Abstract

   Castration‐resistant prostate cancer (CRPC) remains the leading cause
   of mortality among prostate cancer patients. While second‐generation
   androgen receptor (AR) pathway‐targeted therapies, such as Abiraterone,
   have significantly improved survival outcomes, resistance to these
   treatments ultimately emerges, posing a critical challenge.
   Understanding the mechanisms underlying Abiraterone resistance is
   essential for developing strategies to enhance patient outcomes. In
   this study, a super‐enhancer (SE)‐driven transcriptional regulatory
   circuit is identified involving BCL6, NFIB, and SMAD3 that facilitates
   Abiraterone resistance in CRPC. Through comprehensive analyses of SE
   expression profiles in Abiraterone‐resistant CRPC cells and their
   parental counterparts, it is revealed that this circuit plays a pivotal
   role in resistance progression. Mechanistically, BCL6, NFIB, and SMAD3
   synergistically remodel the transcriptional landscape of resistant CRPC
   cells, driving resistance by regulating cholesterol biosynthesis and
   cell cycle pathways. The findings provide critical insights into the
   transcriptional dysregulation underlying Abiraterone resistance and
   highlight potential therapeutic strategies to counteract treatment
   resistance in CRPC, ultimately aiming to improve patient survival and
   quality of life.

   Keywords: abiraterone resistance, castration‐resistant prostate cancer,
   transcriptional regulation
     __________________________________________________________________

   This study identifies a super‐enhancer‐driven transcriptional
   regulatory circuit comprising BCL6, SMAD3, and NFIB that cooperate to
   drive cholesterol synthesis via SREBF2/HMGCR/FDFT1 activation and
   regulate CDK2/CCND3 for cell cycle control. Targeting this circuit with
   BI‐3802 or downstream inhibitors (Fatostatin/Lovastatin) overcomes
   abiraterone resistance in CRPC.

   graphic file with name ADVS-12-e01284-g004.jpg

1. Introduction

   Prostate cancer (PCa) is the most common tumor and the second leading
   cause of death in men in the United States and other Western
   countries.^[ [60]^1 ^] Androgen deprivation therapy (ADT) is the main
   method for the treatment of advanced prostate cancer. However, after an
   average of 2 years of ADT treatment, the tumor often progresses from
   castration‐sensitive prostate cancer (CSPC) to castration‐resistant
   prostate cancer (CRPC) with a poor prognosis.^[ [61]^2 ^]

   Abiraterone has been approved by the FDA as a first‐line treatment for
   metastatic castration‐resistant prostate cancer (mCRPC) both before
   chemotherapy and in cases where chemotherapy has failed. It functions
   by inhibiting CYP17A1, an enzyme critical for androgen precursor
   synthesis, targeting 17α‐hydroxylase and 17,20‐lyase, thereby impeding
   androgen production.^[ [62]^3 ^] However, the development of acquired
   resistance to Abiraterone significantly compromises treatment efficacy
   and patient survival. Studies have demonstrated an association between
   androgen receptor (AR) expression/activity and anti‐androgen resistance
   in CRPC.^[ [63]^4 , [64]^5 , [65]^6 , [66]^7 ^] Aberrant expression and
   splice variants of the AR gene are implicated in resistance mechanisms.
   The AR‐V7 androgen receptor splice variant is currently the most
   well‐defined resistance variant, lacking the androgen‐binding domain
   but retaining an activation domain that promotes tumor growth.^[ [67]^8
   , [68]^9 ^] In light of this, scholars are striving to identify new
   therapeutic targets and develop more effective drugs to eradicate CRPC,
   such as AR‐V7 degraders and the AKT inhibitor Ipatasertib.^[ [69]^10 ,
   [70]^11 ^] Nevertheless, the efficacy of these new treatments remains
   to be definitively established, and resolving Abiraterone resistance in
   CRPC patients is an urgent priority.

   Super‐enhancers (SEs) are large clusters of transcriptional enhancers
   characterized by exceptionally high levels of histone modifications
   associated with active transcription, such as H3K27ac and H3K4me1, as
   well as dense binding of transcription factors and mediator complex
   subunits.^[ [71]^12 , [72]^13 ^] SEs recruit a number of
   transcriptional activators, such as histone modification enzymes,
   transcription factors and cofactors, exhibiting stronger
   transcriptional activation capabilities than typical enhancers.^[
   [73]^14 ^] In contrast to typical enhancers, SEs orchestrate the
   expression of lineage‐defining genes by forming spatially organized 3D
   chromatin loops that facilitate physical interactions between distal
   regulatory elements and target promoters.^[ [74]^15 ^] Notably, SEs
   frequently govern cell identity programs and are enriched for
   disease‐associated variants, highlighting their pathological relevance
   in cancer and genetic disorders.^[ [75]^16 ^] Researches suggest that
   transcriptional regulatory loops driven by super‐enhancers promote
   tumorigenesis. For example, in T‐cell acute lymphoblastic leukemia,
   TAL1, GATA3 and RUNX1 form a core transcriptional regulatory loop that
   activates the MYB gene, enhancing the regulation of oncogenic
   programs.^[ [76]^17 ^] In multiple myeloma, SEs drive the formation of
   a transcriptional regulatory network involving transcription factors
   from the IRF, ETS, MEF2, E‐box and AP‐1 families, collaboratively
   promoting the malignant phenotype of the tumor.^[ [77]^18 ^] Tumor
   types that have been studied in association with core transcriptional
   regulatory loops also include esophageal squamous cell carcinoma,
   esophageal adenocarcinoma and neuroblastoma.^[ [78]^19 , [79]^20 ^]
   Furthermore, our research group previously elucidated the mechanism by
   which a SE‐driven core transcriptional regulatory loop composed of
   TCF4, KLF15 and NKX2.2 promotes the initiation and progression of Ewing
   sarcoma.^[ [80]^21 ^] In recent years, research on SEs in prostate
   cancer has gradually deepened, revealing their important roles in tumor
   initiation, progression and treatment. The AR, as the key driver in
   prostate cancer, can bind to SE regions to regulate the expression of
   downstream genes. The interaction between AR and SEs is of great
   significance for maintaining the proliferation and metastatic potential
   of cancer cells. FOXA1 is a pioneer factor that can reposition AR and
   other transcription factors to new super‐enhancers, altering gene
   expression patterns and promoting prostate cancer progression.^[
   [81]^22 ^] Impeding SWI/SNF‐mediated enhancer accessibility represents
   a promising therapeutic approach for enhancer‐addicted CRPC.^[ [82]^23
   ^]

   Ss are related to chemoresistance in various cancers, including
   small‐cell lung cancer, ovarian cancer, and breast cancer.^[ [83]^24 ,
   [84]^25 , [85]^26 ^] Genes driven by SEs also reduce tumor cell
   sensitivity to chemotherapy. For instance, enzalutamide resistance
   emerges via SE‐driven CHPT1 overexpression in CRPC, regulated by AR
   binding to distinct enhancers and lncRNA‐mediated H3K27ac reader BRD4
   interaction, highlighting SE‐targeted CHPT1 inhibition as a therapeutic
   strategy to overcome enzalutamide resistance.^[ [86]^27 ^]
   Additionally, mounting evidence suggests that inhibiting SE‐driven
   genes is a viable anti‐cancer strategy. For example, the small‐molecule
   inhibitor ML264, which targets KLF5, mitigates colorectal cancer
   progression both in vivo and in vitro.^[ [87]^28 ^] In liposarcoma, the
   bromodomain (BET) protein inhibitor ARV‐825 can inhibit the
   transcriptional regulatory loop composed of FOSL2, MYC and RUNX1—which
   holds regulatory dominance in liposarcoma and overcomes the resistance
   of liposarcoma to the clinical chemotherapeutic drug trabectedin.^[
   [88]^29 ^]

   This study performs ChIP‐Seq targeting H3K27ac on CRPC cell line C42B
   and its Abiraterone‐resistant counterpart C42B‐ABi, identifying
   resistance‐specific SE‐associated genes. The analysis revealed a
   transcriptional regulatory loop driven by super‐enhancers in C42B‐ABi,
   comprising BCL6, NFIB and SMAD3, which are highly expressed in C42B‐ABi
   cells. Our research indicates that the SE‐driven transcriptional
   regulatory loop composed of BCL6, NFIB and SMAD3 is instrumental in
   sustaining CRPC resistance to Abiraterone. These three transcription
   factors directly co‐bind to each promoter and enhancer, thereby
   activating mutual transcription and collaboratively remodeling the
   transcriptome of resistant CRPC cells. Mechanistically, this loop
   promotes resistance through two distinct pathways: 1) regulating the
   cell cycle to facilitate resistance; 2) activating cholesterol
   synthesis to promote the synthesis of testosterone and
   dihydrotestosterone, which sustains resistance. This study delves into
   the transcriptional aberrations occurring during the development of
   Abiraterone resistance in CRPC and posits that targeting this loop and
   its downstream targets presents a potential therapeutic strategy to
   overcome Abiraterone resistance in CRPC.

2. Results

2.1. Identification of the C42B‐ABi‐Specific SE‐Driven TF Loop

   In our previous work, we established the ABi‐resistant cell line
   C42B‐ABi that is significantly tolerated with Abiraterone treatment
   relative to the parent cell line C42B, the IC[50] are over 100 and 9.3
   µm respectively (Figure [89]1A). To rigorously validate the
   authenticity of C42B‐ABi, we conducted systematic comparative analyses.
   The results demonstrated differential abiraterone sensitivity between
   parental C42B and resistant C42B‐ABi cells using both in vitro culture
   systems and subcutaneous xenograft models in mice (Figure [90]S1A–F,
   Supporting Information). Systematically compared RNA‐Seq and
   whole‐exome sequencing (WES) data from these cell lines with published
   studies through comprehensive bioinformatic analysis (Figure [91]S1G–J,
   Supporting Information), confirming C42B‐ABi's resistance‐associated
   signatures. To identify C42B‐ABi specific SE related peaks, we
   performed ChIP‐seq targeting H3K27ac in C42B and C42B‐ABi cell lines,
   the result shows that two cell lines shared 13471 peaks (31.33%), 13097
   peaks are C42B specific (30.47%), and 16422 peaks are C42B‐ABi specific
   (38.2%) (Figure [92]1B). This indicated a different landscape of
   chromatin opening between C42B and C42B‐ABi cells. Interestingly, 674
   C42B‐ABi specific SEs were identified from 16422 C42B‐ABi specific
   peaks (Figure 1C; Table [93]S3, Supporting Information), while 403 SEs
   identified from 13097 C42B specific peaks (Figure [94]S2A, Supporting
   Information). There are 247 C42B‐ABi‐specific SE‐driven genes
   up‐regulated in C42B‐ABi cells. HALLMARK analysis shows that these
   genes are significantly enriched with signaling, survival and response
   to hormone pathways (Figure [95]S2B, Supporting Information), which are
   seen to be tightly involved in C42B resistant to Abiraterone.
   Furthermore, we found 18 of C42B‐ABi specific SE driven genes are
   transcription factors and up‐regulated in C42B‐ABi cells (Figure 1D;
   Table [96]S3, Supporting Information), which could be highly correlated
   with malignant progression of prostate cancer, such as ELF3, reported
   to orchestrate a positive feedback loop with ESE1 activates NF‐κB and
   drives prostate cancer progression.^[ [97]^30 ^] We further note that
   the top‐ranked genes ATF3 and NR2F1 have been implicated in prostate
   cancer. ATF3 directly inhibits AR transcriptional activity and
   downstream targets (e.g., PSA), while NR2F1 loss promotes tumor
   heterogeneity and enzalutamide resistance. However, their roles in
   abiraterone‐resistant CRPC remain unexplored. To unveil whether there
   is a transcriptional regulatory loop in Abiraterone‐resistant CRPC, we
   analyzed the RNA‐seq data of tumor tissues in 13 CRPC patients to
   conduct a Pearson correlation analysis of the 18 transcript factors.
   Our results suggest the potential existence of a transcriptional
   regulatory loop, composed of IRF1, FOS, JUNB, BCL6, SMAD3 and NFIB. In
   contrast, there is no correlation in the expression of the six
   transcription factors within 112 PC tumor tissues (Figure [98]1E).

Figure 1.

   Figure 1
   [99]Open in a new tab

   Screening of SE‐driven TF circuit specific to C42‐ABi cells. A)
   Treatment of C42B and C42B‐ABi cells with abiraterone, MTS assay
   performed to measure cell viability after 72 h. B) In C42B and C42B‐ABi
   cells, employing H3K27ac antibody for ChIP‐Seq to identify specific and
   shared peaks within both cell lines. C) Conducting ROSE analysis using
   the 16422 C42B‐Abi‐specific peaks to identify super enhancers. D) The
   volcano plot illustrates the differential expression of the 44
   transcription factors, which are part of the 674 super enhancers,
   between C42B‐ABi and C42B cells. E) The correlation analysis of the
   expression of 18 transcription factors in tissues from CRPC and CSPC
   patients. F,G) Expression heatmap. Using siRNA (F) or shRNA (G) to
   knock down one of the genes, the mRNA levels of the other indicated
   genes were measured. Statistical significance was determined using
   Two‐way ANOVA Dunnett's multiple comparisons test. n = 3 independent
   experiments. *, P < 0.05; **, P < 0.01;***, P < 0.001. H) Using shRNA
   knock down BCL6, SMAD3, or NFIB in C42B‐ABi, the protein expression of
   these three molecules was analyzed by Western blot. I, The mRNA
   expression of indicated genes in adjacent (n = 4), primary cancer (PC,
   n = 4), and castration‐resistant prostate cancer (CRPC, n = 4) groups.
   Two‐way ANOVA Dunnett's multiple comparisons test, n = 3, *, P < 0.05.
   J, Kaplan‐Meier analyses was conducted by categorizing the cases into
   BCL6, SMAD3, and NFIB simultaneously high expression and low expression
   group. K, Diagram of the SE‐driven TF circuit.

   To verify the specificity of this transcriptional loop, we performed
   ChIP‐seq targeting H3K27ac in five cases of CRPC tumor tissues and
   corresponding para‐cancer tissue, and three cases of CSPC tumor
   tissues. ChIP‐seq data from 15 cases of CRPC PDX model tumor tissues,
   24 cases of CSPC tumor tissues, and 28 cases of para‐cancer tumor
   tissues targeted H3K27ac in the GEO database were integrated.^[
   [100]^31 ^] The result showed that the signal of H3K27ac of the 6
   transcript factors in the enhancer region of the C42B‐ABi cell line or
   CRPC or CRPC‐PDX groups were higher than relative CSPC or adjacent
   groups (Figure [101]S2C, Supporting Information). To validate this
   regulatory circuit, we attenuated the expression of the six
   transcription factors individually using siRNA and measured the
   expression of the remaining transcript factors in C42B‐ABi and C42B
   cell lines. The mutual regulation is only present among BCL6, NFIB and
   SMAD3 in C42B‐ABi cells and does not exist in C42B cells
   (Figure [102]1F, Figure [103]S2D, Supporting Information). The
   validation results of shRNA are consistent with those of siRNA in
   C42B‐ABi (Figure [104]1G). Moreover, knocking down either BCL6, NFIB or
   SMAD3 results in a decrease in the protein levels of these three
   proteins in C42B‐ABi cells (Figure [105]1H). Conversely,
   over‐expressing each of the three proteins individually leads to an
   increase in the three proteins (Figure [106]S2E, Supporting
   Information).

   Importantly, we investigated the differential expression of BCL6, NFIB
   and SMAD3 in the tumor tissue samples and cell lines, with the findings
   indicating that these three genes are most highly expressed in CRPC and
   C42B‐ABi cells (Figure [107]1I, Figure [108]S2F,G, Supporting
   Information). And the correlation between BCL6, NFIB and SMAD3 is only
   present in CRPC tumor tissues and not in CSPC tumor tissues or other
   adenocarcinoma tissues (e.g., esophageal cancer and breast cancer)
   (Figure [109]S2H, Supporting Information). This indicates that the
   transcriptional loop composed of BCL6, NFIB and SMAD3 has better
   specificity and correlation in CRPC. Furthermore, we examined the
   correlation between the expression of BCL6, NFIB and SMAD3 and patient
   survival by proccessing the data of the public article.^[ [110]^32 ^]
   The prognosis of CRPC patients is poorer when all three genes are
   simultaneously highly expressed than when they are expressed at a low
   level (Figure [111]1J), and also poorer than when BCL6 or SMAD3 is
   individually highly expressed (Figure [112]S2I, Supporting
   Information). These results demonstrate that BCL6, NFIB and SMAD3 form
   a specific transcriptional regulatory circuit in C42B‐ABi cell line
   (Figure [113]1K).

2.2. A Direct Transcriptional Activation Effect among BCL6, SMAD3 and NFIB

   To ascertain the regulatory mechanism between BCL6, SMAD3 and NFIB, we
   conducted CUT&Tag using specific antibodies individually targeting the
   three transcription factors and the epigenetic modification marker
   proteins H3K27ac, H3K4me1 and H3K4me3. The results demonstrate that
   BCL6, SMAD and NFIB collectively bind to each other's promoters and
   candidate enhancers (Figure 2A; Figure [114]S3A,B, Supporting
   Information). Importantly, the maximum distance between the motifs of
   the three transcription factors in co‐binding region is 115 bp
   (SAMD3‐E1), and the minimum is only 23 bp (NFIB‐E1), supporting that
   the three transcription factors collectively bind to the same
   chromosomal location (Figure 2B; Figure [115]S3C, Supporting
   Information). Notably, we also compared the binding differences of the
   three transcription factors at representative enhancer regions between
   two cell lines using comparative ChIP‐PCR analysis. In C42B‐ABi cells,
   this regulatory loop exhibits higher enrichment at the SMAD3‐CE2/CE3,
   BCL6‐CE3, and NFIB‐CE1 loci compared to C42B cells. Specifically, in
   C42B cells, NFIB and SMAD3 exhibit no enrichment at these regions,
   while BCL6 shows partial enrichment (Figure [116]S3D–G, Supporting
   Information). These findings provide critical evidence demonstrating
   the specificity relevance of this regulatory circuit in C42B‐ABi cells.

   To validate the co‐binding of three transcription factors at the
   promoter and enhancer regions at the SMAD3 gene locus. We identified a
   promoter where three transcription factors co‐bind, named SMAD3‐P,
   along with three co‐binding candidate enhancers designated as
   SMAD3‐CE1, SMAD3‐CE2, and SMAD3‐CE3 (Figure [117]2A). These sequences
   were individually incorporated into the luciferase reporter vector and
   transfected into C42B‐ABi cells. The findings revealed a substantial
   increase in luciferase activity for SMAD3‐P, SMAD3‐CE1, SMAD3‐CE2, and
   SMAD3‐CE3 compared to the control, while knockdown of BCL6, SMAD3 and
   NFIB resulted in a significant decrease in luciferase activity
   (Figure [118]2C,D). Furthermore, targeted sgRNA sequences were designed
   against the SMAD3‐CE2 and CE3, the sequence with the higher activity,
   and integrated into the pLentiCRISPR V2 vector. The result indicated
   that the destruction of SMAD3‐CE2 and CE3 can not only suppress the
   expression of mRNA and protein levels of SMAD3 but also BCL6 and NFIB
   (Figure [119]2E,F). In conclusion, BCL6, SMAD3 and NFIB activate their
   own as well as each other's transcriptional expression by directly
   binding at mutual promoters and enhancers. To investigate the roles of
   these two enhancers in mediating the regulatory loop's control over
   SMAD3. We performed CUT&Tag experiments using antibodies against BCL6,
   NFIB, SMAD3, and H3K27ac in the control group (Vector) and
   SMAD3‐CE2/CE3 mutant groups (CE2‐sg and CE3‐sg). The results are shown
   in Figure [120]2G. The binding of all three transcription factors
   (BCL6, NFIB, and SMAD3) at the SMAD3‐CE2 and CE3 loci was significantly
   reduced compared to the control group. Concurrently, H3K27ac signals
   (marking chromatin openness) at these loci also showed varying degrees
   of attenuation following SMAD3‐CE2/CE3 mutations. These results
   indicate that mutations in the CE2/CE3 regions disrupt the co‐binding
   of the three transcription factors and alter chromatin accessibility at
   these enhancer loci. This finding is critical for demonstrating
   enhancer‐mediated regulation of SMAD3 by the three transcription
   factors. Furthermore, functional assays revealed that SMAD3‐CE2/CE3
   mutations led to increased sensitivity to abiraterone (Figure [121]2H)
   and impaired cell proliferation (Figure [122]2I). These phenotypic
   changes mimic the effects of SMAD3 knockdown, suggesting that CE2/CE3
   mutations functionally inactivate SMAD3‐mediated regulatory activity.
   Due to the co‐binding of three transcription factors at a shared
   enhancer or promoter elements and their concerted activation of
   transcription, we speculate a potential existence of protein‐level
   interaction among them. Co‐IP assays were performed to verify this, and
   Figure [123]2G illustrates that BCL6, SMAD3 and NFIB exhibit mutual
   protein‐level interactions. These results indicate that BCL6, SAMD3,
   and NFIB compose stable complexes in Abiraterone‐resistant CRPC.

Figure 2.

   Figure 2
   [124]Open in a new tab

   Validating the mechanism of the SE‐driven TF circuit. A) Integrative
   Genomics Viewer (IGV) plots of CUT&Tag showing co‐occupancy of BCL6,
   SMAD3, and NFIB at the promoter and enhancer of the SMAD3 gene locus.
   B) A schematic representation of the nearest distance pattern of BCL6,
   NFIB, ans SMAD3 motifs at SMAD3‐CE2. C, D, Promoter (C) and enhancer
   (D) activities were measured by luciferase reporter assays in C42B‐ABi
   cells in either the presence or absence of knockdown indicated TF.
   Two‐way ANOVA Dunnett's multiple comparisons test. n = 3. ****, P <
   0.0001. E) The mRNA expression of BCL6, SMAD3, and NFIB were measured
   on C42B‐ABi cells transfected either vector or sgRNA targeting
   SMAD3‐CE2. Two‐way ANOVA Dunnett's multiple comparisons test. n = 3.
   ****, P < 0.0001. F) The protein expression of BCL6, SMAD3, and NFIB
   were measured on C42B‐ABi cells transfected either vector or sgRNA
   targeting SMAD3‐CE2. G) C42B‐ABi cells transfected either vector or
   sgRNA targeting SMAD3‐CE2 and SMAD3‐CE3, integrative Genomics Viewer
   (IGV) plots of CUT&Tag showing co‐occupancy of BCL6, SMAD3, and NFIB at
   the enhancers of the SMAD3 gene locus. H) Abiraterone dose‐response
   curves for C42B‐ABi cells transfected either vector or sgRNA targeting
   SMAD3‐CE2 and SMAD3‐CE3. I) C42B‐ABi cells transfected either vector or
   sgRNA targeting SMAD3‐CE2 and SMAD3‐CE3, cell proliferation were
   detected. One‐way ANOVA Dunnett's multiple comparisons test. n = 3.
   ****, P < 0.0001. J) Overexpress Flag‐tagged BCL6, SMAD3, and NFIB in
   C42B‐ABi cells individually, perform co‐IP experiments using a Flag
   antibody, and detect BCL6, SMAD3, and NFIB via Western Blot. IgG serves
   as the negative control.

2.3. BCL6, SMAD3, and NFIB Cooperatively Orchestrate the Transcriptome of
C42B‐ABi Cells

   To unravel the transcriptional regulatory mechanisms of BCL6, SMAD3 and
   NFIB in Abiraterone‐resistant CRPC, we analyzed the epigenomic
   characteristics of their occupancy in C42B‐ABi cells. Based on the
   co‐occupancy of H3K27ac and H3K4me3, 6104 promoters were identified,
   while 30 327 enhancers were identified by H3K27ac and H3K4me1
   co‐occupying. The quantity of enhancers far exceeds that of promoters.
   The number of concurrent bindings of BCL6, SMAD3 and NFIB either in
   promoter or enhancer is classified into four groups: trio, dual, solo,
   and none. Notably, 71.9% of promoters and 48.4% of enhancers are
   co‐binding by at least two factors individually (Figure [125]3A,B;
   Table [126]S3, Supporting Information), indicating a profound
   synergistic occupancy among the three transcription factors. The
   intensity of H3K27ac indicates that chromatin accessibility is highest
   in the trio group for both promoters and enhancers (Figure [127]3C).
   Correspondingly, the expression of target genes is also highest in the
   trio cohort compared with the other three (Figure [128]3D). Moreover,
   the enhancers occupied by the trio group demonstrate a higher
   proportion of overlap with super‐enhancers (Figure [129]3E). All of the
   results indicated the transcription factors BCL6, SMAD3 and NFIB
   synergistically activate the transcription of target genes in C42B‐ABi
   cells.

Figure 3.

   Figure 3
   [130]Open in a new tab

   Transcriptional cooperativity between BCL6, SMAD3, and NFIB. A) The pie
   chart show the co‐binding overview of BCL6, SMAD3, and NFIB at
   promoters and enhancers. B) Heatmaps of CUT&Tag signals of indicated
   factors in C42B‐Abi cells, stratified by different combinatorial
   binding patterns. C) The signal intensity of H3K27ac among the peaks of
   indicated groups. D) mRNA expression of target genes regulated by the
   peaks of indicated groups in C42B‐ABi cells. E) The overlapping of
   indicated groups of peaks with super‐enhancers in C42B‐ABi cells. F)
   Volcano plots of RNA‐Seq after individually knocking down BCL6, SMAD3,
   and NFIB in C42B‐ABi cells. Up regulated in siTFs (BCL6, NFIB, SMAD3),
   log[2] fold‐change>0.584, p‐value < 0.05; down regulated in siTFs
   (BCL6, NFIB, SMAD3), log[2] fold‐change < ‐0.584 and p‐value < 0.05. G)
   Genes down‐regulated by knocking down BCL6 are subjected to GSEA
   enrichment analysis with the RNA‐Seq data of SMAD3 or NFIB knockdown.
   Similarly, down‐regulated genes from the SMAD3 or NFIB knockdown are
   analyzed in the same manner. H) The overlap genes down‐regulated by
   knocking down BCL6, NFIB, and SMAD3 are subjected to GSEA enrichment
   analysis with the RNA‐Seq data of C42B‐ABi and C42B. I) Venn diagram of
   down‐regulated genes following silencing of each single TFs and binding
   by them. J) Gene Ontology pathway enrichment analysis of overlapping
   genes.

   To further substantiate the synergistic regulation of transcripts by
   the three transcription factors. We employed siRNA‐mediated knockdown
   of BCL6, SMAD3 and NFIB for RNA‐Seq analysis in C42B‐ABi cells (Figure
   3F; Table [131]S3, Supporting Information). Interestingly, Gene Set
   Enrichment Analysis (GSEA) showed genes that were down‐regulated by
   silencing of BCL6 were significantly enriched in genes that were also
   down‐regulated following silencing of the other two factors (NFIB or
   SMAD3) (Figure [132]3G). The same phenomenon was also observed in NFIB
   and SAMD3, suggesting that the three transcription factors exhibit
   remarkable similarity in orchestrating transcripts in C42B‐ABi cells.
   Correspondingly, the up‐regulated genes also show significant
   similarity (Figure [133]S4A, Supporting Information). Moreover, the
   genes that are commonly down‐regulated after silencing the three
   transcription factors are significantly enriched in the gene set
   characterized by high expression of C42B‐ABi (Figure [134]3H). These
   results suggest that BCL6, SAMD3 and NFIB collaboratively remodel
   transcription to facilitate ABi‐resistant in CRPC.

   Additionally, we investigated whether there is a regulatory
   relationship between this transcriptional regulatory circuit and AR,
   the driver gene of prostate cancer. First, we knocked down AR and found
   that the mRNA and protein levels of the three transcription factors
   were upregulated, while down‐regulated with AR over expression (Figure
   [135]S4B–D, Supporting Information). We compared the expression levels
   of AR and its target protein PSA between C42B and C42B‐ABi cells and
   found their protein level significantly decreased in the
   Abiraterone‐resistant cells (Figure [136]S4E, Supporting Information),
   which is a consensus about the treatment of androgen receptor signaling
   inhibitors has led to an increase of CRPC de‐differentiate into
   AR‐negative disease.^[ [137]^4 , [138]^6 , [139]^33 ^] And silencing
   BCL6, SMAD3 and NFIB, the protein levels of AR were up‐regulated
   (Figure [140]S4F, Supporting Information), suggesting that there is an
   inverse regulatory relationship between AR and this transcriptional
   regulatory circuit. And these results demonstrated that inhibiting the
   transcriptional loop could recover the sensitivity of Abiraterone via
   increasing the expression of AR in C42B‐ABi cells to some extent.

2.4. The factors of the SE‐Driven TF Loop Promote the Malignant Phenotype of
ABi‐Resistant CRPC

   To investigate the biological processes that the transcriptional
   regulatory network promotes in conferring drug resistance, we employed
   CUT&Tag and RNA‐Seq data for BCL6, SMAD3 and NFIB, and found 247 genes,
   representing downstream target genes jointly transcriptional regulated
   by BCL6, SMAD3 and NFIB (Figure 3I; Table [141]S3, Supporting
   Information). The top two enriched pathways are the “Lipid Biosynthetic
   Process” and “Regulation of Mitotic Cell Cycle” according to Gene
   Ontology pathway enrichment analysis of the 247 genes (Figure [142]3J),
   suggesting a promotion of proliferation with the loop‐mediated
   facilitation of Abiraterone resistance in CRPC.

   Individually silencing BCL6, SMAD3 and NFIB using siRNA or shRNA
   resulted in inhibited proliferation and migration capabilities of the
   C42B‐ABi cells (Figure [143]4A–C). Furthermore, following castration
   surgery in nude mice, subcutaneous implantation of C42B‐ABi cells with
   stable knockdown of BCL6, SMAD3 and NFIB significantly suppressed tumor
   proliferation compared to the control group (Figure [144]4D,E). Tumor
   tissues were isolated for Western blot and Immunohistochemistry
   analysis, revealing a protein‐level inter‐regulatory relationship among
   the three factors (Figure [145]4F,G). We also observed a negative
   regulatory relationship between this loop and AR/PSA in subcutaneous
   tumor tissues (Figure [146]S4G,H, Supporting Information). The findings
   indicate that the SE‐driven TF loop composed of BCL6, SMAD3 and NFIB
   are promoting the malignant progression of Abiraterone‐resistant CRPC.

Figure 4.

   Figure 4
   [147]Open in a new tab

   BCL6, SMAD3, or NFIB loss inhibits cell proliferation and enhances
   sensitivity to abiraterone in C42B‐Abi cells. A) The knockdown
   efficiency of siRNA and shRNA. One‐way ANOVA Dunnett's multiple
   comparisons test. n = 3. **, P < 0.01. B,C), Silencing BCL6, SMAD3, and
   NFIB individually by siRNA or shRNA, cell proliferation (B) and cell
   migration (C) were detected. One‐way ANOVA Dunnett's multiple
   comparisons test. n = 3. **, P<0.01; ***, P < 0.001; ****, P < 0.0001.
   D) Inoculate castrated nude mice with C42B‐ABi cells that have knocked
   down BCL6, SMAD3, or NFIB. At the terminal of the experiments, the mice
   were sacrificed, and then the tumors were excised and weighed. One‐way
   ANOVA Dunnett's multiple comparisons test. n = 6. **, P < 0.01; ****, P
   < 0.0001. E) The tumor sizes were recorded every day, starting from the
   tenth day post‐inoculation. One‐way ANOVA Dunnett's multiple
   comparisons test. n = 3. **, P < 0.01; ***, P < 0.001. F) The indicated
   proteins in tumor tissue were detected by Western blots. G)
   Immunohistochemical (IHC) analysis of the indicated proteins in tumor
   tissue. H) Abiraterone dose‐response curves for C42B‐ABi cells with
   BCL6, SMAD3, or NFIB knockdown. I) Overexpress Flag‐tagged BCL6, SMAD3,
   and NFIB in C42B cells individually and detect the protein level of
   Flag via Western Blot assay. J) Overexpress BCL6, SMAD3, and NFIB in
   C42B cells individually, cell proliferation were detected. One‐way
   ANOVA Dunnett's multiple comparisons test. n = 3. ***, P < 0.001. K)
   Abiraterone dose‐response curves for C42B cells with BCL6, SMAD3, or
   NFIB over‐expression. L) BI‐3802 dose‐response curves for C42B and
   C42B‐ABi cells after 3 days. M) C42B‐ABi cells were treated with
   BI‐3802, and then the indicated proteins were detected by Western
   blots. N) The CI value of BI‐3802 combining with abiraterone were
   smaller than one, which indicated synergy effect on BI‐3802 with
   abiraterone. CI = 1 means additive effect, CI<1 means synergism, and
   CI>1 means antagonism. Fraction affected (Fa) is the proportion of
   cells affected by combination treatment.

Figure 5.

   Figure 5
   [148]Open in a new tab

   The super‐enhancer‐driven transcription factor loop activates the
   cholesterol biosynthesis. A) Reactome enrichment analysis of the genes
   enriched in the“Lipid biosynthetic process”of Figure [149]3J. B) the
   RNA‐Seq data of BCL6, SMAD3, and NFIB knockdown are subjected to GSEA
   enrichment analysis with the gene sets associated with cholesterol
   homeostasis. C,D) Intracellular cholesterol, testosterone (T), and
   dihydrotestosterone (DHT) levels measured following the individual
   knockdown of BCL6, SMAD3, and NFIB in C42B‐ABi cells (C) and tumor
   xenografts in mice (D). One‐way ANOVA Dunnett's multiple comparisons
   test. n = 3. *, P <0.05; **, P < 0.01; ***, P < 0.001. E) Heatmaps
   illustrating the mRNA expression and binding patterns of the gene
   collection enriched in cholesterol homeostasis. F,G) The mRNA (F) and
   protein (G) expression level of indicated genes were detected in
   C42B‐ABi cells following silencing of BCL6, SMAD3, and NFIB. Two‐way
   ANOVA Dunnett's multiple comparisons test. n = 3. *, P <0.05; **, P <
   0.01; ***, P < 0.001; ****, P < 0.0001. H) The protein expression of
   SREBF2, HMGCR, and SCAP were measured following the over‐expression of
   BCL6, SMAD3, or NFIB in C42B‐ABi cells. I,J) Silencing SREBF2 by siRNA
   in C42B‐ABi cells (I), the cell proliferation (J) was detected. One‐way
   ANOVA Dunnett's multiple comparisons test. n = 3. **, P < 0.01; ***, P
   < 0.001; ****, P < 0.0001. K) Abiraterone dose‐response curves for
   C42B‐ABi cells after 3 days of SREBF2 knockdown. One‐way ANOVA
   Dunnett's multiple comparisons test. n = 3. *, P <0.05; **, P < 0.01;
   ***, P < 0.001. L) In vitro MTS proliferation assay of prostate cancer
   cells in the presence of different doses of fatostatin (SREBF2
   inhibitor). M) Intracellular cholesterol, testosterone (T), and
   dihydrotestosterone (DHT) levels measured with fatostatin treatment in
   C42B‐ABi cells. One‐way ANOVA Dunnett's multiple comparisons test. n =
   3. *, P <0.05; **, P < 0.01. N) Overexpress SREBF2 after silencing
   BCL6, SMAD3, and NFIB by shRNA in C42B‐ABi cells, the cell
   proliferation was detected. Two‐way ANOVA Dunnett's multiple
   comparisons test. n = 3. *, P <0.05; ***, P < 0.001; ****, P < 0.0001.

   Subsequently, we investigated the regulatory roles of the three
   transcription factors on Abiraterone resistance. The sensitivity of
   C42B‐ABi to Abiraterone was heightened upon individual knockdown of
   BCL6, SMAD3 and NFIB (Figure [150]4H). Corresponding, over‐expression
   of BCL6, SAMD3 and NFIB in the C42B can promote cell proliferation,
   inhibit the cell migration and the sensitivity to Abiraterone
   (Figure [151]4I–K; Figure [152]S4I, Supporting Information). BI‐3802, a
   BCL6 inhibitor, significantly decreases the cell viability of C42B‐ABi
   and C42B. Due to higher BCL6 expression, compared to C42B, BI‐3802
   exhibits a more notable inhibitory effect on C42B‐ABi (Figure [153]4L).
   Moreover, the protein levels of BCL6, SAMD3 and NFIB were suppressed
   with BI‐3802 treatment (Figure [154]4M). Notably, BI‐3802 can
   synergistically cooperate with Abiraterone to suppress the viability of
   C42B‐ABi cells (Figure [155]4N). Taken together, the transcription
   factor loop components driven by SE, namely BCL6, SMAD3 and NFIB,
   contribute to fortifying the resistance of C42B‐ABi against
   Abiraterone. Conversely, targeted suppression of this loop could
   increase the sensitivity of C42B‐ABi to Abiraterone.

2.5. The SE‐Driven TF Loop Activates Cholesterol Homeostasis

   We have demonstrated the role of the transcriptional regulatory network
   in driving Abiraterone resistance in C42B‐ABi cells. However, the
   specific mechanism through which it operates remains unknown. The
   pathway enrichment analysis suggests that BCL6, SMAD3 and NFIB
   collaboratively regulate the pathway of lipid biosynthetic processes
   (Figure [156]3I). To validate this, we examined the lipid droplet
   abundance within cells following the knockdown of BCL6, SMAD3 and NFIB
   individually, revealing a significant decrease in lipid droplets
   (Figure [157]S5A,B, Supporting Information). The genes enriched in the
   “lipid biosynthetic processes” were further subjected to Reactome
   enrichment analysis, and the results indicated a significant enrichment
   in the “cholesterol biosynthesis pathway” (Figure [158]5A). HALLMARK
   analysis revealed that the overlap genes down‐regulated after the
   knock‐down of three transcription factors are significantly enriched in
   cholesterol homeostasis (Figure [159]5B). Genes associated with
   cholesterol homeostasis are significantly enriched in the gene set with
   high expression in C42B‐ABi compared to C42B cells (Figure [160]S5C,
   Supporting Information). Cholesterol is an important component of cell
   membranes, and it acts as a precursor to bile acids, steroid hormones,
   and 7‐dehydrocholesterol to regulate cell proliferation and survival.
   Several studies have shown that abnormal cholesterol homeostasis,
   including the dynamic processes of uptake, efflux, synthesis, and
   degradation, can promote the malignant progression of tumors.^[
   [161]^34 ^] Our findings suggest that the SE‐driven TF loop may
   regulate cholesterol homeostasis to sustain the resistance of
   Abiraterone. To validate this, we quantified the intracellular levels
   of cholesterol and its synthetic products testosterone and
   dihydrotestosterone, which activate the androgen receptor (AR)
   signaling pathway. The results indicate a significant decrease in
   cholesterol, testosterone, and dihydrotestosterone following the
   individual knockdown of BCL6, SMAD3 and NFIB (Figure [162]5C). A
   similar phenomenon was observed in tumor xenografts in mice; however,
   not in the serum of the mice (Figure 5D; Figure [163]S5D, Supporting
   Information). Elevated levels of cholesterol, testosterone, and
   dihydrotestosterone were observed upon individual overexpression of
   each TF in C42B cells, and all three biomarkers were significantly
   enrichment in C42B‐ABi cells (Figure [164]S5E,F, Supporting
   Information). These results illustrated that the SE‐driven TF loop
   promotes an increase in intratumoral cholesterol, testosterone, and
   dihydrotestosterone.

   To elucidate the specific mechanism by which the SE‐driven TF loop
   regulates cholesterol homeostasis, we compiled genes enriched in
   cholesterol homeostasis and discovered these genes involved in
   cholesterol synthesis, cholesterol uptake, and lipogenesis, with genes
   related to cholesterol synthesis being the most prevalent
   (Figure [165]5E). With low cholesterol levels, SCAP binds to sterol
   regulatory element binding proteins (SREBP) and transports from the ER
   to the Golgi.^[ [166]^35 ^] The SREBP is then proteolytically cleaved
   and enters the nucleus to activate the transcription of target genes,
   including the core regulators of the mevalonate pathway HMGCR, FDFT1,
   and cholesterol uptake receptors LDLR.^[ [167]^36 ^] Thus, we validated
   a subset of genes associated with cholesterol synthesis and uptake. Our
   findings reveal a significant reduction in mRNA levels of these genes
   upon knockdown of BCL6, SMAD3 and NFIB (Figure [168]5F). Moreover,
   there is a substantial decrease observed in the protein abundance of
   SCAP, SREBF2 and HMGCR following the depletion of three transcription
   factors (Figure [169]5G). Conversely, SCAP, SREBF2 and HMGCR were
   up‐regulated when the three transcription factors were over‐expressed
   (Figure [170]5H). CUT&Tag data indicate that BCL6, SMAD3 and NFIB
   collectively bind to the promoter or enhancer elements of these target
   genes (Figure [171]S5G, Supporting Information). These data demonstrate
   that BCL6, SMAD3 and NFIB transcriptionally activate genes associated
   with cholesterol synthesis directly, thereby promoting cholesterol
   biosynthesis.

   To validate the regulation of cholesterol‐mediated SE‐driven TF loop on
   Abiraterone resistance in C42B‐ABi cells, we focused on the SREBF2,
   given its role as a transcription factor governing genes associated
   with cholesterol synthesis. First, in C42B‐ABi cells, employing siRNA
   to silence SREBF2 led to a notable attenuation in cell proliferation
   and heightened sensitivity to Abiraterone (Figure [172]5I–K).
   Fatostatin, the inhibitor of SREBF2, significantly decreased cell
   viability of C42B and C42B‐ABi cells (Figure [173]5L). As expected,
   with the treatment of fatostatin, there was a time‐dependent decrease
   of cholesterol, testosterone, and dihydrotestosterone in C42B‐ABi cells
   (Figure [174]5M). Moreover, the protein level of SREBF2 and its targets
   HMGCR, FDFT1, and SCAP also decreased with fatostatin treatment (Figure
   [175]S5H, Supporting Information). Furthermore, upon individually
   silencing BCL6, SMAD3 and NFIB, overexpression of SREBF2 was
   demonstrated to ameliorate the cell proliferation suppression, as same
   as the down‐regulation of HMGCR and FDFT1, caused by the knockdown of
   these three transcription factors (Figure 5N; Figure [176]S5I,
   Supporting Information). Meanwhile, Lovastatin, the inhibitor of HMGCR
   also inhibits cell viability and synergistically coordinates with
   Abiraterone in suppressing cell proliferation (Figure [177]S5J,K,
   Supporting Information).

2.6. The SE‐Driven TF Loop Mediates Cell Cycle Arrest in C42B‐ABi Cells

   We discovered that the SE‐driven TF loop promotes the progression of
   Abiraterone resistance by regulating the cell cycle (Figure [178]3I).
   To validate this, we individually silenced BCL6, SMAD3 and NFIB in
   C42B‐ABi cells, the cell cycle exhibited a notable blockade in the
   G0/G1 phase (Figure [179]6A). Consistent with the experimental results,
   reactome analysis of genes clustered in “Regulation of Mitotic Cell
   Cycle” in figure [180]3I showed significant enrichment in the G1/S
   phase transition (Figure [181]6B). To reveal the mechanism, we examined
   the genes enriched in the “Regulation of Mitotic Cell Cycle” pathway of
   figure [182]3I according to RNA‐seq results (Figure [183]6C). By
   knocking down three transcription factors, we validated that the mRNA
   level of these genes was reduced (Figure [184]6D). We observed that the
   CDK2 and CCND3, which facilitate the transition of cells from the G1
   phase to the S phase, are significantly down‐regulated after BCL6,
   SMAD3 and NFIB knockdown. Therefore, we hypothesize that silencing the
   loop may reduce the expression of CCND3 and CDK2 by regulating their
   transcription, resulting in the G0/G1 phase arrest in C42B‐ABi cells.

Figure 6.

   Figure 6
   [185]Open in a new tab

   The super‐enhancer‐driven transcription factor loop mediates cell cycle
   arrest in the G0/G1 phase. A) Propidium iodide staining for cell cycle
   analysis. The bar chart represents the statistical data for each phase.
   B) Reactome enrichment analysis of the genes enriched in the“regulation
   of mitotic cell cycle”of Figure [186]3J. C) Heatmap showing the mRNA
   expression of the genes enriched in the “regulation of mitotic cell
   cycle”. D) The mRNA expression levels of CDK2 and CCND3 were assessed
   following the knockdown of BCL6, SMAD3, or NFIB in C42B‐ABi cells. E)
   IGV showing indicated protein binding at the CDK2 and CCND3 loci. F,G)
   The protein expression level of CDK2 and CCND3 were measured following
   the silence (F) or overexpress (G) of BCL6, SMAD3, or NFIB in C42B‐ABi
   cells. H) Silencing SREBF2 by siRNA in C42B‐ABi cells, the cell cycle
   was detected. One‐way ANOVA Dunnett's multiple comparisons test. n = 3.
   **, P < 0.01; ***, P < 0.001. I) C42B‐Abi cells were treated with 10 µm
   lovastatin (the inhibitor of HMGCR) for 12, 24, and 48 h, and the cell
   cycle was detected. One‐way ANOVA Dunnett's multiple comparisons test.
   n = 3. **, P < 0.01; ***, P < 0.001,****, P < 0.0001. J) Overexpress
   SREBF2 after silencing BCL6, SMAD3, and NFIB by shRNA in C42B‐ABi
   cells, the proportion of G0/G1 was detected. One‐way ANOVA Dunnett's
   multiple comparisons test. n = 3. *, P < 0.015, ***, P < 0.001,****, P
   < 0.0001.

   The CUT&Tag data demonstrate direct binding of the three transcription
   factors to the promoter and enhancer elements of the key regulatory
   genes CDK2 and CCND3 (Figure [187]6E). The protein expression of both
   CDK2 and CCND3 was inhibited following the individual knockdown of
   BCL6, SMAD3 and NFIB in C42B‐ABi cells (Figure [188]6F). Conversely,
   overexpression leads to an elevation in the protein expression of CDK2
   and CCND3 (Figure [189]6G). In conclusion, these findings suggest that
   the transcriptional regulatory circuit directly activates CDK2 and
   CCND3 to regulate the cell cycle, thereby promoting the proliferation
   of Abiraterone‐resistant CRPC cells.

   Interestingly, recent literature indicates that there is a tight
   crosstalk between the mevalonate pathway and the cell cycle.^[ [190]^37
   , [191]^38 ^] Such as inhibition of HMGCR with high doses of statins
   primarily leads to cell cycle arrest at the G1 phase due to reduced
   availability of mevalonate‐derived non‐sterol isoprenoids. Furthermore,
   HMGCR inhibition can also block the cell cycle at G2/M because of the
   reduction of cholesterol availability. Inhibition of
   diphosphomevalonate decarboxylase (MVD) slows down S‐phase progression
   by affecting DNA replication. We speculate that cholesterol synthesis
   inhibition may lead to cell cycle arrest at the G0/G1 phase in C42B‐ABi
   cells. Therefore, we used siRNA and statins to respectively inhibit
   SREBF2 and HMGCR, and found the cell cycle was indeed arrested at the
   G0/G1 phase (Figure [192]6H,I). Furthermore, the over‐expression of
   SREBF2 can partially reverse the G0/G1 phase arrest induced by
   knockdown of the SE‐driven TF loop (Figure [193]6J). These results
   indicate that the SE‐driven TF loop can regulate the cell cycle by
   transcriptionally activating cell cycle‐related proteins, and by a
   bypass pathway, which is activates cholesterol metabolism.

3. Discussion

   The transcriptional dysregulation mediated by super‐enhancers in
   prostate cancer has been reported. Such as targeting the SWI/SNF
   complex, which regulates abnormal chromatin remodeling, disrupting the
   transcriptional expression of super‐enhancer‐driven genes like AR,
   FOXA1, ERG and MYC, as well as their regulatory programs on downstream
   genes in prostate cancer.^[ [194]^39 ^] LSD1, BRD4 and FOXA1 form a
   transcriptional regulatory loop, binding to super‐enhancer elements to
   regulate the transcription of target genes. Inhibitors of LSD1 and BET
   exhibit potent synergistic effects, inducing tumor growth inhibition by
   disrupting the activity of multiple driver genes in CRPC.^[ [195]^40 ^]
   A set of aberrantly activated super‐enhancers identified in
   Enzalutamide‐resistant CRPC cells. The super‐enhancer driving oncogene
   CHPT1 catalyzes phosphatidylcholine synthesis to sustain Enzalutamide
   resistance.^[ [196]^27 ^] However, studies on super‐enhancer‐driven
   transcriptional dysregulation in Abiraterone‐resistant CRPC have not
   been reported.

   This study has, for the first time, delineated the specific landscape
   of super‐enhancers in Abiraterone‐resistant CRPC cells by performing
   H3K27ac‐targeted ChIP‐seq on the C42B cell line and its
   Abiraterone‐resistant variant, C42B‐ABi cells. Genes associated with
   super‐enhancers exhibit higher expression levels than those regulated
   by typical enhancers, which is one of the key characteristics
   underlying the significant role of super‐enhancers. By integrating
   RNA‐Seq data from these two cell lines, we identified and validated a
   mutually regulatory loop driven by super‐enhancers. This loop, which is
   highly expressed in C42B‐ABi, is composed of BCL6, NFIB and SMAD3.
   Previous research indicates that SMAD3 is overexpressed in advanced
   prostate cancer tissues and promotes CRPC cell proliferation.^[
   [197]^41 ^] NFIB facilitates tumor cell proliferation and resistance in
   hormone‐dependent breast cancer and colorectal cancer.^[ [198]^42 ^]
   NFIB is also identified to promote EMT and metastasis of CRPC.^[
   [199]^43 ^] Initially identified as an oncogene in B‐cell lymphoma,
   BCL6 drives malignant phenotypes by inhibiting proliferation and DNA
   damage checkpoints, while obstructing terminal B‐cell differentiation.
   However, recent reports identified that BCL6 acts as a pro‐tumor growth
   factor in various malignancies, playing a crucial role in promoting
   resistance to anti‐tumor drugs and chemotherapy. Notably, no previous
   studies have reported the involvement of BCL6, NFIB and SMAD3 in
   Abiraterone‐resistant CRPC. In our study, we found they regulate each
   other's and their transcription by co‐binding to each other's and their
   own enhancer or promoter regions (Figure [200]2). Moreover, they
   collaboratively reshape the transcriptome of C42B‐ABi cells, thereby
   sustaining the Abiraterone resistance of C42B‐ABi cells both in vitro
   and in vivo. The BCL6 inhibitor significantly suppresses the expression
   of BCL6, NFIB and SMAD3 and inhibits the proliferation of C42B‐ABi
   cells. Additionally, BCL6 inhibitor BI‐3802 can synergize with
   Abiraterone to impede the growth of C42B‐ABi cells, suggesting that
   targeting this circuit is a potential strategy for treating
   Abiraterone‐resistant cases.

   The accumulation of lipid droplets, that consist of a central core of
   triglycerides and cholesteryl esters, is caused by aberrant lipid
   metabolism. Lipids and cholesterol serve as precursors for steroid
   hormone synthesis, increasing intratumoral androgen production and
   promoting Enzalutamide resistance in CRPC cells.^[ [201]^44 , [202]^45
   ^] These results suggest that the dysregulation of cholesterol
   homeostasis could be a common signature of resistance to androgen
   deprivation therapy. However, in our study, we found that the SE‐driven
   TF loop sustains drug resistance via increasing intratumoral
   cholesterol, testosterone, and dihydrotestosterone in
   Abiraterone‐resistant CRPC. Cholesterol, as an important component of
   cell membranes, can participate in regulating cell proliferation and
   other processes. At the same time, it serves as a precursor for the
   steroid hormones testosterone and dihydrotestosterone, regulating the
   level of androgen and thus activating the AR signaling. Meanwhile, we
   found that AR expression was lower in C42B‐ABi cells than C42B cells, a
   phenomenon resembling AR‐independent transdifferentiation or NEPC‐like
   dedifferentiation was observed.^[ [203]^6 ^] Knocking down BCL6, SMAD3
   and NFIB could restore AR expression, suggesting that the C42B‐ABi
   cells regained sensitivity to Abiraterone, as our results confirmed
   (Figure [204]4H). These results suggest that TF mediates cell
   sensitivity to Abiraterone by regulating both intratumoral cholesterol
   homeostasis and AR expression.

   Furthermore, our research identified this circuit enhances cholesterol
   synthesis of CRPC cells both in vitro and in vivo. Specifically, the
   circuit directly activates the transcription of key cholesterol
   synthesis genes including SREBF2, HMGCR, SCAP and FDFT1. SREBF2, as a
   transcription factor, activates the expression of HMGCR and SCAP, which
   are validated in C42B‐ABi cells. In C42B‐ABi cells, the overexpression
   of SREBF2 can counteract the proliferation inhibition caused by the
   knockdown of BCL6, NFIB, and SMAD3. Additionally, this circuit
   maintains resistance by regulating the cell cycle, with knockdown of
   these transcription factors resulting in cell cycle arrest at the G0/G1
   phase. Mechanistically, our studies revealed that the circuit directly
   transcriptionally activates the G0/G1 cyclin‐dependent protein CDK2 and
   CCND3, thereby promoting cell cycle progression and the proliferation
   of C42B‐ABi cells (Figure [205]6E,F). Perturbation of cholesterol
   synthesis via SREBF2 knock‐down or HMGCR inhibition leads to cell cycle
   arrest in G0/G1 phase (Figure [206]6H,I). Therefore, we propose that
   the transcriptional regulatory circuit promotes the synthesis of
   cholesterol, testosterone, and dihydrotestosterone, by a SREBF2
   mediated manner activating the transcription of cholesterol synthesis
   genes in some extent, to sustaining Abiraterone resistance. Secondly,
   This circuit affects the progression of cell cycle by regulating
   cholesterol synthesis, thereby promoting cell resistance. Furthermore,
   the statins and SREBF2 inhibitors can augment the cytotoxic effects of
   Abiraterone on C42B‐ABi cells, suggesting that targeting the downstream
   target genes of this circuit could be a potential therapeutic strategy
   for overcoming Abiraterone resistance. In particular, BCL6‐targeted
   inhibitors hold significant promise for clinical application. Multiple
   clinical‐stage inhibitors show promise, BMS‐986458 (Bristol‐Myers
   Squibb): Phase I trial ([207]NCT06090539) as monotherapy and with
   rituximab, ARV‐393 (Arvinas): PROTAC degrader in Phase I
   ([208]NCT06393738) for NHL.

4. Conclusion

   In this study, we have unveiled the mechanisms by which aberrant
   activation of transcriptional regulation sustains Abiraterone
   resistance. Our research identifies and validates a transcriptional
   regulatory circuitry, driven by super‐enhancers composed of BCL6, NFIB
   and SMAD3, that promotes the development of Abiraterone resistance.
   Targeting this circuit or its downstream genes is proposed as a
   potential strategy for treating Abiraterone‐resistant CRPC.

5. Experimental Section

Cell Lines and Patient Samples

   C42B, PC3, 22RV1 and 293T cell lines were obtained from ATCC (American
   Type Culture Collection). Abiterone resistant cell lines, defined as
   C42B‐ABi, are estabolised by stepwise incubate the parental cell line
   C42B with increasing doses of Abiraterone, from 5 to 50 µmol L^−1. To
   maintained the resistance, 10 µmol L^−1 Abiterone routinely
   supplemented in the culture medium. C42B and C42B‐ABi were maintained
   in RPMI 1640 medium (Gbico). 293T was maintained in DMEM (Gbico). All
   media were supplemented with 10% fetal bovine serum (FBS, ExCell) and
   1% penicillin‐streptomycin. All cells were cultured at 37 °C in an
   incubator containing 5% CO[2].

   The adjacent, PC and CRPC tissue in this study were obtained from the
   Department of Urology, the Third Affiliated Hospital of Guangzhou
   Medical University (Guangzhou, Guangdong, China), which was approved by
   the medical research ethics committee review (approve number:
   2024‐008), with the permission of the patients.

Antibodies and Reagents

   The following antibodies purchased from CST were used in this study,
   includes anti‐BCL6 (#14 895), anti‐FLAG (#14 793), Rabbit (DA1E) mAb
   IgG XP Isotype Control (#3900), Mouse (G3A1) mAb IgG1 Isotype Control
   (#5415), anti‐CCND3 (#2936), anti‐CDK2 (#2546). Antibody of HMGCR
   (#13533‐1‐AP), NFIB (#29898‐1‐AP), SMAD3 (#30130‐1‐AP), SCAP
   (#12266‐1‐AP), SREBF2 (#28212‐1‐AP), FDFT1 (#13128‐1‐AP), AR (#5153),
   PSA (#5365) were purchased from Proteintech (Rosemont, USA). Antibody
   of H3K27ac (ab4729) was purchased from Abcam (Cambridge, MA). The
   molecules Abiraterone (S2246, Selleck), BI3802 (S6937, Selleck) were
   purchased from Selleck Chemicals and dissolved in DMSO and stored at
   −20 °C. The final dose of DMSO limited to 0.3% in all experiments.

MTT Cell Proliferation Assay

   The pre‐treated cell suspension was added to a 96‐well plate at
   3000–4000 cells/100 µL medium (10% FBS)/well, and cultured for the
   indicated time course. Add 20 µL of MTT (3‐(4,
   5‐dimethylthiazol–2‐yl)‐2, 5‐diphenyl tetrazolium bromide) to each
   well, incubate at 37 °C for 2 h, and measure the OD value at 490 nm on
   a microplate reader.^[ [209]^46 ^]

Cell Migration Assay

   The edited cells were resuspended in 100 µL of serum‐free medium and
   seeded into a transwell chamber with an 8 µm diameter, at a density of
   100 000 cells. The chamber was then placed in a 24‐well plate and
   supplemented with 600 µL of medium containing 10% FBS. After incubation
   for 24 h, the transwell chamber was removed, fixed and stained with
   crystal violet. Subsequently, images were captured under a microscope
   and cell counts were performed by Image J Software.

Quantification of Total Cholesterol

   1×10^^6 cells were collected from each group, washed once with PBS, and
   then prepare cholesterol extract buffer chloroform: isopropanol: NP‐40
   = 7:11:0.1. 200 µL cholesterol extract buffer was added to each sample,
   then vortex appropriately, centrifuged at 15 000 rpm for 10 min,
   discarded the precipitate, and the supernatant was retained and dried
   under vacuum. Finally, 200 µL assay buffer was added for dissolution.
   For serum total cholesterol measurement, 20 µL of serum was added to
   180 µL Assay Buffer, and the mixture was mixed by blowing. For
   measurement of total cholesterol in tissues, 10 mg of tissue was
   crushed in a homogenizer in cholesterol extract buffer, then following
   the cell extraction steps as mentioned. Cholesterol quantification was
   subsequently performed the steps of the kit (TR13421, Thermo
   Scientific).

ELISA Assay of Testosterone and Dihydrotestosterone

   For sample preparation, 1×10^^6 cells were collected, washed twice with
   PBS, and then suspended in 100 µL PBS. The cells were then subjected to
   repeated freeze‐thaw cycles at −80 °C, sonicated, and centrifuged to
   obtain the supernatant. For mouse serum samples, blood was allowed to
   clot naturally for 20 min, then centrifuged at 3000 rpm for 20 min to
   carefully collect the supernatant. For mouse tissue samples, 10 mg of
   tissue was homogenized with a certain amount of PBS using a
   homogenizer. The supernatant was then obtained by centrifugation. The
   prepared samples were then subjected to the instructions provided in
   the kit for testosterone (Mouse, E‐OSEL‐M0003; Human, E‐OSEL‐H0007;
   Elabscience, Wuhan, China) and dihydrotestosterone (E‐EL‐0031,
   Elabscience, Wuhan, China).

Cell Cycle Assay

   A certain number of cells were harvested, washed twice with PBS, and
   resuspended in pre‐chilled 70% ethanol for 2 h. Cell pellets were
   washed twice with PBS buffer and stained with propidium iodide. Cell
   cycle distribution was detected on a flow cytometer, and the resulting
   data were analyzed on a FlowJo X.

ChIP‐Seq and CUT&Tag Analysis

   ChIP uses the method described before.^[ [210]^21 , [211]^46 ^]
   However, tumor tissue needs to be extracted for some special processing
   before standardized ChIP. The tumor tissue was the size of soybean, and
   it was completely crushed by using a homogenizer under 2–3 mL 1%
   paraformaldehyde (#28908, Thermo) immersion, and then centrifugally
   filtered under a 25 µm filter. The obtained filtrate can continue to be
   carried out according to the ChIP step. CUT&Tag was performed according
   to the manufacturer's instructions (TD903, Vazyme).

   Raw reads were initially aligned to the hg19 reference genome using Bwa
   (version 1.2.2) with default settings [19 451 168]. Subsequently, the
   high‐quality mapped reads were extracted and sorted by SAMtools
   (version 1.3.1) using the “‐q 10” parameter [19 505 943]. PCR
   duplicates were then eliminated using Picard MarkDuplicates tool
   (version 1.136) and blacklist regions were filtered out utilizing
   bedtools (version 2.27.1).

   For histone ChIP‐seq data, MACS2 (Model‐Based Analysis of ChIP‐Seq,
   version 2.1.2) was applied to call peaks with parameters “‐q
   0.01–extsize  =  146 ‐nomodel ‐B” [18 798 982]. Bigwig files were then
   generated via deepTools bamCompare function (version 3.1.3) using
   parameters “–operation subtract –normalizeUsing CPM –extend Reads 146
   –binSize 20” [24 799 436]. For Cut‐tag data, peaks were identified
   through MACS2 with options “‐broad ‐q 0.01”. Bigwig files were
   generated using deepTools bamCoverage function with default settings.

Super‐Enhancer Analysis

   The C42B and C42B‐ABi specific H3K27ac peaks were extracted according
   to the non‐overlapping regions. Rank Order of Super Enhancers (ROSE),^[
   [212]^12 ^] a common tool for SE identification, was utilized to
   identify SEs enriched within each specific peak set. In summary, ROSE
   merged enhancer elements within a 12 kb range and organized them by
   decreasing intensities. SEs were then defined as stitching
   elements with a tangent slope showing an inflection point value of ≥1.

RNA‐Seq Analysis

   Raw reads were mapped to GRCh37 reference genome utilizing HISAT2
   (version 2.2.0) and quantified by htseq‐count program (version 0.11.3)
   at default setting.^[ [213]^47 ^] Then FPKM was employed to normalize
   the raw read counts. Differentially expressed genes were identified by
   the DESeq2 package (version 1.44.0) with p‐value < 0.05 and absolute
   log2 fold‐change (siRNA vs control) > 0.584.^[ [214]^48 ^]

Analysis of TF Co‐Binding in Relation to Gene Expression

   First, intersection analysis was performed using BEDTools to identify
   overlaps between H3K27ac‐enriched regions and binding sites of the
   three transcription factors. This classified H3K27ac regions into three
   groups, regions bound by only one transcription factor (Solo group),
   regions bound by two transcription factors (Dual group), regions bound
   by all three transcription factors (Trio group). Subsequently, the
   annotatePeaks.pl module from HOMER was applied to perform genomic
   region annotation and gene association analysis for all H3K27ac
   regions. Regions annotated as “promoter‐TSS” were defined as promoters,
   while others were classified as enhancers. Finally, expression trends
   of genes associated with H3K27ac regions in each group were calculated
   using C42B‐ABi cell line RNA‐seq data. Differential expression patterns
   between pairwise groups were statistically evaluated using t‐tests.

RT‐qPCR

   Total RNA was extracted using the kit purchased from Accurate Biology
   (Human, China), and purified RNA was reverse transcribed in Maxima H
   Minus cDNA Synthesis Master Mix (Thermo Fisher). Quantitative real‐time
   qPCR was performed on an AB7300 detection system (Applied Biosystems,
   Foster City, CA, USA) using gene‐specific primers and Power SYBR Green
   PCR Master Mix (Applied Biosystems). Expression of each gene was
   normalized to GAPDH and quantified by the Comparative Ct method. RNA
   primer sequences are provided in Table [215]S1 (Supporting
   Information).

Immunoblotting Assay

   Whole cell lysates were prepared in RIPA with added components.^[
   [216]^21 ^] Immunoblots were performed using specific primary and
   HRP‐linked secondary antibodies as indicated.

siRNA Knockdown

   siRNA sequences were purchased from Shanghai Genechem Co or Gemma.
   Cells were transfected with 100 nm siRNA in OPTIMEM‐I (Gibico) for
   24–48 h using Lipofectamine RNAiMAX transfection reagent (Invitrogen).
   siRNA sequences are provided in Table [217]S2 (Supporting Information).

Retroviral Infections

   The BCL6, SMAD3, NFIB, SREBF2 expression vector was amplified based on
   the pCDH‐CMV‐Flag‐EF1‐puro vector, and 3xFLAG‐tag was added by PCR. The
   pLKO.1‐puro vector expressing shRNA targeting BCL6, NFIB and SMAD3 was
   constructed. To produce viral particles, the recombinant viral vector
   and packaging vector were co‐transfected into 293T cells. Viral
   particle‐containing supernatant was collected at 48 h and filtered
   through a 0.45 µm filter after transfection. Cells of interest were
   then infected with the virus under 10 µg mL^−1 polybrene for 24–48 h.
   shRNA sequences are provided in Table [218]S2 (Supporting Information).

Dual Luciferase Reporter Assay

   Experimental procedures were adapted from the previous studies.^[
   [219]^21 ^] Candidate regions were amplified including the SMAD3
   promoter (SMAD3‐P) and three enhancer regions (SMAD3‐CE1, SMAD3‐CE2,
   SMAD3‐CE3). The promoter region was cloned into the pGL3‐Basic firefly
   luciferase reporter vector, and each enhancer region was cloned into
   the pGL3‐Promoter firefly luciferase reporter vector (IGEbio). C42B‐ABi
   cells were transfected with BioT Transfection Reagent (Bioland,
   B01‐00). Co‐transfected with Renilla luciferase control vector as a
   normalization control. Appropriate control vectors (pGL3‐Promoter or
   pGL3‐Basic) were included in each experiment. Forty‐eight hours after
   transfection, lysed cells were collected using the Dual Luciferase
   Reporter Assay Kit (YEASEN, 11402ES) and luciferase activity was
   determined by a full wavelength reader (GloMax Discover).

Lipid Droplet Staining Assay

   Prepare a cell suspension at a density of 5×10^4 to 1×10^5 cells.
   Discard the culture medium and gently rinse the cells with PBS. Fix the
   cells with 4% paraformaldehyde (PFA) at room temperature for 10 min.
   Wash the cells twice with PBS to remove residual PFA. Prepare the
   staining solution by diluting Lipid dye (invitrogen, [220]H34476) in
   PBS at a 1:1000 ratio (e.g., 1 µL dye in 1 mL PBS). Add 200–500 µL of
   the staining solution (enough to fully cover the cells) and mix
   thoroughly by gentle pipetting. Incubate at room temperature in the
   dark for 30 min. Fluorescence was quantified by flow cytometry
   (CytoFLEX) and visualized by microscopy (ZEISS).

Xenograft Assays in Nude Mice

   The animal experiments were subjected to the approval of the
   Institutional Animal Care and Use Committee of Guangdong Zhiyuan
   Biomedical Technology Co., LTD (Guangzhou, Guangdong, China) and
   following the ethical rules of animal experiments (approve number:
   AL20211205). 27 3‐week‐old nude mice were prepared, 24 of which
   underwent bilateral total orchiectomy (castrated group), and the
   remaining three groups were non‐castrated groups. Two weeks after
   orchiectomy, castrated mice were randomly divided into four groups,
   including PLKO, shBCL6, shSMAD3 and shNFIB, then subcutaneously
   inoculated with corresponding tumor cells. When the largest tumor
   diameter was ≈1.0 cm, Peripheral blood was collected from eyeballs for
   hormone detection, and tumors were dissected, weighted, and analyzed.

Immunohistochemical staining

   Paraffin sections of 4% paraformalin‐fixed tumor tissues were prepared
   according to the standard protocol for the detection of BCL6, SMAD3 and
   NFIB expression. Freshly configured DAB buffer (Beyotime, Shanghai,
   China) was used for the color reaction. Morphological HE staining was
   performed with hematoxylin. The prepared slides were scanned and
   photographed.

Gene Set Enrichment Analysis (GSEA)

   For the knockdown of BCL6, NFIB and SMAD3, the down‐regulated genes
   were individually chosen as the library genes. Then GSEAPreranked was
   performed using the fold change values from the knockdown of the
   respective TFs compared to the control as the ranked list. The
   significant terms with q value < 0.05 were shown.

Survival Analysis

   Taking BCL6, NFIB and SMAD3 as a signature, single‐sample GSEA was
   conducted to derive the enrichment score for each sample.^[ [221]^49 ^]
   To obtain the most significant relation between enrichment score and
   survival outcome, the optimal cutpoint was identified using the
   maximally selected rank statistics from the “maxstat”^[ [222]^50 ^] R
   package, followed by performing the log‐ranked test.

Statistical Analysis

   Two‐tailed Student's t‐test was used for the statistics of differences
   in measurement data between groups, Two‐way ANOVA Dunnett's multiple
   comparisons test was used for statistics of differences in measurement
   data between multiple groups, and One‐way ANOVA Dunnett's multiple
   comparisons test or t test was used for statistics of differences in
   count data between groups. All statistics and graphs were created with
   GraphPad Prism 8.0. Differences were considered statistically
   significant at P < 0.05 (*), P < 0.01 (**), P < 0.001 (***) and
   P<0.0001 (****); ns, not significant.

Ethics Approval Statement

   The adjacent, PC and CRPC tissue in this study were obtained from the
   Department of Urology, the Third Affiliated Hospital of Guangzhou
   Medical University (Guangzhou, Guangdong, China), which was approved by
   the medical research ethics committee review (approve number:
   2024‐008), with the permission of the patients. The animal experiments
   were subjected to the approval of the Institutional Animal Care and Use
   Committee of Guangdong Zhiyuan Biomedical Technology Co., LTD
   (Guangzhou, Guangdong, China) and following the ethical rules of animal
   experiments (approve number: AL20211205).

Conflict of Interest

   The authors declare no conflict of interest.

Author Contributions

   L.L.J., J.M.W., G.J.P., and H.C.Z. contributed equally to this work as
   co‐first authors. X.P.S., Z.G.Z., and Y.Y.Z. conceived and directed the
   study as co‐corresponding authors. L.L.J., J.M.W., X.Y.L., and H.C.Z.
   performed the experiments. J.X.F., E.Z.L., and Y.Z.L. analyzed and
   interpreted the data. G.J.P., B.Y.L., Q.M., and L.Z.J. provided
   bioinformatics analysis. W.D., Y.Y.G., S.W.H., A.C.L., and Q.M.M.
   provided statistical and technical support. Z.G.Z. and J.M.W. were
   responsible for providing clinical advice. L.L.J. and H.C.Z. wrote the
   first draft of the manuscript. X.P.S. and Y.Y.Z. revised the
   manuscript. All the authors had full access to the data, and reviewed
   the manuscript before submission.

Supporting information

   Supporting Information
   [223]ADVS-12-e01284-s002.docx^ (1.7MB, docx)

   Supporting Information
   [224]ADVS-12-e01284-s003.pdf^ (5.1MB, pdf)

   Supporting Information
   [225]ADVS-12-e01284-s001.xlsx^ (4.6MB, xlsx)

Acknowledgements