Article
Open access
Published: 04 July 2024

Transposable elements-mediated recruitment of KDM1A epigenetically silences HNF4A expression to promote hepatocellular carcinoma

Nature Communications volume 15, Article number: 5631 (2024) Cite this article

932 Accesses
1 Altmetric
Metrics details

Subjects

Abstract

Transposable elements (TEs) contribute to gene expression regulation by acting as cis-regulatory elements that attract transcription factors and epigenetic regulators. This research aims to explore the functional and clinical implications of transposable element-related molecular events in hepatocellular carcinoma, focusing on the mechanism through which liver-specific accessible TEs (liver-TEs) regulate adjacent gene expression. Our findings reveal that the expression of HNF4A is inversely regulated by proximate liver-TEs, which facilitates liver cancer cell proliferation. Mechanistically, liver-TEs are predominantly occupied by the histone demethylase, KDM1A. KDM1A negatively influences the methylation of histone H3 Lys4 (H3K4) of liver-TEs, resulting in the epigenetic silencing of HNF4A expression. The suppression of HNF4A mediated by KDM1A promotes liver cancer cell proliferation. In conclusion, this study uncovers a liver-TE/KDM1A/HNF4A regulatory axis that promotes liver cancer growth and highlights KDM1A as a promising therapeutic target. Our findings provide insight into the transposable element-related molecular mechanisms underlying liver cancer progression.

UHRF1 inhibition epigenetically reprograms cancer stem cells to suppress the tumorigenic phenotype of hepatocellular carcinoma

Article Open access 28 June 2023

LINC01419-mediated epigenetic silencing of ZIC1 promotes metastasis in hepatocellular carcinoma through the PI3K/Akt signaling pathway

Article 26 March 2021

DNA hypermethylation modification promotes the development of hepatocellular carcinoma by depressing the tumor suppressor gene ZNF334

Article Open access 09 May 2022

Introduction

Transposable elements, or transposons, are mobile DNA units present in the majority of eukaryotic genomes. Although once considered “junk DNA” with no functional significance, recent studies have revealed that TEs provide regulatory sequences that wire transcriptional regulatory networks^1,2,3. While host organisms usually silence TEs, the remnants of TE-derived cis-regulatory elements (CRE) can persist and adapt to control the transcription of host genes^4,5,6,7.

TEs interact with cancer in intricate ways, offering novel insights into the mechanisms of cancer development and progression^8,9. TEs can function as promoters, driving oncogene expression and oncogenesis, which contribute to tumor initiation and maintenance¹⁰. They can establish regulatory circuits involving the KRAB zinc-finger protein family to suppress tumors¹¹. Moreover, TEs can interplay with epigenetic events such as DNA methylation and histone modifications to modulate gene expression patterns and promote tumor growth¹². The involvement of TEs in tumoral epigenetic regulation represents a consequential outcome of TE silencing to maintain genome stability. A primary mechanism for suppressing TEs involves the selective deposition of repressive histone modifications¹³. Tumor cells employ this strategy to facilitate immune evasion^14,15. Additional epigenetic regulators, like the human silencing hub (HUSH) complex and lysine demethylase 1A (KDM1A), employ a similar TE-dependent mechanism to impact tumor development^16,17,18.

KDM1A, a well-established oncogene, exhibits high expression in multiple cancers and correlates with unfavorable prognosis in cancer patients^19,20. In liver cancer, KDM1A promotes cancer growth and drug resistance by regulating signals such as FKBP8/Bcl2 or LINC01134/SP1/p62^21,22,23,24. KDM1A typically functions as part of a complex, such as the CoREST complex, where it selectively removes H3K4me1 and H3K4me2 methylation modifications to inhibit transcription of target genes^25,26. Alternatively, in a complex with hormone receptors, KDM1A specifically regulates H3K9me2 and H3K9me3 methylation modifications^27,28. The selectivity of KDM1A towards histone substrates depends on the composition of the complex^20,28,29,30. HNF4A, a member of the nuclear receptor superfamily, serves as a pivotal regulatory factor in the initial stages of liver cancer development^31,32,33. In normal physiological conditions, HNF4A exhibits high expression in liver tissue and plays a critical role in preserving the differentiation and function of liver cells^31,34. However, HNF4A expression is downregulated in liver cancer and shows tumor-suppressive characteristics³⁵.

Here, we show a mechanism in which KDM1A epigenetically suppresses transposable elements in proximity to the HNF4A gene, resulting in the inhibition of HNF4A expression. This work focuses on liver-specific accessible TEs (liver-TEs) that provide CRE and recruit KDM1A to silence nearby genes. Moreover, we discover a signaling axis in which liver-TEs, KDM1A, and the liver-TE-associated gene (HNF4A) synergistically control the proliferation of hepatocellular carcinoma cells. These findings elucidate the functional and clinical implications of transposable element-mediated molecular events, potentially paving the way for novel therapeutic strategies in hepatocellular carcinoma.

Results

Liver-TEs function as inhibitory cis-regulatory elements for the HNF4A gene

To explore the involvement of transposable elements (TEs) as cis-regulatory elements in liver cancer development, we utilized the TCGA ATAC-seq data from 23 different types of cancer tissues to identify specific accessible TEs within transcriptional regulatory regions (TRR: TSS ± 10 kb) in liver cancers. Our analysis unveiled significant heterogeneity in TE accessibility across distinct tumor types. Through ATAC-seq data analysis, we identified a total of 3,762 TEs exhibiting elevated ATAC-seq intensity within liver cancer samples (Fig. 1a, b and Supplementary Data 1). By using ATAC-seq data from non-malignant human tissue samples in the ENCODE database, we found that the identified TEs were highly specific to healthy liver tissues as well (Fig. 1c). Therefore, we designated this subset of TEs as liver-specific accessible TEs (liver-TEs). Concurrently, the transcriptional regulatory regions harboring liver-TEs (designated as liver-TE-TRR) in liver cancer samples exhibited a noticeable increase in ATAC-seq signal intensity (Supplementary Fig. 1a and Supplementary Data 1). To further elucidate the distinctions in the accessibility of liver-TEs and liver-TE-TRRs between healthy and cancerous liver cells, we conducted ATAC-seq assays in both cell types. The results revealed that liver-TEs and liver-TE-TRRs were more accessible in normal liver cells (Supplementary Fig. 1b).

**Fig. 1: Negative regulation of HNF4A expression by liver-TEs.**

Liver-TEs comprised major TE families, including SINEs (Alu, MIR), LINEs (L1, L2), LTRs (ERV1, ERVL, ERVL-MaLR), and DNA (hAT-Charlie, TcMar-Tigger). Upon comparing the proportions of significant TE families within the human genome and liver-TEs, we observed the highest representation of the Alu family and MIRs (Mammalian-wide interspersed repeats) in liver-TEs, with a notably augmented proportion of MIR family constituents (Supplementary Fig. 1c). Genes located within ±10 kb of liver-TEs were defined as liver-TE associated genes (Supplementary Data 2). Analysis of gene expression profiles using the combined GETx-TCGA database revealed that liver-TE-associated genes exhibited liver tissue specificity, both healthy and cancerous, with a slight downregulation in liver cancer compared to normal liver tissues (Supplementary Fig. 1d). Furthermore, prognosis analysis of liver-TE-associated gene expression demonstrated that high expression of these genes indicated a favorable prognosis for liver cancer patients (Fig. 1d and Supplementary Fig. 1e). Gene function enrichment analysis of liver-TE-related genes indicated that they are associated with liver-specific gene expression, and are enriched for HNF4A signature (Fig. 1e). Intriguingly, the P1-driven HNF4A, a predominant isoform within liver tissues³⁶, exhibits two liver-TE clusters flanking the transcription start site (TSS) (Fig. 1f). To explore the regulatory impact of liver-TEs on HNF4A expression, we utilized CRISPR/Cas9 technology to concurrently remove each cluster of TEs, situated either upstream or downstream of the P1-driven HNF4A TSS, referred to as HNF4A-liver-TEs, in liver cancer cells (Supplementary Fig. 1f). Based on ATAC-seq data, the majority of HNF4A-liver-TEs exhibited enhanced accessibility in normal liver cells (Supplementary Fig. 1g). Through the analysis of Capture Hi-C data, we identified that genomic interactions involving HNF4A-liver-TEs predominantly involve cis-interactions, with a notable concentration in the q13.12 region of chromosome 20 (Supplementary Fig. 1h). Deletion of liver-TEs resulted in an upregulation of HNF4A expression (Fig. 1f-g), thus illustrating the inhibitory role of TE elements in controlling HNF4A expression. To gain deeper insights into the regulatory mechanisms, we employed gene editing techniques to individually excise several long HNF4A-liver-TEs. This approach enabled us to precisely elucidate the specific roles of liver-TEs on the HNF4A gene, revealing that depletion of a single TE element plays a role in regulating HNF4A expression (Supplementary Fig. 1i). Furthermore, liver cancer cells deficient in HNF4A-liver-TEs exhibited diminished growth capacity both in vitro (Fig. 1h) and in vivo (Supplementary Fig. 1j). Similarly, the depletion of HNF4A-liver-TEs in HCC patient-derived organoids resulted in growth retardation (Supplementary Fig. 1k). These observations underscore the significant role of this element in promoting tumorigenesis.

KDM1A is enriched in liver-TEs and is necessary for HCC cell growth

To investigate the epigenetic characteristics of liver-TEs, we analyzed the histone marks enriched in liver-TEs using the RemapEnrich algorithm. The results showed that liver-TEs are enriched in active histone marks such as H3K4me1, H3K4me2, and H3K27ac (Supplementary Fig. 2a, Supplementary Data 3). The enrichment of H3K4 methylation marks in liver-TE regions suggests their presence in relatively accessible chromatin environments. Additionally, our investigation of CpG methylation levels in liver-TEs showed that they were comparable to those in non-liver-TEs (Supplementary Fig. 2b), suggesting that DNA methylation does not significantly influence liver-TE activity.

To elucidate the molecular mechanisms underlying the regulation of gene expression by liver-TE, we performed a RemapEnrich analysis of transcriptional regulators (TR) enrichment on liver-TEs. Our results reveal a substantial enrichment of the histone demethylase KDM1A on liver-TEs (Fig. 2a and Supplementary Data 4). Heatmaps displaying the KDM1A enrichment signals within liver-TEs demonstrate a significant increase in KDM1A signals within these regions compared to similarly sized shuffled TE regions (Fig. 2b). Then, we conducted KDM1A CUT&Tag-seq assays in both normal liver cells (THLE2, THLE3) and HCC cell lines to compare the binding ability of KDM1A within liver-specific TEs. Our results demonstrated a diminished binding capability of KDM1A to these TEs in normal liver cells compared to HCC cells (Supplementary Fig. 2c), suggesting potential aberrant recruitment of KDM1A to liver-specific TEs within liver cancer cells. To test whether KDM1A plays an epigenetic regulatory role within liver-TEs, ATAC-seq and CUT&Tag-seq experiments were conducted. The results revealed that KDM1A knockdown led to an increase in chromatin accessibility (Fig. 2c) and histone mark H3K4me1, H3K4me2, H3K9me2, and H3K27ac within liver-TEs (Fig. 2d). However, the global histone methylation modifications remained largely unaffected by KDM1A knockdown (Supplementary Fig. 2d). These results indicate that KDM1A downregulation enhanced the activation and accessibility of liver-TEs and nearby chromatin. Additionally, we conducted a detailed examination of how KDM1A knockdown influences H3K4me1 and H3K9me2 marks in the transcriptional regulatory regions bound by KDM1A. Our results demonstrate that in regions directly targeted by KDM1A, H3K4me1 exhibited upregulation following KDM1A knockdown, while H3K9me2 showed no significant change (Supplementary Fig. 2e, Top panel). This observation is consistently supported by findings in liver-TE-associated KDM1A-targeted TRRs, where H3K4me1 marks also significantly increased upon KDM1A knockdown (Supplementary Fig. 2e, Down panel). Notably, the reduction of KDM1A in normal liver cells did not lead to an elevation in H3K4me1 modification of liver-TEs, which lack KDM1A binding peaks (Supplementary Fig. 2f). These results strongly suggest that KDM1A may directly regulate H3K4me1, leading us to focus our subsequent analyses on this mark.

**Fig. 2: Inhibition of KDM1A regulates liver-TEs and impedes HCC cell growth.**

To further confirm our findings, we treated HepG2 cells with several KDM1A small-molecule inhibitors and observed similar changes in H3K4me1 as seen in KDM1A knockdown cells (Supplementary Fig. 2g). We found that out of a total of 1923 liver-TE associated genes, 1202 genes were targeted by KDM1A, accounting for approximately 62.5% of the cohort (Fig. 2e). RNA-seq and GSEA assays demonstrated that knockdown of KDM1A resulted in a global elevated expression of liver-TE-associated genes (Fig. 2f), suggesting a global inhibitory role of KDM1A in the regulation of this set of genes and coordinating the inhibitory role of KDM1A toward liver-TEs.

Colony formation assays demonstrated significant inhibition of liver cancer cell growth upon KDM1A downregulation (Fig. 2g), while KDM1A overexpression promoted cell growth (Supplementary Fig. 2h). To assess the relative importance of KDM1A in liver cancer cell growth compared to other histone methylation regulators, we ranked their mean dependency scores derived from RNAi screen results in the DepMap database. The results indicated a high dependency of liver cancer cells on KDM1A (Supplementary Fig. 2i and Supplementary Data 5). Furthermore, pharmacological inhibiting KDM1A with small-molecule inhibitors dramatically attenuating liver cancer cell growth (Supplementary Fig. 2j). Subcutaneous tumorigenicity assays in nude mice demonstrated that KDM1A knockdown significantly inhibited the in vivo tumorigenicity of liver cancer cells (Supplementary Fig. 2k). Strikingly, we conducted experiments using HCC-PDX model to assess the in vivo inhibitory effects of the KDM1A inhibitor SP2509. Our results demonstrated significant retardation of tumor growth upon SP2509 treatment, while no alterations in body weights were observed (Fig. 2h). Subsequently, we analyzed tumor samples from these experiments and observed an increase in the H3K4me1 within liver-TEs following KDM1A inhibition (Fig. 2i). Concurrently, we cultured PDX samples in vitro to establish a patient-derived cell line. Depletion of HNF4A-liver-TEs and KDM1A knockdown exerted notable suppressive effects on tumor cell growth (Supplementary Fig. 2l), consistent with the observed effects in conventional HCC cell lines. Moreover, by utilizing the HCC-PDO model, we investigated the impact of HNF4A-liver-TEs depletion and KDM1A inhibition on the 3D growth of HCC cells, revealing significant inhibition of 3D growth upon KDM1A inhibition (Fig. 2j). Taken together, the above findings indicate that targeting KDM1A could be a promising approach to regulate liver-TE activity and a potential strategy for the treatment of HCC.

KDM1A regulates HNF4A expression via liver-TEs

We demonstrated that HNF4A is a liver cancer suppressor gene negatively regulated by liver-TEs. This leads us to hypothesize that KDM1A can regulate the expression of the HNF4A gene through liver-TEs. This hypothesis was substantiated through subsequent experiments. Real-time PCR and Western blot experiments validated the regulatory role of KDM1A in HNF4A expression, as KDM1A downregulation elevated mRNA and protein levels of HNF4A (Fig. 3a and Supplementary Fig. 3a). Similarly, the application of SP2509, a KDM1A inhibitor, heightened HNF4A expression in liver cancer cells (Fig. 3b). Conversely, the overexpression of KDM1A led to a reduction in HNF4A expression (Fig. 3c and Supplementary Fig. 3b). The HCC-PDX samples treated with SP2509 also exhibited elevated expression levels of HNF4A (Supplementary Fig. 3c). Additionally, gene function enrichment analysis revealed that KDM1A negatively regulated the downstream gene signature of HNF4A (Fig. 3d). Further examination of histone marks around the HNF4A gene revealed that KDM1A knockdown enhanced H3K4me1 modification levels (Fig. 3e and Supplementary Fig. 3d). Given that H3K4me1 is a direct demethylation substrate of KDM1A, our results suggest that KDM1A inhibits HNF4A expression by promoting H3K4me1 demethylation. To ascertain whether KDM1A-mediated regulation of HNF4A expression relies on its catalytic activity, we reintroduced either wild-type or enzymatically inactive K661A mutant KDM1A into KDM1A knockdown cells. The results indicated that while wild-type KDM1A effectively counteracted the elevated HNF4A expression induced by KDM1A knockdown, the K661A mutant failed to produce a similar effect (Supplementary Fig. 3e). Additionally, in addition to reversible KDM1A inhibitors, irreversible inhibitors that selectively block the catalytic activity of KDM1A also increased HNF4A expression (Supplementary Fig. 3f). These findings suggest that the catalytic activity of KDM1A is pivotal for its regulation of HNF4A expression.

**Fig. 3: Epigenetic inhibition of HNF4A expression by KDM1A.**

To explore whether HNF4A-liver-TEs is essential for KDM1A-mediated inhibition of HNF4A expression, we overexpressed KDM1A in HCC cells depleted of HNF4A-liver-TEs, which resulted in the loss of the ability of KDM1A to suppress HNF4A expression (Fig. 3f). Correspondingly, the binding affinity of KDM1A to the transcriptional regulatory region of the HNF4A gene was weakened upon the removal of HNF4A-liver-TEs (Fig. 3g). Furthermore, the tumor-promoting effects of KDM1A were also attenuated by HNF4A-liver-TE deletion (Fig. 3h).

In consideration of the pivotal role played by HNF4A-liver-TEs in mediating the regulatory functions of KDM1A, it was postulated that exposure to HNF4A-liver-TEs may facilitate the recruitment of KDM1A. To interrogate this hypothesis, we employed CRISPRa technology to manipulate HNF4A-liver-TE, observing a consequent increase in the binding affinity of KDM1A (Supplementary Fig. 3g), concomitant with the repression of HNF4A expression (Supplementary Fig. 3h). Subsequent assessment of chromatin state revealed a reduction in chromatin accessibility, alongside decreases in H3K27ac and H3K4me1 modifications, at both HNF4A-liver-TEs and the HNF4A transcription start site (Supplementary Fig. 3i). These findings suggest that despite the action initiated by CRISPRa, the subsequent inhibitory effects of recruited KDM1A prevail. This is evidenced by observations indicating that knockdown of KDM1A in cells with CRISPRa targeting HNF4A-liver-TEs counteracted the loss of chromatin accessibility and reductions in H3K27ac and H3K4me1 (Supplementary Fig. 3i), ultimately alleviating the downregulation of HNF4A expression (Supplementary Fig. 3j). These collective observations highlight the predominant role of KDM1A in mediating the transcriptional inhibition of HNF4A induced by HNF4A-liver-TEs.

In non-malignant liver cells, where KDM1A exhibited lower binding efficiency to liver-TEs (Supplementary Fig. 2c), neither disruption of liver-HNF4A-TEs nor inhibition of KDM1A influenced HNF4A expression (Supplementary Fig. 3k, l). To ascertain whether the promoting effect of KDM1A on liver cancer depends on its regulation of HNF4A, we simultaneously silenced KDM1A and HNF4A in liver cancer cells. We found that the restrained in vitro and in vivo cell growth caused by KDM1A knockdown was partially restored by HNF4A downregulation (Supplementary Fig. 4a, b), indicating that KDM1A promotes the growth of liver cancer cells by inhibiting the expression of HNF4A.

KDM1A interacts with the HNF4A complex to epigenetically silence its downstream genes

Analysis of ChIP-seq experimental data revealed a significant positive correlation in the binding intensities of TRR between KDM1A, HNF4A, and other transcription factors (HNF1A, FOXA3, and GATA4), which play regulatory roles in liver development (Supplementary Fig. 4c–e). These observations suggest a potential interaction between KDM1A and the HNF4A transcription complex. To test this hypothesis, we conducted Co-IP experiments and confirmed that endogenous KDM1A interacts with HNF4A in liver cancer cells (Fig. 4a). Moreover, we observed an interaction between exogenously expressed KDM1A and HNF4A in 293T cells (Fig. 4b). We also identified interactions between HNF4A and other entoderm transcription factors (Supplementary Fig. 4f). Additionally, these transcription factors, including HNF4A, HNF1A, FOXA3, and GATA4, were found to interact with KDM1A (Fig. 4c). Immunofluorescence experiments using laser confocal imaging demonstrated colocalization of KDM1A and HNF4A within the nuclei (Fig. 4d). Further analysis of RNA-seq data in hepatocellular carcinoma with KDM1A downregulation revealed significant enrichment of conserved motifs of HNF4A in the promoters of KDM1A-negative regulatory genes (HOMER: P = 1e−14, Rank 1) (Fig. 4e and Supplementary Data 6). Based on the ChIP-seq results, the binding peaks of KDM1A and other known HNF4A partners were enriched around HNF4A motifs (Fig. 4f). These findings indicate that KDM1A interacts with HNF4A to suppress the expression of HNF4A target genes.

**Fig. 4: Interaction between KDM1A and HNF4A regulates histone methylation.**

To investigate the epigenetic regulation of KDM1A on HNF4A target genes, we performed CUT&Tag-seq experiments for H3K4me1, H3K4me2, H3K9me2, and H3K27ac in HepG2 cells with KDM1A knockdown or inhibition. We observed that the intensities of H3K4me1 surrounding HNF4A motifs and binding sites were upregulated upon KDM1A knockdown or inhibition (Fig. 4g, h). Furthermore, KDM1A downregulation or inhibition was accompanied by enhanced sequencing signals of H3K27ac-CUT&Tag in the aforementioned regions (Fig. 4g, h), indicating that KDM1A repressed the transcription of HNF4A target genes by removing H3K4me1 within TRR and altering chromosomal accessibility.

KDM1A inhibits the expression of the HNF4A target gene

Real-time PCR analysis of several known downstream genes of HNF4A (Fig. 5a, b) and luciferase reporter assays (Fig. 5c) demonstrate that KDM1A knockdown or pharmacological inhibition activates the transcriptional regulatory activity of HNF4A. As anticipated, the downregulation of HNF4A in KDM1A knockdown cells counteracted the elevation of HNF4A target genes induced by KDM1A knockdown (Supplementary Fig. 5a, b). To identify liver cancer suppressor genes related to liver-TE that are controlled by KDM1A-mediated HNF4A repression, we focused on MAT1A, a known HNF4A target gene associated with liver-TE which is directly repressed by KDM1A (Fig. 5d and Supplementary Fig. 5c). Pan-cancer expression profiling showed that MAT1A is liver tissue-specific, highly expressed in normal liver and liver cancer tissues, but downregulated in liver cancers compared to normal liver tissues (Supplementary Fig. 5d), a pattern exemplifies HNF4A downstream signature (Supplementary Fig. 5e). MAT1A is a key rate-limiting enzyme in methionine metabolic pathways and functions as a liver cancer suppressor³⁷. Observation by the IGV browser revealed that the TRR of MAT1A is bound by KDM1A, HNF4A, HNF1A, FOXA3, and GATA4, wherein the H3K4me1 intensities were upregulated upon KDM1A knockdown, accompanied by an upregulation of MAT1A mRNA expression levels revealed by RNA-seq assays (Fig. 5e). Real-time PCR and Western blot analysis further confirmed that inhibition of KDM1A effectively enhanced MAT1A expression (Fig. 5f, g). Mass spectrometry analysis displayed a decrease in the upstream metabolites of methionine and sulfur amino acids (MSO and MET), which are catalyzed by MAT1A, and an increase in the downstream products SAM and SAH upon KDM1A inhibition (Fig. 5h), confirming the negative regulation of MAT1A by KDM1A. This suggests that KDM1A downregulates the activity of the HNF4A-regulated hepatocyte-specific methionine metabolic pathway, ultimately promoting liver cancer.

**Fig. 5: KDM1A negatively regulates HNF4A target genes to attenuate HNF4A function.**

ZMYM3 mediates the binding of liver-TE by KDM1A

We hypothesized that KDM1A requires an adapter protein with specific DNA-binding activity to recognize liver-TEs, as KDM1A lacks a DNA-binding domain. By using an IP-LC-MS approach, we identified eight DNA-binding proteins that are present in KDM1A immunoprecipitants in liver cancer cells (Fig. 6a). To determine which factor has the potential to assist KDM1A for DNA binding, we integrated RNA-seq results from KDM1A knockdown cells, public ChIP-seq results and the Lisa algorithm. The results showed that ZMYM3 was enriched in the gene transcription regulatory regions that were directly negatively regulated by KDM1A (the gene whose TRR region possessed KDM1A binding peaks and which was upregulated upon KDM1A deletion based on RNA-seq assays) (Fig. 6b). ZMYM3 is a zinc-finger protein with a DNA-binding domain, and its interaction with KDM1A may contribute to the recruitment of KDM1A to the target DNA regions. The interaction was confirmed by endogenous Co-IP experiments (Fig. 6c). Additionally, similar to the in vivo impact of KDM1A, ZMYM3 knockdown also impeded tumorigenesis (Supplementary Fig. 6a). Peak intensity analysis showed that ZMYM3, KDM1A and HNF4A were positively correlated in the TRR (Fig. 6d and Supplementary Fig. 6b), indicating their co-occupancy in the genome. Moreover, ZMYM3, KDM1A, and HNF4A exhibited stronger binding ability to TRRs containing liver-TE compared to those without liver-TEs (Fig. 6d and Supplementary Fig. 6c).

**Fig. 6: ZMYM3 is required for the epigenetic regulatory role of KDM1A in liver cancer cells.**

To investigate whether ZMYM3 mediates KDM1A-induced demethylation of liver-TE, we performed CUT&Tag-seq experiments on KDM1A-overexpressing cells with or without ZMYM3 knockdown (Vector+sh-control, KDM1A+sh-control and KDM1A+sh-ZMYM3, respectively, Supplementary Fig. 6d). We observed the alterations in the KDM1A binding intensity and H3K4me1 modification both at KDM1A/ZMYM3 binding sites (Fig. 6e) and at liver-TEs (Fig. 6f). The results showed that ZMYM3 is required for the DNA-binding ability of KDM1A, as well as its specific binding to liver-TEs, as the binding of KDM1A to these regions decreased after ZMYM3 knockdown. Consistently, the H3K4me1 modifications on KDM1A/ZMYM3 binding regions and liver-TEs decreased after KDM1A overexpression, which was recovered by ZMYM3 knockdown. As ZMYM3 binding peaks are also correlated with HNF4A, we tested the roles of ZMYM3 in KDM1A-mediated demethylation of HNF4A binding regions. The results showed that knockdown of ZMYM3 reversed the effects of KDM1A overexpression on HNF4A binding sites (Supplementary Fig. 6e). Functionally, knocking down ZMYM3 inhibits KDM1A-mediated growth of liver cancer cells (Fig. 6g). These results suggested that ZMYM3 is required for the epigenetic regulatory role of KDM1A and is necessary for KDM1A to exert its pro-growth effect in liver cancer cells.

To investigate the mechanism underpinnings of ZMYM3-mediated facilitation of KDM1A DNA binding, particularly within liver transposable elements (TEs), we posited that these liver-TEs might harbor or be in close proximity to specific DNA sequences recognized by the ZMYM3 protein. To evaluate this hypothesis, we utilized the MEME-ChIP program to identify ZMYM3 motifs and mapped their locations across the human genome. We then computed the distances between liver-TEs and the identified ZMYM3 motifs. This analysis revealed a notable overlap or close proximity between the majority of ZMYM3 motifs and liver-TEs (Supplementary Fig. 6f). Importantly, the distances between liver-TEs and ZMYM3 motifs were statistically closer than those observed between a randomly shuffled set of TEs and the motifs (Wilcoxon test, P < 2.2e−16). Subsequently, we provided additional evidence by assessing the KDM1A binding capacity in regions with or without ZMYM3 motifs within or near KDM1A peaks. Our findings revealed a significantly heightened binding capacity of KDM1A in regions containing ZMYM3 motifs (Supplementary Fig. 6g), suggesting a potential role of ZMYM3 in enhancing the DNA-binding capability of KDM1A.

High expression of KDM1A is associated with poor prognosis in hepatocellular carcinoma

We performed immunohistochemical staining and survival analysis of KDM1A using 90 hepatocellular carcinoma tissue samples and found that high expression of KDM1A indicated poor prognosis (Fig. 7a, b). Furthermore, survival analysis using TCGA, ICGC, and GSE14520 datasets showed that the overall survival (OS) and progression-free survival (PFS) of patients in the high KDM1A expression group were significantly shorter than those in the low expression group (Fig. 7c). To corroborate our molecular findings with clinical samples, we conducted immunohistochemical staining on the same set of tissue samples used for KDM1A analysis, to detect ZMYM3 and HNF4A expression (Supplementary Fig. 7a). Subsequent correlation analyses unveiled a negative association between both KDM1A and ZMYM3 with HNF4A expression in HCC samples (Supplementary Fig. 7b). Additionally, we analyzed the clinical correlations of KDM1A expression level, ZMYM3 expression level, liver-TE-related gene expression profiles, and HNF4A downstream gene expression profiles in the TCGA and ICGC hepatocellular carcinoma expression profile datasets. The results showed that the expression levels of KDM1A and ZMYM3 were significantly negatively correlated with the expression profiles of liver-TE-related genes and HNF4A downstream genes (Fig. 7d), suggesting clinical relevance to our molecular-level findings.

**Fig. 7: KDM1A is overexpressed in hepatocellular carcinoma and predicts poor prognosis.**

Discussion

In our study, we revealed the functional role and importance of liver-TEs in liver cancer. We identified liver-TEs that were enriched in histone marks, epigenetic regulators, and adjacent to a set of liver tissue-specific genes that negatively associate with poor prognosis. Liver-TE-associated genes are highly expressed in both liver tissue and liver cancer. The liver-TE-associated genes include known liver cancer suppressor genes, such as HNF4A³⁴. Specifically, we focused on the regulatory role of liver-TEs on the expression of the HNF4A gene. Through the use of CRISPR/Cas9 technology, we selectively removed liver-TEs located within the transcriptional regulatory region of the HNF4A gene (HNF4A-liver-TEs) in liver cancer cells. Remarkably, the deletion of this liver-TE led to a notable upregulation of HNF4A expression and tumor cell growth inhibition, providing compelling evidence for the inhibitory function of this type of DNA elements on HNF4A.

Our findings indicate the enrichment of both the histone marker H3K4me1 and its demethylase KDM1A at the liver-TEs, suggesting a complex regulation of chromatin states in this region. H3K4me1 is typically associated with enhancer regions, and its presence is a hallmark of active enhancers³⁸. However, KDM1A is a demethylase that removes H3K4me1 marks from histones, resulting in gene repression^39,40. The co-enrichment of both H3K4me1 and KDM1A at the liver-TEs may seem contradictory, but we propose several possible explanations for this observation.

One possible explanation is that KDM1A modulates H3K4me1 levels at the liver-TE sites to fine-tune their transcriptional regulatory activity. KDM1A has been shown to function as a transcriptional co-repressor and may act in concert with other chromatin regulators to silence gene expression. In this context, the presence of KDM1A at liver-TE could help prevent aberrant activation of nearby genes by modulating the levels of H3K4me1. Another possibility is that liver-TE represents a transitional state between active and inactive enhancers, where H3K4me1 and KDM1A coexist. In this scenario, the balance between H3K4me1 and KDM1A levels may determine whether the TE-derived CRE is active or repressed. Low levels of KDM1A may allow for H3K4me1 to persist and maintain an active state around TE-derived CRE, whereas high levels of KDM1A may result in H3K4me1 removal and subsequent silencing. Overall, the co-enrichment of H3K4me1 and KDM1A at the liver-TEs suggests a complex regulation of transactivation activity in this region, which is consistent with the high expression of KDM1A in liver cancer and the low expression of liver-TE-related genes in liver cancer compared to normal liver tissue. This provides mechanistic insight into the inherent connection between KDM1A and liver-TE-associated gene expression in liver cancer.

The application of CRISPRa within the context of HNF4A-liver-TE also reveals the significant regulatory influence of KDM1A. While CRISPRa technology is commonly employed to induce targeted gene expression, typically leading to transcriptional activation, our findings present a nuanced scenario. We observe that CRISPRa targeting of HNF4A-liver-TE results in a suppressive chromatin state. One plausible explanation for this observation is the transient exposure of the TE induced by CRIPSRa, which subsequently recruits KDM1A. Consequently, despite the intended activation by CRISPRa, the recruitment of KDM1A leads to the suppression of HNF4A expression. These results highlight the intricate interplay between CRISPRa-based manipulation and the context of target regions. Further investigations into the precise mechanisms underlying these atypical effects of CRISPRa are warranted, as they may offer valuable insights into the development of more refined gene manipulation strategies.

Our study emphasizes the critical role of KDM1A in regulating liver cancer cell growth through two mechanisms involving HNF4A. Firstly, KDM1A promotes demethylation and restricts the accessibility of liver-TE, leading to the suppression of associated genes. Our study demonstrates that KDM1A directly regulates HNF4A by binding to liver-TEs within the HNF4A TRR and suppressing its expression, suggesting the involvement of KDM1A in inhibiting HNF4A expression. Secondly, KDM1A interacts with the HNF4A complex to epigenetically silence downstream genes, impacting both expression levels and transcriptional activity of HNF4A. HNF4A emerges as a significant factor among the transcriptional regulators enriched in liver-TEs, with its target genes highly enriched in liver-TE-associated genes. This interaction and epigenetic regulation contribute to the suppression of liver-TE-associated genes by KDM1A. Our findings shed light on the regulatory mechanisms of liver cancer growth by KDM1A and HNF4A, highlighting their essential roles in liver-TE accessibility and transcriptional activity. The liver-TE/KDM1A/HNF4A regulatory loop exemplifies the intricate interplay between these factors in liver cancer development and progression.

We hypothesized that KDM1A requires an adapter protein with DNA-binding activity to recognize liver-TEs, as KDM1A lacks a DNA-binding domain. To test this hypothesis, we identified that ZMYM3, a zinc-finger protein with a DNA-binding domain, interacts with KDM1A. Subsequently, we demonstrated that ZMYM3 is required for the DNA-binding ability and targeted demethylation activity of KDM1A. Functionally, we found that knocking down ZMYM3 inhibits KDM1A-mediated growth of liver cancer cells. Our findings strongly suggest that ZMYM3 is necessary for the epigenetic regulatory role of KDM1A and is vital for KDM1A to exert its pro-growth effect in liver cancer cells. Previous studies have identified ZMYM3 as a subunit of the CoREST complex, but its precise role in epigenetic regulation has remained unclear^41,42. Our results indicated the functional significance of ZMYM3, highlighting its ability to enhance the DNA-binding properties of KDM1A and direct its regulatory function to specific target sites. This mechanism of action is consistent with the broader role of protein complexes in regulating gene expression, where the specific composition of a complex determines its ability to bind to specific DNA sequences and modulate the activity of associated enzymes.

The research showed that high expression of KDM1A is associated with a poor prognosis in hepatocellular carcinoma patients, indicating its potential as a prognostic biomarker. The expression levels of KDM1A and ZMYM3 were negatively correlated with the expression profiles of liver-TE-related genes and HNF4A downstream genes, underscoring the clinical relevance of the molecular-level findings. Knockdown of KDM1A significantly inhibited liver cancer cell growth both in vitro and in vivo, highlighting the essential role of KDM1A in HCC cell growth. Additionally, the use of a small-molecule inhibitor targeting KDM1A substantially suppressed the growth of liver cancer cells, indicating KDM1A as a promising target for HCC treatment. The use of KDM1A small-molecule inhibitors could be a viable strategy for HCC treatment, and further studies are necessary to explore the clinical application of KDM1A as a prognostic biomarker and therapeutic target in HCC.

In summary, our study provides insights into the potential role of liver-TEs in modulating gene expression in liver cancer. Furthermore, our findings highlight the critical function of KDM1A in the growth of liver cancer cells through liver-TE-mediated mechanisms, specifically by repressing the transcriptional activity of HNF4A (Fig. 8). Overall, our research enhances the understanding of the precise regulation of transactivation activity at the liver-TEs, thereby opening avenues for the identification of novel therapeutic targets for liver cancer.

**Fig. 8: Schematic diagram illustrating the regulatory mechanism of liver-TEs in liver cancer cell proliferation.**

Methods

Ethical statement

Our research complies with all relevant ethical regulations. This study was approved by the Ethics Committee of Renji Hospital, Shanghai Jiao Tong University School of Medicine, and Institutional Animal Care and Shanghai Jiao Tong University Animal Care Commission.

Cell lines and tissue specimens

PLC/PRF/5 (CRL-8024), HepG2 (HB-8065), THLE2 (CRL-2706), THLE3 (CRL-3583), and HEK293T (CRL-3216) cells were acquired from the American Type Culture Collection (ATCC, Manassas, VA, USA). Huh7 (SCSP-526) cells were obtained from the Cell Bank of the Chinese Academy of Sciences (Shanghai, China). MHCC97H (97H) cells were provided by the Liver Cancer Institute of Zhongshan Hospital, Fudan University (Shanghai, China). Cell lines were tested by short tandem repeat (STR) profiling to verify authentication. All cells were cultured at 37 °C in 5% CO₂ in Dulbecco’s modified Eagle medium supplemented with 10% fetal bovine serum. All cells were routinely tested using a mycoplasma detection kit (C0301S, Beytime) to ensure no mycoplasma contamination. A set of commercial tissue microarrays (TMAs) containing 90 HCC tissues and non-tumoral adjacent liver tissues were used for IHC staining. The tissues were sourced from 90 HCC patients, comprising 80 males and 10 females, with ages ranging from 16 to 73 years (median age = 48.5 years). This study was approved by the Ethics Committee of Renji Hospital, Shanghai Jiao Tong University School of Medicine.

Animal studies

BALB/c nude mice (6 weeks old) were obtained from SLAC (Shanghai, China) and cultured under pathogen-free conditions in Shanghai Jiao Tong University Laboratory Animal Center. The mice were housed in a controlled environment with a 12-h light/dark cycle. The ambient temperature was maintained within the range of 20–26 °C, while humidity levels were kept between 40% and 70%. Daily welfare monitoring was conducted to assess their health, behavior, and any signs of distress, ensuring their well-being throughout the study period. All animal experiments were conducted in accordance with the guidelines approved by Institutional Animal Care and Shanghai Jiao Tong University Animal Care Commission. The maximal tumor size/burden permitted is 2000 mm³. To ensure humane endpoints, mice were euthanized using carbon dioxide (CO₂) asphyxiation followed by cervical dislocation to ensure death.

In vivo tumorigenesis assay

For in vivo tumorigenesis assays, 2 × 10⁶ Huh7 cells expressing either vector control or specified shRNA was subcutaneously injected into male nude mice. At the end of the study, tumors were excised, weighed, and photographed.

CRISPR/Cas9 and CRISPRa assays

For CRISPR/Cas9-mediated DNA element deletion, we used the CHOPCHOP software (http://chopchop.cbu.uib.no/) to design a pair of sgRNA flanking target elements. Then, the pair of sgRNA was constructed into a U6-promoter-sgRNA-U6-promoter-sgRNA cassette and cloned into lentiCRISPR_v2 (Addgene #60954) for a single vector expressing two sgRNAs. The sgRNA target sequences used for DNA editing include: HNF4A-liver-TE-UP-1-sgRNA: GTTAGACAAAACTTCTCCAA(TGG); HNF4A-liver-TE-UP-2-sgRNA: GGATGCATACCCTTGGCTCC(TGG); HNF4A-liver-TE-UP-3-sgRNA: CCATCTCTCACTGGCATCCC(TGG); HNF4A-liver-TE-UP-4-sgRNA: TAAAACAGCTGCATATCCAG(TGG); HNF4A-liver-TE-DOWN-1-sgRNA: GAGTGTTGTGTGGCCCCACG(AGG); HNF4A-liver-TE-DOWN-2-sgRNA: AGTGGAGGGGGCTGCACTCC(TGG); HNF4A-liver-TE-DOWN-3-sgRNA: GAAAGATCTGGGCTCAAATC(CGG); and HNF4A-liver-TE-DOWN-4-sgRNA: GGGTGCTTAGCCCTGGTAAA(GGG).

For the CRISPRa assays, we employed the CRISPR-based Synergistic Activation Mediator (SAM) system⁴³, which involves two lentiviral plasmids, one carrying dCas9-VP64 (lenti dCas9-VP64_Blast, addgene: #61425), and the other carrying the MS2-P65-HSF1 and gRNAs expressing cascade (pXPR_502, addgene: #96923). Lentivirus was generated and transduced into target cells, which were then selected with hygromycin, blasticidin, or puromycin. We targeted HNF4A-associated liver-TEs by CRISPRa system, using the following gRNA sequences: AAGGGTAGCCCTGGAGTTAG(AGG).

Identification of liver-TEs

To identify liver-TEs, we obtained RepeatMasker-annotated transposable elements from the UCSC database. We also obtained gene annotation files (gencode.v41) for the hg38 genome from the UCSC database to extract transcription start sites (TSS). We defined transcriptional regulatory regions (TRR) as genomic regions 10 kb upstream and downstream of TSS with a Smith-Waterman score >100. We then restricted TEs for analysis to major TE class/families, including SINE (Alu, MIR), LINE (L1, L2), LTR (ERV1, ERVL, ERVL-MaLR), and DNA (hAT-Charlie, TcMar-Mariner) that overlapped with TRR (TRR-TEs).

To assess chromatin accessibility in TRR-TEs across different tissue types, we downloaded ATAC-seq data (bigWig format) from the NIH GDC data portal (https://gdc.cancer.gov/about-data/publications/ATACseq-AWG)⁴⁴. We used the multiBigwigSummary program (parameters: BED file, –binSize 500) in the deeptools package (3.5.0) to generate the chromatin accessibility landscape, and analyzed RawCounts files using the Rtsne (0.15) to generate a tSNE map. We identified tumor type-specific TE markers using the FindConservedMarkers function, a program embedded in the Seurat package (v4.0.2) and limma (3.44.3) package. We used adj.p value < 0.0001 & log2FC > 10 as the cut-off values for the FindConservedMarkers algorithm and adj.p value < 0.0001 & logFC >2 and B value > 10 for the limma algorithm. We considered TEs filtered by both algorithms as liver-specific accessible TEs (liver-TEs). We defined the TRR regions containing liver-TEs as liver-TE-TRRs. To identify these regions, we utilized the “bedtools intersect” function within the bedtools software (v2.31.1), analyzing bed files containing liver-TEs and TRR regions to get TRRs harboring liver-TEs. Subsequently, the resultant bed files containing liver-TE-TRR regions were processed using the “bedtools merge�� function to create a list of non-overlapping liver-TE-TRRs. In this study, we defined liver-TE-associated genes as genes that have at least one liver-TE located within 10 kb of their transcription start site (TSS). We created a gene set comprising all liver-TE-associated genes identified in our analysis. This gene set was subjected to gene set variation analysis (GSVA) and gene set enrichment analysis (GSEA) based on publicly available HCC transcriptome datasets or RNA-seq results generated from our own experiments.

The ATAC-seq data of non-tumoral tissues were obtained from BioProject PRJNA63443⁴⁵ (https://www.ncbi.nlm.nih.gov/bioproject/PRJNA63443). Samples used in this study are listed in Supplementary Data 7. The uniquely mapped reads were isolated by using samtools (v1.6) under parameters: samtools view -f 2 -q 10 -b. Then, the Bam files including uniquely aligned reads were transformed to bigwig formatted files by the bamCoverage function in Deeptools (v3.5.1) with default parameters. Then, we used the multiBigwigSummary program (parameters: BED file, –binSize 500) in the deeptools package (3.5.0) to analyze the chromatin accessibility landscape.

Reagents and antibodies

SP2509 (S7680), Seclidemstat (S6722), GSK-LSD1 (S7574), ORY-1001 (S7795), and GSK2879552 (S7796) were obtained from Selleck. Puromycin and penicillin/streptomycin were purchased from yeason, China. The primary antibody were as followed: anti-KDM1A (abcam, ab129195[EPR6825], lot1018822-16, 1:100 for CUT&Tag, IF and IHC, 1:1000 for Western blots; ABclonal, A21801[O60341], lot3560844003, 1:200 for IP and 1:1000 for Western blots); anti-ZMYM3 (Proteintech, 25742-1-AP[AB_2880221], lotC10-027S, 1:100 for IHC, 1:200 for IP and 1:1000 for Western blots); anti-HNF4A (R&D, PP-H1415-00[Cl H1415], lotA-2, 1:200 for IF, 1:100 for IHC and 1:1000 for Western blots; ABclonal, A20865[ARC2794], lot3560844003, 1:1000); anti-H3K4me1 (abcam, ab176877[ERP16597], lotGR3208750-3, 1:100 for CUT&Tag and 1:5000 for Western blots); anti-H3K4me2 (abcam, ab32356[Y47], lotGR253788-33, 1:100 for CUT&Tag and 1:5000 for Western blots); anti-H3K27ac (abcam, ab4729, lot1059037-1, 1:100 for CUT&Tag); anti-H3K9me2 (abcam, ab1220[mAbcam 1220], 1:100 for CUT&Tag and 1:5000 for Western blots), anti-FLAG (ABclonal, AE063[ARC5111-02], lot9100026002, 1:200 for IP and 1:5000 for Western blots); anti-MAT1A (ABclonal, A2630[AB_2764502], lot002800101, 1:1000); anti-V5 (ABclonal, AE017[AMC0506], lot9200017002, 1:200 for IP and 1:5000 for Western blots); anti-GAPDH (Santa Cruz biotechnology, sc-32233[6C5], lotG3020, 1:5000); IP lysis buffer (Beyotime Institute of Biotechnology, P0013); RIPA (Beyotime Biotechnology, P0013B) ; Flag M2 affinity gel (Bimake, B26101), Protein G Dynabeads (Thermo Fisher, 10017D); Poly Flag peptides (Bimake, Poly FLAG F4799). Second antibody for Western blot: Goat anti-rabbit (LI-COR, 926–3221, lotD31205-5, 1:10,000); Goat anti-mouse (LI-COR, 926–68020, lotD00310-25,1:10000). Second antibody for Immunofluorescence analysis: Goat anti-rabbit (Thermo Fisher, A11008, lot2420731, 1:1000); Goat anti-mouse (Thermo Fisher, A11020, lot2306811, 1:1000); DAPI (Thermo Fisher, D1306, 1:1000).

Plasmids and stable cell line construction

The ORF of KDM1A (shRNA-resistant, wild-type, or K661A mutant) were cloned into pLVX-IRES-Puro or pLVX-IRES-ZsCreen plasmid. The specific shRNA sequence was designed and inserted into pLKO.1: shKDM1A(sh1: CCGGGCTACATCTTACCTTAGTCATCTCGAGATGACTAAGGTAAGATGTAGCTTTTTG; sh2:CCGGGCCTAGACATTAAACTGAATACTCGAGTATTCAGTTTAATG TCTAGGCTTTTTG);shHNF4A(CCGGTCAGGGTCTGAGCCCTATAAGCTCGAGCTTATAGGGCTCAGACCCTGATTTTTG);shZMYM3(CCGGGTTGTACCGGGCTCAACTATTCTCGAGAATAGTTGAGCCCGGTACAACTTTTTG). HEK293T cells and packaging plasmids (psPAX2, pMD2.G) were used to produce lentivirus. HCC cells were infected with lentivirus in the presence of polybrene. Cells were collected for gene expression assay.

Identification of transcriptional regulators and histone markers enrichment in liver-TEs

To determine the enrichment of transcription regulatory (TR) proteins or histone markers (HM) in liver-TEs, we utilized the Remapenrich package (0.99.0) (https://github.com/remap-cisreg/ReMapEnrich). First, we obtained a bed-formatted file containing the positions of liver-TEs as the query set. For TR enrichment analysis, we downloaded the bed-formatted Remap catalog (2020 version) from the Remap database (https://remap.univ-amu.fr/storage/remap2020/hg38/MACS2 /remap2020_all_macs2_hg38_v1_0.bed.gz)⁴⁶. We then extracted ChIP-peaks from liver cancer HepG2 cells to assemble a sub-catalog for this analysis. For HM enrichment analysis, we downloaded the HM ChIP-seq peaks of HepG2 cells from the GTRD database (http://gtrd.biouml.org:8888/downloads/20.06/bigBeds/hg38/ChIP-seq_HM/Peaks/) and assembled a custom Remap-catalog formatted file. For both TR and HM enrichment analyses, the software generated background regions by randomly shuffling the query regions within TRR for 500 iterations (shuffles=500). Other parameters in the program were set as follows: byChrom = F, fractionQuery = 0.01, fractionCatalog = 0.01, included = 0.9, tail = “both”. The enrichment of TR or HM in liver-TEs was then determined using the Remapenrich package.

To analyze the TRs that enriched the promoters of a selected set of genes, Lisa software (https://github.com/liulab-dfci/lisa) was also used. It uses a gene-centric approach to identify transcription factors and other regulatory proteins by integrating online ChIP-seq data.

Western blot and co-immunoprecipitation (Co-IP)

To perform Western Blot, the cells were lysed with RIPA buffer (Beyotime Biotechnology, P0013B) containing a protease inhibitor cocktail (bimake, B14001) for 1 h on ice. Next, the cell lysates were centrifuged at 13,000 × g for 20 min at 4 °C. The protein concentrations of the supernatants were determined using a BCA protein assay kit (Beyotime Biotechnology, P0011). To separate the protein samples, 100 μg total proteins were loaded onto SDS-PAGE and transferred to nitrocellulose (NC) membranes (Pall Corporation). The NC membranes were then blocked with 5% BSA and incubated with primary antibodies overnight at 4 °C. After washing the membranes three times with TBST, they were incubated with goat anti-rabbit and goat anti-mouse antibodies and visualized using the Odyssey imaging system.

For Co-immunoprecipitation (Co-IP), the cells were lysed with IP lysis buffer containing protease inhibitor cocktail for 1 h on ice. The supernatant proteins were then incubated with 50 μl of Protein G Dynabeads (Thermo Fisher, 10017D) on a rocking platform for 3 h at 4 °C to reduce non-specific binding. After removing the Dynabeads, the supernatants were incubated with 50 μl of Protein G Dynabeads and primary antibody on a rocking platform overnight at 4 °C. The immunoprecipitates were collected using a magnet and washed with TBS three times. The Dynabeads were denatured with a loading buffer and subjected to Western Blot analysis. For Co-IP conducted with Flag M2 affinity gel (Bimake, B26101), the supernatant proteins were incubated with 50 μl of Flag M2 affinity gel on a rocking platform overnight at 4 °C. The subsequent steps were similar to those of Co-IP conducted with Protein G Dynabeads.

LC-MS protein identification

To identify KDM1A interacting proteins, protein complexes were immunoprecipitated from HepG2 cells expressing KDM1A-Flag using an anti-Flag antibody (n = 1). Simultaneously, protein complexes were also immunoprecipitated from HepG2 cells using an anti-KDM1A antibody (n = 1). In both cases, an anti-IgG antibody was used as a control. Subsequently, the immunoprecipitated KDM1A complexes were subjected to LC-MS identification. The purified protein complex was resolved by SDS-PAGE followed by excised, digested, and then subjected to LC-MS (Liquid chromatography-tandem mass spectrometry) analysis. This approach was accomplished by PANOMIX BioMedical Tech. Co. Suzhou, China.

Simple preparation: The gel is cut into small pieces and treated with a destaining solution containing 40% methanol and 50 mM NH₄HCO₃. The gel pieces are heated and shaken for 15 min, with the destaining solution replaced at least once. After destaining, the gel pieces are washed twice with water and dehydrated with 75% ACN. The gel pieces are then washed with NH₄HCO₃ and subjected to protein digestion. To digest the proteins, NH₄HCO₃ is removed, and trypsin is added to the gel pieces. The gel pieces are ground and centrifuged, and the mixture is incubated overnight at 37 °C. The resulting peptides are extracted with ACN, centrifuged, and the supernatant is collected and dried.

Liquid chromatography (LC) was performed using an Easy-nLC 1200 system (Thermo Fisher Scientific, USA). Samples were dissolved in 80 μL of mobile phase A (0.1% formic acid aqueous solution), vigorously mixed, and then 5 μL of the sample was loaded into a 20 μL quantitative loop using an autosampler. The sample in the quantitative loop was loaded onto a C18 reverse-phase pre-column by pump A at a maximum pressure of 280 bar. Subsequently, the sample was eluted onto a C18 analytical column using a gradient of 11%–37% mobile phase B (0.1% formic acid in 80% acetonitrile) at a flow rate of 600 nl/min for separation.

Mass spectrometry data were acquired using an Orbitrap Fusion Lumos mass spectrometer (Thermo Fisher Scientific, USA). The ion source employed was a nanospray ionization source (NSI) with a spray voltage of 2200 V and an ion transfer capillary temperature of 320 °C. Mass spectrometric data were collected in data-dependent acquisition (DDA) mode under positive ion mode. In the first stage, full scans were performed using the Orbitrap with a scan range of m/z 350–1550, at a resolution of 120,000, with automatic gain control (AGC) set to 5e5 ions, and a maximum injection time of 50 ms. For the second stage, fragmentation of precursor ions was achieved by high-energy collision dissociation (HCD) at 32%, with fragment ions detected in the Orbitrap at a resolution of 15,000. The first mass was set to 100, with an AGC target of 5e4 ions and a maximum injection time of 22 ms. A precursor ion selection window of 1.6 Th was employed, and MS/MS acquisition was performed for ions with charge states ranging from 2 to 7. Dynamic exclusion was set to exclude the same precursor ion from MS/MS acquisition for 30 s after being selected once for MS/MS acquisition.

The Sequest algorithm integrated in Proteome Discoverer software (v2.2) was used for database searching. PD software was used for qualitative analysis based on Sequest search results and the first-step spectrum screening. Protein quantification values were the sum of the peptide quantification values. Search parameters were as follows: MS tol: 20 ppm, MS/MS tol: 0.05 Da, max missed cleavages: 2. The database used was Uniprot (Human, version 202002).

Quantitative real-time PCR (RT-PCR)

Total RNA was extracted with the RNAiso Plus kit (Takara Bio Inc) according to the manufacturer’s instructions. RNA was reverse transcribed to cDNA with reverse transcriptase (yeason, 14605ES08). Quantitative real-time PCR (RT-PCR) was performed with qPCR SYBR Green Master Mix (yeason, 11199ES03) and analyzed by Viia7 Real-Time PCR System (AB Applied Biosystems). The sequences of primers used for RT-PCR are provided in Supplementary Data 8. Data are normalized to GAPDH expression in each sample.

Immunofluorescence analysis

Cells were seeded on coverslips and cultured for 24 h. Then cells were fixed with methanol and permeabilized with 0.1% Triton X-100 for 10 min. Cells were washed with PBS three times and a blocking buffer (5% goat serum) was applied for 30 min at room temperature. The cells were incubated with the primary antibody overnight at 4 °C. After washing three times in PBS, the cells were incubated with a second antibody for Immunofluorescence analysis for 1 h. Cell nuclei were counterstained using DAPI. The slides were observed under a Leica SP8 confocal microscope (Leica Microsystems, Wetzlar, Germany).

Immunohistochemistry

The tissue microarrays underwent deparaffinization, rehydration, immersion in 3% hydrogen peroxide solution for 15 min, antigen retrieval in citrate buffer (pH 6.0) for 25 min at 95 °C, and cooling for at least 60 min at room temperature. They were then blocked with 10% normal goat serum for 30 min. Next, the slides were incubated with primary antibody overnight at 4 °C and visualized using the PV-9000 Polymer Detection System (GBI, USA) following the manufacturer’s instructions. After washing with PBS, the slides were counterstained with hematoxylin. Protein expression levels were determined based on staining intensity and percentage of immunoreactive cells, where staining intensity was graded as 0 (negative), 1 (weak positive), 2 (moderate positive), and 3 (strong positive), and immune reaction cell percentage were classified as 0 (0%), 0.5 (1~10%), 1 (11~20%), 2 (21~50%), 3 (51~80%), 4 (81~100%). The final score for a sample was calculated as the mean of tumor cell staining intensity scores multiplied by the percentage of positive cell scores. The prognostic value of certain proteins in patients with different subtypes was evaluated using the R software packages survival (3.2-13) and survminer (0.4.9). The best cut-off value for the IHC staining score was estimated using the R package maxstat (0.7-25).

Cell viability assays

In the colony formation assays, cells were placed in 6-well plates and allowed to incubate for a period of 14 days. The resulting colonies were stained with crystal violet (Beyotime, C0121) for visualization. Each well was seeded with 1000 cells for gain of function assays, 2000 cells for loss of function assays, and 10,000 cells for in vitro pharmacological assays.

Luciferase reporter assay

HNF4-Luc Reporter plasmid was purchased from Genomeditech. Renilla luciferase Reporter plasmid was acquired from Inovogen Tech. Co. 5.0 × 10⁴ cells were seeded in a 12-well plate and co-transfected with the above vectors when the cell confluence reached 80%. After 24 h, the cells were treated with a Luciferase Reporter Assay Substrate Kit (Beyotime Institute of Biotechnology, RG027) following the manufacturer’s instructions, and the luciferase activity was measured using a SpectraMax reader (Molecular Devices, Shanghai, China).

Publicly available transcriptome and epigenome datasets

The HCC dataset from The Cancer Genome Atlas (TCGA) database⁴⁷ (https://portal.gdc.cancer.gov/) was obtained using the TCGAbiolinks package (2.16.4). FPKM-normalized RNA-seq data were converted to TPM values using the TCGA dataset for further analysis. Additionally, the GSE14520⁴⁸ (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE14520) and GSE54236⁴⁹ (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE54236) HCC transcriptome datasets were obtained from the Gene Expression Omnibus (GEO) database using the GEOquery package (2.56.0). The HCC dataset from the International Cancer Genome Consortium (ICGC) database was obtained from the ICGC data portal (https://dcc.icgc.org/)⁵⁰. From the UCSC Xena database (https://xenabrowser.net/), we downloaded a combined TCGA and Genotype-Tissue Expression (GTEx) dataset⁵¹ (https://xenabrowser.net/datapages/?cohort=TCGA%20TARGET%20GTEx&removeHub=https%3A%2F%2Fxena.treehouse.gi.ucsc.edu%3A443) for analyzing the overall expression of liver-TE associated genes in different types of normal tissues/cancer tissues.

To investigate the prognostic relationship of liver-TE-associated genes, we used our created liver-TE gene signature to calculate the GSVA score of liver-TE using the GSVA package. Next, survival analyses were performed on transcriptome datasets containing survival information using the survival (3.2-11) and survminer (0.4.9) packages in R based on the GSVA score in each sample. Cut-off values were estimated using the maxstat package (0.7–25). Furthermore, the maxstat method was also utilized to stratify KDM1A expression for Kaplan–Meir survival analysis.

Several publicly available ChIP-seq raw data were downloaded from the ENCODE database (https://www.encodeproject.org/), including ENCFF750VZA (KDM1A-ChIP-seq in HepG2 cells), ENCFF000PJT (HNF4A-ChIP-seq in HepG2 cells), ENCSR848YWD (ZMYM3-ChIP-seq in HepG2 cells), ENCFF442RQA (HNF1A-ChIP-seq in HepG2 cells), ENCFF492CBJ (FOXA3-ChIP-seq in HepG2 cells), and ENCFF163SRP (GATA4-ChIP-seq in HepG2 cells)^45,52. To determine protein binding regions based on these data, adapters were trimmed using Trim-galore (0.6.5-1), and reads were aligned to the human genome (UCSC hg38) using BWA (0.7.17). The aligned reads were then processed by MACS2 (v2.2.7.1) for peak calling, and bigwig files for peak visualization in Integrative Genomics Viewer (IGV) were generated using Deeptools (v3.5.1). For the KDM1A-ChIP-seq data, we categorized genes with KDM1A peaks (identified by MACS2, with a score >200) within a ±10 kb vicinity of their TSS as KDM1A target genes. To analyze the methylation status of CpG sites in liver-TEs, we downloaded a BED file (GSM1204463⁵³, https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSM1204463) from the GEO database. This file contained the frequency of methylated reads in each detected CpG site in HepG2 cells, which were obtained using a Reduced Representation Bisulfite Sequencing (RRBS) assay. We shifted the reference genome from hg19 to hg38 using the liftOver software and used BEDTools (v2.31.1) to extract CpG sites within liver-TEs for comparison of methylation levels with a similarly sized sample of randomly selected TEs within TRR regions.

Capture Hi-C data were obtained from the ArrayExpress database (E-MTAB-7144)⁵⁴, originating from HepG2 cells. The raw sequencing data underwent preprocessing using the HiC-Pro software (3.1.0). This included alignment to the human hg38 genome, removal of duplicate reads, filtering for valid interactions, and generation of binned interaction matrices. cis-interactions involving liver-TEs were selected using the bedtools (v2.31.1), and visualized by the IGV genome browser.

ChIP-PCR

ChIP assays were conducted with a minimum of 1 × 10⁷ cells per sample. The cells were initially crosslinked with 1% formaldehyde for 10 min at room temperature, and the crosslinking reaction was stopped by 0.125M glycine. Subsequently, the cells were washed with cool PBS and collected by centrifugation at 700 × g for 5 min at 4 °C. Nuclei were isolated using a nuclear extraction kit (Solarbio, EX2650). The isolated nuclei were suspended in a 20% SDS lysis buffer and incubated on ice for 10 min, followed by sonicating to fragment the DNA into sizes ranging from 200 to 1000 base pairs. The resulting nuclear lysates were then clarified by centrifugation at 9400 × g for 10 min at 4 °C. 1% aliquot of the supernatant was reserved for use as the whole-cell extract (WCE) DNA control, while the remaining supernatant was diluted with a dilution buffer consisting of Triton X-100, EDTA, NaCl, Tris-HCl pH 8, and a protease inhibitor. The diluted supernatant was then pre-cleared using protein G Dynabeads (Thermo Fisher, 10017D) on a rocking platform for 3 h at 4 °C to reduce non-specific binding. After removal of the Dynabeads, the supernatants were incubated overnight at 4 °C with 50 μl of Protein G Dynabeads and the primary antibody on a rocking platform. Protein/DNA immunoprecipitates were subsequently collected using a magnet and subjected to a series of washes, including a low salt wash buffer, high salt wash buffer, LiCl wash buffer, and TE buffer. Reversal of protein/DNA cross-links in both WCE DNA and immunoprecipates was carried out by Elution buffer (0.5 M NaHCO₃, SDS%) via incubating at 65 °C for 6 h. The immunoprecipitated protein and RNA were then digested with Proteinase K and RNAse A at 55 °C for 2 h. DNA Purification was carried out according to the manufacturer’s instructions for the Qiaquick PCR Purification Kit (Qiagen, 28106). The antibodies used in ChIP-PCR assays were anti-Flag antibody (ABclonal, AE063, 1:100) and anti-H3K4me1 antibody (abcam, ab176877, 1:100). The Primers used in evaluating Flag-KDM1A occupancy at the HNF4A-liver-TE region are as following: Forward: GGGCCCCAAGTCTATGGTTC and Reverse: AGGCACCCACAAAGCTTCAA. The primers used in detecting H3K4me1 levels at the HNF4A TRR regions were primer pair 1: GTTCTCCACAGGGAGGTAG, GGTGAGCACCTGCTGAGCTG, and primer pair 2: GCTCGGCTGACCTCAG, ACAAGCAGACACTGCCGCA.

RNA-seq

Total RNA was extracted using TRIzol Reagent (Life Technologies, CA, USA) according to the manufacturer’s protocol, and RNA integrity was assessed using an Agilent Bioanalyzer 2100 (Agilent Technologies, CA, USA) to obtain RIN values. RNA with RIN values >7 was purified using an RNAClean XP Kit (Beckman Coulter, Inc. CA, USA) and RNase-Free DNase Set (QIAGEN, GmBH, Germany). RNA libraries were prepared for sequencing using a VAHTS Universal V6 RNA-seq Library Prep Kit for Illumina (Vazyme, Nanjing, China), and sequencing was performed on an Illumina HiSeq 2500 system.

Raw RNA-seq data were subjected to quality control using FastQC (v0.11.9) and trimmed using Trim_galore (0.6.5-1). SortMeRNA (4.2.0) was used to remove rRNA reads to generate clean data. The clean reads were aligned to the human reference genome (hg38) using the STAR (2.7.6a) aligner to generate BAM files. Duplicates were removed using Samtools (1.7), and the unique mapped counts were obtained using featureCounts (2.0.1). Differential gene expression analysis was performed using EdgeR (3.30.3), and gene set enrichment analysis (GSEA) and gene function enrichment were performed using the ClusterProfiler package (3.16.1) based on the EdgeR results. The aligned reads (BAM files) were transformed into bigwig files using Deeptools software (v3.5.1) for visualization in Integrative Genomics Viewer (IGV).

CUT&Tag-seq

We used the Hyperactive® Universal CUT&Tag Assay Kit for Illumina (Vazyme #TD903) to prepare the DNA library. Briefly, fifty thousand cells were gently resuspended with NE buffer and incubated on ice for 10 min. The nuclei were isolated and conjugated to 10 μL pre-activated concanavalin A-coated magnetic beads. The bead-bound nuclei were then incubated with 50 μL Antibody Buffer containing anti-biotin rabbit mAb (1:100 dilution) overnight at 4 °C. Subsequently, the nuclei were resuspended with 100 μL dig-wash buffer containing antibody (1:100 dilution) to bind with the primary antibody or rabbit IgG (control). The nuclei were washed three times with 200 μL of dig-wash buffer and tagmented with pA/G-Tn5 adapter complex. The tagmented DNA was collected using DNA Extract Beads for library preparation. PCR was performed for library amplification for 15 cycles, and the library was purified using VAHTS DNA Clean Beads (Vazyme #N411). The tagmented DNA was sequenced using the Illumina novaseq 6000 platform.

The sequencing reads obtained from the Illumina platform were subjected to quality control using FastQC (v0.11.9) to assess the quality of the reads. Adapter trimming and read filtering were performed using Trim-galore (0.6.5-1) with the following parameters: –phred33 –length 35 –stringency 3. The filtered reads were then aligned to the human genome (UCSC hg38) using Bowtie2 version 2.2.6 with parameters: –local –very-sensitive –no-mixed –no-discordant –phred33. The aligned reads were processed by MACS2 (v2.2.7.1) to call peaks and transformed to bigwig formatted files by the bamCoverage function in Deeptools (v3.5.1) with default parameters. To visualize the genomic tracks and peak profiles, the processed data were visualized using the Integrative Genomics Viewer (IGV), Ngsplot (v2.63) or Deeptools plotheatmap function. The differential peaks between groups were determined by the macs2 bdgdiff function with paramter: –cutoff 2.

The primary antibody used in CUT&Tag-seq assays were as followed: anti-KDM1A (abcam, ab129195, 1:100); anti-H3K4me1 (abcam, ab176877, 1:100); anti-H3K4me2 (abcam, ab32356, 1:100); anti-H3K27ac (abcam, ab4729, 1:100) and anti-H3K9me2 (abcam, ab1220, 1:100).

ATAC-seq

The ATAC-seq library preparation was conducted using the Hyperactive ATAC-Seq Library Prep Kit for Illumina (Vazyme, TD711) according to the manufacturer’s guidelines. Each sample comprised a total of 5 × 10⁵ cells. Initially, cells were washed with 500 μl PBS and then centrifuged at 500 × g for 5 min at room temperature. Subsequently, the cell pellets were resuspended in 50 μl of cold lysis buffer and incubated on ice for 10 min to isolate the nuclei. Following centrifugation at 500 × g at 4 °C for 5 min, the nuclei were subjected to a transposition reaction by incubating them with a 50 μl Tn5 transposome/Transposition reaction mix at 37 °C for 30 min. The tagmentation process was performed within the transposition reaction system. The fragmented/transposed DNA was purified using VAHTS DNA Clean Beads. The purified DNA underwent two washes with 200 μl of fresh 80% ethanol and was finally eluted in 26 μl of Nuclease-free ddH₂O. The library amplification protocol followed the program: 72 °C for 3 min; 95 °C for 3 min; 12–15 cycles of 98 °C for 10 s, 60 °C for 5 s; 72 °C for 1 min; and then held at 12 °C. The amplified ATAC-Seq library was further purified using VAHTS DNA Clean Beads. The purified DNA was washed twice with 200 μl of fresh 80% ethanol and eluted in 22 μl of Nuclease-free ddH₂O. Finally, all ATAC-seq libraries were sequenced using the Illumina NovaSeq 6000 platform.

The sequencing data obtained from the Illumina platform underwent quality control assessment using FastQC (v0.11.9) to evaluate read quality. Subsequently, adapter trimming and read filtering were performed using Trim-galore (v0.6.5-1) with the following parameters: –phred33 –length 35 –stringency 3. The resulting high-quality reads were aligned to the reference genome (hg38 for humans) using Bowtie2 (v2.2.5) with the parameters: –very-sensitive -x 2000. The uniquely mapped reads were isolated by using samtools view function (samtools v1.6) under parameters: -f 2 -q 10 -b.

Patient-derived xenografts (PDX) and organoids (PDO)

The HCC patient-derived xenografts (HCC-PDX, No. LI0024, female) were obtained from Shanghai GeneChem Organism by implanting tumor sample fragments (20–30 mm³) derived from an HCC patient into female nude mice (6 weeks old). After one month, twelve mice with tumors reaching an average size of approximately 163 mm³ were selected and randomly divided into 2 groups: six mice in the control group received intraperitoneal injection of 10 μl/g vehicle (10% DMSO, 90% corn oil), and six mice in the test group received SP2509 treatment (i.p., 10 μl/g, 25 mg/kg, twice a week for 3 weeks). Subsequently, tumor xenograft volumes and mice body weights were measured every three days for three weeks. The formula for calculating tumor volume was V = 0.5 × a × b², where a and b represent the long and short diameters of the tumor, respectively. On the 21st day, mice were euthanized. The dissected tumor samples from PDX models were then subjected to the following real-time PCR assays and CUT&Tag-seq assays. The informed consent from all participants was acquired. The use of female mice in the PDX assay is due to the fact that the PDX donor for this experiment was female. Female mice were chosen to maintain experimental consistency and to avoid sex-specific factors that could confound the results. This experimental protocol was approved by the Institutional Animal Care and Use Committee (IACUC) of Shanghai GeneChem Organism and the Ethics Committee of Renji Hospital, Shanghai Jiao Tong University School of Medicine.

HCC patient-derived organoids (HCC-PDO, male) were acquired from D1 Medical Technology (Shanghai, China) through a Material Transfer Agreement (MTA). The organoids were cultured on Matrigel-coated plates to facilitate 3D growth and were maintained in the organoid culture medium (D1 Medical Technology, No. K21103). To evaluate the organoid-forming capacity of the HCC patient-derived cells, organoids were enzymatically dissociated into single cells. Subsequently, these cells were seeded onto Matrigel-coated plates for 3D growth, with a seeding density of 2000 cells per plate. The resulting organoids were then quantified. This study was approved by the Ethics Committee of Renji Hospital, Shanghai Jiao Tong University School of Medicine.

DepMap database analysis

To investigate the function of histone methylation-related enzymes in hepatocellular carcinoma (HCC) cell viability, we obtained normalized RNAi screening data from the DepMap database (19Q3) via the website: https://depmap.org/portal/download. We selected genes annotated by Gene Ontology (GO) terms GO_HISTONE_DEMETHYLASE_ACTIVITY or GO_HISTONE_LYSINE_N_METHYLTRANSFERASE_ACTIVITY and liver cancer cell lines for analysis. Ultimately, we ranked the mean dependency scores of 62 histone demethylases and methyltransferases in 26 HCC cell lines to evaluate and compare their significance for HCC cell viability. A gene with a negative dependency score is essential for cell growth.

Motif identification

Motifs significantly enriched in the promoter regions of KDM1A negatively regulated genes were identified using the findMotifs.pl script in the HOMER package (v4.11), with default parameters.

To identify motifs for protein binding or consensus sequences within each family of liver-TEs, BED-formatted files containing genomic regions (MACS2 called peaks, overlapped binding regions, or liver-TE regions) were analyzed using the MEME-ChIP program (v5.4.1), with default parameters. During the MEM-ChIP process, the software MEME and STREME were used to identify novel motifs or consensus sequences. The newly identified motifs were then matched to known motifs in the Cis_BP database or a given set of motifs in MEME format using TOMTOM. The FIMO program was used to screen motifs and find their localizations. Finally, all screened GTF files were merged into a single BED file for Ngsplot visualizing of CUT&Tag-seq or ATAC-seq intensity profiles surrounding specific motifs.

LC-MS Metabolite Identification

HepG2 cells treated with SP2509 (n = 2) and vehicle-treated cells (control, n = 2) were subjected to metabolite identification. The untargeted metabolism assays were performed by PANOMIX BioMedical Tech. Co. Suzhou, China. For sample preparation, cells were mixed with glass beads and a mixed solution of acetonitrile, methanol, and water. The mixture was rapidly frozen in liquid nitrogen and thawed, followed by centrifugation to collect the supernatant. The supernatant was then dried and redissolved in an acetonitrile and 2-amino-3-(2-chloro-phenyl)-propionic acid solution. The sample was filtered and transferred for LC-MS analysis.

Liquid Chromatography (LC) analysis was performed on a Vanquish UHPLC System (Thermo Fisher Scientific, USA) with an ACQUITY UPLC ® HSS T3 (150 × 2.1 mm, 1.8 μm) (Waters, Milford, MA, USA) column. Mass spectrometry (MS) was performed on an Orbitrap Exploris 120 instrument (Thermo Fisher Scientific, USA) with an ESI ion source. Full MS-ddMS2 mode with data-dependent MS/MS acquisition was conducted simultaneously. The following parameters were applied: sheath gas pressure of 30 arb, aux gas flow of 10 arb, spray voltage of 3.50 kV for ESI (+) and −2.50 kV for ESI (-), capillary temperature of 325 °C, MS1 range of m/z 100–1000, MS1 resolving power of 60,000 FWHM, 4 data-dependent scans per cycle, MS/MS resolving power of 15,000 FWHM, normalized collision energy of 30%, and automatic dynamic exclusion time.

Metabolite data analysis was performed using MS-DIAL (version 4.8) (http://prime.psc.riken.jp/Metabolomics_Software/MS-DIAL/index.html), an open-source software platform for metabolomic data processing and analysis. The raw LC-MS data files were converted to mzXML format using ProteoWizard MSConvert (version 3.0.22015) and then imported into MS-DIAL for peak alignment and identification. The parameters are as follows: Alignment: Retention time tolerance: 0.05 min; MS‘ tolerance: 0.015 Da; Identification: Retention time tolerance: 100 min; Accurate mass tolerance (MS1): 0.1 Da; Accurate mass tolerance (MS2): 0.5 Da; Identification score cut off: 50%.

Statistics & reproducibility

Statistical analyses were conducted by GraphPad Prism 8.0 software or R statistical packages version 4.1.3. Significant differences between groups were examined by Two-sided Student’s t-test or Wilcoxon test. P values < 0.05 were considered significant.

No statistical method was used to predetermine the sample size. No data were excluded from the analyses. All mice were randomized before experiments. All experiments, including ChIP-PCR, RT-PCR, western blotting, luciferase reporter assays, immunofluorescence (IF), co-immunoprecipitation (Co-IP), metabolic assays, and in vitro/in vivo functional studies, were independently replicated two or three times with consistent results. The number of replicates performed is detailed in the respective figure legends. Immunohistochemistry (IHC) experiments were conducted on a cohort of 90 HCC tissue samples. The main findings from the high-throughput DNA/RNA sequencing data were validated using other independent sequencing datasets from public repositories. For studies with cell cultures, biological replicates are referred to as independent dishes of cells receiving the same treatment that were processed on the same days or on different days. For each of the following experiments, data collection and analysis were performed blinded to group allocation to ensure unbiased results: MS assays, CUT&Tag-seq, ATAC-seq, and RNA-seq Experiments: Samples were collected in our lab, and the assay procedures were conducted by a third-party company that did not have access to the sample group information. The analyses were performed by a specialist who did not have access to the exact group information until the analysis was complete. The assays and subsequent data analysis were conducted without knowledge of which samples belonged to which groups. In vitro/ in vivo functional Studies: Cells or animals used in functional assays were conducted by at least two persons. Experimental manipulations and outcome measurements were separately performed without revealing the group identities. IHC Staining: Tissue samples were anonymized. The immunohistochemistry staining and subsequent scoring were performed by researchers blinded to the clinical information and group allocation.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability

The sequencing data generated in this study have been deposited in the Gene Expression Omnibus (GEO) repository under the accession codes: GSE228075 (RNA-seq), GSE228072 (RNA-seq) GSE255638 (ATAC-seq), GSE255639 (ATAC-seq), GSE228074 (CUT&Tag-seq), GSE228071 (CUT&Tag-seq), GSE228069 (CUT&Tag-seq), GSE228255 (CUT&Tag-seq), GSE255634 (CUT&Tag-seq), GSE255636 (CUT&Tag-seq), GSE255640 (CUT&Tag-seq), GSE255641 (CUT&Tag-seq), and Sequence Read Archive (SRA) under accession codes: SRP502998 (CUT&Tag-seq), SRP503102 (CUT&Tag-seq and ATAC-seq). The raw data of LC-MS assays are available at integrated proteome resources (iProX) database under the project ID: PXD043284. The raw data of MS-based metabolomics have been deposited in the National Genomics Data Center under the accession code: PRJCA026639. The published data reused in this study includes GSE14520⁴⁸ (Expression profile by array), GSE54236⁴⁹ (Expression profile by array), and GSM1204463⁵³ (Bisulfite-Seq) from GEO database, and E-MTAB-7144 (Capture Hi-C) from ArrayExpress database⁵⁴. Several publicly available ChIP-seq raw data were downloaded from the ENCODE database (https://www.encodeproject.org/)⁴⁵, including:ENCFF750VZA⁵² (ChIP-seq), ENCFF000PJT⁵² (ChIP-seq), ENCSR848YWD⁵² (ChIP-seq), ENCFF442RQA⁵² (ChIP-seq), ENCFF492CBJ⁵² (ChIP-seq), ENCFF163SRP⁵² (ChIP-seq). The ATAC-seq data of non-tumoral tissues were obtained from BioProject PRJNA63443⁴⁵. The LIHC RNA-seq data from The Cancer Genome Atlas (TCGA) database was derived from the TCGA data portal [https://portal.gdc.cancer.gov/]⁴⁷. We downloaded ATAC-seq data (bigWig format) from the NIH GDC data portal [https://gdc.cancer.gov/about-data/publications/ATACseq-AWG]⁴⁴. The RNA-seq data of HCC samples from the International Cancer Genome Consortium (ICGC) database⁵⁰ was obtained from the ICGC data portal [https://dcc.icgc.org/]. From the UCSC Xena database, we downloaded a combined TCGA and Genotype-Tissue Expression (GTEx) dataset [https://xenabrowser.net/datapages/?cohort=TCGA%20TARGET%20GTEx&removeHub=https%3A%2F%2Fxena.treehouse.gi.ucsc.edu%3A443]⁵¹. The remaining data are available within the Article, Supplementary Information or Source Data file. Source data are provided with this paper.

References

Sakashita, A. et al. Endogenous retroviruses drive species-specific germline transcriptomes in mammals. Nat. Struct. Mol. Biol. 27, 967–977 (2020).
Article CAS PubMed PubMed Central Google Scholar
Modzelewski, A. J., Gan Chong, J., Wang, T. & He, L. Mammalian genome innovation through transposon domestication. Nat. Cell Biol. 24, 1332–1340 (2022).
Article CAS PubMed PubMed Central Google Scholar
Ye, M. et al. Specific subfamilies of transposable elements contribute to different domains of T lymphocyte enhancers. Proc. Natl Acad. Sci. USA 117, 7905–7916 (2020).
Article ADS CAS PubMed PubMed Central Google Scholar
Fueyo, R., Judd, J., Feschotte, C. & Wysocka, J. Roles of transposable elements in the regulation of mammalian transcription. Nat. Rev. Mol. Cell Biol. 23, 481–497 (2022).
Article CAS PubMed PubMed Central Google Scholar
Senft, A. D. & Macfarlan, T. S. Transposable elements shape the evolution of mammalian development. Nat. Rev. Genet. 22, 691–711 (2021).
Article CAS PubMed Google Scholar
Jiang, J. C. & Upton, K. R. Human transposons are an abundant supply of transcription factor binding sites and promoter activities in breast cancer cell lines. Mob. DNA 10, 16 (2019).
Article PubMed PubMed Central Google Scholar
Lynch-Sutherland, C. F., Chatterjee, A., Stockwell, P. A., Eccles, M. R. & Macaulay, E. C. Reawakening the developmental origins of cancer through transposable elements. Front. Oncol. 10, 468 (2020).
Article PubMed PubMed Central Google Scholar
Grundy, E. E., Diab, N. & Chiappinelli, K. B. Transposable element regulation and expression in cancer. FEBS J. 289, 1160–1179 (2022).
Article CAS PubMed Google Scholar
Burns, K. H. Transposable elements in cancer. Nat. Rev. Cancer 17, 415–424 (2017).
Article CAS PubMed Google Scholar
Jang, H. S. et al. Transposable elements drive widespread expression of oncogenes in human cancers. Nat. Genet. 51, 611–617 (2019).
Article CAS PubMed PubMed Central Google Scholar
Ito, J. et al. Endogenous retroviruses drive KRAB zinc-finger protein family expression for tumor suppression. Sci. Adv. 6, eabc3020 (2020).
Article ADS CAS PubMed PubMed Central Google Scholar
Mehdipour, P. et al. Epigenetic therapy induces transcription of inverted SINEs and ADAR1 dependency. Nature 588, 169–173 (2020).
Article ADS CAS PubMed Google Scholar
Wu, K. et al. SETDB1-mediated cell fate transition between 2C-like and pluripotent states. Cell Rep. 30, 25–36.e6 (2020).
Article CAS PubMed Google Scholar
Griffin, G. K. et al. Epigenetic silencing by SETDB1 suppresses tumour intrinsic immunogenicity. Nature 595, 309–314 (2021).
Article ADS CAS PubMed PubMed Central Google Scholar
Zhang, S. M. et al. KDM5B promotes immune evasion by recruiting SETDB1 to silence retroelements. Nature 598, 682–687 (2021).
Article ADS PubMed PubMed Central Google Scholar
Gu, Z. et al. Silencing of LINE-1 retrotransposons is a selective dependency of myeloid leukemia. Nat. Genet. 53, 672–682 (2021).
Article CAS PubMed PubMed Central Google Scholar
Tunbak, H. et al. The HUSH complex is a gatekeeper of type I interferon through epigenetic regulation of LINE-1s. Nat. Commun. 11, 5387 (2020).
Article ADS CAS PubMed PubMed Central Google Scholar
Sheng, W. et al. LSD1 ablation stimulates anti-tumor immunity and enables checkpoint blockade. Cell 174, 549–563.e19 (2018).
Article CAS PubMed PubMed Central Google Scholar
Kim, D., Kim, K. I. & Baek, S. H. Roles of lysine-specific demethylase 1 (LSD1) in homeostasis and diseases. J. Biomed. Sci. 28, 41 (2021).
Article CAS PubMed PubMed Central Google Scholar
Hosseini, A. & Minucci, S. A comprehensive review of lysine-specific demethylase 1 and its roles in cancer. Epigenomics 9, 1123–1142 (2017).
Article�� CAS PubMed Google Scholar
Lv, S. et al. Lysine demethylase KDM1A promotes cell growth via FKBP8-BCL2 axis in hepatocellular carcinoma. J. Biol. Chem. 298, 102374 (2022).
Article CAS PubMed PubMed Central Google Scholar
Ma, L. et al. LSD1-demethylated LINC01134 confers oxaliplatin resistance through SP1-induced p62 transcription in HCC. Hepatology 74, 3213–3234 (2021).
Article CAS PubMed Google Scholar
Huang, M. et al. Targeting KDM1A attenuates Wnt/beta-catenin signaling pathway to eliminate sorafenib-resistant stem-like cells in hepatocellular carcinoma. Cancer Lett. 398, 12–21 (2017).
Article CAS PubMed Google Scholar
Lei, Z. J. et al. Lysine-specific demethylase 1 promotes the stemness and chemoresistance of Lgr5(+) liver cancer initiating cells by suppressing negative regulators of beta-catenin signaling. Oncogene 34, 3188–3198 (2015).
Article CAS PubMed Google Scholar
Lee, M. G., Wynder, C., Cooch, N. & Shiekhattar, R. An essential role for CoREST in nucleosomal histone 3 lysine 4 demethylation. Nature 437, 432–435 (2005).
Article ADS CAS PubMed Google Scholar
Yang, M. et al. Structural basis for CoREST-dependent demethylation of nucleosomes by the human LSD1 histone demethylase. Mol. Cell 23, 377–387 (2006).
Article CAS PubMed Google Scholar
Garcia-Bassets, I. et al. Histone methylation-dependent mechanisms impose ligand dependency for gene activation by nuclear receptors. Cell 128, 505–518 (2007).
Article CAS PubMed PubMed Central Google Scholar
Metzger, E. et al. LSD1 demethylates repressive histone marks to promote androgen-receptor-dependent transcription. Nature 437, 436–439 (2005).
Article ADS CAS PubMed Google Scholar
Lin, Y. et al. The SNAG domain of Snail1 functions as a molecular hook for recruiting lysine-specific demethylase 1. EMBO J. 29, 1803–1816 (2010).
Article CAS PubMed PubMed Central Google Scholar
Thambyrajah, R. et al. GFI1 proteins orchestrate the emergence of haematopoietic stem cells through recruitment of LSD1. Nat. Cell Biol. 18, 21–32 (2016).
Article CAS PubMed Google Scholar
Ning, B. F. et al. Hepatocyte nuclear factor 4 alpha suppresses the development of hepatocellular carcinoma. Cancer Res. 70, 7640–7651 (2010).
Article CAS PubMed Google Scholar
Walesky, C. et al. Hepatocyte nuclear factor 4 alpha deletion promotes diethylnitrosamine-induced hepatocellular carcinoma in rodents. Hepatology 57, 2480–2490 (2013).
Article CAS PubMed Google Scholar
Ning, B. F. et al. Hepatocyte nuclear factor 4alpha-nuclear factor-kappaB feedback circuit modulates liver cancer progression. Hepatology 60, 1607–1619 (2014).
Article CAS PubMed Google Scholar
Xu, Q. et al. HNF4alpha regulates sulfur amino acid metabolism and confers sensitivity to methionine restriction in liver cancer. Nat. Commun. 11, 3978 (2020).
Article ADS CAS PubMed PubMed Central Google Scholar
Lau, H. H., Ng, N. H. J., Loo, L. S. W., Jasmen, J. B. & Teo, A. K. K. The molecular functions of hepatocyte nuclear factors - in and beyond the liver. J. Hepatol. 68, 1033–1048 (2018).
Article CAS PubMed Google Scholar
Tanaka, T. et al. Dysregulated expression of P1 and P2 promoter-driven hepatocyte nuclear factor-4alpha in the pathogenesis of human cancer. J. Pathol. 208, 662–672 (2006).
Article CAS PubMed Google Scholar
Lu, S. C. et al. Methionine adenosyltransferase 1A knockout mice are predisposed to liver injury and exhibit increased expression of genes involved in proliferation. Proc. Natl Acad. Sci. USA 98, 5560–5565 (2001).
Article ADS CAS PubMed PubMed Central Google Scholar
Heintzman, N. D. et al. Distinct and predictive chromatin signatures of transcriptional promoters and enhancers in the human genome. Nat. Genet. 39, 311–318 (2007).
Article CAS PubMed Google Scholar
Whyte, W. A. et al. Enhancer decommissioning by LSD1 during embryonic stem cell differentiation. Nature 482, 221–225 (2012).
Article ADS CAS PubMed PubMed Central Google Scholar
Wang, J. et al. Opposing LSD1 complexes function in developmental gene activation and repression programmes. Nature 446, 882–887 (2007).
Article ADS CAS PubMed Google Scholar
Shapson-Coe, A., Valeiras, B., Wall, C. & Rada, C. Aicardi-Goutieres Syndrome associated mutations of RNase H2B impair its interaction with ZMYM3 and the CoREST histone-modifying complex. PLoS ONE 14, e0213553 (2019).
Article CAS PubMed PubMed Central Google Scholar
Hiatt, S. M. et al. Deleterious, protein-altering variants in the transcriptional coregulator ZMYM3 in 27 individuals with a neurodevelopmental delay phenotype. Am. J. Hum. Genet. 110, 215–227 (2023).
Article CAS PubMed Google Scholar
Konermann, S. et al. Genome-scale transcriptional activation by an engineered CRISPR-Cas9 complex. Nature 517, 583–588 (2015).
Article ADS CAS PubMed Google Scholar
Corces, M. R. et al. The chromatin accessibility landscape of primary human cancers. Science 362, eaav1898 (2018).
Article ADS PubMed PubMed Central Google Scholar
Consortium, E. P. An integrated encyclopedia of DNA elements in the human genome. Nature 489, 57–74 (2012).
Article ADS Google Scholar
Cheneby, J. et al. ReMap 2020: a database of regulatory regions from an integrative analysis of Human and Arabidopsis DNA-binding sequencing experiments. Nucleic Acids Res. 48, D180–D188 (2020).
CAS PubMed Google Scholar
Cancer Genome Atlas Research Network. Comprehensive and integrative genomic characterization of hepatocellular carcinoma. Cell 169, 1327–1341.e23 (2017).
Article Google Scholar
Roessler, S. et al. A unique metastasis gene signature enables prediction of tumor relapse in early-stage hepatocellular carcinoma patients. Cancer Res. 70, 10202–10212 (2010).
Article CAS PubMed PubMed Central Google Scholar
Villa, E. et al. Neoangiogenesis-related genes are hallmarks of fast-growing hepatocellular carcinomas and worst survival. Results from a prospective study. Gut 65, 861–869 (2016).
Article CAS PubMed Google Scholar
Zhang, J. et al. International Cancer Genome Consortium Data Portal–a one-stop shop for cancer genomics data. Database 2011, bar026 (2011).
Article PubMed PubMed Central Google Scholar
Goldman, M. J. et al. Visualizing and interpreting cancer genomics data via the Xena platform. Nat. Biotechnol. 38, 675–678 (2020).
Article CAS PubMed PubMed Central Google Scholar
Partridge, E. C. et al. Occupancy maps of 208 chromatin-associated proteins in one human cell type. Nature 583, 720–728 (2020).
Article ADS CAS PubMed PubMed Central Google Scholar
Ziller, M. J. et al. Charting a dynamic DNA methylation landscape of the human genome. Nature 500, 477–481 (2013).
Article ADS CAS PubMed PubMed Central Google Scholar
Chesi, A. et al. Genome-scale capture C promoter interactions implicate effector genes at GWAS loci for bone mineral density. Nat. Commun. 10, 1260 (2019).
Article ADS PubMed PubMed Central Google Scholar

Download references

Acknowledgements

This work was supported by the National Natural Science Foundation of China (82372737 BS.W., 82203441 TT.J., 82273283 YZ.L., 82172905 BS.W., 81972209 BS.W., 82060041 YZ.L., and 82303079 XL.X.) and Shanghai Natural Science Foundation (21ZR1461500 BS.W.).

Author information

These authors contributed equally: Tiantian Jing, Dianhui Wei, Xiaoli Xu.

Authors and Affiliations

State Key Laboratory of Systems Medicine for Cancer, Shanghai Cancer Institute, Renji Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, 200032, China
Tiantian Jing, Dianhui Wei, Xiaoli Xu, Chengsi Wu, Lili Yuan, Yiwen Huang & Boshi Wang
Department of Medical Oncology, Fudan University Shanghai Cancer Center; Department of Oncology, Shanghai Medical College, Fudan University, Shanghai, 200032, China
Yizhen Liu
Institute of Health and Medical Technology, Hefei Institutes of Physical Science, Chinese Academy of Sciences, Hefei, 230031, China
Yanyi Jiang
University of Science and Technology of China, Hefei, 230026, China
Yanyi Jiang

Authors

Tiantian Jing
View author publications
You can also search for this author in PubMed Google Scholar
Dianhui Wei
View author publications
You can also search for this author in PubMed Google Scholar
Xiaoli Xu
View author publications
You can also search for this author in PubMed Google Scholar
Chengsi Wu
View author publications
You can also search for this author in PubMed Google Scholar
Lili Yuan
View author publications
You can also search for this author in PubMed Google Scholar
Yiwen Huang
View author publications
You can also search for this author in PubMed Google Scholar
Yizhen Liu
View author publications
You can also search for this author in PubMed Google Scholar
Yanyi Jiang
View author publications
You can also search for this author in PubMed Google Scholar
Boshi Wang
View author publications
You can also search for this author in PubMed Google Scholar

Contributions

TT.J., BS.W., DH.W., YY.J., XL.X., and YZ.L. designed the research, analyzed data, and wrote the manuscript; TT.J., DH.W., XL.X., CS.W., LL.Y., YW.H., YZ.L., YY.J., and BS.W. performed the experiments; TT.J., BS.W., DH.W., and YY.J. revised and edited the paper. YZ.L., YY.J., and BS.W. supervised the project.

Corresponding authors

Correspondence to Yizhen Liu, Yanyi Jiang or Boshi Wang.

Ethics declarations

Competing interests

The authors declare no competing interests.

Peer review

Peer review information

Nature Communications thanks Özgen Deniz, Derek Janssens, Gangadhara Sareddy, and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Supplementary Information

Peer Review File

Description of Additional Supplementary Files

Supplementary Data 1

Supplementary Data 2

Supplementary Data 3

Supplementary Data 4

Supplementary Data 5

Supplementary Data 6

Supplementary Data 7

Supplementary Data 8

Reporting Summary

Source data

Source Data

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and permissions

About this article

Cite this article

Jing, T., Wei, D., Xu, X. et al. Transposable elements-mediated recruitment of KDM1A epigenetically silences HNF4A expression to promote hepatocellular carcinoma. Nat Commun 15, 5631 (2024). https://doi.org/10.1038/s41467-024-49926-2

Download citation

Received: 12 September 2023
Accepted: 25 June 2024
Published: 04 July 2024
DOI: https://doi.org/10.1038/s41467-024-49926-2

Comments

By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.