CAND1 is upregulated and correlate to poor prognosis of HCC
HCC data from The Cancer Genome Atlas (TCGA) were used to analyze CAND1 expression. CAND1 expression in HCC tumor tissues was higher than in normal tissues (Fig. 1A). Microarray data (GSE76427 and GSE14520 datasets) from the Gene Expression Omnibus (GEO) database showed similar results (Fig. 1B, P<0.001). TCGA data also showed that high CAND1 expression was significantly associated with an advanced HCC clinical stage and a poor prognosis (Fig. 1C-D). Quantitative RT-PCR confirmed the overexpression of CAND1 mRNA in 74 HCC tumor tissues (T) from Zhongnan Hospital of Wuhan university compared to the paired adjacent non-tumor tissues (NT) (Fig. 1E, P<0.001). Immunohistochemical and Western blot verified the increased CAND1 protein levels in 9 out of 10 HCC tumor tissues compared to NT (Fig. 1F-G). Furthermore, this 74-patient cohort showed that CAND1 expression was correlated with tumor diameter (P=0.017), TNM stage (P=0.041) and Vascular invasion (P=0.043, Supplementary Table. 1). By univariate Cox regression analysis, high CAND1 expression was associated with poor prognosis (Fig. 1H). Multivariate Cox regression analysis confirmed that CAND1 and TNM stage were the independent prognostic factors (Fig. 1I). Overall, these observations demonstrate that CAND1 overexpression is tightly associated with HCC and may serve as an unfavorable prognosis factor.
CAND1 promotes cell proliferation, migration, and invasion in vitro
The expression of CAND1 protein was relatively high in the Hep3B and Huh7 cell lines and relatively low in the HCCLM3 and HepG2 cell lines (Supplementary Fig. 1). We knocked down CAND1 mRNA level (Fig. 2A) and protein level (Fig. 2B) in Hep3B and Huh7 cells by lentivirus-mediated shRNA technique. shCAND1#2 and shCAND1#3 showed potent knockdown efficiency and were used for subsequent experiments. CAND1 knockdown reduced the cell proliferation potential and the colony formation of Hep3B and Huh7 cells (Fig. 2C-D). Transwell assays showed that CAND1 knockdown inhibited cell invasion (Fig. 2E). Wound healing assays showed that CAND1 knockdown significantly reduced cell migration (Fig. 2F). Collectively, our findings suggest that knockdown of CAND1 inhibits cell proliferation, migration, and colony formation in HCC cells. We also overexpressed CAND1 in HCCLM3 and HepG2 cells, which promoted cell proliferation (Fig. 2G-H). Consistently, CAND1 overexpression promoted cell migration and invasion (Fig. 2I-J). These results suggest that CAND1 promotes HCC cell proliferation, colony formation, migration, and invasion.
CAND1 regulates HCC lipid metabolism in vitro
To elucidate the underlying mechanisms, we performed RNA-Seq analysis in Control and CAND1 knockdown Hep3B and Huh7 cells. Volcano plot revealed 2513 differentially expressed genes (DEGs), of which 1061 genes were upregulated and 1452 genes were downregulated in CAND1 knockdown cells (Fig. 3A). Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis of DEGs showed that metabolic pathways with the largest number of DEGs (233 genes) were among the top 10 pathways (P <0.001) (Fig. 3B). Particularly, many of DEGs enriched in metabolic pathways were involved in lipid metabolism. Major genes responsible for the synthesis of fatty acid, such as fatty acid synthase (FASN), acetyl-CoA carboxylase 1(ACC1) and ATP citrate lyase (ACLY) were downregulated in CAND1 knockdown group whereas fatty acid oxidation-associated genes (PPARα, PPARδ and PPARγ) in CAND1-knockdown cells were not affected (Fig. 3C). Consistent to gene expression pattern, significant decreases of lipid accumulation and intracellular triglyceride and cholesterol were observed in CAND1-knockdown cells (Fig. 3D-E) whereas CAND1 overexpression showed opposite effects (Fig. 3F-G). Furthermore, a quantitative lipidomic analysis showed the total content of all classes of lipids in CAND1-knockdown cells was clearly lower than that in the Control cells (Fig. 3H). Notably, there are significant differences in the levels of triglycerides (TGs), phosphatidylcholine (PC), cholesterol (CE) and other lipid components (Fig. 3I, Supplementary Fig. 2). These results are consistent to previous studies on CAND1 function in adipogenesis (17, 23) and suggest that CAND1 may promote lipid biosynthesis in HCC.
CAND1 promotes tumor growth, metastasis and lipid accumulation in vivo
To assess the in vivo function of CAND1, subcutaneous tumor xenograft mouse models were established. The weight of the tumors from the CAND1-knockdown group was lower than that from the control group (Fig. 4A). The growth rate of subcutaneous tumors was slower with shCAND1 cells than control group (Fig. 4B). Immunohistochemical analysis of samples showed that CAND1, Ki-67, FASN, ACC1 and ACLY expression were significantly reduced in the shCAND1 group (Fig. 4C). In addition, CAND1 knockdown result in decreases in lipid accumulation (Fig. 4D). Triglyceride and cholesterol, which forms the main component of the lipid droplets, were significantly reduced in the shCAND1 group (Fig. 4E). We also performed splenic injection of Huh7 cells and examined liver metastasis four weeks after injection. The shCAND1 group showed significantly decreased tumor nodules in the liver compared to Control group (Fig. 4F). In tail vein metastasis model mice, the number of lung metastatic nodules was significantly lower in the mice injected with the CAND1-knockdown cells than in those injected with the control cells (Fig. 4G-H). These results demonstrated that CAND1 promote HCC growth and metastasis in vivo.
CAND1 impairs SCFFBXO11 assembly to recruit hnRNPA2B1
CAND1 has not been implied to involve in transcriptional regulation. We speculated that CAND1 might somehow regulate certain transcription factors to increase FASN, ACC1, ACLY and other fatty acid synthesis-related genes. Thus, we performed IP of CAND1 followed by MS (IP/MS) and identified CUL1 as a potential interacting protein (Supplementary Table. 2). Co-immunoprecipitation confirmed CAND1 interaction with CUL1 (Fig. 5A-B), which was known to regulate the assembly and dissociation of SCFs (9, 16). In fact, CAND1 knockdown significantly increased F-box proteins binding to CUL1 measured by IP with an anti-CUL1 antibody followed by mass spectrometry and Western blotting, of which FBXO11 was the most significantly changed F-box protein (Fig. 5C-D) (Supplementary Table. 3). As part of the Skp1-Cul1-FBXO11 (SCFFBXO11) complex, FBXO11 specifically recognizes ubiquitylated substrates. Therefore, we performed IP of FBXO11 followed by MS and identified hnRNPA2B1 in the FBXO11 co-precipitation (Supplementary Table. 4). Previous studies reported hnRNPA2B1 as RNA regulators of ACLY and ACC1 (24), which is consistent with our RNA-Seq results and lipid accumulation phenotypes. We confirmed interaction of FBXO11 and hnRNPA2B1 by IP followed by Western blot (Fig. 5E). Reverse IP and western blotting showed that hnRNPA2B1 was only recognized by FBXO11, but not other F-box proteins (Fig. 5F). Overexpression of HA-FBXO11 and Flag-hnRNPA2B1 followed by co-IP assay confirmed the interaction (Fig. 5G). Furthermore, we verified the direct interaction by GST-pulldown assay with purified GST-hnRNPA2B1 and HA-FBXO11 fusion proteins expressed in E. coli (Fig. 5H). In addition, immunofluorescence staining showed the subcellular nuclear colocalization of FBXO11 (red) and hnRNPA2B1 (green) (Fig. 5I). These results demonstrate that CAND1 impairs SCFFBXO11 assembly to recruit hnRNPA2B1.
To examine the effect of SCFFBXO11 on hnRNPA2B1, shFBXO11#1 and shFBXO11#2 were used for efficient knockdown of mRNA of FBXO11 (Supplementary Fig. 3) and protein level of FBXO11 (Supplementary Fig. 4). hnRNPA2B1 expression was increased in FBXO11-knockdown cells compared to that in the Control group. Consistent with the expression of hnRNPA2B1, the expression of the lipid metabolism genes FASN, ACC and ACLY was increased when FBXO11 expression was knocked down (Fig. 5J). When FBXO11 was overexpressed, the expression of hnRNPA2B1, FASN, ACC and ACLY was decreased (Fig. 5K). In addition, the overexpression of FBXO11 reduced the lipid accumulation in tumor cells (Supplementary Fig. 5). Moreover, hnRNPA2B1 and FASN, ACC, as well as ACLY, expression was upregulated and decreased when CAND1 was overexpressed and knocked down, respectively (Fig. 5L-M). In summary, our results indicate that FBXO11 mediates the effect of CAND1 knockdown on hnRNPA2B1 expression.
FBXO11 mediates ubiquitination and degradation of hnRNAPA2B1
FBXO11 is known to influence the degradation of several proteins(25, 26). Hence, we evaluated the effect of FBXO11 expression on the rate of hnRNPA2B1 degradation. Tthe degradation rate of hnRNPA2B1 was significantly accelerated when FBXO11 was overexpressed (Fig. 6A). In contrast, FBXO11 knockdown decelerated the hnRNPA2B1 degradation rate (Fig. 6B). Moreover, hnRNPA2B1 degradation was significantly faster in CAND1-knockdown cells (Fig. 6C). These results showed that CAND1 and FBXO11 regulated the protein expression of hnRNPA2B1 by affecting its degradation rate. FBXO11 belongs to the ubiquitin‐proteasome system family and may mediate hnRNPA2B1 degradation through the ubiquitin‐proteasome system. As expected, hnRNPA2B1 protein level was rapidly increased upon the addition of the proteasome inhibitor MG132 (Fig. 6D). FBXO11 overexpression significantly decreased the level of hnRNPA2B1 protein, which was normalized by MG132 treatment (Fig. 6E). Consistently, hnRNPA2B1 ubiquitination was significantly increased in CAND1-knockdown cells (Fig. 6F) but significantly decreased after CAND1 overexpression (Fig. 6G). FBOX11 showed opposite effects (Fig. 6H-I). We further confirm the promoting effect of FBXO11 on hnRNPA2B1 ubiquitination in 293T cells transfected with Flag-hnRNPA2B1 and HA-FBXO11 (Fig. 6J and Supplementary Fig. 6). FBXO11 overexpression decreased the endogenous hnRNPA2B1 protein expression level in a dose-dependent manner (Fig. 6K). Similarly, overexpression of FBXO11 increased the hnRNPA2B1 ubiquitination level in a dose‐dependent manner (Fig. 6L), whereas CAND1 overexpression reduced the hnRNPA2B1 ubiquitination level (Fig. 6M). In summary, these results suggest that hnRNPA2B1 is regulated by FBXO11-mediated ubiquitination.
FBXO11 physically interacts with the RRM_2 region of hnRNPA2B1 via its F-box domain and promotes K48- and K27-linked ubiquitination of hnRNPA2B1
FBXO11 mainly contains an F-box domain, which can link target proteins to a ubiquitin ligase, as well as nineteen parallel beta-helix repeat (PbH1) motifs. To investigate whether the F-box mutations interfere with FBXO11 activity, we generated a mutant FBXO11 plasmid containing complementary DNA constructs in the F-box domain (ΔFBXO11) located at amino acids (aa) 153-199. Overexpression of FBXO11 increased the ubiquitination of hnRNPA2B1, but the mutant FBXO11 did not (Fig. 7A). Moreover, mutant FBXO11 failed to interact with hnRNPA2B1 (Fig. 7B-C). hnRNPA2B1 proteins possess two RNA recognition motif (RRM) domains, RRM_1 located in 21-104 aa (amino acids) and RRM_2 (112-191 aa), a disordered region (193-353 aa) and one nuclear localization signal motif (9-15 aa) (www.uniprot.org). Co-IP and Western analysis of full length and serial truncations of hnRNPA2B1 with a Flag-tag demonstrated that a segment of 81 aa in the RRM_2 region of hnRNPA2B1 mediated hnRNPA2B1 interaction with FBXO11 (Fig. 7D-E).
To identify the exact site of hnRNPA2B1 that was ubiquitinated by FBXO11, we searched a database of ubiquitination sites (ubibrowser.bio-it) and found a total of four sites where hnRNPA2B1 may be ubiquitinated by FBXO11 (Supplementary Fig. 7). Mutant hnRNPA2B1 protein carrying alanine residues that substituted for four lysine residues (K120A, K137A, K168A and K173A) was not degraded through FBXO11 ubiquitination (Fig. 7F). In fact, mutation of either site of 4 lysine residues abolished hnRNPA2B1 degradation by FBXO11 (Fig. 7G). Moreover, replacement of the four lysine residues or any one of these four lysine residues caused a marked decrease in hnRNPA2B1 ubiquitylation levels (Fig. 7H) and FBXO11 specifically promoted the addition of K27- and K48-linked ubiquitin to hnRNPA2B1 (Fig. 7I). Then, the K27-mutant (K27R) and K48-mutant (K48R) ubiquitin vectors were used to examine the K27- and K48-linked ubiquitination of hnRNPA2B1 (Fig. 7J). When both the K27 and K48 ubiquitination sites were mutated (K27R and K48R), FBXO11-mediated ubiquitination of hnRNPA2B1 was abolished (Fig. 7K). Therefore, these data suggest that FBXO11 promotes the degradation of hnRNPA2B1 through K27- and K48-linked ubiquitination.
CAND1 promoted HCC and lipid synthesis by impairing the FBXO11-mediated hnRNPA2B1 degradation
Knocking down CAND1 inhibited Huh7 and Hep3B cell proliferation. However, the inhibition of proliferation was reversed by hnRNPA2B1 expression (Fig. 7L-M). Moreover, overexpression of hnRNPA2B1 reversed the effects of CAND1 knockdown on tumor cell migration and invasion (Fig. 7N-O). Western blot analysis confirmed that CAND1 knockdown downregulated hnRNPA2B1 expression, and treatment with shFBXO11 reversed the effect of CAND1 knockdown on hnRNPA2B1 expression (Fig. 7P). Furthermore, the increase in hnRNPA2B1 protein expression and lipid synthesis caused by CAND1 overexpression was reversed by FBXO11 (Fig. 7Q-R). These findings demonstrate that CAND1 functions in HCC by impairing FBXO11-mediated hnRNPA2B1 degradation.
Targeting CAND1 efficiently suppresses HCC
We used PDX mouse model to test whether targeting CAND1 was an effective strategy against HCC. Intratumoral injection of shCAND1 potently downregulated CAND1 expression, inhibited tumor growth, and decreased tumor volume without alteration of mouse body weight (Fig. 8A-C). Injection of shCAND1 also decreased IFP in the center of the tumors (Fig. 8D). Moreover, the survival of nude mice was significantly improved in the group injected with shCAND1 compared to that in the control group (Fig. 8E). PDX tumors with shCAND1 injection showed smaller sizes and lower weights than control group (Fig. 8F-G). Results of immunohistochemistry validated the descent of CAND1, hnRNPA2B1, fatty acid biosynthesis related proteins (FASN, ACC1 and ACLY) and lipid accumulation due to intratumoral injection of shCAND1 (Fig. 8H). We also used myr-AKT/NRASV12-induced mouse liver cancer model to test the strategy, which exhibited abnormal lipid metabolism. Mice with injection of shCAND1 lentivirus displayed decreased tumor numbers and tumor sizes as well as reduced lipid accumulation compared to control mice (Fig. 8I). In addition, hnRNPA2B1, FASN, ACC1 and ACLY were also significantly downregulated in the mice injected with shCAND1 lentivirus (Fig. 8J). These results suggest that targeting CAND1 is a potent strategy against HCC.