ABL1, Overexpressed in Hepatocellular Carcinomas, Regulates ABL1, Overexpressed in Hepatocellular Carcinomas, Regulates Expression of NOTCH1 and Promotes Development of Liver Expression of NOTCH1 and Promotes Development of Liver Tumors in Mice Tumors in Mice

BACKGROUND & AIMS: We investigated whether ABL proto-oncogene 1, non-receptor tyrosine kinase (ABL1) is involved in development of hepatocellular carcinoma (HCC). METHODS: We analyzed clinical and gene expression data from The Cancer Genome Atlas. Albumin -Cre (Hep WT ) mice and mice with hepatocyte-speci ﬁ c disruption of Abl1 (Hep Abl – / – mice) were given hydrodynamic injections of plasmids encoding the Sleeping Beauty transposase and transposons with the MET gene and a catenin b 1 an N-terminal Some with the ABL1 inhibitor nilotinib or vehicle (control) daily for 4 weeks. We knocked down ABL1 with short hairpin RNAs in Hep3B and HCC cells and analyzed their proliferation and growth as xenograft tumors in mice. We performed RNA sequencing and gene set enrichment analysis of tumors. We knocked down or overexpressed NOTCH1 and MYC in HCC cells and analyzed proliferation. We measured levels of phosphorylated ABL1, MYC, and NOTCH1 by immunohistochemical analysis of an HCC tissue microarray. RESULTS: HCC tissues had higher levels of ABL1 than non-tumor liver tissues, which correlated with shorter survival times of patients. Hep WT mice with the MET and catenin b 1 transposons liver and survived a 64 days; Hep Abl – / – mice with these transposons tumors that were 50% and survived a slows growth of liver tumors in mice. Inhibitors of ABL1 might be used for treatment of HCC.


BACKGROUND & AIMS:
We investigated whether ABL protooncogene 1, non-receptor tyrosine kinase (ABL1) is involved in development of hepatocellular carcinoma (HCC). METHODS: We analyzed clinical and gene expression data from The Cancer Genome Atlas. Albumin-Cre (Hep WT ) mice and mice with hepatocyte-specific disruption of Abl1 (Hep Abl-/mice) were given hydrodynamic injections of plasmids encoding the Sleeping Beauty transposase and transposons with the MET gene and a catenin b1 gene with an N-terminal truncation, which induces development of liver tumors. Some mice were then gavaged with the ABL1 inhibitor nilotinib or vehicle (control) daily for 4 weeks. We knocked down ABL1 with short hairpin RNAs in Hep3B and Huh7 HCC cells and analyzed their proliferation and growth as xenograft tumors in mice. We performed RNA sequencing and gene set enrichment analysis of tumors. We knocked down or overexpressed NOTCH1 and MYC in HCC cells and analyzed proliferation. We measured levels of phosphorylated ABL1, MYC, and NOTCH1 by immunohistochemical analysis of an HCC tissue microarray. RESULTS: HCC tissues had higher levels of ABL1 than non-tumor liver tissues, which correlated with shorter survival times of patients. Hep WT mice with the MET and catenin b1 transposons developed liver tumors and survived a median 64 days; Hep Abl-/mice with these transposons developed tumors that were 50% smaller and survived a median 81 days. Knockdown of ABL1 in human HCC cells reduced proliferation, growth as xenograft tumors in mice, and expression of MYC, which reduced expression of NOTCH1. Knockdown of NOTCH1 or MYC in HCC cells significantly reduced cell growth. NOTCH1 or MYC overexpression in human HCC cells promoted proliferation and rescued the phenotype caused by ABL1 knockdown. The level of phosphorylated (activated) ABL1 correlated with levels of MYC and NOTCH1 in human HCC specimens. Nilotinib decreased expression of MYC and NOTCH1 in HCC cell lines, reduced the growth of xenograft tumors in mice, and slowed growth of liver tumors in mice with MET and catenin b1 transposons, reducing tumor levels of MYC and NOTCH1. CONCLU-SIONS: HCC samples have increased levels of ABL1 compared with nontumor liver tissues, and increased levels of ABL1 correlate with shorter survival times of patients. Loss or inhibition of ABL1 reduces proliferation of HCC cells and slows growth of liver tumors in mice. Inhibitors of ABL1 might be used for treatment of HCC. H epatocellular carcinoma (HCC) is the major form of liver cancer. It is the sixth most common malignancy globally and ranks fourth in total cancer-related deaths annually. 1 The 5-year overall survival of patients with a new diagnosis of HCC is <18%, and a majority of HCC patients present with advanced disease, so treatment options are limited. 2 Currently, first-line therapeutic agents for advanced HCC, either sorafenib or lenvatinib, increase survival by only approximately 3 months. 3 Recently a number of drugs, including regorafenib, cabozantinib, and nivolumab, have been approved by the US Food and Drug Administration for second-line treatment of HCC. 3 However, these drugs only offer a further increase in overall survival of 3-5 months. Therefore, it is imperative to develop new and more effective therapeutic strategies and agents to treat HCC, but achieving this goal requires a better understanding of the molecular signaling pathways that drive or mediate the development of the disease.
Abelson tyrosine-protein kinase 1 (ABL1) is a nonreceptor tyrosine kinase of the Abelson-like family. It is mostly known for its involvement in leukemias harboring the Philadelphia chromosome, which results from the translocation of the short arms of chromosomes 9 and 22, creating a fusion of the BCR gene to the second exon of the ABL1 gene, resulting in the production of the BCR-ABL fusion protein. 4 Recent evidence has shown that ABL1 also plays an important role in the development of solid tumors, such as melanoma, breast cancer, ovarian cancer, and lung cancer, by an independent mechanism not involving any fusion oncoproteins. 4 We previously reported that ABL1 is overexpressed and activated in human HCC specimens. 5 However, the role of ABL1 in hepatocarcinogenesis must be understood because it is critical for determining whether ABL1 is a suitable candidate target in the treatment of HCC.
We report that overexpression of ABL1 correlates with poor prognosis in HCC. We investigated the role of ABL1 in HCC growth using in vitro and mouse models. We found that ABL1 inhibition impaired HCC growth and extended overall survival of mice with HCC. Mechanistically, we found that inhibition of ABL1 suppresses HCC cell growth by decreasing NOTCH1 expression through the regulation of c-MYC. Collectively, our data strongly suggest that ABL1 is involved in the pathogenesis of HCC and that its inhibition could be a promising novel strategy to treat this disease.

Mice and Treatments
All animals received humane care according to the Guide for the Care and Use of Laboratory Animals. The procedures for all animal experiments detailed were approved by the Institutional Animal Care and Use Committee of Loyola University Chicago. All mice were housed in micro-isolator cages in a room illuminated from 7:00 AM to 7:00 PM (12:12-hour light-dark cycle) and were given access to water and chow ad libitum.
To generate mice with hepatocyte-specific Abl1-deficiency, Abl1 flox/flox mice (Jackson Laboratory, Bar Harbor, ME; cat #013224) were backcrossed to C57BL/6J mice for 5 generations and then were mated with Albumin-Cre mice (Jackson Laboratory; cat #003574). The resulting offspring Alb-Cre; Abl1 flox/þ mice were then mated to generate the Albumin-Cre and Alb-Cre; Abl1 flox/flox littermates. The age-(6-8 weeks old) and sex-matched Albumin-Cre and Alb-Cre; Abl1flox/flox mice were injected with plasmids, encoding the Sleeping Beauty transposase (HSB2) and transposons with GFP (pT3-GFP) or MET gene and catenin b1 gene with the Nterminal truncation (referred to here as MET/CAT), as described previously. 10 Four weeks after MET/CAT injection, C57BL/6J wild-type mice were given with vehicle (30% captisol), nilotinib (20 mg/kg) or sorafenib (25 mg/kg) by oral gavage daily for 4 weeks or before being sacrificed (some mice treated with vehicle had to be euthanized earlier due to tumor burdens). Age-and sex-matched mice were allocated to different

BACKGROUND AND CONTEXT
The oncogene ABL1 might be involved in the pathogenesis of hepatocellular carcinoma (HCC).

NEW FINDINGS
HCC samples have increased levels of ABL1 compared with non-tumor liver tissues, and loss or inhibition of ABL1 reduces proliferation of HCC cells and slows growth of tumors in mice.

LIMITATIONS
This study was performed using human tissue samples, cell lines, and mice. Further studies in humans are needed.

IMPACT
Inhibitors of ABL1 might be developed for treatment of HCC treatment groups. Both male and female mice were used in the experiments. Six-to-eight-week-old mice were used for hydrodynamic injections.
Xenograft model. Huh7 cells (5 Â 10 6 in 100 mL serum-free medium) were injected into the left or right flanks of the 8-to 12-week-old SCID-bg mice. Three weeks postinjection, some mice were given with vehicle (30% captisol) or nilotinib (20 mg/kg) by oral gavage daily for 10 days. Tumor volumes were measured daily using a caliper until the day of sacrifice.

Western Blotting
Western blotting was performed as described previously. 6 Information on primary antibodies is provided in Supplementary Table 2.

Quantitative Real-Time Polymerase Chain Reaction
Cellular or tissue messenger RNA (mRNA) was extracted using Zymo mini-columns and quantitative real-time polymerase chain reactions were performed as described previously. 6 Primers used for real-time polymerase chain reaction are listed in Supplementary Table 3.

Immunohistochemical Staining
IHC was performed as described previously. 10 Human tissue microarrays LV801, LV807, and LV8012 were purchased from US Biomax (Rockville, MD). LV801 and LV807 contain a total of 66 cases of HCC and 50 cases of HCC adjacent normal or normal liver tissues. LV8012 contains 80 cases of HCC (TNM stage II-IV). The IHC signals were quantified visually. The staining was scored as À (0, negative), þ (1, weak signal), þ þ (2, moderate signal), and þ þ þ (3, strong signal) by 2 independent observers, including a pathologist from Loyola University Chicago; a sample was rated as positive if it showed at least 1% of cells with a staining score !1þ. For IHC on mouse samples, cells with positive staining were scored in at least 5 fields at 400Â or 200Â magnification and reported as mean ± SD. Information on primary antibodies for IHC is provided in Supplementary Table 2.

Terminal Deoxynucleotidyl Transferase-Mediated Deoxyuridine Triphosphate Nick-End Labeling Staining
Terminal deoxynucleotidyl transferase-mediated deoxyuridine triphosphate nick-end labeling (TUNEL) staining was performed as described previously. 6 The TUNEL-positive cell number was scored in at least 5 fields at 400Â magnification/ mouse and reported as mean ± SD. Three or more mice were used in each group.

RNA Sequencing and Analysis
RNA from scrambled control and ABL1-KD Huh7 cells was extracted using RNeasy Plus Micro Kit (Qiagen, Germantown, MD). RNA sequencing was performed by Novogene Corporation (Hong Kong). Gene set enrichment analysis (GSEA) was performed using the 3.0 GSEA software. The RNA sequencing data were deposited into the National Center for Biotechnology Information's Gene Expression Omnibus database (GEO GSE133294. https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi? acc¼GSE133294).

Chromatin Immunoprecipitation Assay
Huh7 cell were cultured as described above and chromatin immunoprecipitation assays were performed as described previously 11 using a c-MYC antibody (#5605; Cell Signaling, Danvers, MA). The primers used are listed in Supplementary  Table 3.

Human Sample Analysis
Alterations of ABL1, NOTCH1, and c-MYC mRNA were analyzed from publicly available data from The Cancer Genome Atlas (TCGA). 12

Statistical Analysis
Statistical analyses were performed using GraphPad Prism software, version 8.0. Variation is indicated using standard error presented as mean ± SD. Statistical significance was calculated using the 2-tailed Student t test, except for the experiments involving repeated measures, which were analyzed using 2-way analysis of variance. P < .05 was considered significant. Means ± SDs are shown in the Figures where applicable.

High Expression of ABL1 in Human Hepatocellular Carcinomas Is Positively Correlated With Shorter Survival Times of Patients
We previously reported that ABL1 is overexpressed in human HCC specimens with a small sample size. 5 To confirm the results with a larger sample size, we analyzed the TCGA database and confirmed that ABL1 was expressed at higher levels in HCCs compared to normal liver tissues ( Figure 1A). In addition, we found that higher ABL1 expression is positively correlated with poorer prognosis in human HCC patients from the TCGA database (Figure 1B  and Supplementary Figure 1), which suggests that it could be a good prognostic factor. We also performed 2 additional HCC tissue microarrays, which contain 66 cases of HCC and 50 cases of HCC adjacent normal or normal liver tissues. Consistent with previous results, 5 we found that ABL1 protein levels were significantly higher in tumors compared to normal liver tissues ( Figure 1C and D). Kinase activity is critical for the functions of ABL1. 13 Phosphorylation of Tyr412, which is located in the kinase activation loop of ABL1, is required for its kinase activity. 14, 15 We found that the level of p-ABL1 (p-Tyr412) is largely absent in normal liver tissues, but is abundant in HCC specimens ( Figure 1C and E). It is notable that 86% of HCC specimens with high p-ABL1 (p-Tyr412) staining also express high levels of ABL1 ( Figure 1D and E). In general, these data indicate that ABL1 is overexpressed and activated in human HCCs, and that these factors correlate with shorter survival times of patients.

Deletion of Abl1 in Hepatocytes Does Not Affect Morphology, Histology, Proliferation, or Apoptosis in Mouse Livers
To investigate the role of ABL1 in liver tumorigenesis, we generated mice with hepatocyte-specific deletion of Abl1 (Albumin-Cre; Abl1 flox/flox ). Albumin-Cre; Abl1 flox/flox mice (referred to here as Hep Abl1-/-) express Cre recombinase from the albumin promoter, which is specifically expressed in hepatocytes. Hep Abl1-/mice are viable, fertile, and visually indistinguishable from wild-type (Albumin-Cre, referred to here as Hep WT ) mice, suggesting that Abl1 is not required for normal liver development. We confirmed that Cre recombinase expression in hepatocytes removes the Abl1 allele between 2 loxP sites (Supplementary Figure 2A) and showed that ABL1 expression was decreased in both whole liver tissue and hepatocytes (Supplementary Figure 2B). There was no significant difference in morphology or histology of livers between Hep WT and Hep Abl1-/mice (Supplementary Figure 2C and D). Furthermore, Abl1 deficiency did not affect cell proliferation or apoptosis in mouse livers (Supplementary Figure 2E-H). These results suggest that deletion of Abl1 in hepatocytes does not affect mouse liver homeostasis.

Deficiency of Abl1 in Hepatocytes Suppresses MET/CAT-Induced Hepatocellular Carcinoma Development
MET can bind directly to ABL1 and activate it in mouse mammary tumors and breast cancer cells. 16 To determine the role of ABL1 in HCC development, we used the MET (MET)/bcatenin (CAT)-driven HCC model, which is useful for studying the functions of genes in hepatocarcinogenesis because of its clinical relevance and efficiency of HCC induction. 10,17 We found that levels of phosphorylation of both ABL1 on Tyr412 and CRKL, a direct target of ABL kinases 18 commonly used to assess ABL kinase activity, was increased in MET/CATinduced liver tumors, suggesting that ABL1 is activated in MET/CAT-induced HCC (Figure 2A). Expression of ABL1 was also increased in MET/CAT-induced liver tumors (Figure 2A). In contrast, expression or phosphorylation of ABL1 was not altered in the diethylnitrosamine-induced HCC model (Supplementary Figure 3). These observations suggest that the MET/CAT model is suitable for studying the role of ABL1 in hepatocarcinogenesis. We hydrodynamically injected age-and sex-matched Hep WT and Hep Abl1-/mice with plasmids encoding the Sleeping Beauty transposase (HSB2) and transposons with the MET/CAT oncogenes. Comparable transfection efficiency was observed in wild-type and Abl1 KO mouse livers (Supplementary Figure 4). Intriguingly, we found that the overall tumor load and tumor sizes in Hep Abl1-/mice was decreased significantly compared to Hep WT mice ( Figure 2B). The relative liver weight, including tumor vs body weight, in Hep Abl1-/mice was decreased by 50% compared to Hep WT mice ( Figure 2C). Importantly, the Hep WT mice with a liver tumor burden died at the age of 55-70 days (median survival 64 days) compared to a median survival of 81 days in Hep Abl1-/mice ( Figure 2D). These data indicate that Abl1 deficiency in hepatocytes suppresses MET/CAT-induced HCC growth and prolongs survival of mice with HCC.
ABL1 regulates cell survival and proliferation. 4 Suppression of HCC development by the deletion of Abl1 could be due to increased apoptosis or decreased proliferation of tumor cells. We first analyzed proliferation in the MET/CAT-injected livers from Hep WT and Hep Abl1-/mice by Ki67 staining. The number of Ki67-positive cells was significantly decreased in tumors but not in tumorfree areas in Abl1-deficient livers compared to wild-type livers ( Figure 2E and F). We did not find significant differences in apoptosis in livers of Hep WT mice compared to those of Hep Abl1-/mice ( Figure 2G and H). These results demonstrate that Abl1 deficiency in hepatocytes decreases tumor cell proliferation but not cell survival in MET/CATinduced HCC.

Knockdown of ABL1 Reduces Human Hepatocellular Carcinoma Cell Proliferation and Suppresses Tumor Growth
We further investigated the role of ABL1 in human HCC. To achieve this goal, we used shRNA to knock down ABL1 expression in 2 HCC cell lines that show high levels of ABL1 mRNA, Hep3B, and Huh7 (Supplementary Figure 5). Western blotting showed that ABL1 was successfully knocked down by 2 shRNAs in both cell lines ( Figure 3A). Importantly, knockdown of ABL1 significantly reduced cell growth in both Hep3B and Huh7 cells ( Figure 3B). Consistently, ABL1 knockdown decreased growth as xenograft tumors in mice ( Figure 3C-E). Tumors grown from ABL1-KD cells also displayed less cell proliferation compared to those grown from scrambled-RNA control cells ( Figure 3F and G). However, there was no significant difference in apoptosis in tumors from the 2 groups ( Figure 3H and I). Collectively, these results indicate that knockdown of ABL1 reduces human HCC cell proliferation and suppresses tumor growth.

Knockdown of ABL1 Inhibits Hepatocellular Carcinoma Cell Proliferation by Decreasing NOTCH1 Expression
To determine the molecular mechanisms by which ABL1 promotes HCC cell proliferation, we performed RNA sequencing analysis using scrambled-control and ABL1-KD cells and proceeded to do GSEA. NOTCH signaling is one of the most significantly down-regulated gene pathways that result from ABL1 knockdown ( Figure 4A and B). NOTCH signaling has been shown to play an important role in tumor cell growth in many types of cancer, 19 including HCC. We therefore hypothesized that knockdown of ABL1 inhibits HCC cell proliferation by inhibiting signaling along the NOTCH pathway. We used real-time polymerase chain reaction assays to confirm that ABL1 knockdown decreased mRNA expression for a number of NOTCH signaling pathway genes, including NOTCH1, NOTCH3, JAG1, LFENG, and DTX1 ( Figure 4C). We also found that expression of NOTCH downstream targets, including CyclinD1, NRARP, HES1, and HES2, were reduced by knockdown of ABL1 in HCC cells ( Figure 4D), suggesting that NOTCH activity was suppressed in ABL1-KD cells.
Because NOTCH signaling receptors NOTCH1 and NOTCH3 play critical roles in HCC cell growth, [19][20][21] we focused on testing whether knockdown of ABL1 inhibits cell proliferation by decreasing the expression of either NOTCH1 or NOTCH3 in HCC cells. We first performed Western blotting to determine whether the expression of NOTCH1 and/ or NOTCH3 proteins were also decreased by knockdown of ABL1 in HCC cells. Intriguingly, expression of NOTCH1 but not NOTCH3 protein was significantly lower in ABL1-KD cells compared to scrambled-RNA control cells ( Figure 4E), suggesting a possible post-transcriptional, translational, or post-translational mechanism by which NOTCH3 might be regulated in ABL1-KD cells. We therefore focused on NOTCH1 and hypothesized that knockdown of ABL1 inhibits cell proliferation by decreasing NOTCH1 in HCC cells. We found that expression of NOTCH1 protein was higher in MET/CAT-induced liver tumors compared to control mouse liver, and was reduced in liver tumors when Abl1 was deleted in hepatocytes ( Figure 4F). Consistently, expression of NOTCH1-targeted genes was also downregulated in Abl1-KO mouse HCCs (Supplementary Figure 6A-C). To further determine whether decreased NOTCH1 expression could inhibit HCC cell proliferation, we knocked down NOTCH1 using shRNA in Huh7 and Hep3B cells ( Figure 4G). We tested 5 distinct shRNAs (data not shown) and only 1 shRNA efficiently decreased NOTCH1 expression in HCC cells ( Figure 4G). We found that knockdown of NOTCH1 significantly reduced HCC cell growth ( Figure 4H). In addition, NOTCH1 overexpression promoted cell proliferation and rescued the phenotype caused by ABL1 knockdown in HCC cells ( Figure 4I and J). Overall, our data demonstrate that knockdown of ABL1 inhibits HCC cell growth by decreasing NOTCH1 expression.

Knockdown of ABL1 Decreases NOTCH1 Expression Through Regulation of c-MYC in Hepatocellular Carcinoma Cells
We next investigated the molecular mechanisms by which ABL1 knockdown decreases NOTCH1 expression in HCC cells. As NOTCH1 mRNA levels were reduced by ABL1 knockdown, we reasoned that this knockdown affected the transcription and/or post-transcriptional processing of NOTCH1 mRNA. MicroRNAs (miRs) have been shown to directly target NOTCH1 and reduce NOTCH1 mRNA levels. 22 We therefore hypothesized that ABL1 knockdown might decrease NOTCH1 expression by increasing the expression of some miR(s). Using TargetScan, we identified several miRs, including miR-150-5p, miR-34-5p, and miR-146b-5p, which can potentially directly target NOTCH1. However, we found that only the expression of miR-146b-5p was increased when ABL1 was knocked down in HCC cells (Supplementary Figure 7A-C). To determine whether ABL1 knockdown decreases NOTCH1 expression through increasing miR-146b-5p expression in HCC cells, we used miR146b-5p mimics or inhibitors. We confirmed that the miR146b-5p mimic effectively induced expression of miR146b-5p and inhibitors of miR146b-5p sufficiently decreased expression of miR146b-5p (Supplementary Figure 7D and E). However, although we had expected that overexpression of miR146b-5p by the mimic miR would decrease expression of NOTCH1 in HCC cells; instead we found that NOTCH1 expression was slightly increased (Supplementary Figure 7F). Consistent with this observation, we found that miR146b-5p inhibition decreased the expression of NOTCH1 in HCC cells (Supplementary Figure 7G). These data would appear to suggest that ABL1 knockdown might not decrease NOTCH1 expression by increasing the expression of miRs.
We therefore tested whether ABL1 knockdown reduces the transcription of NOTCH1 by regulating transcription factors. ABL1 can regulate phosphorylation of the transcription factor c-MYC and its transcriptional activity. 23 c-MYC plays a critical role in HCC development. 24 Interestingly, GSEA indicated that MYC's target gene set was down-regulated in ABL1-KD cells ( Figure 5A). Consistently, c-MYC protein level was much lower in ABL1-KD HCC cells compared to scrambled-RNA cells ( Figure 5B). Expression of c-MYC protein was higher in MET/CAT-induced liver tumors compared to control mouse liver, and was reduced in liver tumors when Abl1 was deleted in hepatocytes (Supplementary Figure 8). Phosphorylation of c-MYC on Ser62, which is regulated by ABL1 23 and is critical for c-MYC stabilization, 25 was also decreased by ABL1 knockdown ( Figure 5B). In addition, using proximity ligation assay, 26 we found a strong interaction between ABL1 and c-MYC in HCC cells ( Figure 5C, Supplementary Figure 9). Although c-MYC has been shown to be a direct transcriptional target downstream of NOTCH1 in T-cell acute lymphoblastic leukemia, 27 we found that c-MYC protein was not significantly affected by either knockdown or overexpression of NOTCH1 in HCC cells (Supplementary Figure 10). On the other hand, enhanced c-MYC expression may increase the level of NOTCH1 mRNA through regulation of NRF2. 28,29 We therefore hypothesized that knockdown of ABL1 decreases NOTCH1 expression through regulation of c-MYC in HCC cells. We found that both mRNA and protein expression of NOTCH1 was decreased when c-MYC was knocked down in HCC cells ( Figure 5D and E). In line with this, c-MYC knockdown inhibited HCC cell growth ( Figure 5F). We analyzed the promoter of human NOTCH1 and identified a putative binding site for c-MYC (Supplementary Figure 11A). The chromatin immunoprecipitation assay revealed that c-MYC directly binds to the promoter of NOTCH1 in HCC cells (Supplementary Figure 11B and C). Moreover, overexpression of activated c-MYC promoted cell proliferation and restored decreased NOTCH1 expression and cell growth caused by ABL1 knockdown in HCC cells ( Figure 5G and H). Considered altogether, these data indicate that knockdown of ABL1 decreases NOTCH1 expression through regulation of c-MYC in HCC cells.

Expression of p-ABL1, c-MYC, and NOTCH1 Is Positively Correlated in Human Hepatocellular Carcinoma
Our data indicate that ABL1 regulates the phosphorylation of c-MYC, leading to increased c-MYC protein levels, which results in enhanced NOTCH1 mRNA expression and promotes HCC cell growth ( Figure 6A). To determine whether the ABL1/c-MYC/NOTCH1 axis is also relevant in human HCC, we examined the levels of p-ABL1 (Y412) (an indicator of ABL1 activity), c-MYC, and NOTCH1 in human HCC specimens using tissue microarrays. The expression of both c-MYC and NOTCH1 proteins was significantly correlated with levels of p-ABL1 in human HCC specimens ( Figure 6B-D). In addition, NOTCH1 protein expression correlated with c-MYC in human HCC ( Figure 6E). We further analyzed the TCGA database and found that NOTCH1 mRNA level positively correlated with ABL1 mRNA level ( Figure 6F). However, c-MYC mRNA level had no correlation with either ABL1 or NOTCH1 mRNA level in the TCGA HCC samples (Supplementary Figure 12), suggesting that ABL1 may not regulate c-MYC at transcript level in human HCC. Overall, these data indicate that the ABL1/c-MYC/NOTCH1 axis is important in human HCC.

ABL1 Inhibitors Suppress Hepatocellular Carcinoma Growth in Preclinical Models
The above results encouraged us to examine whether ABL1 inhibitors could be useful to treat HCC in preclinical models, which would provide a translational basis for the potential clinical use of ABL1 inhibitors in the treatment of HCC patients. Several ABL kinase inhibitors are already used clinically or are under investigation in clinical trials for treating chronic myelogenous leukemia and other cancers,4 but their efficacy in HCC remains unknown. Nilotinib, an ATP-competitive inhibitor of ABL kinases, has been approved by the US Food and Drug Administration to treat Philadelphia chromosome-positive chronic myelogenous leukemia. 30 GNF-5, a newer allosteric inhibitor that targets the myristate pocket of ABL kinases, has been tested in preclinical leukemia models. 31 Because ABL1 knockdown inhibits HCC cell growth (Figure 3), we hypothesized that ABL inhibitors would be efficacious in treating HCC. First, we confirmed that both nilotinib and GNF-5 effectively inhibit ABL kinase activity, as indicated by phosphorylation of CRKL in HCC cells (Supplementary Figure 13A and B). Further, we found that both nilotinib and GNF5 inhibited HCC cell growth in vitro ( Figure 7A, Supplementary Figures 13C and 14A). Importantly, the sensitivity of nilotinib positively correlated with the expression of ABL1 ( Figure 7B and Supplementary Figure 14B). We also found there is a positive correlation between ABL1 and c-MYC expression in human HCC cell lines (Supplementary Figure 14B and C). However, there is no significant correlation between ABL1 and NOTCH1 in these HCC cell lines (Supplementary Figure 14B and C ), suggesting other mechanisms may also regulate NOTCH1 expression. It is notable that NOTCH1 expression is positively correlated with ABL1 in HCC cells without p53 mutations (Hep3B, Huh7, SK-Hep1 and SNU-423), but there is no correlation in p53-mutant HCC cells (SNU-449, SNU-475, PLC/PRF/5, and SNU-387), suggesting that p53 mutations may affect NOTCH1 expression. Consistent with the functional inhibition of ABL1 using shRNAs, nilotinib effectively decreased the expression of c-MYC and NOTCH1 ( Figure 7C). Moreover, nilotinib significantly suppressed tumor growth ( Figure  7D-G) and proliferation (Supplementary Figure 15A-D) in an HCC xenograft model. To further test the efficacy of ABL inhibitors in mouse models with intact immune responses, we injected wild-type C57BL/6 mice with MET/CAT to induce HCC and gavaged these mice with nilotinib or vehicle solution by oral gavage daily for 4 weeks. We found that nilotinib treatment dramatically inhibited MET/CAT-induced tumor growth ( Figure 7H and I) and proliferation (Supplementary Figure 16), decreased the expression of MYC and NOTCH1 ( Figure 7J), and prolonged the survival of animals with HCC ( Figure 7K). In contrast, sorafenib was not effective in inhibiting MET/CAT-induced HCC growth (Supplementary Figure 17). Taken together, these data indicate that ABL1 inhibitors effectively suppress HCC growth in preclinical models.

Discussion
Tyrosine kinases have been shown to be good targets for treating cancer, including HCC. Currently, 4 of every 5 drugs approved to treat advanced HCC are tyrosine kinase inhibitors. 2 Despite the success of tyrosine kinase inhibitors in treating HCC, many patients do not respond. There is significant genetic and biological heterogeneity in HCC. A better understanding of tyrosine kinases in the development HCC is necessary in order to predict response to therapy and to develop new therapies. Only a few tyrosine kinases have been shown to play key roles in HCC initiation and progression. 32 In this study, we found that ABL1 is overexpressed and activated in human HCCs, and ABL1 inhibition effectively suppresses HCC growth in human and mouse preclinical models, suggesting that inhibition of ABL1 may be a promising new strategy to treat HCC. We showed that ABL1 inhibitors such as nilotinib significantly inhibit HCC growth in vitro and in vivo. Our findings provide translational support for the development of a clinical trial to assess the safety and efficacy of nilotinib in treating HCC.
How ABL1 is activated in HCC remains unclear. MET can bind directly to ABL1 and activate it in breast cancer cells. 16 We found that ABL1 is activated in MET/CAT-induced HCC tumors, suggesting that ABL1 might be activated by MET in human HCC. In addition, we found that ABL1 expression was also increased in MET/CAT-induced liver tumors. Consistently, ABL1 mRNA expression is highly correlated with the gene signature of activated MET in human HCC (Supplementary Figure 18). 17 Therefore, it is possible that ABL1 mRNA expression can be up-regulated by activated MET in HCC. SP1 can modulate ABL1 expression at the transcription level, 33 and MET can induce the phosphorylation of SP1 and enhance its transcriptional activity. 34 Thus, MET activation may increase ABL1 mRNA expression through the regulation of SP1. This hypothesis will be tested in our future studies.
For the first time, we demonstrated that the NOTCH signaling pathway is suppressed by ABL1 knockdown in HCC cells. NOTCH signaling is a crucial determinant of tumor cell growth in many types of cancer, including HCC. 19 However, NOTCH inhibitors such as gamma secretase inhibitors (GSIs) cause gastrointestinal toxicity due to its off-target effects. 35 Our data suggest a possible new strategy to inhibit the NOTCH pathway, which would be to inhibit ABL1. As functional inhibition of ABL1 showed low and tolerable toxicity, ABL1 inhibitors might be useful to treat HCC cases that exhibit up-regulated NOTCH signaling. It remains to be determined whether the regulation of the NOTCH signaling pathway by ABL1 is tissue-or cell-context-dependent. Our findings might provide a broader application for using ABL1 inhibitors in other type of cancers with activated NOTCH. For example, NOTCH signaling is known to play a critical role in intrahepatic cholangiocarcinoma. 36  cholangiocarcinoma. It would be intriguing to test whether ABL1 inhibitors might be useful in inhibiting intrahepatic cholangiocarcinoma growth in future studies. ABL kinases function by regulating more than 100 different targets in a cell-context-specific manner, making it difficult to select appropriate downstream targets of ABL1 for study. 4 ABL1 has been reported to phosphorylate AKT and ERK to regulate cancer cell proliferation and survival. 37 However, ABL1 knockdown did not affect phosphorylation of either AKT or ERK in HCC cells (Supplementary Figure 19), which supports the concept that ABL1's regulation of different targets is cell-context-dependent. We found that ABL1 knockdown decreases c-MYC protein expression and further demonstrated that decreased c-MYC results in reduced NOTCH1 expression in HCC cells. These data reveal a novel mechanism by which ABL1 promotes cell growth in HCC. c-MYC is a well-known proto-oncogene that plays a critical role in many cancers, including liver cancer. 38 However, it is still currently "untargetable" clinically. Our results suggest that inhibition of ABL1 might be useful to treat HCC cases demonstrating high expression of c-MYC. Although c-MYC is a direct transcriptional target downstream of NOTCH1 in T-ALL, 27 our data suggest that c-MYC may not be regulated by NOTCH1 in HCC cells. This is not particularly surprising, as NOTCH1 regulates different targets in tissue-and cell-context-dependent manners. 39 It remains unknown whether c-MYC regulates NOTCH1 in other contexts, as this mechanism has not been reported previously. It will be instructive to examine whether this regulation occurs in other cancers. Besides the c-MYC/ NOTCH1 axis, we also identified other signaling pathways regulated by ABL1 in HCC cells through RNA sequencing analysis. The gene sets that are most significantly downregulated by ABL1 knockdown include hypoxia, the p53 pathway, glycolysis, and androgen response (Supplementary Figure 20), which have been shown to regulate tumor growth and progression. [40][41][42][43] It is possible that ABL1 promotes HCC development through regulation of these signaling pathways, and we plan to investigate this in our future studies.
In conclusion, our study shows that ABL1 plays a critical role in the development of HCC by regulating the c-MYC/NOTCH1 axis. Inhibition of ABL1 represents a promising new strategy to treat HCC in patients who overexpress this protein-tyrosine kinase. A number of ABL kinase inhibitors have been developed and are being used clinically for leukemia and gastrointestinal stromal tumors.
These inhibitors might also prove to be useful in treating HCC, especially in patients showing overexpression and/or activation of ABL1.

The Cancer Genome Atlas Data Analysis
The data for all boxplots and correlation plots were retrieved from the UCSC Xena platform (https://xenabrowser. net/), specifically from the TCGA Liver Cancer (LIHC) study. For the first variable, the data type selected was "Genomic" and the assay type selected was "Gene Expression." For the second variable, the data type selected was "Phenotypic." To create the boxplots in R software, version 3.6.0, gene expression values were first separated into the categories "normal tissue," consisting of solid tissue normal samples, and "tumor tissue," consisting of primary tumor and recurrent tumor samples. These sorted data were then plotted using the function boxplot. P values were generated using Welch's t test on 2-sample, unpaired data. Normality required for all t tests was assumed by central limit theorem and visual inspection of boxplots. The correlation plots were created using GraphPad Prism software, version 8.0.1. The Pearson's coefficient of correlation, R, and its respective P value (2-tailed) were also calculated using GraphPad Prism software, version 8.0.1.
The data for Kaplan-Meier plots were retrieved from the pathology atlas section of the Human Protein Atlas (https:// www.proteinatlas.org/humanproteome/pathology).
Liver cancer TCGA RNA sample, description, and FPKM data were used for each individual gene. To create the Kaplan-Meier plots, Python software, version 3.0 was used to modify the original data from the Pathology Atlas. Specifically, censored data were generated by dividing patients into 2 groups in which the censored group consisted of patients who had not died by the end of the study time. Patient data were also separated into 2 different groups based on the expression level indicated by FKPM values, for which the dividing threshold was provided by the Pathology Atlas. Binary censored data, binary expression group data, and survival times were imported into R software, version 3.6.0, where Kaplan-Meier plots were created using the survival, survminer, dplyr packages and the Surv, survfit, and ggsurvplot functions. P values were generated by including the command "pval ¼ TRUE" in the ggsurvplot function from the survminer library. Based on the FPKM value of each gene, patients were classified into 2 groups and association between prognosis (survival) and gene expression (FPKM) was examined. The best expression cutoff refers the FPKM value Supplementary References that yields maximal difference with regard to survival between the 2 groups at the lowest log-rank P value. Best expression cutoff was selected based on survival analysis. If FPKM values were greater than or equal to the best expression cutoff value, they were defined as high. If they were lower than this they were defined as low. We also used the median (50%) group cutoff method and found similar results.

MET/CAT-Induced Hepatocellular Carcinoma Model
For the MET/CAT-driven HCC model, 50 mg of total plasmids, encoding the Sleeping Beauty transposase (HSB2) and transposons with GFP (pT3-GFP) or MET gene and catenin b1 gene with the N-terminal truncation (referred to here as MET/CAT) (22.5 mg pT3-EF1a-MET, 22.5 mg pT3-EF1a-DN90-b-catenin, and 5 mg HSB2) were injected hydrodynamically into age-and sex-matched mice as described previously. 1,2 All mice were maintained on the standard diet until being euthanized.

Proximity Ligation Assay
Proximity ligation assay was performed using the Duolink In Situ Red Starter Kit Mouse/Rabbit (MilliporeSigma) according to the manufacturer's instructions. In brief, cells were seeded onto an 8-well-Nunc Lab-Tek II CC2 Chamber Slide System (Thermo Fisher) at 17.5 Â 10 3 /well overnight, then fixed with 4% paraformaldehyde for 30 minutes at room temperature and washed in phosphate-buffered saline, followed by permeabilization with 0.1% Triton X-100 for 10 minutes. After washing with Wash Buffer A (Milli-poreSigma) followed by blocking with Duolink Blocking Buffer (MilliporeSigma) for 30 minutes at room temperature, cells were incubated with primary antibodies (ABL1, # 2862, 1:100; Cell Signaling and c-MYC, #5605, 1:100; Cell Signaling) overnight at 4 C. The following day, cells were washed repeatedly in Wash Buffer A, followed by incubation with appropriate Duolink secondary antibodies (Milli-poreSigma) for 1 hour at 37 C. According to the manufacturer's protocol. After washing with Wash Buffer A at room temperature, ligation, and amplification steps of the proximity ligation assay were performed according to the manufacturer's protocol. After final washes with Wash Buffer B at room temperature, slides were mounted with Corning 24 Â 50 mm Rectangular #1 Cover Glass (Corning, Corning, NY) using Duolink In Situ Mounting Medium with 4 0 ,6diamidino-2-phenylindole (MilliporeSigma).