Mutated KLF4(K409Q) in meningioma binds STRs and activates FGF3 gene expression

Abstract

Krüppel-like factor 4 (KLF4) is a transcription factor that has been proven necessary for both induction and maintenance of pluripotency and self-renewal. Whole-genome sequencing defined a unique mutation in KLF4 (KLF4^K409Q) in human meningiomas. However, the molecular mechanism of this tumor-specific KLF4 mutation is unknown. Using genome-wide high-throughput and focused quantitative transcriptional approaches in human cell lines, primary meningeal cells, and meningioma tumor tissue, we found that a change in the evolutionarily conserved DNA-binding domain of KLF4 alters its DNA recognition preference, resulting in a shift in downstream transcriptional activity. In the KLF4^K409Q-specific targets, the normally silent fibroblast growth factor 3 (FGF3) is activated. We demonstrated a neomorphic function of KLF4^K409Q in stimulating FGF3 transcription through binding to its promoter and in using short tandem repeats (STRs) located within the locus as enhancers.

Keywords: Cancer; Molecular biology; Transcriptomics.

Graphical abstract

Figure 1

Both WT and mutant KLF4 recognize and bind to the same DNA consensus (A) KLF4 consensus DNA-binding site schematically bound to three protein ZFs. According to conventional recognition code, ZF1 binds nucleotides at positions 7 to 9. K409Q mutation is located within ZF1 (shown as red asterisk). N: any nucleotide; Y: pyrimidine (C or T); K: Keto (G or T). (B) Expression of FLAG-tagged KLF4 proteins in HEK293 cells. Western blot analysis was done using HEK293 whole-cell lysates and indicated antibodies. Anti-KLF4 (Full) and anti-KLF4 (C-term) antibody were raised against full-length recombinant protein and C-terminus peptide, respectively. Anti-Sp1 antibody was used as the loading control. (C) EMSA using nuclear extracts from HEK293 cells transfected with KLF4- or KLF4^K409Q-expressing plasmids and a 20 bp long oligonucleotide probe spanning KLF4 consensus as described in experimental model and subject details. Arrows indicate specific bands. Antibodies used for super-shifts are shown above the gel.

Figure 2

KLF4^K409Q activates FGF3 expression (A) Venn diagram showing the overlap of DEGs in HEK293 or A549 cells by RNA-seq analysis. (B) Scatter-plot of the log₂ (fold change) of all genes called as significant in KLF4 or KLF4^K409Q RNA-seq analysis. Positions of dots corresponding to FGF3, CALML5, ALPG, and TRH genes are marked by black arrows. (C) Time course of FGF3 and TRH mRNA expression in HEK293 cells transfected with KLF4- or KLF4^K409Q-expressing plasmids by RT-qPCR analysis. Number of copies on the y-axis is presented as calculated copy number per 1,000 copies of GAPDH mRNA in the same sample (see also Figure S1).

Figure 3

KLF4^K409Q binds to and drives transcription from the FGF3 promoter (A) Activity of luciferase (LUC) gene under the control of FGF3 and TRH promoters in HEK293 cells co-transfected with KLF4- or KLF4^K409Q-expressing plasmids. Luciferase activity was measured 48 h posttransfection. Data are represented as mean ± SD from at least four independent experiments. (B) Quantitative DNase I footprinting analysis of minimal FGF3 promoter region (−276 to +84 bp) relative to transcription start site (TSS) with increasing amounts of recombinant KLF4 and KLF4^K409Q DBD proteins (left panel) and similar analysis of minimal TRH promoter region (−270 to +110 bp) (right panel). Positions of protein-binding sites and TATA-boxes are marked by open boxes. Each binding site name reflects the position of the middle nucleotide in the KLF4 site consensus (Y5 in Figure 1A). (C) The alignment of 10 bp KLF4-binding consensus sequence with newly identified strongest binding sites FGF3-201 and TRH-122 in FGF3 and TRH minimal promoters. The 9 bp KLF4-binding sequence in both probes is bolded. ZF1-binding nucleotide triplet is set out by spaces. N: any nucleotide; Y: pyrimidine (C or T); K: Keto (G or T). (D) EMSA analysis of KLF4-binding sites from FGF3 and TRH minimal promoters. Protein/DNA binding was tested in nuclear extracts from HEK293 cells with overexpressed WT or mutated KLF4 proteins (top gel panel) and BL21 E. coli lysates expressing MBP-KLF4 full-length proteins (bottom gel panel). Sequences of tested probes FGF3-201 and TRH-122 are shown in (C), and their mutants with a single base pair change within the ZF1-binding nucleotide triplet are shown in red. Only protein/DNA complexes are shown (see also Figure S2).

Figure 4

KLF4 and KLF4^K409Q exhibit distinct binding specificity in vitro and in vivo (A) Venn diagram showing the overlap of the peaks in KLF4 and KLF4^K409Q sets by ChIP-seq analysis of HEK293 cells with ectopically expressed proteins. (B) Genomic (left) and epigenetic (right) context for the three ChIP-seq peak sets: KLF4, KLF4^K409Q, and Shared. Categories for epigenetic context are defined by the ENCODE SCREEN project (https://screen.encodeproject.org/). (C) Heatmap of read density of KLF4 and KLF4^K409Q ChIP-seq at ranges ±2 kb around consensus peak sets. The read depth was normalized across all six biological replicates for shared consensus peaks. (D) Scatter plot and boxplots of log₂ average normalized read depth for peaks in each of three consensus sets: KLF4, KLF4^K409Q, and Shared. (E ) Heatmap of normalized read depth per replicate for the top 10% most variable peaks in each set. Rows represent peaks. Columns represent replicates. (F) Motifs discovered de novo from differentially bound sequences within each set in (E). Sequence letter height is correlated with conservation. (G) EMSA analysis using FGF3-201 and TRH-122 sites and their mutants as probes and purified recombinant DBD of MBP-KLF4 or MBP-KLF4^K409Q proteins (left panel) or BL21 E. coli lysates expressing MBP-KLF4 full-length proteins (right panel), as indicated. Free probes and DNA/protein complexes are marked by arrows. (H) Sequence of DNA probes used in EMSA in (G) and mutated nucleotides (in red) are aligned with the KLF4 DNA-binding consensus below. Nucleotides corresponding to the aligned consensus are bolded and nucleotide triplicates binding different ZFs are interspaced. N: any nucleotide; Y: pyrimidine (C or T); K: Keto (G or T). See also Tables S2 and S3.

Figure 5

KLF4^K409Q-binding regions in the FGF3 locus appear to be vast short tandem repeats (STRs) (A) ChIP-seq analysis of FGF3 locus. Bigwig tracks display log₂ ratio of KLF4 ChIP-seq coverage relative to input. One representative replicate for each ChIP-seq condition (KLF4 in blue and KLF4^K409Q in red) is labeled on the left side. Six independent biological replicates for each ChIP-seq condition were analyzed and are shown in Figure S3. Schematic position and direction of FGF3 gene transcription are shown below the tracks. Bottom track shows the FGF3 locus alignment with heatmap of RefSNPs database. Promoter and STR regions are depicted above the plots (see also Figures S3 and S6). (B) Sequences of FGF3 locus STRs. Schematic drawing (up-to-scale) of FGF3 locus shown as a thick grey line. FGF3 gene and its direction of transcription is shown by black arrow. Exons presented as thin black boxes. Promoter region and STRs are shown as wider open boxes. STR sequences are shown above and below the locus scheme in blow-out windows. Tandem copies of 19 bp repeats in −52 kb STR are labeled by alternating black and green letters. Nucleotides corresponding to the 10 bp KLF4^K409Q consensus site are shown above the sequence window with bold letters (Y: pyrimidine, R: purine). Tandem copies of 4 bp repeats in IN2.1 and IN2.2 STRs from FGF3 intron 2 are marked by red, blue, and black letters. STR lengths (bp) and direction of FGF3 transcription are marked above each window with number and arrow (see also Figures S4 and S7).

Figure 6

FGF3 locus STRs bind KLF4^K409Q and enhance KLF4^K409Q-specific FGF3 promoter activity (A) Activity of luciferase (LUC) gene under the control of minimal FGF3 promoter (from −236 to +84 bp from TSS), alone or with different STRs positioned as enhancers. Reporter vectors were transiently co-transfected with KLF4- or KLF4^K409Q-expressing plasmids into HEK293 cells. Luciferase activity was measured 48 h posttransfection. Data are represented as mean ± SD from at least four independent experiments. (B) Quantitative DNase I footprinting analysis of the 441 bp DNA fragment from STR-52 kb upstream of FGF3 TSS (FGF3-52 kb) with increasing amounts of recombinant KLF4 and KLF4^K409Q DBD proteins as indicated. G + A ladder is shown as probe sequence marker. Open bars mark areas of KLF4 protein binding. The wide long grey box denotes the binding area of KLF4^K409Q protein. (C) Quantitative DNase I footprinting analysis of the 463 bp DNA fragment from STR in FGF3 intron 2 (FGF3-IN2.1) was performed and marked as in (B). Both coding (+) and template (−) DNA strands were labeled and tested (as indicated) (see also Figure S4).

Figure 7

Meningeal cells display KLF4^K409Q-dependent FGF3 transcription and proliferate in response to FGF3 (A) FGF3 and TRH mRNA expression in HBL-52 meningioma cells transfected with KLF4- or KLF4^K409Q-expressing plasmids. Total RNA was purified 48 h posttransfection, and mRNA was amplified by RT-qPCR using GAPDH as internal control. Copy numbers were calculated per 10⁶ copies of GAPDH mRNA in the same sample. (B) PCA Plot of RNA-seq meta-analysis in HMCs transduced with virus expressing KLF4 or KLF4^K409Q proteins (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE156211. Clustering of samples shows batch effect: instead of clustering by conditions, the samples clustered by their Sample ID (i.e., 1, 2, 3, 4). See also Figure S5A. (C) Scatter plot showing activated and repressed DE genes in meta-analysis of RNA-seq in HMCs. FGF3 and TRH are marked by arrows. See also Venn diagrams in Figure S5B. (D) Proliferation response of meningioma cell line HBL-52 to recombinant FGF3, FGF1, and EGF. Cells were treated with human recombinant FGF3 or FGF1 at 0.1 μg/mL in the presence of 1 μg/mL of heparin or EGF at 5 ng/mL final concentrations in complete medium. Cell proliferation was measured on different days poststimulation (as marked) by Absorbance (OD value at 450 nm) using a Cell Counting Kit-8 [CCK-8] as described in Method details. Assay was set up in 96-well plates (starting at 1,000 cells/well), with eight replicates for each condition and repeated three times. Data are represented as mean ± SD (see also Figure S5, Tables S4 and S5).