ + ductular LPCs in our models, we suggest that YAP and TAZ are also activated in response to the microenvironment changes following autophagy and PTEN deletion in the liver. Building on these findings, we observed that only the combined deletion of YAP and TAZ prevented the emergence of hepatocyte-derived LPCs that initiate tumorigenesis in autophagy-deficient livers. Our study uncovered a role for autophagy in suppressing the emergence of hepatocyte-derived ductular LPCs that can give rise to HCCs via concomitant activation of YAP and TAZ.

Animal experiments
Male and female animals were housed in a pathogen-free environment and kept under standard conditions with a 12-hour day/night cycle and access to food and water ad libitum. All in vivo experiments were carried out under guidelines approved by the Glasgow University Animal Welfare and Ethical Review Body and in accordance with U.K. Home Office guidelines under license P54E3DD25. As described previously (82) Experimental cohort (males and females) sizes were based on previous similar studies that have given statistically significant results while also respecting the limited use of animals in line with the 3R system: replacement, reduction, and refinement. All treatment studies were randomized but did not involve blinding. Genotyping was performed by Transnetyx. To lineage trace the ductular cell origin, we crossed our model with the Rosa26-mtdTomato-mEGFP mouse (the Jackson laboratory, stock 007576, RRID: IMSR_JAX:007576) (85).

Animal tissue harvesting and serum analysis
Mice were euthanized by CO 2 inhalation followed by cervical dislocation, and blood was harvested by cardiac puncture in accordance with U.K. Home Office guidelines. Tissues were weighed and stored immediately at −80°C or in paraffin blocks after fixation in 10% formalin (in PBS) for 24 hours, followed by dehydration in 70% ethanol before embedding. Blood samples (EDTA-plasma and serum) were stored at −80°C following 10-min centrifugation at 900g at 4°C. Serum was sent to the Veterinary Diagnostic Services (University of Glasgow) for ALT, AST, ALP, and GGT analyses.

Plasma AFP levels
Plasma AFP levels were assessed using the enzyme-linked immunosorbent assay (ELISA) kit (catalog no. ab210969) according to the manufacturer's instruction. Each sample was analyzed in triplicate.

Quantification and statistical analysis
For IHC studies, five representative pictures were taken per mouse and were analyzed using Fiji software. For all in vivo studies, data are shown as means ± SD. Sample normality was assessed by Shapiro-Wilk test. Statistical significances were determined by two-tailed unpaired Student's t test for two-group comparison, two-way analysis of variance (ANOVA) with Tukey or Dunnett for multiple group comparison, and log-rank (Mantel-Cox) test for survival comparison using GraphPad Prism software. Results were considered statistically different when *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001 with ns indicating no significance.

SUPPLEMENTARY MATERIALS
Supplementary material for this article is available at http://advances.sciencemag.org/cgi/ content/full/7/23/eabf9141/DC1 View/request a protocol for this paper from Bio-protocol.