ISL1 predicts poor outcomes for patients with gastric cancer and drives tumor progression through binding to the ZEB1 promoter together with SETD7

Abstract

ISL1, a LIM-homeodomain transcription factor, serves as a biomarker of metastasis in multiple tumors. However, the function and underlying mechanisms of ISL1 in gastric cancer (GC) have not been fully elucidated. Here we found that ISL1 was frequently overexpressed in GC FFPE samples (104/196, 53.06%), and associated with worse clinical outcomes. Furthermore, the overexpression of ISL1 and loss-of-function of ISL1 influenced cell proliferation, invasion and migration in vitro and in vivo, including GC patient-derived xenograft models. We used ChIP-seq and RNA-seq to identify that ISL1 influenced the regulation of H3K4 methylation and bound to ZEB1, a key regulator of the epithelial-mesenchymal transition (EMT). Meanwhile, we validated ISL1 as activating ZEB1 promoter through influencing H3K4me3. We confirmed that a complex between ISL1 and SETD7 (a histone H3K4-specific methyltransferase) can directly bind to the ZEB1 promoter to activate its expression in GC cells by immunoprecipitation, mass spectrometry, and ChIP-re-ChIP. Moreover, ZEB1 expression was significantly positively correlated with ISL1 and was positively associated with a worse outcome in primary GC specimens. Our paper uncovers a molecular mechanism of ISL1 promoting metastasis of GC through binding to the ZEB1 promoter together with co-factor SETD7. ISL1 might be a potential prognostic biomarker of GC.

Fig. 1. ISL1 expression in primary GC tissues.

a–c The expression of ISL1 in primary GC (T) and corresponding surgical margin (N) tissues was examined by RT-PCR (a), western blotting (b), and immunohistochemical staining (c). Original magnification: 200× in c. d Association between ISL1 expression status and TNM stage. e, f Kaplan–Meier survival curves of disease-free survival (DFS) and overall survival (OS) for patients with ISL1-negative vs. ISL1-positive staining in GC tissues

Fig. 2. ISL1 knockdown attenuated GC cell growth, migration, and invasion in vitro and in vivo.

a Cell proliferation was monitored with an IncuCyte system every 6 h. b, c Plate colony formation assay and soft agar colony formation assay with SGC7901 and MGC803 cells. d, e Assessment of invasion of SGC7901 and MGC803 cells using Matrigel or Boyden chambers. f, g Wound healing assay were assayed by live cell imaging on an IncuCyte system. h The effect of ISL1 on distant metastatic colonization through blood circulation. Each bar in the bar charts of a–e represents the mean ± SD from three independent experiments of six replicate wells. *P < 0.05, **P < 0.01. Tumor weights in h are shown as the mean ± SD

Fig. 3. ISL1 overexpression promoted cell growth, migration, and invasion in MKN28 cells.

a Cell proliferation was monitored with an IncuCyte system every 6 h. b Plate colony formation assay. c Soft agar colony formation assay. d Assessment of invasion of MKN28 cells using Matrigel or Boyden chambers. e Wound healing assay by live-cell imaging on an IncuCyte system. Each bar in the bar charts of a–e represents the mean ± SD from three independent experiments of six replicate wells. *P < 0.05, **P < 0.01

Fig. 4. ISL1 promoted tumor growth in cell lines and GC patient-derived xenograft mouse models.

a–c Photographs showing tumor formation in nude mice injected with stable cell lines with ISL1 knockdown (a and b) and ISL1 overexpression (c). d Immunohistochemical staining of ISL1. F₀, primary GC tissue; F₃, third generation of tumor graft. e Photographs showing tumor formation in NOD/SCID mice injected with a mixture of minced tumor graft (F₃) and lentivirus expressing either ISL1-shRNA1# and ISL11-shRNA2# for ISL1-positive tumor graft (e) or ISL1 for ISL1-negative tumor graft (f). The tumor volume and weight in a–f are shown as the mean ± SD. *P < 0.05, **P < 0.01

Fig. 5. ISL1 enhances gastric cancer tumorigenesis through ZEB1.

a Pie diagram summarizing the genomic occupancy of ISL1-bound regions as revealed by ChIP-seq. b GO functional clustering of upregulated genes allowed the identification of cellular functions directly regulated by ISL1. c Genome-wide analysis of downstream targets of ISL1. Overlap of RNA-seq and ChIP-seq results revealed 1039 genes as potential direct targets of ISL1. d Western blotting analysis of ZEB1 and EMT-associated proteins. e The mRNA expression of ISL1 and ZEB1 was analyzed by real-time RT-PCR in 36 paired GC samples with GAPDH as the reference gene. f GEPIA results indicated a correlation between ISL1 and ZEB1 gene expression in stomach adenocarcinoma samples from the TCGA. g Kaplan–Meier survival curves of survival time for patients with high vs. normal ZEB1 expression. ISL1 expression was assessed in 36 paired human GC tissues (P = 0.0196). h Kaplan–Meier survival reanalysis of overall survival. The data were obtained from publicly available gene expression datasets (GSE14210, GSE15459, GSE22377, GSE29272, GSE51105)

Fig. 6. ISL1 promoted ZEB1 expression by H3K4me3 modification through binding SETD7.

a Consensus binding site (TAAT) for ISL1 on the ZEB1 promotor. b Luciferase reporter assay was performed in 293FT cells c, d Sonicated chromatin from the SGC7901 cells with ISL1 knockdown was immunoprecipitated with ISL1 (c) and H3K4me3 antibodies (d). The resulting input and ChIP DNA were characterized with qPCR primers specific for ZEB1 genomic loci to calculate the percentage of coprecipitated DNA relative to the input. e, f Immunoprecipitation and mass spectrometry analysis of cofactors of ISL1. For the coimmunoprecipitation assay, normal IgG served as the negative control. g ChIP and re-ChIP assays. The second ChIP was performed using an antibody against SETD7, and the resulting input and ChIP DNA were characterized with qPCR primers specific for ZEB1 genomic loci to calculate the percentage of coprecipitated DNA relative to the input. h Schematic depicting how ISL1 activate ZEB1 expression in gastric cancer cells. Each bar in the column chart data of b–d represents the mean ± SD from three independent experiments in triplicate