SP1 undergoes phase separation and activates RGS20 expression through super-enhancers to promote lung adenocarcinoma progression

Significance

As a transcription factor, specificity protein 1 (SP1) has been reported to play a regulatory role in cancers. In this study, we demonstrated that SP1 undergoes phase separation and that its zinc finger 3 is essential for both SP1 condensation and the malignant phenotypes of lung adenocarcinoma (LUAD) cells. Using the Cleavage Under Targets & Release Using Nuclease (CUT&RUN) approach, we identified regulator of the G protein signaling 20 (RGS20) as a potential target gene regulated by SP1 through super-enhancer (SE) mechanisms. The demethylase inhibitor glycogen synthase kinase J4 (GSK-J4) can disrupt SP1 phase-separated condensation and abolish SP1-induced RGS20 expression. Overall, this research significantly contributes to the understanding of SP1-mediated transcriptional regulation and provides insights into how SP1 promotes the malignant progression of LUAD through liquid–liquid phase separation and SE mechanisms.

Abstract

Lung adenocarcinoma (LUAD) is the leading cause of cancer-related death worldwide, but the underlying molecular mechanisms remain largely unclear. The transcription factor (TF) specificity protein 1 (SP1) plays a crucial role in the development of various cancers, including LUAD. Recent studies have indicated that master TFs may form phase-separated macromolecular condensates to promote super-enhancer (SE) assembly and oncogene expression. In this study, we demonstrated that SP1 undergoes phase separation and that its zinc finger 3 in the DNA-binding domain is essential for this process. Through Cleavage Under Targets & Release Using Nuclease (CUT&RUN) using antibodies against SP1 and H3K27ac, we found a significant correlation between SP1 enrichment and SE elements, identified the regulator of the G protein signaling 20 (RGS20) gene as the most likely target regulated by SP1 through SE mechanisms, and verified this finding using different approaches. The oncogenic activity of SP1 relies on its phase separation ability and RGS20 gene activation, which can be abolished by glycogen synthase kinase J4 (GSK-J4), a demethylase inhibitor. Together, our findings provide evidence that SP1 regulates its target oncogene expression through phase separation and SE mechanisms, thereby promoting LUAD cell progression. This study also revealed an innovative target for LUAD therapies through intervening in SP1-mediated SE formation.

Lung cancer is a prevalent and deadly type of malignancy worldwide (1). Among the different subtypes of lung cancer, lung adenocarcinoma (LUAD) accounts for approximately 40% of lung cancer cases (2). Despite recent advancements in targeted therapies and immunotherapies, the recurrence and metastasis of tumors continue to contribute to the low survival rates of lung cancer patients (3). Hence, gaining a comprehensive understanding of the molecular mechanisms involved in the development and progression of LUAD is crucial for improving the diagnosis and treatment of this specific type of cancer.

Cellular liquid–liquid phase separation (LLPS) refers to the formation of membraneless liquid phase compartments in eukaryotic cells through the aggregation of biological macromolecules such as nucleic acids and proteins (4). This phenomenon plays a crucial role in various biological and pathological processes, including transcription, tumorigenesis, chromatin organization, the DNA damage response, ferroptosis, and autophagy. Researchers have demonstrated that LLPS can generate specialized condensates to carry out various functions (5–10). In 2017, Sharp and Young proposed the hypothesis that super-enhancers (SEs) are involved in gene regulation through a phase separation mechanism (11). Subsequent investigations using advanced microscopic imaging corroborated the formation of phase separation aggregates at SEs, which could enrich or compartmentalize transcriptional coactivators at the chromatin loci of target genes. Many transcription factors (TFs) capable of undergoing phase separation contain intrinsically disordered regions (IDRs) that play crucial roles in the protein condensation process (12). The cocondensation of master TFs with transcription mediators, coactivators, and other essential components of the transcriptional machinery can result in the assembly of SEs with the participation of many enhancer elements on chromatin and robustly activate downstream target genes, especially those with proliferative or oncogenic activities (13–15). Thus, intervention in LLPS processes that promote oncogene activation represents a promising direction for identifying alternative cancer therapeutics (16).

As one of the first identified TFs, specificity protein 1 (SP1) belongs to the SP/Kruppel-like factor family (17). Previous studies have indicated that SP1 regulates various biological processes, including cancer cell proliferation, invasion, and metastasis (18–20). Consistently, increased SP1 expression has been reported in a number of cancers compared with matched normal tissues and is associated with reduced survival or other adverse clinical outcomes in lung cancer patients (21, 22). As a master TF, SP1 activates the expression of numerous tumor-related genes, such as epidermal growth factor receptor and matrix metallopeptidases (23, 24). Recently, increasing evidence has suggested that phase separation plays a crucial role in transcription initiation. Therefore, an important inquiry has arisen regarding whether SP1 may undergo phase separation, promote SE formation when regulating the transcription of cancer-stimulating genes, and subsequently promote LUAD progression.

In this study, we identified multiple IDRs in the SP1 protein sequence and observed SP1 phase-separated condensation. However, our data indicated that zinc finger 3 (ZF3) of the C-terminal DNA-binding domain (DBD), rather than the N-terminal IDRs, is essential for SP1 phase separation. Deletion of ZF3 also abolished the ability of SP1 to promote LUAD cell proliferation, migration, invasion, and xenograft tumor growth. Through CUT&RUN-seq analysis, we found a significant association between SP1 enrichment and SE formation and identified the regulator of the G protein signaling 20 (RGS20) gene as the most likely target of SP1. We verified the crucial role of RGS20 in SP1-regulated LUAD cell malignancy. Additionally, we identified that GSK-J4, an inhibitor of H3K27me3 demethylase, could dampen SP1 phase separation and negate its proliferative activity. This study both expanded our understanding of SP1-regulated transcription and provided insights into the design of advanced strategies for LUAD treatment.

Results

SP1 Expression in LUAD Tissues Is Associated with Tumor Progression and Poor Patient Prognosis.

To evaluate the contribution of SP1 to LUAD progression, we first analyzed publicly available SP1 expression datasets from the Gene Expression Omnibus (https://www.ncbi.nlm.nih.gov/geo/). Our results revealed that SP1 expression in LUAD tissue samples was significantly greater than that in normal lung tissue samples (SI Appendix, Fig. S1A). Consistently, analysis of SP1 protein levels in the CPTAC database (https://cptac-data-portal.georgetown.edu/datasets) also revealed a significant increase in SP1 protein levels in LUAD tissues compared to normal tissues (SI Appendix, Fig. S1B). Using the Kaplan–Meier plotter database (https://kmplot.com/analysis/) (25), we observed that SP1 expression was positively correlated with the progression and negatively correlated with the overall survival of LUAD patients (SI Appendix, Fig. S1 C and D). Next, we carried out an SP1 immunohistochemical (IHC) study using a tissue array consisting of 126 samples from LUAD patients and 30 normal lung tissue samples. The results indicated that the level of the SP1 protein in LUAD samples significantly increased and was positively correlated with clinical stage (SI Appendix, Fig. S1E). Additionally, Kaplan–Meier plot analyses revealed that patients with relatively high SP1 levels exhibited shorter disease-free survival and overall survival than those with low SP1 levels (P = 0.019 and 0.006, respectively; SI Appendix, Fig. S1 F and G). We also carried out western blot analysis to examine SP1 levels in a human bronchial epithelial (HBE) cell line and six LUAD cell lines. Most cancerous cell lines, except H1650, showed relatively high SP1 expression compared with that in HBE cells (SI Appendix, Fig. S1H). Together, our results indicated that SP1 expression is elevated in LUAD samples and is associated with a poor patient prognosis.

To investigate the biological role of SP1 in lung cancer development, we manipulated SP1 expression in LUAD cells. We first designed three SP1 siRNAs (si-SP1-1, 2, 3), as well as a control siRNA (si-cont) (the target sequences are listed in SI Appendix, Table S1), and used them to transfect PC-9 and H1299 cells that showed high levels of endogenous SP1. Western blot analysis revealed that compared with si-cont, si-SP1-2 efficiently reduced the SP1 level in both cell lines (SI Appendix, Fig. S2A). Functionally, SP1 knockdown mediated by si-SP1-2 inhibited the viability, proliferation, migration, and invasion of LUAD cells (SI Appendix, Fig. S2 B–E). Next, we tested the effect of increased SP1 expression on H1650 cells that showed low endogenous SP1 levels. Exogenous SP1 was introduced into H1650 cells by lentiviral infection (SI Appendix, Fig. S3A). Consistent with the results of the knockdown study, ectopic SP1 expression increased H1650 cell viability, proliferation, migration, and invasion compared to an empty vector (SI Appendix, Fig. S3 B–E). Overall, our data demonstrated that SP1 promotes the malignant properties of LUAD cells.

SP1 Undergoes LLPS in Cells.

To investigate the mechanism underlying SP1-promoted LUAD progression, we asked whether SP1 possesses any special feature that could contribute to its functionality. Recent studies have indicated that many TFs may form phase-separated condensates to exert their transcriptional regulatory effects. Therefore, we asked whether SP1, as a master TF, could also undergo phase separation. Biomolecular condensates formed through LLPS can maintain locally elevated concentrations of resident proteins, often displaying intracellular punctate structures that can be detected through microscopy (26, 27). In our study, we first examined the subcellular distribution of endogenous SP1 and observed the formation of nuclear puncta in H1299 and PC-9 cells (Fig. 1A). Next, we ectopically expressed an enhanced green fluorescent protein (EGFP)-SP1 fusion protein in these two LUAD cell lines. As predicted, we observed distinct condensate puncta (Fig. 1 B, Top row). Notably, these EGFP-SP1 droplets largely diffused in the presence of 1,6-hexanediol (1,6-hex), a chemical that disrupts the weak hydrophobic interactions required for protein LLPS (Fig. 1 B, Bottom). To evaluate the molecular dynamics of these SP1 puncta, we carried out a fluorescence recovery after photobleaching (FRAP) assay. Photobleached spots in EGFP-SP1 puncta showed rapid fluorescence recovery (Fig. 1C), indicating rapid molecular motility in these condensates. Adenosine triphosphate (ATP) reportedly acts as a biological hydrotrope and promotes protein solubility (28). Moreover, ATP exhibited dual effects on protein phase separation and regulated molecular dynamics in the condensates (29). In our FRAP experiment, photobleached EGFP-SP1 puncta showed retarded fluorescence recovery in a cellular environment with ATP depletion by glucose deprivation and oligomycin/2-deoxyglucose treatment (Fig. 1D). Collectively, these findings strongly suggest that SP1 protein localized in the nucleus forms liquid–liquid phase-separated condensates.

Fig. 1.

SP1 nuclear condensation and FRAP studies. (A) Immunofluorescence staining of endogenous SP1 protein in H1299 and PC-9 cells using an SP1 antibody. (B) EGFP-SP1 punctum formation in live H1299 and PC-9 cells treated with 5% 1,6-hex or vehicle. Nuclei were visualized by Hoechst staining. (C) FRAP studies of EGFP-SP1 puncta in H1299 and PC-9 cells. Red squares depict photobleached puncta, and green squares depict control puncta. FRAP quantification is shown on the Right. (D) Representative images of FRAP studies of EGFP-SP1 in H1299 and PC-9 cells under ATP depletion conditions. FRAP quantification is shown on the Right.

The C-terminal Disordered Regions of SP1 Determine Its Phase-Separated Condensation.

IDRs of proteins are conformationally flexible, allowing protein interactions with multiple partners through both intramolecular and intermolecular mechanisms. These structurally heterogeneous protein domains possess diverse functions (30, 31). Recent studies have indicated that the low complexity of IDRs contributes to LLPS (32–35). Using the PONDR algorithm (http://www.pondr.com/) (36), we identified five potential IDR segments in the SP1 protein sequence (Fig. 2 A, Top row). Due to the relatively high molecular weight (130 kDa) of the EGFP-SP1 protein, it was difficult to obtain highly purified full-length recombinant protein. Therefore, to evaluate the ability of these IDRs to undergo phase separation, we generated six His×6-tagged SP1 mutants with EGFP at their N termini. Three of these mutants containing the N-terminal (with IDRs 1 and 2), middle (IDR 3), and C-terminal (IDRs 4 and 5) regions were designated EGFP-SP1-N, M, and C, respectively (Fig. 2 A, Lower). The other three mutants included EGFP-SP1ΔN, ΔM, and ΔC, with these regions individually deleted. The recombinant proteins of these SP1 mutants were purified after being expressed in bacteria (SI Appendix, Fig. S4A). Among these six SP1 mutants, EGFP-SP1-C, EGFP-SP1ΔN, and ΔM were able to generate droplets in vitro in the presence of polyethylene glycol 8000 (PEG 8000), which is commonly used as a crowding agent to simulate the intracellular environment (37) (SI Appendix, Fig. S4B). Under the same conditions, EGFP-SP1-M produced very small droplets, while EGFP-SP1-N and EGFP-SP1ΔC did not form any detectable condensates. However, when these six SP1 mutants were transfected into LUAD cells, EGFP-SP1ΔN and ΔM retained the ability to form nuclear puncta, but EGFP-SP1-C exhibited a diffuse green fluorescence signal in the nucleus, and EGFP-SP1-N, M, and EGFP-SP1ΔC were dispersed in both the nucleus and cytoplasm without forming any condensate (SI Appendix, Fig. S4C). Thus, the IDRs in the C terminus of SP1 determine its nuclear localization and nuclear punctum formation.

Fig. 2.

Analysis of IDRs of SP1 and examination of their droplet formation. (A) SP1 domain structure, IDR graph, and IDR mutants. The SP1 domain structure is shown in the Top panel. BTD: Buttonhead domain. The graph of IDRs in the SP1 protein (Middle) was analyzed using the PONDR algorithm. Scores > 0.5 indicate disorder. The black lines depict the predicted IDRs. In the Bottom panel, schematic diagrams of six EGFP-SP1 mutants are presented. SP1-N, M, and C contain amino acids 1 to 269, 266 to 530, and 515 to 785, respectively. (B) Representative fluorescence and differential interference contrast (DIC) images of EGFP-SP1-C droplets at different protein concentrations in a buffer containing 125 mM NaCl and 10% PEG-8000 (the same conditions are used hereafter, if not specified). The quantification of droplet numbers and areas is shown in the Bottom panel. (C) Turbidity visualization of EGFP-SP1-C droplet formation. Tubes containing EGFP-SP1-C (10 μM, the same concentration hereafter, if not specified) in the buffer containing PEGs with increasing molecular weights. (D) Phase diagram of turbidity changes of EGFP-SP1-C droplet formation in buffers containing different PEGs. The absorbance was measured at 600 nm. (E) Representative fluorescence and DIC images of EGFP-SP1-C droplets at different NaCl concentrations. The quantification of droplet numbers and areas is shown in the Right panel. (F and G) EGFP-SP1-C droplet formation at different temperatures (F) and in the presence or absence of 5% 1,6-hex (G). Quantification is shown in the Right panel. In F and G, the data are presented as the mean ± SD, and the statistical significance was calculated by one-way ANOVA with Tukey’s post hoc test (F) or Student’s t test (G). *P < 0.05, **P < 0.01, and ***P < 0.001.

We further characterized the in vitro phase separation of EGFP-SP1-C and found that the numbers and sizes of its droplets were generally proportional to the protein concentration (Fig. 2B). The turbidity of the EGFP-SP1-C solutions also increased with increasing molecular weight of the PEGs in the buffers (Fig. 2C). As quantified by the optical density at 600 nm, the turbidity of the samples monotonically increased with increasing PEG molecular weight and EGFP-SP1-C concentration in the solutions (Fig. 2D). Additionally, EGFP-SP1-C condensation varied at different sodium chloride (NaCl) concentrations, and maximal droplet formation based on their numbers and sizes was observed at 150 mM NaCl (Fig. 2E), close to its physiological level. Moreover, the ability of EGFP-SP1-C to form droplets was greater at 37 °C than at 25 °C and markedly decreased at 4 °C (Fig. 2F). Furthermore, 1,6-hex effectively reduced the numbers and sizes of the EGFP-SP1-C droplets (Fig. 2G). Collectively, our findings suggest that SP1 undergoes phase-separated condensation in vitro under conditions close to physiological conditions.

SP1-ZF3 Is Required for Phase Separation and Essential for the Tumorigenic Properties of LUAD Cells.

The C-terminal region of SP1 contains several featured motifs or domains, including a buttonhead (BTD) domain that plays a critical role in promoter selective transcriptional synergy (38), a ZF domain consisting of three ZFs responsible for DNA binding, and domain D, which is required for transcriptional activation (39). To investigate the contribution of special C-terminal regions to SP1 phase separation, we generated the EGFP-SP1 mutants ΔBTD, ΔZF1, ΔZF2 and ΔZF3 with these regions individually deleted (SI Appendix, Fig. S5A) and expressed them in H1299 and PC-9 cells. Compared with the wild-type (WT) SP1, all four deletion mutants showed reduced nuclear condensation with an increased fluorescent background signal (SI Appendix, Fig. S5B). In particular, ZF3 deletion abolished the ability of SP1 to form nuclear condensates; moreover, the SP1ΔZF3 mutant was partially distributed in the cytoplasm, although the majority of the protein was still in the nucleus. These data suggested that ZF3 plays a crucial role in both the nuclear localization and phase-separated condensation of SP1.

SP1 has been reported to promote malignant progression in various cancers. To examine the contribution of phase separation to its biological function, we first generated a shRNA targeting the 3UTR of the SP1 messenger RNA (mRNA) (sh-SP1-3UTR) that silences only endogenous SP1 but not ectopically expressed SP1 without the 3′-UTR sequence. Since deletions of BTD and ZF3 exhibited the most pronounced adverse effects, we concentrated on these two mutants in subsequent functional studies.

H1299 and PC-9 cells were infected with lentivirus carrying EGFP-SP1 WT, ΔBTD, and ΔZF3, as well as an empty vector as a control, with simultaneous knockdown of endogenous SP1 by sh-SP1-3UTR. The silencing of endogenous SP1 and the expression of exogenous SP1 WT or mutants were verified by western blot analysis (Fig. 3A). In a cell viability study using a CCK8 assay, loss of ZF3 completely abolished the ability of SP1 to restore the viability of both cell lines compared to that of WT SP1 (Fig. 3B), indicating the indispensability of ZF3 for SP1 function. However, EGFP-SP1ΔBTD retained a certain degree of SP1 activity and partially restored the loss of cell viability caused by SP1 silencing at most time points, especially in PC-9 cells. Consistent with the results of the viability assays, EGFP-SP1ΔBTD and ΔZF3 showed a reduced ability to reverse the SP1 knockdown-induced retardation of LUAD cell proliferation, migration, and invasion (Fig. 3 C–E).

Fig. 3.

Effects of different SP1 deletions on its biological functions. (A) Expression of SP1 deletion mutants in H1299 and PC-9 cells with simultaneous endogenous SP1 knockdown. A lentivirus carrying sh-SP1 targeting the 3′-UTR of the SP1 mRNA (sh-SP1-3UTR) was used to knock down endogenous SP1, and exogenous SP1 WT and mutants were also generated by lentiviral infection. Western blotting was used to analyze endogenous and exogenous SP1 expression. (B) Viability analyses of H1299 or PC-9 cells expressing WT or mutant SP1 with endogenous SP1 knockdown. CCK-8 assays were used to determine cell survival. (C) Cell proliferation analyses using EdU in H1299 or PC-9 cells expressing WT or mutant SP1 with endogenous SP1 knockdown. (D) Wound healing assays were performed in cells expressing WT or mutant SP1 with endogenous SP1 knockdown. The quantification is shown in the Right panel. (E) Transwell assays were used to evaluate the migration (Top) and invasion (Bottom) of H1299 or PC-9 cells with endogenous SP1 knockdown expressing WT or mutant SP1. The quantification is shown in the Bottom panel. In B–E, the data are presented as the mean ± SD and were compared by one-way ANOVA with Tukey’s post hoc test. *P < 0.05, **P < 0.01, ***P < 0.001, ns: not significant.

The H3K27me3 Demethylase Inhibitor GSK-J4 Blocks SP1 Condensation.

A previous study in osteosarcoma cells indicated that GSK-J4, an H3K27 demethylase inhibitor, disrupted HOXB8 and FOSL1 protein condensation by directly binding to their IDRs (40). When testing the effect of GSK-J4 on SP1 phase separation in LUAD cells, we observed that GSK-J4 could abolish EGFP-SP1 puncta formation in the nuclei of both H1299 and PC-9 cells (SI Appendix, Fig. S6A). Consistently, we observed reduced in vitro EGFP-SP1-C droplet formation upon treatment with GSK-J4 (SI Appendix, Fig. S6B). In these LUAD cells, GSK-J4 treatment reversed the increase in cell viability and proliferation caused by ectopic SP1 expression (SI Appendix, Fig. S6 C and D). Consistently, GSK-J4 also reduced the mobility and invasion of LUAD cells expressing exogenous SP1 (SI Appendix, Fig. S6 E and F). Together, our findings suggested that GSK-J4 attenuated the malignant properties of LUAD cells, likely through interfering with SP1 phase separation.

SP1 Condenses at SEs of the RGS20 Gene Locus to Promote Its Transcription.

A number of previous studies have indicated that TFs undergo phase-separated condensation that compartmentalizes transcription coactivators and the transcriptional machinery to form enhancers or even SEs and subsequently activate target gene expression (41–45). To investigate the potential of SP1 to regulate this type of gene expression in LUAD, we conducted CUT&RUN-seq assays in H1299 cells using an SP1 antibody and an H3K27ac antibody and analyzed the enrichment of SP1 binding and H3K27ac modification, which are well-defined markers of enhancers and SEs (46). Based on the H3K27ac dataset, we carried out rank ordering of SEs (ROSE) analyses, which can determine chromatin regions containing putative SEs versus typical enhancers, and identified a total of 156 SEs in H1299 cells (Fig. 4A). Compared to typical enhancers in the CUT&RUN dataset, SE elements showed specific enrichment of H3K27ac signals in terms of distance and density (Fig. 4 B, Left and Middle). We also observed enrichment of SP1 in the regions close to the transcription start sites (TSSs) of 6621 genes (Fig. 4 B, Right). Based on these data, we identified 38 genes with overlapping enrichment of both the SP1 signal and SE potential (Fig. 4C), suggesting potential regulation of these genes by SP1 through SE formation. Among these 38 genes, 13 were excluded due to lack of expression or correlation with noncoding RNAs since we focused on protein-coding genes in the present study. Additionally, to prioritize these genes for further investigation, we selected genes that met two criteria: first, elevated expression in LUAD samples versus normal lung tissues (SI Appendix, Fig. S7 A and B) with prognostic significance, and second, a significant positive correlation with a poor clinical outcome in LUAD patients (SI Appendix, Fig. S7 B and C). We identified three differentially expressed and potentially SE-associated genes, ERRFI1, RGS20, and CSNK1A1, as candidates for further analysis (Fig. 4D). To further elucidate the most likely target genes potentially regulated by SP1 through the SE mechanism, we treated H1299 and PC-9 cells with inhibitors of extra-terminal domain proteins JQ-1 and I-BET-762, which have been demonstrated to inhibit SEs but have minimal effects on typical enhancers (41). In response to the treatments, the expression of the RGS20 and CSNK1A1 genes, but not that of ERRFI1, decreased in both cell lines (SI Appendix, Fig. S8A). Next, we treated the cells with the phase separation inhibitors 1,6-hex and GSK-J4 or silenced endogenous SP1 using sh-SP1-3UTR. Under these conditions, the RGS20 gene always exhibited significantly reduced expression, and CSNK1A1 showed decreased levels in most cases, but the changes in ERRFI1 were inconsistent (SI Appendix, Fig. S8 B and C). Importantly, the downregulation of RGS20 in H1299 and PC-9 cells after treatment with these inhibitors or SP1 knockdown was verified by western blot analysis using an RGS20 antibody (SI Appendix, Fig. S8D).

Fig. 4.

Analysis of the correlation of SP1 with enhancer formation and identification of RGS20 as a target gene through the SE mechanism. (A) Hockey stick plots based on the input-normalized H3K27ac signals in H1299 cells. SEs were defined as those located above the inflection point of the curve, while the remaining enhancers were considered typical enhancers. (B) Heatmaps for genes with differential expression between SE, typical enhancer, and SP1 based on the signals of their closest H3K27ac and SP1 peaks. (C) Venn diagram showing overlapping genes between the SP1 and SE peaks based on CUT&RUN data obtained from H1299 cells. (D) Venn diagram of genes/mRNAs containing LUAD-specific SEs, genes associated with unfavorable clinical outcomes of LUAD patients, and genes overexpressed in LUAD samples. (E) Schematic view across the RGS20 gene locus (chr 8: 53797807–54018308) with genomic and epigenetic information. Graphical active regulatory regions were generated using the ENCODE database. The red box on the RGS20 gene represents an enhancer. (F) Reporter assays to examine the effects of a potential enhancer on RGS20 promoter activity. The 1,511 bp enhancer sequence was subcloned downstream of the RGS20 promoter and luciferase coding sequence to create an enhancer reporter vector. In the Left panel, the relative genomic positions of the RGS20 promoter and potential enhancer and a schematic diagram of the enhancer reporter are presented. A control reporter was generated using a DNA fragment of the same length upstream of the TSS to replace the enhancer. The enhancer reporter and control vectors were cotransfected with the SP1 expression plasmid into H1299 and PC-9 cells, and luciferase activity was subsequently measured (Right). (G and H) The effects of SP1 mutations on RGS20 promoter activity were evaluated. In G, the RGS20 promoter reporter was constructed by inserting the Gaussia luciferase (Gluc)-encoding sequence downstream of the RGS20 promoter. The RGS20 reporter and pCMV-SEAP vectors were cotransfected with the SP1 WT and mutant vectors into H1299 (Left) and PC-9 (Right) cells in triplicate in 24-well plates. Gluc activity in each well was measured and normalized against SEAP activity. In H, the SP1 WT and mutant vectors were transfected into H1299 (Left) and PC-9 (Right) cells with simultaneous knockdown of endogenous SP1 by sh-SP1-3UTR, followed by RT–qPCR to quantify RGS20 mRNA levels, with GAPDH as a control. In F–H, the data are presented as the mean ± SD, and the statistical significance was calculated by one-way ANOVA with Tukey’s post hoc test. *P < 0.05, **P < 0.01, ***P < 0.001, ns: not significant.

To investigate whether SP1-mediated SE formation contributes to elevated RGS20 expression in cancers, we scanned the genomic locus of the RGS20 gene and its adjacent regions for gene activation markers. Within a 220 kb range of the RGS20 promoter in the genomes of several LUAD cell lines, we identified potential enhancer regions based on the enrichment of H3K27ac and other enhancer markers, including H3K4me1, H3K4me3, and DNase accessible sites (Fig. 4E). Moreover, we selected a potential enhancer element located within the RGS20 gene and inserted it downstream of the corresponding luciferase coding sequence in an RGS20 promoter reporter (Fig. 4 F, Left). In the generated enhancer reporter, the relative promoter–enhancer position was in accordance with that in the genome. Compared with cells cotransfected with an empty vector, cells cotransfected with an SP1 expression vector and an enhancer reporter exhibited increased luciferase activity in LUAD cells (Fig. 4 F, Right). Additionally, the ability of SP1ΔBTD to stimulate the reporter was comparable to that of WT SP1, while SP1ΔZF3 lost this ability and exhibited similar activity to that of the empty vector (Fig. 4G). Consistently, when ectopically expressed in H1299 and PC-9 cells, both WT SP1 and SP1ΔBTD, but not SP1ΔZF3, steadily increased the expression of endogenous RGS20 according to RT–qPCR (Fig. 4H). Together, our data support that SP1-promoted RGS20 expression likely occurs through SE formation and depends on SP1 phase separation.

The RGS20 Gene Is a Primary Target of SP1 that Promotes LUAD Cell Proliferation and Tumor Metastasis.

As we investigated the mechanism of SP1-mediated RGS20 expression, we further evaluated the importance of this regulation in SP1-promoted LUAD cell malignancy. For this purpose, we knocked down endogenous SP1 by sh-SP1-3UTR and simultaneously expressed exogenous RGS20 in H1299 and PC-9 cells (Fig. 5A). In both cell lines, SP1 silencing significantly decreased cell viability and proliferation, which could be generally restored by ectopically introduced RGS20 (Fig. 5 B and C). Similarly, the reduced mobility and invasion of LUAD cells caused by SP1 depletion were also reversed by exogenous RGS20 (Fig. 5 D and E). Thus, our results indicated that SP1 primarily targets the RGS20 gene to promote various malignant properties of LUAD cells and subsequently contributes to lung cancer development.

Fig. 5.

Evaluation of the role of SP1-mediated RGS20 expression in promoting LUAD cell malignancy. In A–E, H1299 and PC-9 cells were infected with lentivirus carrying sh-SP1 to knock down endogenous SP1 without or with ectopic RGS20 expression, and the cells were analyzed for SP1 and RGS20 expression by western blot (A), cell viability (B), DNA synthesis activity based on EdU incorporation (C), cell migration by wound healing assay (D), and cell migration and invasion by Transwell assay (E). In B–E, the data are presented as the mean ± SD and were compared by one-way ANOVA with Tukey’s post hoc test. *P < 0.05, **P < 0.01, ***P < 0.001, ns: not significant.

SP1-ZF3 Is Essential for Xenograft Tumor Growth in LUAD.

To examine the effect of SP1 phase separation on tumor growth, we employed a mouse xenograft tumor model. PC-9 cells with endogenous SP1 silenced by sh-SP1-3UTR and infected with lentivirus expressing SP1 WT, ΔBTD, ΔZF3 or an empty vector were inoculated into the right axillae of nude mice. Tumor growth was monitored by measuring the tumor diameter to determine the tumor volume, and the mice were killed on the 32nd day to collect xenograft tumors. Based on the tumor growth curves, images, and weights of the excised tumors, both SP1 WT and ΔBTD could rescue the tumor growth retardation caused by endogenous SP1 knockdown, although the latter showed relatively low effects (Fig. 6 A–C). However, SP1ΔZF3 lost this ability, and tumors harboring this mutant showed comparable growth to that of the vector control group. RT–qPCR analysis revealed that the levels of these SP1 mRNAs were elevated in the tumors harboring SP1 WT, ΔBTD, and ΔZF3 expression plasmids (Fig. 6 D, Top). However, compared with those in the SP1 WT group, the tumors in the SP1ΔZF3 group exhibited significantly reduced RGS20 expression (Fig. 6 D, Bottom). Consistently, according to the results of the IHC staining studies, the SP1-WT tumors exhibited a strong signal for Ki-67, a marker of cell proliferation, and the signal intensity in the SP1ΔBTD tumors was relatively low (Fig. 6E). Compared to the samples in these two groups, tumors harboring either SP1ΔZF3 or an empty vector exhibited greatly reduced Ki-67 intensity (Fig. 6E). Consistent with the RT–qPCR results, SP1 WT samples also exhibited increased RGS20 staining compared with that in the control group.

Fig. 6.

Mouse model study to examine the effects of BTD and ZF3 deletion on SP1-promoted xenograft tumor formation. (A–C) Xenograft tumor growth curves (A), excised tumor images (B), and weights (C). PC-9 cells with endogenous SP1 silenced by sh-SP1-3UTR and infected with lentivirus carrying an empty vector, SP1 WT, SP1ΔBTD or SP1ΔZF3 were used for implantation into the right axillary area of nude mice. Tumor size was measured every 2 d after tumor cell implantation, and tumor growth curves were plotted at 8-d intervals (A). Tumors were excised on the 32nd day and photographed (B). Tumor weight was measured (C). (D) Determination of SP1 and RGS20 levels in xenograft tumors using RT–qPCR. (E) IHC analyses of SP1, RGS20, and Ki67 expression in xenograft tumor samples. The IHC images are shown in the Top panel, while the quantification of the staining is presented in the Bottom panel. (F) A schematic diagram showing how SP1 activates RGS20 expression through phase separation and SE mechanisms and subsequently promotes LUAD development. Various transcriptional coactivators can be recruited to SP1-mediated nuclear condensates, which can be disrupted by GSK-J4 to impede LUAD progression. In A, C, D, and E, the data are presented as the mean ± SD and were compared by one-way ANOVA with Tukey’s post hoc test. *P < 0.05, **P < 0.01, ***P < 0.001. ns: not significant.

Overall, we identified a unique mechanism by which SP1 undergoes phase-separated condensation regulated by ZF3, which can promote SE formation to activate RGS20 gene expression and subsequently aggravate LUAD development. GSK-J4 disrupted SP1 phase separation to compromise its target gene activation, leading to retarded LUAD progression (Fig. 6F).

Discussion

The high mortality rate of lung cancer patients is primarily attributed to tumor recurrence and metastasis, which have a complex pathogenesis involving multiple factors. Despite extensive ongoing research, effective treatments for LUAD are still lacking. Thus, it is crucial to identify new treatment targets and develop effective therapeutic strategies. Recent studies have provided ample evidence for the involvement of phase separation mechanisms in tumorigenesis, including the assembly of the transcriptional machinery through phase-separated condensation, which contributes to the activation of various oncogenes or proliferative genes (5). In this study, we identified SP1-mediated transcription through a phase separation mechanism in LUAD cells, which depends on its ZF3. Our data support the notion that SP1 promotes SE formation and subsequently activates the expression of the RGS20 gene, thereby driving the malignant progression of LUAD cells. Furthermore, our study revealed that the small molecule GSK-J4 disrupted SP1 phase separation and blocked the cancer-promoting effects of SP1 in LUAD. Therefore, modulating the LLPS process of SP1 holds promise as an effective approach for developing innovative cancer therapeutic strategies.

Many studies from recent years have demonstrated the contribution of phase separation to the transcriptional activation of cancer-related genes, which has greatly contributed to our understanding of gene expression mechanisms at the molecular level. The general model shows a TF with two functional regions: an activation domain (AD) and a DBD. Among many previously characterized TFs, such as OCT4 and GCN4, an AD may harbor one or more IDRs that undergo phase separation with transcriptional cofactors to compartmentalize them into membraneless condensates, leading to the formation of enhancers or SEs (37). We previously reported that a histidine cluster in the N-terminal IDR of YY1 plays a key role in forming an enhancer cluster through a phase separation mechanism, which subsequently promotes oncogene FOXM1 expression and aggravates breast cancer (44). However, the contribution of DBDs to TF phase separation has rarely been reported. Recently, we revealed that ZF3–ZF5 of MAZ, but not the N-terminal IDR, are responsible for its phase separation (47). Similarly, KLF4 and peroxisome proliferator-activated receptor γ (PPARγ), which are both ZF-containing TFs, undergo phase-separated condensation dependent on their DBDs, not the predicted IDRs (48, 49). Additionally, the C4-type ZF domain of PPARγ can independently form phase-separated condensates (49).

In this study, we identified that the region essential for SP1 phase separation was located at its C terminus, which contains three ZFs and is responsible for DNA binding. Our further characterization identified ZF3 as the key element for both SP1 nuclear punctum formation in cells and its oncogenic activity in LUAD cell xenograft tumor formation. Thus, our study provides evidence to support the contribution of DBDs to the phase separation of TFs.

We identified that SP1 phase separation depended on its ZF3, which is also one of the three ZFs that bind to DNA, indicating that the chromatin contact of SP1 is potentially required for and may even promote its condensation. Consistently, studies from our group and others demonstrated that the phase separation of multiple TFs could be enhanced by their DNA contact (44, 47–50). Therefore, the defective oncogenic activity of SP1ΔZF3 could result from both abolished phase separation and reduced DNA-binding affinity. The SP1ΔZF3 mutant exhibited only partial cytoplasmic translocation (SI Appendix, Fig. S5A), while SP1ΔC, lacking all three ZFs responsible for DNA binding, showed an almost uniform distribution between the nucleus and cytoplasm (SI Appendix, Fig. S4A), consistent with previous studies indicating the essential role of the three ZFs in the nuclear localization of SP1 (51–53). Additionally, a previous study reported that the nuclear localization signals of 90% of DNA-associated proteins overlapped with their DNA-binding regions (54). Notably, increased SP1 expression and its nuclear accumulation are reportedly associated with insulin-induced glycosylation and phosphorylation (55, 56). As many as 12 serine and threonine residues in the C terminus of SP1 are involved in its glycosylation and phosphorylation (57). Although the current study did not involve the treatment of cells with insulin or other hormones, future exploration may be needed to determine whether modifications of these residues contribute to the subcellular localization and phase separation of SP1. In our functional studies, BTD deletion could also generate a mutant with compromised oncogenic activity and phase separation ability of SP1 but to a significantly reduced extent compared to those of SP1ΔZF3. The underlying mechanism for this observation deserves future exploration.

Using CUT&RUN technology and ROSE analysis, we determined that SP1 binds to the TSS regions of 6,621 genes in LUAD cells. Among them, 38 genes also showed H3K27ac enrichment that overlapped with these SP1-binding SE elements. After correlating the differential expression of these genes with the clinical data of LUAD patients, we restricted the extent of these candidates to three genes, namely, ERRFI1, RGS20, and CSNK1A1. In subsequent enhancer reporter assays, we identified RGS20 as the most likely target gene of SP1 through an enhancer regulatory mechanism. It is known that enhancer formation and function are cell type specific (58). Different cell types may exhibit distinct enhancer patterns, while a specific profile of enhancers also determines the unique gene expression pattern of each cell type. The development of the ROSE algorithm was mostly based on multiple myeloma cells and tested in different cancer cells, including the small cell lung cancer cell line H2171 (59). However, the LUAD cell line H1299 we used is a type of non–small cell lung cancer; thus, the ROSE analysis based on our CUT&RUN datasets could miss some SP1-targeted genes through the SE mechanism due to differences in the pathological backgrounds of the cell lines. Moreover, we focused only on genes that may either promote LUAD development or significantly correlate with lung cancer progression. Therefore, RGS20 is not the only gene regulated by SP1 through the SE mechanism. In particular, CSNK1A1, a promising therapeutic target in lung cancer therapies (60), also showed a similar response to RGS20 in most reporter assays. Whether SP1 regulates other cancer-related genes, especially CSNK1A1, through the SE mechanism needs future investigation.

As a bona fide oncogene, RGS20 is overexpressed in various cancers and plays a stimulating role in cancer cell proliferation, invasion, and metastasis (61, 62). Mechanistically, RGS20 stimulates the PI3K/AKT and NF-κB signaling pathways (61, 63). Previous studies revealed that RGS20 levels are regulated by microRNAs and ubiquitination (64, 65), but the transcriptional regulation of RGS20 gene expression has not been reported. In this study, we provided evidence demonstrating the activation of RGS20 expression through SP1 phase separation and SE formation mechanisms, which expanded our understanding of the molecular mechanism underlying RGS20 overexpression in cancers.

Research on TF phase separation is still in its infancy, and the exploration of inhibitors targeting oncogenic TF condensates has only recently been initiated. Although specific inhibitors of phase separation are still lacking, promising proofs of concept for condensate-targeted drug discovery have emerged and started to accumulate. Among these compounds, oxaliplatin can disrupt the interaction network of nucleoli, leading to impaired ribosomal RNA processing, reduced RNA polymerase I-mediated transcription, and subsequent cell death (66). JQ1 and IBET may release the Mediator complex from enhancer elements and inhibit gene transcription (67). Cisplatin can selectively reduce MED1 condensates and thereby preferentially target actively transcribed genes, especially those with SEs (68). In this study, we observed that GSK-J4, an inhibitor of H3K27me3 demethylase, disrupted SP1 phase separation and abolished SP1-driven RGS20 expression in LUAD. GSK-J4 is a derivative of GSK-J1 but has better cell membrane permeability. They are inhibitors of Jumonji histone demethylases by competitively inhibiting two cofactors, 2-oxoglutarate and Fe²⁺, required for enzymatic activity (69). SP1 contains 20 lysine residues, but none of them have been reported to be methylated. Importantly, GSK-J4 inhibited purified recombinant EGFP-SP1 droplet formation in vitro (Fig. 5B), suggesting that its ability to disrupt phase separation was irrelevant to possible SP1 methylation. SP1 can apparently regulate the expression of multiple genes through the formation of enhancers or SEs. The inhibitory effects of GSK-J4 on SP1 may not be limited to RGS20, and additional targets may also contribute to the anticancer effects of GSK-J4 in LUAD.

Intracellular phase separation is inevitably affected by the tissue microenvironment, a complex ecosystem involved in cell‐to‐cell and cell‐to‐extracellular-matrix interactions and regulated by various signaling pathways (70). Currently, in-depth research exploring how the tumor microenvironment contributes to protein phase separation in cancer cells is still lacking. Nevertheless, as discussed above, SP1 glycosylation and phosphorylation are induced by insulin (55), and its expression and phosphorylation are also modulated by EGFR-regulated signaling pathways (71, 72), suggesting that SP1 phase separation is susceptible to the tumor microenvironment, which deserves future exploration.

In conclusion, our findings demonstrate that SP1 undergoes phase separation in LUAD cells and that the C terminus of ZF3 plays an essential role in its condensation. SP1 condensates may compartmentalize transcriptional cofactors and transcriptional machinery, leading to SE formation to activate downstream genes, such as RGS20, and subsequently promote LUAD progression. Thus, targeting SP1 phase separation may serve as a promising strategy in cancer therapies.

Materials and Methods

Phase Separation Assay in Cells.

Live cell imaging was performed by transfecting plasmids expressing EGFP fusion proteins into cells, which were subsequently cultured for 24 h. After the nuclei were stained with Hoechst (Solarbio C0030), images were captured using a GE Delta Vision Elite (Boston) with a 60× objective. The cells were incubated in a 5% CO₂ chamber at 37 °C during image acquisition.

In Vitro Phase Separation Assay.

The purified recombinant EGFP fusion protein was diluted to the desired concentrations in a buffer containing 125 mM NaCl and 10% PEG 8000. Each protein sample was immediately dropped onto a slide and covered with a coverslip. The slide was then imaged using a 60× objective lens from GE Delta Vision Elite (Boston).

FRAP after Assay.

SE analysis.

To identify potential SEs, the ROSE algorithm (http://younglab.wi.mit.edu/super_enhancer_code.html) was used. SEs were defined as a set of H3K27ac peaks between 2.5 and 12.5 kb from the TSS of each gene. Each potential SE showed the highest level of H3K27 acetylation by plotting the inflection point and selecting the value for which the slope of the fitted curve exceeded 1. The elements that fell below the point on this curve where the slope was 1 were considered typical enhancers.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

The CUT&RUN data generated in this study have been deposited in the NCBI Gene Expression Omnibus (GEO, https://www.ncbi.nlm.nih.gov/geo/; with an accession number GSE270008) (73). The RNA-seq data used in this research could be downloaded from the NCBI GEO under accession number GSE10072 (74). The other ChIP-seq data reused are available in ENCODE database (https://www.encodeproject.org/) under accession numbers ENCSR000BPE (A549-SP1) (75), ENCSR783SNV (A549-H3K27ac) (76), ENCSR769FOC (PC-9-H3K27ac) (77), ENCSR636PIN (A549-H3K4me1) (78), ENCSR913MGR (PC-9-H3K4me1) (79), ENCSR441JWF (PC-9-H3K4me3) (80). The DNase-seq data used in this research could be downloaded from ENCODE database under accession number ENCSR940NLN (81). Other data used to evaluate the conclusions of this study are presented in the paper and the SI Appendix.