Abstract

   Poor response to 5‐fluorouracil (5‐FU) remains an obstacle in the
   treatment of gastric cancer (GC). Super enhancers (SEs) are crucial for
   determining tumor cell survival under drug pressure. SE landscapes
   related to 5‐FU‐resistance are mapped to GC using chromatin
   immunoprecipitation‐sequencing (ChIP‐Seq). SiRNA transcription factors
   (TFs) screen determines master TF Nuclear Receptor Subfamily 3 Group C
   Member 1 (NR3C1) activated by SE. High NR3C1 expression driven by SE
   correlated with 5‐FU resistance in patient‐derived organoids (PDOs).
   Phase separation formed by NR3C1 is observed using fluorescence
   recovery after photobleaching (FRAP). NR3C1 protein and Mediator
   promoted SE‐related gene transcription via phase separation. SEs and
   NR3C1 co‐binding patterns are explored using Cleavage Under Targets and
   Tagmentation (CUT&Tag) sequencing. 5‐FU‐related genes driven by NR3C1
   are identified using epigenetic reader inhibitor JQ1 and NR3C1 specific
   inhibitor Cort108297. NR3C1 knockdown increases 5‐FU sensitivity and
   alters the SE landscape through enhancer reprogramming, reducing
   downstream 5‐FU‐related target genes. JQ1 and Cort108297 both improve
   5‐FU efficacy in PDOs and patient‐derived xenografts (PDXs) by
   destroying SEs or inhibiting NR3C1. In conclusion, SE‐driven NR3C1
   promotes 5‐FU resistance in GC. SE destruction and NR3C1 inhibition
   lead to enhancer reconstruction and reduce 5‐FU‐related gene
   transcription, providing alternative therapeutic strategies for
   improving 5‐FU sensitivity.

   Keywords: 5‐Fluorouracil, gastric cancer, NR3C1, phase separations,
   super enhancers
     __________________________________________________________________

   The patients with GC have poor response to 5‐FU. We identified that
   SE‐driven master TF NR3C1 increased 5‐FU resistance by forming phase
   separation droplets and promoting 5‐FU‐related gene transcription.
   NR3C1 knockdown changed SEs landscape through enhancer reprogramming.
   The epigenetic reader inhibitor JQ1 and NR3C1 specific inhibitor
   Cort108297 improved 5‐FU effectiveness in PDX models of GC.

   graphic file with name ADVS-12-2409050-g005.jpg

1. Introduction

   The morbidity and mortality associated with gastric cancer (GC) remain
   high worldwide, especially in Middle and East Asian countries.^[ [56]^1
   ^] Refractory GC is characterized by significant molecular
   heterogeneity, lack of effective therapeutic targets, and poor response
   to emerging targeted therapies and immunotherapies.^[ [57]^2 , [58]^3
   ^] Chemotherapy remains a pivotal component in the comprehensive
   treatment of advanced GC. Fluorouracil (5‐FU) is a basic drug used in
   cancer chemotherapy with classic regimens such as FLOT (fluorouracil +
   oxaliplatin + docetaxel + leucovorin) and SOX (oxaliplatin +
   fluorouracil derivatives: S‐1).^[ [59]^4 ^] However, only some patients
   are sensitive to 5‐FU‐based chemotherapy.^[ [60]^2 , [61]^3 ^]
   Sensitivity to chemotherapy is determined by metabolic enzymes (such as
   thymidylate synthetase),^[ [62]^5 ^] transporters (such as ABC
   proteins),^[ [63]^6 ^] and other intrinsic characteristics.^[ [64]^7 ,
   [65]^8 ^] Many attempts have been made to overcome this resistance,
   namely by targeting known factors such as the downregulation of
   multidrug resistance proteins (MDR1 and P‐gp),^[ [66]^9 ^] but none
   have proven effective. Exploring the mechanisms underlying 5‐FU
   resistance and developing effective interventions targeting these
   mechanisms may offer a promising alternative to solving drug
   resistance.

   Super enhancers (SEs) are clusters of enhancers located in close
   proximity (full length <12.5 kb) that bind transcription factors and
   co‐regulators and intensively activate gene transcription.^[ [67]^10 ^]
   It is believed that the construction of active SE components and
   collaborating regulators, or SE landscape, is cell‐specific and
   spatial‐temporal specific, which is crucial for precise transcriptional
   regulation in cancer cells.^[ [68]^10 , [69]^11 ^] SEs can adapt within
   a certain range to regulate specific gene expression patterns at
   different spatial‐temporal levels, allowing cells to respond to
   different microenvironments, such as drug pressure.^[ [70]^10 , [71]^11
   ^] It has been found that SEs can promote chemotherapy resistance in
   triple‐negative breast cancer, and when combined with bromodomain and
   extra‐terminal domain (BET) inhibitors, chemotherapy efficacy is
   enhanced, predominantly due to SEs disruption and transcription
   deregulation.^[ [72]^12 ^]

   Among SEs and SE‐related genes (SRGs), a cluster of SE‐driven
   transcription factors (TFs) can bind their own SEs, activate their
   self‐expression through positive feedback mechanisms, and form core
   transcription regulatory circuitries (CRCs).^[ [73]^13 ^] TFs in CRCs
   regulate numerous downstream target genes and play a central regulatory
   role in maintaining cellular identity and specific status.^[ [74]^13 ^]
   For instance, in limbal stem cells, a CRC composed of three master TFs
   is essential for maintaining corneal epithelial identity and
   homeostasis.^[ [75]^14 ^] Furthermore, it has been demonstrated that
   the master TF in osteosarcoma CRC promotes tumor growth and metastasis
   and induces sensitivity to chemotherapy.^[ [76]^15 ^] Although several
   studies have described the SE atlas of GC,^[ [77]^16 ^] most have
   focused on the role of SE in GC oncogenesis, while few studies have
   explored the impact of CRCs and master TFs on chemotherapy sensitivity
   in GC.

   In this study, we characterized SEs and CRCs of 5‐FU sensitive and
   resistant GC cells. We found that high expression of the SE‐driven
   Nuclear Receptor Subfamily 3 Group C Member 1 (NR3C1) promoted 5‐FU
   resistance in GC cells and organoids. Inhibition of NR3C1 expression
   restored 5‐FU sensitivity. NR3C1 drove the formation of
   phase‐separation droplets around SEs and participated in the activation
   of several target genes related to 5‐FU resistance. In addition, we
   found that the epigenetic reader inhibitor JQ1 and the NR3C1 specific
   inhibitor Cort108297 both suppressed NR3C1 function by disrupting SEs
   or directly inhibiting NR3C1, thereby improving 5‐FU sensitivity in GC.
   These findings suggested a possible drug combination scheme for
   patients with GC undergoing 5‐FU‐basic chemotherapy.

2. Results

2.1. Characterization of SEs and CRCs Related to 5‐FU Resistance in GC

   In seven GC cell lines, we found that MKN45 was the most sensitive to
   5‐FU (IC[50]:3.274 µm), and HS‐746T was the most resistant to 5‐FU
   (IC[50]:26.50 µm), which was defined as primary 5‐FU resistance (Figure
   [78]1A). Using chromatin immunoprecipitation sequencing (ChIP‐Seq) with
   an H3K27Ac antibody combined with the rank ordering of super‐enhancers
   (ROSE) algorithm,^[ [79]^17 ^] the SEs and SRGs of MKN45
   (5‐FU‐sensitive) and HS‐746T (5‐FU‐resistant) cells were calculated and
   identified (Figure [80]1B,C). In HS‐746T cells, there were 1077 SEs
   identified with a threshold of 11 265.8 (Figure [81]1B). In MKN45
   cells, there were 756 SEs identified with a threshold of 19 750.8
   (Figure [82]1C).

Figure 1.

   Figure 1
   [83]Open in a new tab

   Super enhancer landscape and CRC of 5‐FU‐resistant GC cells. A) 5‐FU
   Sensitivity of common gastric cancer cell lines. B) Super enhancer (SE)
   curve of 5‐FU spontaneously resistant HS‐746T gastric cancer cells. C)
   SE curve of 5‐FU‐sensitive MKN45 gastric cancer cells. D) KEGG
   enrichment analysis of SE‐related genes (SRGs) in HS‐746T cells. E) GO
   enrichment analysis of SRGs in HS‐746T cells. F) Peak graphs of SE and
   Typical enhancers (TE) in HS‐746T. G) Transcriptional levels of SRGs
   and TE‐related genes (Student's t‐tests, two‐tailed). H) Differential
   expressed genes (DEGs) in HS‐746T and MKN45 cells. I) Venn map of SRGs,
   autoregulatory transcription factors (TFs), and HS‐746T high expression
   genes. J) Three CRCs of HS‐746T. NS: not significant, *p < 0.05, **p <
   0.01, ***p < 0.001, ****p < 0.0001.

   As SEs influence cancer cells by directly activating the transcription
   of SRGs,^[ [84]^10 ^] pathway enrichment of SRGs may provide clues for
   understanding the biological processes hidden in 5‐FU resistant GC
   cells. Hence, SRGs in HS‐746T (5‐FU‐resistant; Figure [85]1D,E; Figure
   [86]S1A, Supporting Information) and MKN45 cells (5‐FU‐sensitive,
   Figure [87]S1B,C, Supporting Information) were analyzed using the Kyoto
   Encyclopedia of Genes and Genomes (KEGG) and Gene Ontology (GO)
   enrichment analyses, respectively. SRGs in both cell types were
   enriched in cancer‐related pathways, such as gastric cancer, colorectal
   cancer, and adherens junction. We further analyzed the effect of the
   SRGs on 5‐FU resistance in HS‐746T cells. SRGs were enriched in the
   cell cycle pathway related to 5‐FU resistance.^[ [88]^18 ^] The
   inhibition of cyclin‐dependent kinase 4/6 (CDK4/6) displayed a
   synergistic effect with 5‐FU in GC.^[ [89]^19 ^] In terms of cellular
   metabolism, SRGs were enriched in central carbon metabolism in cancer
   (CCM). The CCM is mainly involved in glycolysis, the pentose phosphate
   pathway, and the tricarboxylic acid cycle (TCA cycle).^[ [90]^20 ^]
   Abnormal CCM contributes to cancer stemness maintenance and redox
   imbalance,^[ [91]^20 ^] thereby enhancing resistance to chemotherapy
   drugs.^[ [92]^21 ^] Cellular signaling pathways, including the
   PI3K‐AKT, ErbB, and FoxO pathways, were enriched. The activation of the
   PI3K‐AKT pathway inhibits 5‐FU‐induced apoptosis and promotes 5‐FU
   resistance in GC, whereas the AKT pathway inhibition effectively
   restores 5‐FU sensitivity^[ [93]^22 ^] (Figure [94]1D). In the GO
   analysis, Wnt‐related signaling pathways were enriched (Figure [95]1E;
   Figure [96]S1A, Supporting Information). These pathways are
   significantly correlated with the maintenance of cancer stemness and
   5‐FU resistance in GC.^[ [97]^23 ^]

   Because pathway enrichment analysis cannot reflect gene transcription
   changes, we focused on SRGs’ transcription changes. The H3K27Ac peak
   values of SEs were significantly higher than those of typical enhancers
   (TE) (Figure [98]1F; Figure [99]S1D, Supporting Information), which was
   consistent with previous reports.^[ [100]^10 ^] By combining ChIP‐Seq
   with RNA‐Seq, we found that the expression levels of SRGs were also
   significantly higher than those of the TE‐related genes
   (Figure [101]1G; Figure [102]S1E, Supporting Information). The H3K27Ac
   peak distribution was mainly gathered around promoters (<3 kb), which
   supported the enhancer‐promoter interaction reported in other studies
   ^[ [103]^24 ^] (Figure [104]S1F,G, Supporting Information).

   There are nearly a thousand SEs and the range of SRGs is further
   extended. CRC, which is constructed from SE‐driven TFs, has been
   proposed to play a central regulatory role in maintaining a specific
   cellular status.^[ [105]^13 ^] Thus, we attempted to determine the
   specific CRC in HS‐746T (5‐FU‐resistant). Differentially expressed
   genes (DEGs) were identified between 5‐FU‐reluctant HS‐746T cells and
   5‐FU‐sensitive MKN45 cells and highly expressed genes in HS‐746T cells
   (HS‐746T upDEGs, red box, Figure [106]1H) were used for subsequent
   analyses. Combining SRGs, HS‐746T upDEGs, and autoregulated TFs (based
   on prior knowledge of TF motifs), the CRCmapper algorithm calculated
   HS‐746T‐specific SE‐driven potential CRCs
   ([107]https://bitbucket.org/young_computation/crcmapper/src/master/),
   and the three most probable CRCs were identified (the score in the
   table reflects the probability of each circuitry, Figure [108]1I,J).
   The determinants underlying 5‐FU resistance were narrowed down to the
   TFs listed in the CRCs.

2.2. Identification of Master TF NR3C1 that Correlated with 5‐FU Resistance

   To explore the contribution of each TF in CRC, small interfering RNA
   (siRNA) screening was performed on HS‐746T cells. The inhibitory
   effects of each siRNA on TF transcription in CRC (Figure [109] 2A), and
   on the whole CRC transcription (all TF mRNA levels in CRC as a set;
   Figure [110]2B) were assessed. The results showed that NR3C1 knockdown
   (siNR3C1) and MYC knockdown (siMYC) inhibited the entire CRC set to the
   greatest extent (Figure [111]2B). Simultaneously, the effects of each
   interference on 5‐FU sensitivity were examined, showing that siNR3C1
   importantly improved the 5‐FU sensitivity of GC cells (siNR3C1:10.16
   µm, p < 0.0001; siNC: 27.37 µm, Figure [112]2C). SiFOXK1, siGLI2,
   siMYC, and siSTAT3 displayed similar effects on 5‐FU sensitivity,
   whereas the knockdown of some TF did not affect 5‐FU sensitivity, such
   as TEAD1 (Figure [113]2C; Figure [114]S2, Supporting Information).
   Correlation analysis between the expression levels of the CRC set and
   the 5‐FU IC[50] value showed that the higher the expression of the
   whole set, the higher the 5‐FU IC[50] value (R^2 = 0.7487, p < 0.0001;
   Figure [115]2D). NR3C1 knockdown inhibited the entire CRC set the most
   significantly, making GC more sensitive to 5‐FU (Figure [116]2D).
   Therefore, we focused on NR3C1 to explore the molecular mechanisms
   underlying SE‐driven, TF‐mediated resistance to 5‐FU in GC.

Figure 2.

   Figure 2
   [117]Open in a new tab

   SE‐CRC‐driven NR3C1 correlated with 5‐FU resistance in GC cell. A) With
   each siRNA of CRC‐TF (x‐axis), the heatmap of each gene mRNA level
   (y‐axis: fold change of gene mRNA level to GAPDH) in CRC. B) With each
   siRNA of CRC‐TF, the transcription level of the whole CRC gene set
   (ANOVA, two‐tailed, mean ± SD). C) With each siRNA of CRC‐TF, the 5‐FU
   sensitivity of si‐TF gastric cancer cells. D) Correlation analysis
   between the transcription level of CRC and 5‐FU IC[50] of gastric
   cancer cells (Spearman correlation test). E) IC[50] of 5‐FU in MKN45‐R
   and parental cells (ANOVA, two‐tailed). F) The transcriptional levels
   of NR3C1 in MKN45 versus MKN45‐R, and HS‐746T versus HS‐746T‐PR cells
   (Student's t‐tests, two‐tailed, mean ± SD). G) The protein levels of
   NR3C1 in MKN45 versus MKN45‐R, and HS‐746T versus HS‐746T‐PR cells
   (Student's t‐tests, two‐tailed, mean ± SD). H) Around NR3C1, IGV map of
   H3K27Ac ChIP‐seq and RNA‐Seq of MKN45, MKN45‐R, HS‐746T, and HS‐746T‐PR
   cells. Yellow area showed the region of SE1, SE2, and SE3. I) SE curve
   of MKN45‐R cells. J) Around NR3C1, IGV map of 5‐FU sensitive and
   resistant gastric cancer cells in [118]GSE186521. K) Disease‐free
   survival curves of the 384 patients in the gastric cancer cohort in
   GEPIA, with patients stratified based on NR3C1 expression levels (n =
   384, Cox regression analysis). L) Flow chart of master TF NR3C1
   identification from SRGs and CRC TFs. NS: not significant, *p < 0.05,
   **p < 0.01, ***p < 0.001, ****p < 0.0001.

   5‐FU resistant cells, MKN45‐R, were screened and constructed by
   long‐term culture in 5‐FU medication, and IC[50] value of MKN45‐R
   increased more than eight times (IC[50] of Resistant vs Parental was
   26.06 µm vs 3.14 µm, Figure [119]2E). 5‐FU partially resistant cells,
   HS‐746T‐PR, were also screened and constructed, and the IC[50] value
   was increased by about two‐fold (Partly resistant: 58.97 µm, Parental:
   26.29 µm, Figure [120]S3A, Supporting Information). The mRNA and
   protein expression of NR3C1 were significantly increased in 5‐FU
   resistant cells (MKN45‐R and HS‐746T‐PR) compared with those of their
   parental cells (MKN45 and HS‐746T, Figure [121]2F,G). The SE landscapes
   of MKN45‐R and HS‐746T‐PR were described, and regions around NR3C1 were
   displayed in Integrative Genomics Viewer (IGV)^[ [122]^25 ^] graphs to
   determine the impact of long‐term medication on SE characteristics
   (Figure [123]2H,I; Figure [124]S3B–D, Supporting Information). HS‐746T
   cells had three SE regions, SE1 (chr5: 143233723‐143262337), SE2 (chr5:
   143275273‐143391380) and SE3 (chr5: 143486735‐143509814), near the
   NR3C1 gene. MKN45 cells had no enhancers in these regions, whereas
   MKN45‐R cells generated new enhancers in the SE1 and SE2 regions (top
   Figure [125]2H). HS‐746T‐PR cells also showed increased peak
   intensities in these regions (top Figure [126]2H). Additionally,
   RNA‐seq analysis revealed that NR3C1 expression levels in MKN45‐R and
   HS‐746T‐PR cells were higher than those in their parental cells (bottom
   Figure [127]2H).

   For external validation, we combined 5‐FU sensitivity data from the
   Genomics of Drug Sensitivity in Cancer (GDSC) database^[ [128]^26 ^]
   with previous H3K27Ac ChIP‐Seq data of GC cells from Patrick Tan's
   study ([129]GSE186521).^[ [130]^27 ^] Consistently, all 5‐FU resistant
   cells (top three cell lines: HS‐746T, MKN7, and RERFGC18) had SEs or
   enhancers around the SE regions of NR3C1, while only a few enhancers
   were present around the SE regions of NR3C1 in 5‐FU sensitive cells
   (bottom four cell lines: AGS, IM95, HGC27, and SNU16) (Figure [131]2J).
   Moreover, NR3C1 expression correlated with disease‐free survival (DFS)
   in patients with GC, with a hazard ratio (HR) of 1.8 (p = 0.0032) in
   Gene Expression Profiling Interactive Analysis (GEPIA)
   ([132]http://gepia.cancer‐pku.cn/),^[ [133]^28 ^] indicating that
   patients with high NR3C1 expression had shorter DFS and worse prognosis
   (Figure [134]2K). The flow chart of master TF identification in GC is
   displayed in Figure [135]2L.

2.3. High NR3C1 Expression Promotes 5‐FU Resistance of GC Cells

   With protein expression tested in seven gastric cancer cell lines
   (Figure [136]S3E, Supporting Information), NR3C1 was knocked down using
   three shRNA lentiviruses in HS‐746T cells and overexpressed with the
   NR3C1 vector in MKN45 and HS‐746T cells (HS‐746T‐OE and MKN45‐OE cells)
   (Figure [137]3A). The effect of shRNA2 was weak; therefore, shRNA1 and
   shRNA3 were selected for subsequent experiments (HS‐746T‐sh‐1 and
   HS‐746T‐sh‐3). In HS‐746T cells, NR3C1 knockdown significantly
   decreased the IC[50] values (sh‐1:8.52 µm, sh‐3:12.55 µm, NC: 28.1 µm,
   p < 0.0001), and NR3C1 overexpression increased the IC[50] values (OE:
   42.12 µm, Vector: 26.87 µm, p < 0.0001; Figure [138]3B,C; Figure
   [139]S4A, Supporting Information). In MKN45 cells, NR3C1 overexpression
   significantly increased the IC[50] value (OE: 18.42 µm, Vector: 3.18
   µm, p < 0.0001). Subsequently, shNR3C1 lentivirus was transinfected
   into 5‐FU resistant cells (MKN45‐R‐sh‐1, HS‐746T‐PR‐sh‐1), and found
   that NR3C1 knockdown effectively reduced IC[50] values of 5‐FU
   resistant cells (MKN45‐R‐NC: 26.39 µm, MKN45‐R‐sh‐1:1.86 µm, p <
   0.0001; HS‐746T‐PR‐NC: 59.38 µm, HS‐746T‐PR‐sh‐1:7.04 µm, p < 0.0001),
   restoring the 5‐FU sensitivity of resistant cells (Figure [140]3D,E;
   Figure [141]S4B,C, Supporting Information).

Figure 3.

   Figure 3
   [142]Open in a new tab

   High NR3C1 expression promoted the 5‐FU resistance of GC cells. A)
   NR3C1 expression knockdown by shRNA in HS‐746T (top) and NR3C1
   overexpression in HS‐746T and MKN45 (bottom), detected by Western
   blotting (Student's t‐tests, two‐tailed, mean ± SD). B) NR3C1 knockdown
   and overexpression in HS‐746T (left) and NR3C1 overexpression in MKN45
   (right), detected by quantitative reverse transcription PCR (qRT‐PCR)
   (Student's t‐tests, two‐tailed, mean ± SD). C) IC[50] values of 5‐FU in
   NR3C1 knockdown HS‐746T cells (left), and IC[50] of 5‐FU in NR3C1
   overexpressed MKN45 cells (right) (ANOVA, two‐tailed). D) NR3C1
   knockdown detection in protein (left) and mRNA (right) levels in
   MKN45‐R (ANOVA, two‐tailed, mean ± SD). E) The 5‐FU IC[50] values of
   MKN45, MKN45‐R, MKN45‐R NC, and MKN45‐R sh‐1 cells (ANOVA, two‐tailed).
   F) NR3C1 expression level of organoids cultured from gastric cancer
   tissues of patients 1–8 (P1‐P8). G) The 5‐FU IC[50] values and NR3C1
   immunofluorescence levels of organoids P1‐6. Scale: 100 µm. H)
   Correlation analysis of NR3C1 expression level and 5‐FU IC[50] of
   gastric cancer organoids (Spearman correlation test). I) NR3C1
   knockdown or overexpression in organoids P7, and 5‐FU IC[50] values of
   indicated groups (ANOVA, two‐tailed). J) Photos around the IC[50]
   values of the organoid P7 in each group. Scale: 100 µm. K) NR3C1
   knockdown or overexpression in organoids P8, and 5‐FU IC[50] values of
   indicated groups (ANOVA, two‐tailed). L) Photos around the IC[50]
   values of the organoid P8 in each group. Scale: 100 µm. M) The 5‐FU
   IC[50] values of HS‐746T, HS‐746T‐OE, MKN45, and MKN45‐OE with
   Cort108297 usage (ANOVA, two‐tailed). NS: not significant, *p < 0.05,
   **p < 0.01, ***p < 0.001, ****p < 0.0001.

   NR3C1 expression and IC[50] values were tested in eight patient‐derived
   organoids (PDOs) constructed from patients with GC (numbered P1‐P8)
   (Figure [143]3F,G, Figure [144]S4D,E, Supporting Information),
   indicating that NR3C1 expression was positively correlated with 5‐FU
   IC[50] values in the organoids (R ^2 = 0.596, p = 0.0248,
   Figure [145]3H). Furthermore, NR3C1 expression was manipulated using
   lentivirus and control virus in the P7 and P8 organoids, respectively
   (P7‐sh, P7‐OE, P8‐sh, and P8‐OE). The IC[50] values of NR3C1 knockdown
   PDOs were significantly decreased (P7‐sh: 14.65 µm, P7‐NC: 44.75 µm, p
   < 0.0001; P8‐sh: 21.56 µm, P8‐NC: 66.24 µm, p < 0.0001), while IC[50]
   values of NR3C1 overexpressed PDOs were significantly increased (P7‐OE:
   148.3 µm, P7‐Vector: 47.19 µm, p < 0.0001; P8‐OE: 140.6 µm, P8‐Vector:
   69.89 µm, p < 0.0001) (Figure [146]3I–L; Figure [147]S5, Supporting
   Information). These results indicated that NR3C1 knockdown improved
   5‐FU sensitivity in GC organoids, and NR3C1 overexpression led to 5‐FU
   resistance. Cort108297 (C297) is a selective NR3C1 antagonist with no
   affinity for other steroid receptors.^[ [148]^29 ^] When C297 inhibited
   NR3C1 function, 5‐FU IC[50] values in HS‐746T and MKN45 cells decreased
   significantly (Figure [149]3M).

2.4. NR3C1 Forms Droplets with SEs by Liquid‐Liquid Phase Separation (LLPS)

   NR3C1 is a master TF regulated by SE in CRC, indicating that it binds
   to the promoters and enhancers of other SE‐associated TFs. Therefore,
   NR3C1 may be an important component of the central protein
   transcription machinery at transcription start sites (TSS) and SE
   sites. In the nucleus, the central transcription complex formed by TFs
   and co‐regulators condenses to form LLPS droplets, inside which SEs,
   target gene promoter regions and TFs are highly condensed, inducing
   intensive transcriptional tornados.^[ [150]^30 , [151]^31 ^]

   Intrinsically disordered region (IDR) is the physical substantial of
   LLPS droplet formation.^[ [152]^31 ^] We estimated the IDRs of CRC TFs
   using PONDR ([153]http://www.pondr.com/)^[ [154]^32 ^] and found that
   some TFs, such as EHF, and PPARG in the CRC did not show IDRs, while
   other TFs, including NR3C1, had IDRs (Figure [155]4A; Figure [156]S6,
   Supporting Information). After the NR3C1 fluorescent protein was
   purified and smeared on slides, confocal microscopy revealed that the
   NR3C1 protein formed droplets in real‐time (Figure [157]4B). The
   quantity and size of the droplets are affected by the protein
   concentration, salt concentration, and the presence or absence of the
   phase‐change inhibitor 1,6‐hexadiol (1,6‐Hex) in the tubes.^[ [158]^31
   ^] With double dilution of protein, droplets were reduced in number and
   size, and droplets were almost invisible with double protein dilution
   plus high salt concentrations (Figure [159]4B). As the protein
   concentration decreased, the number and size of droplets reduced, and
   droplet formation was rare with 1,6‐Hex addition (Figure [160]4C,D). As
   a control for 1,6‐hexadiol,^[ [161]^33 , [162]^34 ^] 2,5‐hexadiol did
   not interfere with droplet formation (Figure [163]4C,D; Figure
   [164]S7A, Supporting Information). This phenomenon indicated that NR3C1
   protein could form LLPS droplets in the tubes.

Figure 4.

   Figure 4
   [165]Open in a new tab

   NR3C1 formed phase separation. A) Diagram of the protein disordered
   region (IDR) in NR3C1. B) The purified EGFP‐NR3C1 protein formed
   droplets, which were inhibited by reducing protein concentration and
   high salt solution (ANOVA, two‐tailed, mean ± SD). Scale: 5 µm. C)
   Statistical plots of the relative fluorescent area of droplets with
   different protein concentrations (ANOVA, two‐tailed, mean ± SD). D)
   Pictures of droplets with different protein concentrations and with the
   use of 1, 6‐hexadiol which destroys the droplets. Scale: 5 µm. E) The
   mCherry‐NR3C1 protein droplets in the nucleus of HS‐746T. Left: scale:
   10 µm, right: scale: 1 µm. F) FRAP curves of mCherry‐NR3C1 in the
   nucleus of HS‐746T, and FRAP curves with si‐BRD4 and si‐MED1
   respectively. G) Time series of FRAP images of mCherry‐NR3C1 protein
   with si‐BRD4 and si‐MED1. Top: scale: 10 µm, bottom: scale: 1 µm. H)
   Anti‐MED1 as a co‐immunoprecipitated antibody, detection of its binding
   to BRD4 and NR3C1. I) Anti‐BRD4 as a co‐immunoprecipitated antibody,
   detection of its binding to MED1 and NR3C1. J) Anti‐NR3C1 as a
   co‐immunoprecipitated antibody, detection of its binding to BRD4 and
   MED1. NS: not significant, *p < 0.05, **p < 0.01, ***p < 0.001, ****p
   <0.0001.

   Inside the cells, specific proteins can form locally concentrated
   droplets at the SEs in the nucleus, which are dynamic and fluid.^[
   [166]^31 ^] After NR3C1‐mCherry plasmids were transfected into the
   cells, NR3C1 formed locally concentrated red droplets in the nucleus
   that fused dynamically (Figure [167]4E; Figure [168]S7B and Movies
   [169]S1 and [170]S2, Supporting Information). Fluorescence recovery
   after photobleaching (FRAP) experiments indicated that the NR3C1
   droplets recovered successfully after being bleached (Figure [171]4F;
   Figure [172]S7C, Supporting Information). DNA fluorescence in situ
   hybridization (FISH) combined with NR3C1 immunofluorescence showed that
   NR3C1 co‐localized with NR3C1 SEs (Figure [173]S8, Supporting
   Information). This showed that NR3C1 formed locally concentrated
   droplets in the SEs. BRD4 and MED1 are the core components of the
   central transcription complex at the SEs and TSS. After BRD4 knockdown
   (siBRD4) and MED1 knockdown (siMED1), NR3C1 droplets did not recover,
   suggesting that droplet formation by NR3C1 was mediated by BRD4 and
   MED1 (Figure [174]4F,G). Co‐immunoprecipitation experiments revealed
   pairwise interactions among NR3C1, BRD4, and MED1 (Figure [175]4H–J).
   Moreover, both ChIP‐PCR and ChIP‐qPCR showed that NR3C1, BRD4, and MED1
   had binding sites in the SE region of NR3C1 (Figure [176]S9A,B,
   Supporting Information). Therefore, NR3C1 formed droplets and promoted
   LLPS in SEs with BRD4 and MED1.

2.5. NR3C1 Knockdown Induces Enhancer Remodeling in GC Cells

   As a TF, the binding pattern of NR3C1 revealed by CUT&Tag‐seq indicated
   the potential function of NR3C1 to the downstream targets (Figure
   [177]5A). In combination with H3K27Ac ChIP‐Seq, CUT&Tag‐seq showed that
   37.9% of the SE region had NR3C1 binding sites, while only 14.2% of the
   TE region had NR3C1 binding sites, showing that the NR3C1 and SE
   regions had a higher overlap rate (Figure [178]5B). This indicated that
   NR3C1 was more likely to be highly enriched in the SE regions of GC
   cells.

Figure 5.

   Figure 5
   [179]Open in a new tab

   NR3C1 knockdown induced enhancer reconstruction of GC cells. A)
   Heatmaps of HS‐746T ChIP‐seq and NR3C1 CUT&Tag sequencing. B) The
   proportion of NR3C1 binding in SE and TE regions, respectively. C) and
   D) SE curves of the control group (left) and NR3C1 knockdown group
   (right). E) ChIP‐seq heatmaps in the control group (left) and NR3C1
   knockdown group (right). F) Peak graphs of SE in the control group
   (left) and NR3C1 knockdown group (right). G) Pie chart showing the
   percentages of decreased or increased H3K27ac signals across the SEs of
   wild‐type cells upon NR3C1 loss. H) Pie chart showing the percentages
   of decreased and increased H3K27ac signals close to the loci of
   down‐regulated genes upon NR3C1 loss. I) GSEA of SE‐associated gene set
   in control versus shNR3C1 cells (Fisher's Exact test, one‐tailed). J)
   Motif de novo analysis of NR3C1 CUT&Tag sequencing (Hypergeometric
   test).

   Hence, we postulated that NR3C1 knockdown might change SE
   characteristics probably in GC cells, and described the SE landscape
   based on H3K27Ac ChIP‐Seq and RNA‐Seq (Figure [180]5C–E). The results
   showed that the control group had 913 SE regions (threshold:
   9678.8223), whereas the number of SE regions was reduced to 686 in the
   NR3C1 knockdown group (threshold: 7902.489) (Figure [181]5C,D). The
   H3K27Ac peak values of SEs and TEs in NR3C1 knockdown cells both
   significantly decreased (Figure [182]5F; Figure [183]S10A, Supporting
   Information), and 82.5% of the SEs have weakened H3K27Ac signals
   (Figure [184]5G). This suggested that NR3C1 knockdown significantly
   inhibited H3K27Ac signaling and changed the SE landscape of GC. RNA‐Seq
   analysis showed that 91.5% of the downregulated DEGs (NR3C1 knockdown
   vs NR3C1 control) had decreased H3K27Ac levels near the gene locus
   after NR3C1 knockdown (Figure [185]5H). Gene Set Enrichment Analysis
   (GSEA) revealed that the expression of the SRG set was significantly
   decreased in the NR3C1 knockdown group (NES = −2.97, p adjust = 1 ×
   10^−10, Figure [186]5I), whereas the expression of the TE‐related gene
   set did not change (p adjust = 0.2, Figure [187]S10B, Supporting
   Information). As demonstrated previously, NR3C1 participated in the
   transcription protein complex in SE regions through LLPS, indicating
   that NR3C1 bound to SEs to synergistically enhance downstream gene
   transcription and that NR3C1 knockdown could significantly inhibit the
   expression of SRGs by altering SE composition.

   Multiple TFs usually bind to each other and form large protein
   complexes that act as transcription‐initiating machines at SEs,
   enhancers, and promoters.^[ [188]^16 ^] TFs inside the complex often
   cooperate to exert synergistic effects.^[ [189]^16 ^] In addition to
   the NR3C1 motif, the binding motifs of Jun‐AP1, MYBL2, KLF5, and GATA4
   were found to be TF partners of NR3C1 using the hypergeometric
   optimization of motif enrichment (HOMER)^[ [190]^35 ^] algorithm
   (Jun‐AP1: p = 1 × 10^−2056; MYBL2: p = 1 × 10^−2029; KLF5: p = 1 ×
   10^−1773; GATA4: p = 1 × 10^−1095, Figure [191]5J). Studies have shown
   that Jun‐AP1 mediates 5‐FU resistance in GC by promoting the expression
   of nucleotide excision repair (NER) genes.^[ [192]^36 ^] In addition to
   accelerating the cell cycle and promoting GC cell proliferation,^[
   [193]^37 ^] MYBL2 promotes the transcription of specific genes that
   mediate GC resistance to DNA‐damaged chemotherapy drugs.^[ [194]^38 ^]
   KLF5 has been shown to be an SE‐related core TF in a subtype of GC.^[
   [195]^39 ^] KLF5 not only inhibits the apoptosis of GC cells,^[
   [196]^40 ^] but also mediates GC resistance to DNA‐damaging
   chemotherapy drugs.^[ [197]^41 ^] GATA4 and KLF5 also cooperate to
   jointly maintain the oncogenic transcriptional regulatory network of GC
   and promote epithelial‐to‐mesenchymal transition (EMT) and GC
   metastasis.^[ [198]^42 , [199]^43 ^] Together, these results indicated
   that NR3C1 synergistically interacted with other TFs in common motifs
   and initiated downstream transcription to mediate 5‐FU resistance in GC
   cells.

2.6. Master TF NR3C1 Promotes 5‐FU Resistance Through Downstream Target Genes

   In 5‐FU resistant cells, based on the SE landscape and KEGG analysis,
   we found that SRGs were enriched in the MAPK and Ras signaling
   pathways, which maintained cell survival and proliferation. SRGs were
   also enriched in estrogen signaling pathways and gastric acid
   secretion, which are related to GC development (Figure [200]6A; Figure
   [201]S11A, Supporting Information). In GO enrichment analysis, SRGs
   were enriched in transcription co‐regulator binding,
   nucleotide‐excision repair pathways, regulation of signaling receptor
   activity, and protein tyrosine phosphatase activity (Figure [202]6B;
   Figure [203]S11B, Supporting Information). These results suggested that
   SRGs activated a series of oncogenic signaling pathways, promoted cell
   repair mechanisms to recover from DNA drug‐induced damage, and
   maintained cell proliferation in 5‐FU resistant cells.

Figure 6.

   Figure 6
   [204]Open in a new tab

   SE‐driven NR3C1 induces 5‐FU resistance in GC cells through activating
   downstream target gene clusters. A) KEGG enrichment analysis of SRGs in
   MKN45‐R. B) GO enrichment analysis of SRGs in MKN45‐R. C) KEGG
   enrichment analysis of NR3C1 binding genes based on CUT&Tag‐Seq. D)
   GO‐MF enrichment analysis of NR3C1 binding genes based on CUT&Tag‐Seq.
   E) Within RNA‐Seq of NR3C1 control group and knockdown group, GO
   enrichment analysis of differentially highly expressed genes in the
   control group. F) KEGG enrichment analysis of differentially highly
   expressed genes in the knockdown group in RNA‐Seq. G) Venn diagram of
   SRGs, NR3C1‐bound genes, differentially down‐regulated gene sets
   (left), and identification of a set of genes that are known to be
   relative to 5‐FU sensitivity (right). H) IGV map of H3K27Ac peak and
   NR3C1 CUT&Tag binding peaks around ABCC1 in HS‐746T, HS‐746T‐PR,
   HS‐746T NC, and HS‐746T‐sh‐1, MKN45 and MKN45‐R groups. I) The mRNA
   levels of 5‐FU‐related gene sets in MKN45, MKN45‐R, MKN45‐R NC, and
   MKN45‐R KD groups (ANOVA, two‐tailed, mean ± SD). J) The mRNA levels of
   5‐FU‐related gene sets in MKN45+NS, MKN45‐R+NS, MKN45+JQ1, MKN45‐R+JQ1
   groups (NS: normal saline, ANOVA, two‐tailed, mean ± SD). K)
   Synergistic effects of the combined use of JQ1 and 5‐FU in MKN45. Two
   drugs were double‐diluted (JQ1 in 4 gradients, 5‐FU in 5 gradients).
   The color scale for drug inhibition values. L) Fa (Fraction of
   affected)‐CI (Combination index) plots in MKN45 for the combination of
   JQ1 and 5‐FU. CI values of <0.7 indicate synergistic effects and CI
   values of <0.3 indicate strong synergistic effects. M) The mRNA levels
   of 5‐FU‐related gene sets in MKN45+NS, MKN45‐R+NS, MKN45+C297,
   MKN45‐R+C297 groups (ANOVA, two‐tailed, mean ± SD). NS: not
   significant, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

   In addition to several oncogenic pathways, such as ErbB and Rap1
   signaling pathways, NR3C1‐bound genes were enriched in the Wnt
   signaling pathway, which is strongly associated with cancer cell
   stemness and multidrug resistance (Figure [205]6C). Several pathways
   mediating the cell response to 5‐FU, including response to chemicals,
   cellular response to stimulus, kinase activity, drug binding,
   transcription factor binding, nucleotide binding, and nucleoside
   phosphate binding, were also enriched (Figure [206]6D). These results
   suggested that NR3C1 affected cell response to stimuli, such as
   external drugs, and activated multiple signaling pathways to maintain
   cell stemness and promote DNA drug resistance.

   Moreover, DEGs (NR3C1 knockdown vs NR3C1 control) enriched in pathways
   like nucleoside‐triphosphatase regulator activity, DNA‐binding
   transcription factor binding, cell growth, and response to xenobiotic
   stimulus, were upregulated in the NR3C1 high‐expression group (control)
   (Figure [207]6E; Figure [208]S11C, Supporting Information). DEGs
   enriched in the purine metabolism and the Hippo signaling pathway were
   upregulated in the NR3C1 low‐expression group (NR3C1 knockdown)
   (Figure [209]6F; Figure [210]S11D, Supporting Information). These
   results suggested that NR3C1 responded to external stimuli and affected
   transcription regulation to maintain cell proliferation. Moreover,
   NR3C1 knockdown affected nucleotide metabolism.

   Therefore, the intersection of SRGs, NR3C1‐binding gene sets, and
   down‐DEGs of the NR3C1 knockdown group were obtained to identify the
   direct target genes regulated by SE‐NR3C1 and related to drug
   resistance (Figure [211]6G). Among the 49 genes at this intersection,
   seven genes were previously reported to be correlated with 5‐FU‐related
   chemotherapy resistance, including FAAP24, ABCC1, EDN1, CAV1, CCND1,
   SRSF3, and SLC3A2. The multiple sequencing data in the IGV graphs
   indicated that NR3C1 bound to the SE regions of these genes and that
   the SE peaks were stronger in 5‐FU resistant cells than in parental
   cells (Figure [212]6H; Figure [213]S12A–F, Supporting Information).
   Taking ABCC1 as an example, near the SE regions of ABCC1, NR3C1 had
   several binding sites. Furthermore, the peak intensity of ABCC1‐SE in
   HS‐746T‐PR was higher than that in HS‐746T, whereas that in
   HS‐746T‐sh‐1 was lower than that in the control group. The peak
   intensity of ABCC1‐SE was consistently higher in MKN45‐R cells than in
   MKN45 cells (Figure [214]6H). Both ChIP‐PCR and ChIP‐qPCR showed that
   NR3C1, BRD4, and MED1 had binding sites in the SE region of ABCC1,
   which were markedly inhibited by the BRD4 inhibitor, JQ1 (Figure
   [215]S9A,B, Supporting Information). Moreover, dual‐luciferase assays
   indicated that NR3C1 bound to the ABCC1 SE region and ABCC1 SE promoted
   luciferase transcription as an enhancer (Figure [216]S9C, Supporting
   Information). The transcription levels of the seven 5‐FU‐related genes
   in the 5‐FU resistant cells (MKN45‐R) were higher than those in the
   parental cells (MKN45). Their transcription was inhibited when NR3C1
   was knocked down (MKN45‐R‐sh‐1), indicating that NR3C1 knockdown led to
   a loss of its promotional effect on 5‐FU‐related genes
   (Figure [217]6I).

   FAAP24^[ [218]^44 , [219]^45 ^] and SRSF3^[ [220]^46 , [221]^47 ^] have
   been reported to participate in homologous recombination repair and
   promote 5‐FU resistance in cancer cells, in addition to the multidrug
   resistance protein ABCC1.^[ [222]^48 ^] In addition, downstream target
   genes promote 5‐FU resistance through ferroptosis via SLC3A2 and
   through the activation of Wnt signaling via CAV1.^[ [223]^49 , [224]^50
   ^] Overexpression of CCND1 enhanced 5‐FU resistance in GC.^[ [225]^51 ,
   [226]^52 ^] The results described above (Figure [227]6G–I) indicated
   that receiving drug signals, NR3C1 activated downstream target genes
   related to DNA repair to eliminate abnormal DNA produced by the 5‐FU
   attack, maintained cell homeostasis, and prevented cell death.

   The epigenetic reader inhibitor JQ1 is a Bromo‐ and extra‐terminal
   domain family inhibitor (BETi), that inhibits BRD4 binding to
   chromatin.^[ [228]^53 ^] JQ1 causes SE destruction, increases
   apoptosis, and hampers SE‐induced oncogene transcription.^[ [229]^53 ^]
   ChIP‐PCR and ChIP‐qPCR showed that JQ1 significantly inhibited NR3C1,
   BRD4, and MED1 binding to both the NR3C1 SE and ABCC1 SE regions
   (Figure [230]S9A,B, Supporting Information). In MKN45 cells, the
   results showed that JQ1 had a significant inhibitory effect on the
   transcription of five 5‐FU‐related genes, except for EDN1 and CAV1
   (Figure [231]6J). MKN45‐R cells were treated with JQ1 and the results
   showed that JQ1 significantly inhibited all seven 5‐FU‐related genes
   (Figure [232]6J). The inhibitory effect of JQ1 on HS‐746T and
   HS‐746T‐PR suggested that drug‐resistant cells had a higher level of
   transcriptional addiction and were more dependent on SE‐driven
   transcriptional activation. Therefore, JQ1 exerted a more significant
   inhibitory effect on HS‐746T‐PR than on HS‐746T (Figure [233]S13A,B,
   Supporting Information). Furthermore, the combination of JQ1 and 5‐FU
   exerted a synergistic effect under unfixed drug proportions in MKN45
   cells (Figure [234]6K). The combination index (CI) of the two drugs
   under fixed drugs proportion (JQ1:5‐FU = 1.25:1) showed that JQ1 and
   5‐FU had a synergistic effect with the CI <0.7 (Figure [235]6L).
   Similar findings were observed in HS‐746T with a CI <0.5 (Figure
   [236]S13C,D, Supporting Information). The NR3C1 inhibitor C297 had an
   inhibitory effect on five 5‐FU‐related genes (WT+NS group vs WT+C297
   group), except for FAAP24 and EDN1 (Figure [237]6M). In 5‐FU‐resistant
   cells, the transcription of all seven 5‐FU‐related genes was
   significantly inhibited by C297 treatment (Figure [238]6M; Figure
   [239]S13E, Supporting Information). Together, these results indicated
   that both JQ1 and C297 could inhibit NR3C1 downstream 5‐FU‐related
   genes, especially in 5‐FU‐resistant cells.

2.7. SE Destruction and NR3C1 Inhibition Improve 5‐FU Sensitivity In Vivo

   Based on the findings in vitro, two GC patient‐derived xenograft (PDX)
   models were constructed and treated with the drugs (5‐FU, 5‐FU+JQ1, and
   5‐FU+C297) (Figure [240]7A–I). Drugs were administered for five
   consecutive days as a course of treatment and rested for two days
   before the next course. The animals were euthanized on the 40th day.
   The results showed that both 5‐FU+JQ1 and 5‐FU+C297 significantly
   inhibited tumor growth, leading to smaller tumors and slower tumor
   growth (Figure [241]7A–C,F–H), and fewer Ki‐67 positive cells compared
   with 5‐FU alone (Figure [242]7E; Figure [243]S14, Supporting
   Information). Moreover, the combination of 5‐FU+JQ1 and 5‐FU+C297 had a
   higher inhibitory effect on NR3C1 high‐expression (NR3C1^Hi) PDXs than
   on NR3C1 low‐expression (NR3C1^Lo) PDXs (Figure [244]7D,E,I). These
   findings suggested that, in preclinical trials, NR3C1^Hi tumors benefit
   more from 5‐FU+JQ1 and 5‐FU+C297 than NR3C1^Lo tumors. For both drug
   combinations, a slight difference was observed in the inhibition rates
   of 5‐FU+C297 and 5‐FU+JQ1 (Figure [245]7J). These findings highlighted
   the clinical translational value of 5‐FU+JQ1 and 5‐FU+C297 as potential
   drug combinations for the treatment of 5‐FU‐resistant tumors.

Figure 7.

   Figure 7
   [246]Open in a new tab

   JQ1 and C297 reverses 5‐FU resistance in PDX models of GC. A) Images of
   NR3C1^Hi and NR3C1^Lo PDX tumors in 5‐FU group and 5‐FU+JQ1 groups (n =
   3). B) Tumor growth curve of PDX tumors (ANOVA, two‐tailed). C) Tumor
   volume of each group at endpoint (Student's t‐tests, two‐tailed, mean ±
   SD). D) Inhibition rate of 5‐FU+JQ1 in each group at endpoint
   (Student's t‐tests, two‐tailed, mean ± SD). E) H&E, KI‐67 and NR3C1
   staining of NR3C1^Hi and NR3C1^Lo PDX tumors. Scale bars: 50 µm. F)
   Images of NR3C1^Hi and NR3C1^Lo PDX tumors in 5‐FU and 5‐FU+C297 groups
   (n = 3). G) Tumor growth curve of PDX tumors (ANOVA, two‐tailed). H)
   Tumor volume of each group at endpoint (Student's t‐tests, two‐tailed,
   mean ± SD). I) Inhibition rate of 5‐FU+JQ1 in each group at endpoint
   (Student's t‐tests, two‐tailed, mean ± SD). J) Inhibition rate of
   5‐FU+JQ1 and 5‐FU+C297 in each group. NS: not significant, *p < 0.05,
   **p < 0.01, ***p < 0.001, ****p < 0.0001.

3. Discussion

   SEs have intrinsic cellular properties with lineage specificity, and
   many efforts have been made to explore their roles in maintaining cell
   identity^[ [247]^31 , [248]^54 ^] and in tumor development.^[ [249]^55
   , [250]^56 , [251]^57 ^] Studies on GC have focused on revealing SE
   characteristics in different subtypes^[ [252]^16 , [253]^27 ^] and
   analyzing the molecular mechanisms by which SEs promote cell
   proliferation and metastasis.^[ [254]^27 , [255]^39 ^] Recent studies
   have shown SE‐mediated resistance to platinum and molecular inhibitors
   in lung and ovarian cancers.^[ [256]^58 , [257]^59 ^] As a
   chemotherapeutic drug, 5‐FU still has an unsatisfactory response in the
   treatment of patients with advanced GC. Few studies have elaborated on
   the role of SEs in 5‐FU resistance in GC. This study described the SE
   landscape and CRC involving SE‐driven TFs in 5‐FU‐reluctant GC cells.
   NR3C1 played a prominent role in transcriptional regulation and NR3C1
   knockdown improved 5‐FU sensitivity. This study also found that new
   enhancers emerged in 5‐FU resistant cells to activate NR3C1
   transcription, suggesting that GC cells promoted NR3C1 transcription
   through inherent SEs or through the emergence of new SEs to resist 5‐FU
   (Figure [258]2).

   SEs and their target genes can be brought together through the flexible
   rotation of chromatin without a physical distance limit.^[ [259]^30 ,
   [260]^31 , [261]^32 ^] This spatial DNA structure is present in protein
   LLPS droplets, which contain many concentrated TFs, the transcriptional
   central controller‐Mediator complex, RNA polymerase II (RNA pol II),
   and the important transcription regulatory factor BRD4. BRD4 and TFs
   recognize specific DNA regions, the Mediator complex and RNA pol II
   initiate transcription, and TFs further promote transcription.^[
   [262]^30 , [263]^31 ^] The NR3C1 protein was observed to form dynamic
   droplets in tubes and nuclei (Figure [264]4), consistent with previous
   reports that the DNA‐binding domain of NR3C1 directed the selective
   partitioning of co‐regulators and protein condensation.^[ [265]^60 ^]
   Moreover, we found that NR3C1 bound to BRD4 and MED1 and that NR3C1
   droplet formation was inhibited by siBRD4 and siMED1 (Figure [266]4),
   suggesting an important role for the NR3C1‐BRD4‐MED1 complex integrity
   in LLPS of the NR3C1 protein. Furthermore, the NR3C1 binding region
   significantly overlapped with more SEs than TEs, and NR3C1 knockdown
   significantly reduced the SE composition of cells and inhibited the
   overall transcription of SRGs (Figure [267]5), indicating that NR3C1
   participated in SE composition and maintenance via NR3C1‐BRD4‐MED1
   droplets.

   The NR3C1 gene encodes the glucocorticoid receptor that binds to
   glucocorticoid response elements. NR3C1 can activate gene transcription
   in a hormone‐independent manner.^[ [268]^61 ^] Dexamethasone and its
   analogs are commonly used to alleviate inflammation and adverse
   reactions to 5‐FU and other chemotherapeutic drugs for GC and other
   cancers. In addition to inhibiting the immune‐inflammatory response,^[
   [269]^61 ^] NR3C1 mediates resistance to platinum and small‐molecule
   drugs in lung and ovarian cancers.^[ [270]^62 , [271]^63 ^] Recently,
   it was reported that glucocorticoids bound to NR3C1 and promoted
   neutrophil extracellular traps and cancer metastasis.^[ [272]^64 ^] In
   GC, we found that SE‐driven NR3C1 served as the master TF promoting
   5‐FU resistance, bringing out the caution of dexamethasone application
   in patients with GC undergoing chemotherapy.

   The BRD4 inhibitor JQ1 is a broad‐spectrum epigenetic modifier that
   obstructs transcription complex formation in SEs, and it has been
   reported that JQ1 can re‐sensitize drug‐resistant patients.^[ [273]^65
   , [274]^66 , [275]^67 ^] This study preliminarily demonstrated the
   phenomenon of JQ1 sensitizing 5‐FU resistant cells. The clinical safety
   of combining the broad‐spectrum inhibitor JQ1 with chemotherapeutic
   drugs needs to be tested in future clinical trials. As reported in this
   study, the NR3C1 specific inhibitor Cort108297 also re‐sensitized 5‐FU
   resistant cells and PDXs, and it may be an alternative for the 5‐FU
   combination, requiring further exploration. The oriented release of
   NR3C1 inhibitor inside the tumor may also solve concerns regarding
   tumor chemoresistance caused by steroid administration. Multiple
   clinical trials on BRD4 inhibitors and NR3C1‐specific inhibitors have
   been carried out (such as [276]NCT04309968: SYHA1801 in advanced solid
   tumors, [277]NCT03205176: AZD5153 in refractory solid tumors, and
   [278]NCT03776812: relacorilant in recurrent platinum‐resistant ovarian,
   fallopian tube, or primary peritoneal cancer), with drug efficacy
   results eagerly awaited. BRD4 inhibitors and NR3C1‐specific inhibitors
   require further clinical trials to test their efficacy in GC.

   This study aimed to elucidate the mechanism of the key TF NR3C1 but
   fell short of delving deeper into the molecular mechanism of core
   regulatory circuitry (CRC) underlying 5‐FU resistance. In this study,
   NR3C1 was depleted using shRNA viruses. Specifically, shRNA‐1 and
   shRNA‐3 were selected because of their strong inhibitory effects on
   NR3C1 expression. However, it is better to use sgRNAs (CRISPR tools)
   that can completely deplete NR3C1 expression. In the future, NR3C1
   sgRNAs will be used for the further exploration of NR3C1 function in
   GC.

4. Conclusions

   In conclusion, we characterized the SE landscapes of 5‐FU sensitive and
   resistant GC cells and found that SE‐driven NR3C1 mediated 5‐FU
   resistance. Together with BRD4 and MED1, NR3C1 formed phase separation
   in the nucleus and promoted the transcription of downstream genes,
   leading to 5‐FU resistance. NR3C1 knockdown significantly changed
   cellular SE composition and inhibited SE function (Figure [279]8 ). The
   epigenetic reader inhibitor JQ1 and NR3C1 specific inhibitor Cort108297
   effectively reversed 5‐FU resistance in GC by destroying SE or
   inhibiting NR3C1. Our findings provide a promising drug combination
   scheme for patients with GC treated with 5‐FU‐basic chemotherapy to
   enhance chemotherapy efficiency.

Figure 8.

   Figure 8
   [280]Open in a new tab

   SE‐driven NR3C1 formed phase separation promoting 5‐FU resistance,
   which was destroyed by JQ1 and Cort108297.

5. Experimental Section

Antibodies and Reagents

   The antibodies against Ki‐67(ab16667), MED1 (ab243893), and GAPDH
   (ab8245) were purchased from Abcam (Cambridge, UK). Antibody against
   NR3C1 (12 041), BRD4 (13 440), H3K27Ac (8173), and SimpleChIP Plus
   Enzymatic Chromatin IP Kit (Magnetic Beads) (9005) were purchased from
   Cell Signaling Technology (Boston, MA, USA). Antibodies against NR3C1
   (sc‐393232) were purchased from Santa Cruz (Dallas, TX, USA). A Cell
   counting Kit‐8 (CCK‐8, CK04) was obtained from Dojindo Molecular
   Technologies (Kumamoto, Japan). CellTiter‐Glo Luminescent Cell
   Viability Assay was obtained from Promega (Madison, WI, USA). JQ1
   (T2110) and Cort108297 ([281]T15000) were obtained from TargetMol
   (Boston, MA, USA).

Cell Culture

   Human gastric cancer cells AGS, HS‐746T, and NCI‐N87 were purchased
   from the American Type Culture Collection (ATCC, USA), and MGC803,
   MKN28, MKN45, and HGC27 cells were purchased from Shanghai Institutes
   for Biological Sciences, Chinese Academy of Sciences. Cells were
   authenticated by STR profiling and free of mycoplasma contamination.
   Cells cultured in RPMI‐1640 and DMEM medium (Meilunbio, Dalian, China)
   containing 10% fetal bovine serum (Gibco, Grand Island, NY, USA), 100 U
   mL^−1 of penicillin, and 100 µg mL^−1 of streptomycin, in a humidity
   culture incubator at 37 °C with 5% CO[2].

Dose‐Effect Curves

   Cells were plated into 96‐well plates in advance. Medium with
   5‐FU/JQ1/C297 at concentrations of different gradients was added to
   wells and the medium was replaced every day with fresh medium with
   drugs. After 4 days of culture, the CCK‐8 kit was used according to the
   manufacturer's instructions. Drug concentrations were converted to
   Log10C values, and cell viability was normalized based on OD450 values
   of cells with 0 mg mL^−1 drug. After applying GraphPad Prism to conduct
   a nonlinear regression analysis, the IC[50] values were calculated.

ChIP‐Seq

   ChIP was performed using 10 × 10^6 cross–linked cells, and sequencing
   libraries were prepared as previously described.^[ [282]^68 ^] The
   following antibodies were used for ChIP: rabbit anti‐H3K27Ac (CST).
   ChIP‐seq libraries were sequenced on Illumina HiSeq. Analysis was
   performed as previously described with some modifications.^[ [283]^68
   ^] ChIP‐Seq was performed by Genefund Biotech (Shanghai, China).

   The SEs were identified with the rank ordering of super‐enhancers
   (ROSE) algorithm to obtain binding peaks (call peak) within 12.5 kb^[
   [284]^17 ^] using ChIP‐Seq data. SRGs and TE‐related genes were
   calculated by “ROSE_geneMapper”.

Library Construction and RNA‐Seq

   Total RNA was extracted from the tissue using TRIzol Reagent according
   to the manufacturer's instructions (Invitrogen) and genomic DNA was
   removed using DNase I (TaKara). Then RNA quality was determined by 5300
   Bioanalyser (Agilent) and quantified using the ND‐2000 (NanoDrop
   Technologies). Only high‐quality RNA samples (OD260/280 = 1.8–2.2,
   OD260/230 ≥ 2.0, RIN ≥ 6.5, 28S:18S ≥ 1.0, >1 µg) were used to
   construct the sequencing library. RNA‐seq transcriptome library was
   prepared using Illumina Stranded mRNA Prep (San Diego, CA, USA),
   SuperScript double‐stranded cDNA synthesis kit (Invitrogen, CA), and
   Phusion DNA polymerase (NEB) according to the manufacturer's protocol.
   After quantified by Qubit 4.0, the paired‐end RNA‐seq sequencing
   library was sequenced with the NovaSeq 6000 sequencer (Illumina).

   RNA purification, reverse transcription, library construction, and
   sequencing were performed at Shanghai Majorbio Bio‐pharm Biotechnology
   Co., Ltd. (Shanghai, China). Analysis was performed as previously
   described with some modifications.^[ [285]^69 ^]

   Differential gene expression analysis was performed using the DESeq2
   package in R to compare gene expression. Raw read counts obtained from
   high‐throughput sequencing were first processed to remove low‐quality
   reads and align them to the human reference genome (GRCh38). The DESeq2
   package was then utilized to construct a count table and perform size
   factor normalization. Differential expression was assessed using a
   negative binomial generalized linear model. Genes were considered
   significantly differentially expressed with a log2 fold change (log2FC)
   ≥ 1 and a p‐value ≤ 0.05.

Enrichment Analysis of Related Genes

   Functional‐enrichment analysis including Kyoto Encyclopedia of Genes
   and Genomes (KEGG) enrichment and Gene Ontology (GO) were performed at
   Bonferroni‐corrected p‐value ≤ 0.05 by Goatools and KOBAS.^[ [286]^70
   ^] Gene set enrichment analysis (GSEA) analysis was performed to
   determine the significantly enriched pathways based on the MSigDB
   database (c2.cp.kegg.v7.2.symbols.gmt gene sets).

Lentivirus Vectors and Small Interfering RNAs

   For the knockdown of NR3C1, target shRNA sequences were subcloned into
   PGMLV‐HU6‐CMV‐MCS‐ZsGreen1‐PGK‐Puro vector (Genomeditech, Shanghai,
   China). The shRNA sequences of NR3C1 and siRNA sequences of MED1, BRD4,
   and TFs in the CRC are listed in Table [287]S1 (Supporting
   Information). Human full‐length cDNA of NR3C1 was subcloned into
   PGMLV‐CMV‐MCS‐EF1‐mScarlet‐T2A‐Blasticidin lentivirus vector
   (Genomeditech). The transfection process and cell selection were
   performed as previously described.^[ [288]^69 ^]

Western Blotting, RNA Extraction, and Quantitative Real‐Time PCR (qRT‐PCR)

   Western blotting, RNA extraction, and qRT‐PCR were performed as
   previously described.^[ [289]^69 ^] The sequences of the primers are
   listed in Table [290]S2 (Supporting Information). The transcription
   level was normalized to the internal control and determined by a
   2^−ΔΔCT method.

Establishment and Culture of Patient‐Derived Organoids (PDO) from GC Patients

   PDO establishment was performed as previously described.^[ [291]^69 ^]
   The tumor tissues from GC patients were digested with shaking and
   filtered through a 70 µm cell strainer (BD Biosciences). After
   centrifuging, Matrigel was added to the pellet and plated in the
   24‐well plate. The shNR3C1 lentivirus, NR3C1 overexpressed lentivirus,
   and control lentivirus were transfected into organoids. Fluorouracil
   was added and refreshed every day for ten days and CellTiter‐Glo
   Luminescent Cell Viability Assay was used to evaluate the viability.

Fluorescence Recovery after Photobleaching (FRAP)

   FRAP was performed as previously described with some modifications.^[
   [292]^15 , [293]^31 ^] A lentiviral overexpression plasmid for NR3C1
   was generated by cloning the full‐length ORF of the human NR3C1 gene
   ([294]NM_000176.3) into the vector, followed by a 6 amino acid GS
   linker sequence “GSGSGS” and mCherry. FRAP was performed on the ZEISS
   LSM880 Airyscan microscope. The bleach spot was centered on a cluster
   and images were taken at 1 s intervals for 34 s to measure the
   fluorescence recovery in the cluster. The integrated intensity of the
   cluster was determined as a function of time, and background intensity
   was corrected for overall photo‐bleaching based on a reference region
   within the same cell, and was normalized to pre‐bleach intensity.

Co‐Immunoprecipitation (Co‐IP) and Immunoprecipitation‐mass spectrometry
(IP‐MS)

   Immunoprecipitation was performed using Pierce Co‐Immunoprecipitation
   Kit (88804, Thermo Fisher) according to the manufacturer's protocol.
   Antibody against MED1, BRD4, and NR3C1 (10 µg) was covalently
   cross‐linked to magnetic protein A/G beads as the immunoprecipitation
   antibody. Lastly, the antigen was eluted and subjected to protein blot
   analysis with antibodies against MED1, BRD4, and NR3C1. IP‐MS was
   performed as previously described.^[ [295]^69 ^] Protein‐Chip Gold
   Array (Bio‐Rad) with a Bio‐Rad Protein Chip System Series 4000 mass
   spectrometer was used for analysis.

CUT&Tag‐Seq

   CUT&Tag assay was performed as described previously with
   modifications.^[ [296]^71 ^] Briefly, 10^5 cells were washed twice
   gently with wash buffer. 10 µL Concanavalin A coated magnetic beads
   (Bangs Laboratories) were added per sample and incubated at RT for 10
   min. Remove unbound supernatant and resuspended bead‐bound cells with
   dig wash buffer and a 1:50 dilution of primary antibody or IgG control
   antibody (normal mouse IgG: Millipore cat.no.12‐371) incubated on a
   rotating platform overnight at 4 °C. The primary antibody was removed
   using a magnet stand. The secondary antibody (Rabbit Anti‐Mouse IgG
   H&L: Abcam, ab6709) was diluted 1:100 in dig wash buffer and cells were
   incubated for 60 min. Cells were washed using the magnet stand 2–3
   times in dig wash buffer. A 1:100 dilution of pA‐Tn5 adapter complex
   was prepared in dig‐med buffer (0.01% Digitonin; 20 mm HEPES pH 7.5;
   300 mm NaCl; 0.5 mm Spermidine; 1 × Protease inhibitor cocktail) and
   incubated with cells for 1 h. Cells were washed 2–3 × for 5 min in 1 mL
   Dig‐med buffer. Then cells were resuspended in a tagmentation buffer
   (10 mm MgCl2 in Dig‐med Buffer) and incubated at 37 °C for 1 h. DNA was
   purified using phenol‐chloroform‐isoamyl alcohol extraction and ethanol
   precipitation.

   To amplify libraries, 21 µL DNA was mixed with 2 µL of a universal i5
   and a uniquely barcoded i7 primer. A volume of 25 µL NEBNext HiFi 2 ×
   PCR Master mix was added and mixed. The sample was placed in a
   Thermocycler with a heated lid using the following cycling conditions:
   72 °C for 5 min (gap filling); 98 °C for 30 s; 14 cycles of 98 °C for
   10 s and 63 °C for 30 s; final extension at 72 °C for 1 min and hold at
   8 °C. Library clean‐up was performed with XP beads (Beckman Counter).

Combination Index

   Combination index experiments were performed as described previously.^[
   [297]^68 ^] JQ1 and 5‐FU were added with serial two‐fold dilutions.
   CompuSyn was used to generate the combination index values.

Patient‐Derived Xenograft (PDX) Tumor Models

   BALB/c male nude mice (4–6 weeks old, purchased from SPF (Beijing)
   Biotechnology Co., Ltd., Beijing, China) were housed in specific
   pathogen‐free cages and used to construct xenograft models. Animal
   experiments were performed in accordance with the institution's
   guidelines and animal research principles, and daily care has been
   provided. After acclimatization for 3 days in an approved facility
   (water and food at libitum and controlled environmental conditions),
   experiments were conducted.

   PDXs were established with tumors of patients with GC from Ruijin
   Hospital as described previously.^[ [298]^69 ^] For PDX
   transplantation, tumor fragments of 1–2 mm were subcutaneously
   engrafted into nude mice under anesthesia. When tumors reached
   over100 mm^3, the mice of two cases were separated into groups for
   treatment with 5‐FU (20 mg kg^−1, intraperitoneally), and 5‐FU+JQ1 (50
   mg kg^−1, intraperitoneally). Dexamethasone (0.1 mg kg^−1,
   intraperitoneally) or C297 (20 mg kg^−1, intraperitoneally) was given
   1 h before chemotherapy administration. Treatment was administered 5
   days consecutively per week. Tumor volumes were measured using calipers
   every 4 days and calculated with the formula: volume (mm^3) = length ×
   width × width/2. If volumes were over 4000 mm^3 or the diameter in
   either dimension was greater than 20 mm (humane endpoint), mice were
   euthanized. Otherwise, mice were euthanized under anesthesia after the
   whole treatment, and tumors were removed and weighed.

Hematoxylin and Eosin (H&E) Analysis and Immunohistochemistry (IHC)

   Tumors collected from mice were fixed with 4% paraformaldehyde.
   Paraffin‐embedded samples were stained with H&E (Sigma). The stained
   sections were viewed and photographed under a microscope. IHC assay was
   performed as previously described according to the manufacturer's
   protocol (Immunostain SP kit, DakoCytomation, USA). Anti‐NR3C1 antibody
   (1:100), and anti‐Ki67 antibody (1:3000) were used.

DNA‐FISH Combined with Immunofluorescence

   Immunofluorescence was performed as previously described.^[ [299]^68 ^]
   After incubating the cells with the secondary antibodies, cells were
   washed, fixed with 4% PFA, and washed again. Cells were incubated in
   70% ethanol, 85% ethanol, and then 100% ethanol. A probe hybridization
   mixture was made by mixing 7 µL of FISH Hybridization Buffer, 1 µL of
   FISH probes, and 2 µL of water. 5 µL of mixture was added on a slide
   and a coverslip was placed on top. The coverslip was sealed using
   rubber cement. Genomic DNA and probes were denatured at 78 °C for 5 min
   and slides were incubated. The coverslip was removed from the slide and
   incubated in a pre‐warmed wash buffer. Air dry slides and Hoechst stain
   nuclei. Images were acquired at the Zeiss LSM880 Airscan confocal
   microscope with 63 × objective.

   DNA FISH probes were custom‐designed and generated by Wuhan Boerfu
   Biotechnology Co., Ltd. to target NR3C1 super‐enhancers. Design Input
   Region – hg38 chr5:143275273‐143391380.

Specimens and Ethics

   The study received approval from the Human Research Ethics Committee of
   Ruijin Hospital (Approval No. 2017‐6), in adherence to ethical
   guidelines outlined in the Declaration of Helsinki. Excluding unpaired
   tumor‐normal samples and samples of incomplete clinical information,
   specimens of tumor and adjacent non‐tumor tissues from gastric cancer
   patients with intact clinical information who underwent D2 gastrectomy
   were collected. Patients who underwent preoperative treatment were
   excluded from our study. Animal experiments were conducted in
   compliance with animal use guidelines and approved by the Laboratory
   Animal Ethics Committee of Ruijin Hospital (Approval No.
   SYXK2023‐0038).

Statistical Analysis

   All statistical calculations were performed using GraphPad Prism
   (version 8.0). All results were presented as the mean ± standard
   deviation (SD) from triplicate experiments. The sample size for each
   statistical analysis was indicated in the figure legends. Student's
   t‐tests (two‐tailed) were applied to assess the statistical
   significance for gene expression (qPCR and WB), ChIP‐qPCR, dual
   luciferase assays, droplet formation in tubes, and tumor weight. The
   Spearman correlation test was performed to calculate correlation
   efficient of mRNA level and IC[50] values. Cox regression analysis was
   performed for NR3C1 survival analysis. Comparisons among multiple
   groups were analyzed by ANOVA. Statistical tests for RNA‐Seq were
   conducted using R (version 4.2.1), and the DESeq2 package (version
   1.38.3) was employed for the differential expression analysis.
   Differences with a p‐value < 0.05 were considered statistically
   significant.

Ethics Approval Statement

   This study was approved by the Human Research Ethics Committee of
   Ruijin Hospital (Approval No. 2017‐6), and adhered to the ethical
   guidelines outlined in the Declaration of Helsinki. Animal experiments
   were conducted in compliance with the animal use guidelines and
   approved by the Laboratory Animal Ethics Committee of Ruijin Hospital
   (Approval No. SYXK2023‐0038).

Conflict of Interest

   The authors declare no conflict of interest.

Author Contributions

   The conceptualization was done by B.L. Data duration was handled by
   J.Y. Formal analysis was conducted by J.Y., M.C., and Q.S. Methodology
   was developed by H.F., F.L., Z.X., and B.Y. Investigation was carried
   out by J.Y., M.C., and Q.S. Resources were provided by C.H., C.Y.,
   Z‐g.Z., J.X., J.L., L.S., and W.D. Software development was done by
   M.C., Q.S., Y.C., and W.D. Supervision was provided by B.L. and Y‐Y.L.
   Validation was performed by F.L. and M.C. The original draft was
   written by J.Y., and the review and editing were done by B.L., Y.Y.L.,
   and Y.C. All the authors read and approved the final manuscript.

Supporting information

   Supporting Information
   [300]ADVS-12-2409050-s004.docx^ (9.2MB, docx)

   Supplemental Movie 1
   [301]Download video file^ (6.5MB, avi)

   Supplemental Movie 2
   [302]Download video file^ (6MB, avi)

   Supporting Information
   [303]ADVS-12-2409050-s003.zip^ (73.8MB, zip)

Acknowledgements