TNF‐α/NF‐κB signaling epigenetically represses PSD4 transcription to promote alcohol‐related hepatocellular carcinoma progression

Abstract Background Chronic alcohol consumption is more frequently associated with advanced, aggressive hepatocellular carcinoma (HCC) tumors. Alcohol adversely impacts ER/Golgi membrane trafficking and Golgi protein N‐glycosylation in hepatocytes; these effects have been attributed (in part) to dysregulated adenosine diphosphate‐ribosylation factor (ARF) GTPase signaling. Here, we investigated the role of the ARF GTPase guanine exchange factor PSD4 in HCC progression. Methods R‐based bioinformatics analysis was performed on publicly available array data. Modulating gene expression was accomplished via lentiviral vectors. Gene expression was analyzed using quantitative real‐time PCR and immunoblotting. PSD4 promoter methylation was assessed using quantitative methylation‐specific PCR. Phospho‐p65(S276)/DNMT1 binding to the PSD4 promoter was analyzed via chromatin immunoprecipitation. We constructed ethanol/DEN‐induced and DEN only‐induced transgenic murine models of HCC. Results We identified PSD4 as a hypermethylated, suppressed gene in alcohol‐related HCC tumors; however, PSD4 was not dysregulated in all‐cause HCC tumors. Certain HCC cell lines also displayed varying degrees of PSD4 downregulation. PSD4 overexpression or knockdown decreased and increased cell migration and invasiveness, respectively. Mechanistically, PSD4 transcription was repressed by TNF‐α‐induced phospho‐p65(S276)’s recruitment of DNA methyltransferase 1 (DNMT1), resulting in PSD4 promoter methylation. PSD4 inhibited pro‐EMT CDC42 activity, resulting in downregulation of E‐cadherin and upregulation of N‐cadherin and vimentin. Hepatocyte‐specific PSD4 overexpression reduced ethanol/DEN‐induced HCC tumor progression and EMT marker expression in vivo. Conclusions PSD4 is a hypermethylated, suppressed gene in alcohol‐related HCC tumors that negatively modulated pro‐EMT CDC42 activity. Furthermore, we present a novel phospho‐NF‐κB p65(S276)/DNMT1‐mediated promoter methylation mechanism by which TNF‐α/NF‐κB signaling represses PSD4 transcription in HCC cells.


| INTRODUCTION
The most common type of liver cancer is hepatocellular carcinoma (HCC), which accounts for between 80 and 90% of cases. 1,2 In 2012, there were 782,000 newly diagnosed HCC cases and 746,000 related deaths globally. 2 Chronic alcohol consumption is recognized as a key risk factor in HCC development. 3 Moreover, chronic alcohol consumption is more frequently associated with advanced and aggressive HCC tumors. 4 Although alcohol is recognized as a key risk factor in HCC, its mechanism(s) of action in promoting HCC progression are currently unknown.
Alcohol alters methyl group transfer catalysis during methylation reactions, suggesting that abnormal DNA methylation may be associated with alcohol-mediated carcinogenesis. 5 Indeed, alcohol-induced alterations in DNA methylation have been linked to alcohol-related HCC. 5 DNA methylation occurs due to DNA methyltransferases (DNMT) that conjoin a methyl group to the 5' carbon position of the cytosine ring. 6 Tumor suppressor genes are often silenced by DNMT-based promoter hypermethylation; consequently, DNMT upregulation has been linked to oncogenesis. 6 DNMT1, DNMT3A, and DNMT3B are upregulated in both alcoholic liver disease (ALD) tissue and HCC tumors. 6,7 Furthermore, elevated DNMTs are associated with a poor prognosis and often used as a predictor of survival. 6 However, the role of DNA methylation in regulating gene expression in alcohol-related HCC remains largely at the profiling stage. 5,8 Alcohol adversely impacts ER/Golgi membrane trafficking and Golgi protein N-glycosylation in hepatocytes; these effects have been attributed (in part) to dysregulated adenosine diphosphate-ribosylation factor (ARF) GTPase signaling. 9,10 ARF GTPases control key cellular processes, most notably membrane trafficking, proliferation/cell division, motility, and gene transcription. 9,10 ARF GTPase activity is negatively regulated by their bound GDP status (preventing GTP binding) by guanine nucleotide exchange factors (GEFs) and positively regulated by GTPase-activating proteins (GAPs). 9,10 The Cancer Genome Atlas next-generation sequencing (NGS) data reveal that ARF GTPase signaling is significantly altered in several types of cancer. 9,10 However, the role of ARF GTPases (and their associated GEFs and GAPs) in alcohol-related HCC remains largely unexplored.
Here, using R-based bioinformatics analysis, we identify the ARF GTPase GEF PSD4 (EFA6B) as a hypermethylated, suppressed gene in alcohol-related HCC tumors. We also demonstrate that PSD4 functions as tumor suppressor in HCC cells via negatively modulating pro-EMT CDC42 activity. Furthermore, we present a novel phospho-NF-κB p65(S276)/DNMT1-mediated promoter methylation mechanism by which TNF-α/NF-κB signaling represses PSD4 transcription in HCC cells. These findings reveal that PSD4 may be a promising therapeutic target for alcohol-related HCC.

| METHODS
The Supporting Methods details the bioinformatics analysis, lentiviral constructs, qPCR, Western blotting, immunoprecipitation, immunofluorescence, cell assays, quantitative methylation-specific PCR (qMSP), chromatin immunoprecipitation (ChIP) and Re-ChIP assays, and the transgenic alcoholic diethylnitrosamine (DEN) murine model of HCC. Data are represented as means ± standard deviations (SDs) unless otherwise specified. SPSS was used to perform all statistical analysis. For comparison of two groups, a Student's t test was used. For comparison of multiple groups, a one-way ANOVA with Bonferroni's post-hoc testing was used. For comparison of qualitative variables, a Pearson χ 2 or Fisher exact test was used. p < 0.05 was considered significant for all analyses.
Confirming our analysis, we found PSD4 mRNA downregulation in alcohol-related HCC tumors within the GSE59261 K E Y W O R D S alcohol, EFA6B, HCC, p65, PSD4 cohort ( Figure 1I). Interestingly, PSD4 mRNA expression was not dysregulated in all-cause HCC tumors from the TCGA LIUD cohort ( Figure 1J). However, below-median PSD4 expression was associated with worse overall survival in the TCGA LIUD cohort ( Figure 1K). This evidence indicates that PSD4 is specifically downregulated in alcohol-related HCC and may play a role in improving HCC patient survival.

| PSD4 inhibits HCC cell proliferation, migration, and invasiveness via CDC42
To examine the role of PSD4 in HCC tumorigenesis, we assessed the impact of PSD4 overexpression or silencing on HCC cell proliferation, migration, and invasiveness in vitro. We selected HepG2 and PLC/PRF/5 cell lines, which displayed the highest and lowest PSD4 expression, respectively, for further experiments. Based on their endogenous PSD4 expression levels, PLC/PRF/5 were transfected with a lentiviral vector (LvPSD4) to increase PSD4 expression, while HepG2 cells were transfected with one of two PSD4 shRNAs (shPSD4.1 or shPSD4.2) to decrease PSD4 expression ( Figure S2A,B). LvPSD4 PLC/PRF/5 cells had decreased cell proliferation, migration, and invasiveness. In contrast, shPSD4 HepG2 cells displayed increased cell proliferation, migration, and invasiveness (Figure 2A-C). Collectively, these data demonstrate that PSD4 inhibits HCC cell proliferation, migration, and invasiveness.
PSD4 is an ARF guanine nucleotide exchange factor (GEF) of ARF6; functionally, the EFA6B-ARF6 dyad is necessary for tight junction maintenance. 11,12 Based on a literature search for downstream targets of PSD4, we discovered Fayad et al.'s recent work describing the ARF6/ CDC42/PAK1 axis as a regulatory target of PSD4 in breast cancer cells. 13,14 Consequently, we hypothesized that modulating PSD4 expression may affect ARF6/CDC42/PAK1 axis activity in HCC cells. Notably, PSD4 overexpression in PLC/PRF/5 cells did not impact ARF6 expression but reduced binding of CDC42 to PAK1's CDC42 binding domain (PAK1[CRIB]) and PAK1 phosphorylation ( Figure 2D). Conversely, PSD4 knockdown in HepG2 cells slightly decreased ARF6 protein levels but significantly enhanced CDC42/PAK1[CRIB] binding and PAK1 phosphorylation ( Figure 2D). These findings confirm that PSD4 negatively regulates CDC42/PAK1 axis activity in HCC cells.
To determine whether CDC42 mediates PSD4's inhibitory effects on HCC cell proliferation, migration, and invasiveness, PSD4 and CDC42 lentiviral vectors were transfected into PLC/PRF/5 and HepG2 cells.

| PSD4 expression suppressed by TNF-αinduced phospho-p65(S276) in HCC cells
Having uncovered CDC42 as a key downstream mediator of PSD4, we next investigated potential upstream regulators of PSD4 in HCC cells. Our CemiTool R analysis revealed that PSD4's gene module M2 contains the NF-κB p100 (NFKB2) as a key hub gene ( Figure 1B), a key regulator of canonical and non-canonical NF-κB signaling. 15,16 Additionally, considering that PSD4 downregulation is unique to alcoholrelated HCC ( Figure 1H,J) and ALD is associated with more pronounced canonical TNF-α/NF-κB signaling, 17,18 we hypothesized that enhanced TNF-α/NF-κB signaling may regulate PSD4 expression in HCC cells. Accordingly, we observed that TNF-α reduced both mRNA and protein levels of PSD4 in a time-dependent manner in HepG2 cells ( Figure 3A,B).
The ENCODE ChIP-seq database has identified the PSD4 promoter as a target of p65 (RELA), 23 which we confirmed through our ChIP assays. Briefly, p-p65(S276) was enriched at approx. 2.5 kbp upstream of the PSD4 TSS; this signal was enhanced by overexpression of WT p65 (but not the S276A mutant) in HepG2 cells and reduced by p65 knockdown in PLC/PRF/5 cells under TNF-α conditions ( Figure 4C,D). Furthermore, ChIP/Re-ChIP assays highlighted co-occupancy of p-p65(S276) and DNMT1 at this PSD4 promoter region under TNF-α conditions ( Figure 4E). We also observed an increase in DNMT1 binding by overexpression of WT p65 (but not the S276A mutant) in HepG2 cells, and a decrease in DNMT1 binding by p65 knockdown in PLC/PRF/5 cells under TNF-α conditions ( Figure 4F,G). This combined evidence suggests that TNF-α-induced p-p65(S276) recruits DNMT1 to the PSD4 promoter, which enhances DNA methylation and PSD4 transcriptional repression.
We next explored whether DNMT1 inhibition by 5-AZA could restore PSD4 expression in WT p65-overexpressing HepG2 cells and, consequently, the metastatic phenotype. 5-AZA decreased DNMT1 protein levels but increased PSD4 gene and protein levels in WT p65-overexpressing HepG2 cells under TNF-α conditions ( Figure 4H,I). Our data demonstrate that PSD4 downregulation in WT p65-overexpressing HCC cells is dependent upon DNMT1 activity.

| Construction and characterization of hepatocyte-specific PSD4 overexpression transgenic mice
To investigate the effects of PSD4 in vivo, we constructed hepatocyte-specific Psd4-overexpressing transgenic mice by placing the murine Psd4 cDNA under the control of the hepatocyte-specific murine albumin (Alb) promoter. This Alb-PSD4 transgene construct ( Figure S5A) was injected into zygotic pronuclei isolated from E0.5-day pregnant female mice. After brief culturing, these transgenic zygotes were transferred into pseudo-pregnant females. Following birth, tail clippings from transgenic Alb-PSD4 (TG Alb-PSD4 ) and non-transgenic (non-TG) neonate littermates were screened for the Alb-PSD4 transgene by qPCR ( Figure S5B). To validate hepatocyte-specific expression of M2-FLAG-PSD4 in TG Alb-PSD4 mice, hepatocytes and non-heptocyte Kupffer cells were isolated from liver tissue samples. Immunoblotting of each cell fraction with an anti-M2-FLAG antibody confirmed that the Alb-PSD4 transgene construct was solely expressed in TG Alb-PSD4 hepatocytes ( Figure S5C).
Following birth, mice were maintained on a regular chow diet to measure growth rates, liver/body weight ratios, and heart/body weight ratios. The differences in growth rates were not statistically significant, and TG Alb-PSD4 mice and non-TG mice achieved statistically similar adult weights for males and females ( Figure S5D). There was also no differences in the liver/body weight ratios or heart/body weight ratios of TG Alb-PSD4 mice in comparison to non-TG littermates at 4 months ( Figure S5E,F).

| In vivo alcohol-induced HCC tumorigenesis and EMT markers suppressed by hepatocyte-specific PSD4 overexpression
To investigate whether hepatocyte-specific PSD4 overexpression has an impact on alcohol-induced HCC tumor progression, we utilized a previously reported murine model of ethanol/DEN-induced HCC ( Figure 5A). By gross observation, ethanol/DEN non-TG mice displayed larger liver tumor masses when compared to DEN non-TG mice ( Figure 5B), an effect reduced in TG Alb-PSD4 mice. Furthermore, ethanol/ DEN non-TG mice displayed dramatic increases in liver/ body weight ratios, tumor volumes, maximum tumor diameters, and PCNA+cell counts relative to DEN non-TG mice ( Figure 5C-F), effects reduced in TG Alb-PSD4 mice. Similar to human ALD, 24 the NF-κB-associated inflammatory markers Tnf, IL-1β, and Il-6 were elevated in the livers of ethanol/DEN mice relative to DEN mice, a factor unaffected by TG Alb-PSD4 status ( Figure 5G). We confirmed Tnf-α/p-p65(S276) axis upregulation in the livers of ethanol/DEN mice relative to DEN mice, a factor unaffected by TG Alb-PSD4 status ( Figure 5H). mRNA and protein levels of EMT genes E-cadherin, N-cadherin, and vimentin were dysregulated in the livers of ethanol/DEN non-TG mice relative to DEN non-TG mice ( Figure 5I,J), effects rescued in TG Alb-PSD4 mice. These data demonstrate that hepatocyte-specific PSD4 overexpression reduces alcohol-induced HCC tumorigenesis and EMT marker expression in mice.

| DISCUSSION
Here, we demonstrate that PSD4 is downregulated in alcoholrelated HCC tumors and certain HCC cell lines. Moreover, below-median PSD4 expression in HCC tumors is associated with inferior survival outcomes. Functionally, PSD4 inhibited HCC cells proliferation, migration, and invasiveness in vitro and suppressed alcohol-induced HCC tumor progression in vivo. At the molecular level, PSD4 transcription is epigenetic silenced via enhanced PSD4 promoter methylation by TNF-α-induced p-p65(S276)/DNMT1 complex binding.
TNF-α promotes the migratory and invasive capabilities of HCC cells. 25 Furthermore, TNF-α induces EMT in HCC cells by downregulating the epithelial marker E-cadherin (CDH1) but upregulating the mesenchymal markers N-cadherin and vimentin. 25 As observed here, TNF-α-induced p-p65(S276) has previously been implicated in methylation-associated transcriptional repression of tumor suppressor genes. For example, enhanced TNF-α signaling induces methylation-associated transcriptional repression of the tumor suppressor BRMS1 via promoting p-p65(S276)/DNMT1 complex formation in NSCLC cells. Moreover, p-p65(S276) recruits DNMT1 to methylate the CRMP4 promoter and transcriptional repress CRMP4 in prostate cancer cells. 26 Therefore, p-p65(S276)/DNMT1-mediated transcriptional repression appears to be a key theme across various cancer types and deserves further investigation.
Other than the ARF GTPase ARF6, little is known about the downstream targets of PSD4. 11 Fayad et al.'s recent work describes the ARF6/CDC42/PAK1 axis as a regulatory target of PSD4 in breast cancer cells. 13,14 Notably, we found that PSD4 did not impact ARF6 expression but did inhibit CDC42-PAK1[CRIB] binding and PAK1 phosphorylation, indicating suppression of CDC42/PAK1 signaling. CDC42 is a Rho GTPase that plays a key oncogenic role via regulating cytoskeletal dynamics, cell proliferation, EMT, cell migration, and invasiveness. 27 CDC42's effector PAK1 promotes Akt/β-catenin activity in HCC cells, thereby promoting HCC cell proliferation. 28 Additionally, PAK1 upregulates the EMT inducer Snail in HCC cells, promoting HCC cell migration and invasiveness. 28 Accordingly, recent evidence now supports that CDC42/PAK1 activity downregulates E-cadherin expression and upregulates N-cadherin and vimentin expression in HCC cells. 29 This cadherin switching-E-cadherin downregulation coupled with N-cadherin upregulation-is a key EMT indicator. 30 Moreover, vimentin upregulation is also associated with EMT. 31 Here, we found that PSD4 reduced cadherin switching and vimentin expression in HCC cells in a CDC42-dependent manner, indicating PSD4's suppression of EMT via CDC42.
A number of in vivo HCC mouse models are based on intraperitoneal injection of diethylnitrosamine (DEN), a carcinogenic DNA alkylator. 32 Here, consistent with previous work by Brandon-Warner et al., Ambade et al., and Yan et al., [33][34][35] the addition of a chronically administered ethanol liquid diet promotes liver inflammatory markers and HCC tumor progression in DEN-exposed male mice. This murine model displays face validity, as alcoholic steatohepatitis potentiates liver inflammation and HCC risk in humans. 36 Consistent with our in vitro findings, PSD4 suppressed ethanol/DEN-induced HCC progression along with reducing cadherin switching and vimentin expression in HCC tumor cells in vivo. Given that PSD4 downregulation is specific to alcohol-related HCC, our evidence suggests that targeting PSD4 may be especially effective in HCC patients with a history of chronic alcohol use.
In conclusion, this study identifies PSD4 as a hypermethylated, suppressed gene in alcohol-related HCC tumors. We also demonstrate that PSD4 functions as tumor suppressor in HCC cells via negatively modulating pro-EMT CDC42 activity. Furthermore, we present a novel p-p65(S276)/DNMT1mediated promoter methylation mechanism by which TNF-α/ NF-κB signaling represses PSD4 transcription in HCC cells. These findings reveal that PSD4 may be promising therapeutic target for alcohol-related HCC.