Fig 1.
The structure of the KDM2A and KDM2B proteins and mRNAs.
A. The protein structure of KDM2A and KDM2B is very similar. They contain the N-terminal Jumonji-C demethylase domain (JmjC), the DNA binding domain (CXXC), the zinc finger PHD domain (PHD), the F-box domain (F-box), and the C-terminal leucine rich repeats domain (LRR). The KDM2A proteins contain also the HP1a interaction motif (in red). B. The exon structures of the KDM2A and KDM2B genes are also very similar. The full length KDM2A and KDM2B mRNA isoforms contain 21 and 23 exons, respectively, and the transcription of both KDM2A-SF and KDM2B-SF mRNAs starts in the 12th intron of the KDM2A and KDM2B genes, respectively.
Fig 2.
KDM2A-SF and KDM2B-SF repress the canonical Wnt signaling luciferase reporter.
A. The wild type TCF/LEF reporter (TOP5) contains five TCF/LEF consensus binding sites (CTTTGAT) that drive the expression of the luciferase gene. The FOP5 construct contains five mutant TCF/LEF binding sites (CTTTGCC) instead and serves as the background activity control. B. Both KDM2A-LF and KDM2A-SF strongly repressed the TOP5 reporter activated with BIO. The reporter activity is expressed as the fold change ratio between the normalized luciferase signal of TOP5 and FOP5. C. KDM2B-SF, but not KDM2B-LF, also strongly repressed the activated reporter. D. Repression of the TOP5 reporter by KDM2A-LF is not dependent on the activity of its JmjC demethylase domain, but on the CXXC DNA binding domain, PHD domain and HP1a interaction motif. The HP1a motif of KDM2A-SF is not necessary for the repressive effect. E. KDM2B-LF was not able to repress the activated TOP5 reporter, whereas KDM2B-SF strongly repressed it in a DNA binding domain dependent manner independently of the PHD domain. (*p < 0.05; **p < 0.01).
Fig 3.
KDM2A-SF and KDM2B-SF repress axin 2 and cyclin D1.
A. Both KDM2A-LF and KDM2A-SF repressed axin 2 (AXIN2) in a DNA binding domain dependent manner. B. When activated with BIO, cyclin D1 (CCND1) is also repressed by both KDM2A-LF and KDM2A-SF, but independently of the DNA binding domain. C. Similarly to KDM2A-SF, KDM2B-SF repressed activated axin 2, whereas KDM2B-LF had no effect on its transcription. D. cyclin D1 was also repressed by KDM2B-SF, but not by KDM2B-LF. The mRNA levels were determined by Q-RT-PCR and are related to GAPDH. Similar patterns were obtained also with HPRT and RPL32 (*p < 0.05; **p < 0.01).
Fig 4.
KDM2A-SF and KDM2B-SF repress the axin 2 promoter.
A. The axin 2 promoter luciferase reporter constructs contain either a 1kb axin 2 intron 1 region (pAXIN2-1kb) or a 2.5kb axin 2 intron 1 region (pAXIN2-2.5kb). B. KDM2A-LF, KDM2A-SF and KDM2B-SF all repressed the pAXIN2-2.5kb reporter, whereas KDM2B-LF did not. The activity of the pAXIN2-2.5kb reporter is expressed as the fold change ratio between the activity of pAXIN2-2.5kb and that of pAXIN2-1kb. Similar pattern was observed when the results were expressed as the fold change between the activity of pAXIN2-2.5kb and the empty luciferase plasmid. (*p < 0.05; **p < 0.01).
Fig 5.
KDM2A-SF and KDM2B-SF bind to the repressed axin 2 and cyclin D1 promoter regions.
A. KDM2A-LF and KDM2A-SF both bind to the AXIN2 promoter. B. Only KDM2B-SF, but not KDM2B-LF, binds to the AXIN2 promoter. C. The levels of H3K4me3 on the AXIN2 promoter rose after the treatment with BIO and then decreased in the presence of the KDM2A isoforms. D. KDM2A-SF binds to the AXIN2 promoter with a higher efficiency than KDM2A-LF. E. KDM2B-SF, but not KDM2B-LF, binds to the cyclin D1 (CCND1) promoter. F. The KDM2A isoforms negatively affected the H3K4me3 levels also on the cyclin D1 promoter. G. The negative control region did not co-immunoprecipitate with the KDM2A/B isoforms. (*p < 0.05; **p < 0.01).
Fig 6.
Both the canonical and the demethylase domain deficient isoforms of KDM2A and KDM2B interact with TCF7L1.
The myc-tagged TCF7L1 was overexpressed in HEK293T cells together with the FLAG-tagged KDM2A/B isoforms and the proteins immunoprecipitated with anti-FLAG antibody were analyzed by western blot. All the four FLAG-tagged KDM2A/B isoforms co-immunoprecipitated with TCF7L1. The input corresponds to 5% of the whole cell extract that was used for each immunoprecipitation reaction.