ELK4 Promotes Colorectal Cancer Progression by Activating the Neoangiogenic Factor LRG1 in a Noncanonical SP1/3‐Dependent Manner

Abstract Although the MAPK/MEK/ERK pathway is prevalently activated in colorectal cancer (CRC), MEK/ERK inhibitors show limited efficiency in clinic. As a downstream target of MAPK, ELK4 is thought to work primarily by forming a complex with SRF. Whether ELK4 can serve as a potential therapeutic target is unclear and the transcriptional regulatory mechanism has not been systemically analyzed. Here, it is shown that ELK4 promotes CRC tumorigenesis. Integrated genomics‐ and proteomics‐based approaches identified SP1 and SP3, instead of SRF, as cooperative functional partners of ELK4 at genome‐wide level in CRC. Serum‐induced phosphorylation of ELK4 by MAPKs facilitated its interaction with SP1/SP3. The pathological neoangiogenic factor LRG1 is identified as a direct target of the ELK4‐SP1/SP3 complex. Furthermore, targeting the ELK4‐SP1/SP3 complex by combination treatment with MEK/ERK inhibitor and the relatively specific SP1 inhibitor mithramycin A (MMA) elicited a synergistic antitumor effect on CRC. Clinically, ELK4 is a marker of poor prognosis in CRC. A 9‐gene prognostic model based on the ELK4‐SP1/3 complex‐regulated gene set showed robust prognostic accuracy. The results demonstrate that ELK4 cooperates with SP1 and SP3 to transcriptionally regulate LRG1 to promote CRC tumorigenesis in an SRF‐independent manner, identifying the ELK4‐SP1/SP3 complex as a potential target for rational combination therapy.

(E) Heatmaps of ChIP-seq data for ELK4, SP1 and SP3 in HCT116 cells.All peaks in each heatmap are centered ± 2.0 kb from the ELK4 peaks in ELK4-bound promoters.
(F) Normalized read density (per bp per peak) for ELK4, SP1 and SP3 plotted in the region ± 2.0 kb from the ELK4-bound promoters.
(G) Heatmaps of ChIP-seq data for ELK4 and SRF peaks in the ELK4 (left) and SRF (right) targeting promoters.
(H) Normalized read density (per bp per peak) for ELK4 and SRF plotted in the region ± 2.0 kb from the ELK4-bound (left) or SRF-bound (right) promoters.
(I and J) ELK4-overexpressing LoVo cells were transfected with control, SP1 and SP3 siRNAs for 2 days and then analyzed by CCK8 (I) and Transwell assays (J).(C) U0126 suppressed the serum-induced interactions between ELK4 and SP1 and SP3 in HCT116 cells.HCT116 cells were transfected with the indicated plasmids for 24 h and were then cultured in the presence or absence of U0126 (10 μM) for 12 h before co-IP assay.
(D, E) Selumetinib (D) and Ulixertinib (E) suppressed the serum-induced interactions between ELK4 and SP1 and SP3 in HCT116 cells.HCT116 cells were transfected with the indicated plasmids for 24 h and were then cultured in the presence or absence of Selumetinib (D) and Ulixertinib (E) for 12 h before co-IP assay.
(F) λ-protein phosphatase (λ-PPase) decreased the interaction between ELK4 and SP1 /SP3.HCT116 cells were transfected with the indicated plasmids for 24 h and the cell lysates were treated with λ-PPase for 1h and then subjected to immunopreciation.
(G) HCT116 cells were transfected with the indicated plasmids for 24 h and were then cultured in the presence or absence of U0126 (10 μM) for 12 h.
(H) HCT116 cells were transfected with ELK4 wild-type and C-terminal deletion plasmids for 24 h and then subjected to co-IP assay.

(
A) KEGG enrichment analysis of the overlapping target genes bound by ELK4, SP1 and SP3.(B) Cumulative frequency distribution of the smallest distance between ELK4 and SP3 (left) or SP1 (right) binding sites associated with each shared target gene.Distances were calculated by using the peak coordinates of each factor nearest one another.(C) Genome Browser representation of ELK4 (red), SP1 (green) and SP3 (purple) peaks in MFSD12 and VCP genes.
(K) Representative images and statistical analysis of HUVEC tube formation incubated with CM from ELK4-overexpressing LoVo cells transfected with control, SP1 and SP3 siRNAs.Two-way ANOVA (I) and one-way ANOVA (J and K) were performed to assess the statistical significance.The data are presented as the mean ± S.D. values.** P < 0.01, *** P < 0.001.Supplementary Figure 4. Related to Figure 4.
(A) Coimmunoprecipitation (Co-IP) to evaluate the interactions between FLAG-ELK4 and HA-SP1 and HA-SP3 in HCT116 cells.(B) Proximity ligation assay (PLA) to evaluate the interactions of ELK4 with SP1 and SP3 in SW480 and RKO cells.The red dots indicate protein interactions (scale bars:10 μm).

(
A and B) The domain organization maps of the full-length SP1 (A) and SP3 (B) proteins are shown schematically.Co-IP analysis was performed to map the domains participating in the interactions of the HA-SP1 and HA-SP3 constructs with Flag-ELK4.(C) Immunofluorescence assay of the full-length SP1/3 and truncated SP1/3 in HCT116 cells.(D) Immunofluorescence assay of the endogenous SP1 in serum-starved HCT116 cells treated serum or serum plus U0126 treatment.Supplementary Figure 6.Related to Figure 5. (A) qPCR analysis of the mRNA levels of LRG1, STC3 and LBHD1 in HCT116 cells with ELK4 or SP1/3 knockdown.(B) GO analysis based on the newly identified ELK4-SP1/3 co-regulated genes.(C) qPCR (up) and western blot (down) analysis of the LRG1 mRNA and protein level in LoVo cells with ELK4 knockdown or overexpression.(D) ELISA analysis of the LRG1 protein level in the condition medium from HCT116 cells with ELK4 or SP1/3 knockdown.(E) qPCR, western blot analysis of the LRG1 mRNA and protein level in HCT116 cells with ELK4 overexpression.ELISA analysis of the LRG1 protein level in the condition medium from HCT116 cells with ELK4 overexpression.(F) Immunohistochemical staining indicated the expression of LRG1 in xenografts derived from ELK4 knockdown or ELK4-overexpressing HCT116 cells (scale bars = 50 μm).One-way ANOVA (A, C and D) and Student's t test (C and E) were performed to assess the statistical significance.The data are presented as the mean ± S.D. values.** P < 0.01, *** P < 0.001.(A) ChIP-qPCR analysis of ELK4 and SP1/3 binding to the LRG1 gene locus.(B) Schematic depiction of the LRG1 luciferase reporter construct containing the LRG1 enhancer, as identified by the ChIP-seq data for ELK4, SP1 and SP3.(C) Schematic depiction of the LRG1 luciferase reporter construct containing both the LRG1 enhancer and promoter, as identified by the ChIP-seq data for ELK4, SP1 and SP3, H3K27ac and H3K4me1.(D) Luciferase assay of HCT116 cells cotransfected with the LRG1 reporter containing the enhancer (with or without the promoter) and the empty vector or the indicated plasmids of ELK4, SP1 and SP3.(E) Luciferase assay of HCT116 cells with the LRG1 reporter (promoter+enhancer) with or without mutation of ELK4 binding sites or SP1/3 binding sites.(F) ELISA analysis of the LRG1 protein level in the condition medium from ELK4 knockdown HCT116 cells overexpressing LRG1.(G) Western blot analysis of the p-SMAD1/5, SMAD1 and SMAD5 in HUVEC treated with the CM from ELK4 knockdown HCT116 cells or with rescued LRG1 expression (left) and ELK4 overexpression HCT116 cells or with LRG1 knockdown (right).(H) Western blot analysis of the p-SMAD1/5, SMAD1 and SMAD5 in HUVEC treated with the CM from ELK4 knockdown HCT116 cells with 500 ng/ml rLRG1.(I) Immunohistochemical staining indicated the expression of CD31, HIF-1α and CD105 in xenografts derived from ELK4 knockdown HCT116 cells overexpressing LRG1 (scale bars = 50 μm).Student's t test (A and D) and One-way ANOVA (E, F and I) were performed to assess the statistical significance.The data are presented as the mean ± S.D. values.* P < 0.05, ** P < 0.01, *** P < 0.001.Supplementary Figure 8. Related to Figure 6.(A) Western blot analysis of SP1 protein in HCT116 and RKO cells treated with MMA or MMA+MG132.(B) Western blot analysis of indicated protein levels in HCT116, HT29, LoVo, RKO and SW480 cells treated with vehicle, 50 nM MMA, 5 μM U0126 or the combination of MMA and U0126.(C) Cell proliferation was assessed following 48 h of exposure to the indicated concentrations of MMA, Selumetinib or the combination in HCT116, HT29, LoVo, RKO and SW480 cells.CI values for the various combinations were calculated using CompuSyn.A CI < 1.0 indicates a synergistic effect.(D) Colony formation assays showed a synergistic response to the combination of Selumetinib and MMA in HCT116, HT29, LoVo, RKO and SW480 cells (E) Representative western bolts of cleaved PARP1 in HCT116, HT29, LoVo, RKO and SW480 cells treated with vehicle, 50 nM MMA, 100 nM Selumetinib or the combination of MMA and Selumetinib.(F) Representative western bolts of cleaved PARP1 in control and SP1 knockdown HCT116 cells treated with vehicle, 50 nM MMA, 5 μM U0126 or the combination of MMA and U0126.Supplementary Figure 9. Related to Figure 6.(A) The body weight of nude mice treated with vehicle, 0.5 mg/kg MMA, 10 mg/kg U0126 or the combination of MMA and U0126 (n = 5 mice per group).(B) Representative western blot analysis of indicated protein levels in HCT116 cellderived and CRC patient-derived xenografts harvested from nude mice treated with vehicle, 0.5 mg/kg MMA, 10 mg/kg U0126 or the combination of MMA and U0126.(C) Immunohistochemical staining indicated the expression of Ki67 and cleaved PARP1 in CDXs treated with vehicle, 0.5 mg/kg MMA, 10 mg/kg U0126 or the combination of MMA and U0126 (scale bars = 50 μm).(D) Representative images of control and LRG1 overexpression HCT116 cell-derived xenografts harvested from nude mice treated with vehicle, 0.5 mg/kg MMA, 10 mg/kg U0126 or the combination of MMA and U0126 (n = 5 mice per group).Xenograft weight were measured for statistical analysis.One-way ANOVA (C, D) were performed to assess the statistical significance.The data are presented as the mean ± S.D. values.* P < 0.05, ** P < 0.01, *** P < 0.001.Supplementary Figure 10.Related to Figure 7. (A) Western blot analysis of the protein level of ELK4 in 24 pairs of CRC and adjacent normal tissues.(B) qPCR analysis of the mRNA levels of ELK4 in the two normal colonic epithelial cell lines and 7 CRC cell lines.(C) Kaplan-Meier analysis of disease-free survival (DFS) for CRC patients in the TCGA CRC database stratified by ELK4 mRNA level.(D) Positive correlations between LRG1 expression and ELK4, SP1 and SP3 expression in the CRC TMA cohort.Pearson correlation analysis was used to evaluate the associations.(E) Correlations of the mRNA level of LRG1 with those of ELK4 and SP3 in GSE20916.Pearson correlation analysis was used to evaluate the association.(F and G) The risk score distribution (F), and OS time and status (G) of patients in TCGA-COAD.The dotted line indicates the cutoff value for dividing the patients into the low-and high-risk groups.Supplementary Tables