ZBTB48 interacts with the m6A/m6Am demethylase enzyme FTO

We first investigated the function (s) of ZBTB48 by exploring its protein-protein interaction (PPI) partners in our previously published affinity-purification followed by mass spectrometry (AP-MS) data in HEK293 cells [44]. Consistent with its known function(s) as a DNA-binding factor, proteins involved in multiple DNA-related processes, including DNA repair, DNA recombination, and chromosome organization, were identified as interaction partners of ZBTB48 (Additional file 1: Fig. S1A). Unexpectedly, we also identified the m6A/m6Am demethylase FTO as an interaction partner, suggesting a potential role for ZBTB48 in m6A/m6Am metabolism. We assessed the specificity of its interaction with FTO by filtering ZBTB48 AP-MS data against similar data for numerous GFP purifications and several DNA-binding transcription factor purifications (total controls: n=218) using the “Significance Analysis of INTeractome” (SAINTexpress) algorithm [45]. While the number of ZBTB48 interaction partners was then reduced to only 12 proteins that passed our statistical threshold of a Bayesian false discovery rate (FDR) ≤ 0.01 (Fig. 1A; Additional file 2: Table S1), FTO remained one of the highest-confidence interaction partners, indicating that FTO association with ZBTB48 is robust. Moreover, our analysis of the ‘Contaminant Repository for Affinity Purification’ (CRAPome) database did not identify FTO as a ‘frequent flyer’ protein in AP-MS experiments (Additional file 1: Fig. S1B) [46], excluding the possibility that FTO might be a general non-specific background protein in proteomic studies.

Fig. 1

ZBTB48 interacts with FTO and binds to RNA in cells. A Left, Dot plot representation of ZBTB48 protein-protein interactions. Right, Interaction of ZBTB48 with FTO detected by co-immunoprecipitation. Cell lysates were treated with Benzonase prior to IPs. Note: The inputs and IPs were loaded twice on 4-12% Bis-Tris SDS polyacrylamide gels and probed with the indicated antibodies. B CLIP autoradiography of 32P-labeled ZBTB48-RNA complexes was performed after RNase I treatment at various dilution. The Western blot in the bottom panel shows the recovery of ZBTB48. C Left, Schematic representation of CLIP RNA over-digestion assay. Right, Autoradiograph of immunopurified 32P-labeled ZBTB48-RNA complexes after RNase I and/or DNase I over-digestion. The bottom panel Western blot indicates the recovery of GFP-ZBTB48 protein. D Bar chart representing the distribution (%) of ZBTB48-bound RNA types. E Immunofluorescence (IF) analysis to examine the localization of ZBTB48 in HEK293 cells. Experiment was performed using an anti-ZBTB48 antibody. For nuclear counterstaining, DAPI was used. F Standardized metaplot profile showing the density of ZBTB48 as average crosslinks per million. Black arrow indicates transcription start site (TSS). G Bar chart representing the distribution of ZBTB48 mRNA CITS peaks per Kb of mRNA (FDR ≤ 0.01). H Enriched sequence motif found in ZBTB48 iCLIP-seq peaks. E-value represents the significance of the motif against randomly assorted sequences. I iCLIP-seq signal density of ZBTB48 around either ZBTB48 ChIP-seq peaks or random sites. Shaded area indicates a standard error of mean (SEM). See also Additional file 1: Fig. S1, S2; Additional file 2, 3, 4: Tables S1, S2, and S3

To validate the interaction between ZBTB48 and FTO, co-immunoprecipitation (co-IP) experiments were performed. Consistent with our AP-MS analysis, FTO co-immunoprecipitated with GFP-ZBTB48 (Additional file 1: Fig. S1C). To further confirm that endogenous ZBTB48 interacts with FTO, co-IPs were performed using antibodies against the endogenous ZBTB48. To test for possible indirect associations mediated by DNA or RNA, cell extracts were treated with a promiscuous nuclease (Benzonase) prior to IPs. In these experiments endogenous FTO co-immunoprecipitated with ZBTB48 (Fig. 1A). Conversely, endogenous ZBTB48 was pulled down with FLAG-FTO (Fig. 1A). To roughly map the FTO-interaction region of ZBTB48, we engineered HEK293 cells expressing ZBTB48 truncation mutants and subjected them to co-IPs. Both the full-length protein and the N-terminal region of ZBTB48 containing its BTB domain immunoprecipitated FTO, whereas the C-terminal region harbouring its zinc fingers (ZnFs) did not (Additional file 1: Fig. S1D). These results suggested that the BTB domain or an unstructured region of ZBTB48 might participate in its interaction with FTO. We conclude that ZBTB48 forms a direct or indirect PPI with FTO. This interaction is highly specific, since FTO did not co-purify with 120 other C2H2-ZFPs [44].

ZBTB48 binds to mRNAs in cells

Because ZBTB48 directly or indirectly interacted with the RNA demethylase FTO, we examined whether ZBTB48 binds to RNA in cells. By utilizing CLIP followed by radio-labeling of RNA, SDS-PAGE, and autoradiography, we observed that ZBTB48 crosslinks robustly to RNA in cells, specifically upon UV exposure. Importantly, the bulk of the radioactive signal migrated above the position of ZBTB48 when a low concentration of RNase I was used and increasingly migrated at the position of ZBTB48 as the RNase I concentration was increased (Fig. 1B). Moreover, the radiolabelled RNA that crosslinked to ZBTB48 was substantially reduced when CLIP samples were over-digested with RNase I, but not DNase I, indicating that ZBTB48 binds directly to RNA in cells (Fig. 1C). To further confirm that the RNA signal emerged from crosslinking of RNA to ZBTB48 and was not due to RNA-binding proteins or RNA interacting with ZBTB48 in extracts from UV-treated cells, we mixed UV-untreated cells expressing GFP-ZBTB48 with UV-treated wildtype cells before lysis and used them as a control in CLIP assays. Consistently, radioactive signal was observed only when the GFP-ZBTB48-expressing cells were treated with UV (Additional file 1: Fig. S1E). In contrast, neither GFP-alone nor the ‘mixed’ control samples displayed any appreciable radiolabeled RNA signal (Additional file 1: Fig. S1E). We then subjected ZBTB48 truncation mutants to CLIP assays and found that the C-terminal region containing its ZnFs strongly crosslinked to RNA, whereas the N-terminal region yielded a much weaker signal (Additional file 1: Fig. S1F). These results suggested that, like its DNA binding, ZBTB48 binds to RNA through its ZnFs.

To identify ZBTB48-bound RNAs in cells, we performed iCLIP-seq experiments using a previously validated anti-ZBTB48 antibody [40] (Additional file 1: Fig. S1G), in parallel with sequencing of size-matched input (SMI) samples. Biological replicates highly correlated with each other, indicating the reproducibility of our data (Additional file 1: Fig. S1H). In addition to the SMI samples, iCLIP-seq data from GFP-alone samples were employed to control for background signal. We defined ZBTB48 RNA-binding sites by identifying ‘crosslinking-induced truncation sites’ (CITS, FDR ≤ 0.01; Methods). The majority (~80%) of the identified CITS peaks fall within protein-coding transcripts, indicating that ZBTB48 predominantly targets mRNAs in cells (Fig. 1D). Without length normalization across various transcript features, ~70% of crosslinking peaks were identified as intronic (Additional file 1: Fig. S1I; Additional file 3: Table S2), consistent with introns being much longer than exons. The peaks within introns suggest binding to pre-mRNA and, consistently, immunofluorescence analysis showed that ZBTB48 is predominantly nuclear (Fig. 1E). When we examined the average RNA-binding density after normalizing to the lengths of various features along target mRNAs using a composite analysis, most ZBTB48 crosslinks accumulated within 5’ and 3’UTRs (Fig. 1F), whereas ~24% of the peaks reside within coding sequences (CDS) (Fig. 1G), suggesting that ZBTB48 might preferentially bind to the untranslated regions of target mRNAs. To examine if ZBTB48 has any sequence preferences, we searched for enriched motifs and identified a U-rich consensus sequence (_UU_ _) as the most significantly enriched motif. GGGAGG and _ _CCAGC were identified as the second and third most over-represented motifs, respectively (Fig. 1H, Additional file 1: Fig. S1J).

Since ZBTB48 binds to telomeric DNA repeats and can also function as a transcription factor for a few genes [40, 41], we examined whether its DNA- and RNA-binding sites are correlated. ChIP-seq experiments were performed with two biological replicates to identify ZBTB48 DNA-binding sites (Additional file 1: Fig. S2A; Additional file 4: Table S3). Even though our ChIP-seq data were generated using GFP-ZBTB48 in HEK293 cells, and we employed enrichment over input to identify significant peaks, the ChIP peaks we identified were enriched around previously published ChIP sites that were detected for endogenous ZBTB48 in U2OS cells using ZBTB48 knockout cells as a control (Additional file 1: Fig. S2B) [40]. Conversely, previously published ChIP sites were enriched around our identified ZBTBTB sites (Additional file 1: Fig. S2B). These observations indicate that our data captured the reported ZBTB48 DNA-binding profile. When the RNA-crosslink sites were plotted around the ChIP-seq peaks, ZBTB48 RNA-binding density was not enriched around its DNA-binding sites (Fig. 1I), and only ~5% of ChIP-seq peaks overlapped with the iCLIP-seq peaks when the peaks were extended by 50 nucleotides on each side (Additional file 1: Fig. S2C). Thus, ZBTB48 appears to have distinct, mostly non-overlapping, DNA- and RNA-binding sites, although there is a small degree of enrichment specifically in 3’ and 5’ UTRs (Additional file 1: Fig. S2D).

In addition to our use of a previously validated antibody [40], as well as our additional validations of ZBTB48 antibody using ZBTB48 knockdown cells (see below), we independently confirmed RNA-binding by ZBTB48 by performing RNA-immunoprecipitation (RIP) followed by qRT-PCR experiments utilizing GFP-ZBTB48-expressing cells. As shown in Additional file 1: Fig. S2E, precipitation of GFP-ZBTB48 successfully pulled down 8/8 of the target transcripts, as defined by our iCLIP-seq experiments, that we examined, whereas precipitation of GFP-alone did not. In contrast GFP-ZBTB48 did not show significant binding to two non-targets (Additional file 1: Fig. S2E). We conclude that ZBTB48 binds directly to mRNAs in cells and that binding is mostly distinct from its DNA-binding activity.

ZBTB48 and FTO RNA-binding sites coincide

Since ZBTB48 associated with FTO, we next examined whether their RNA-binding sites coincide on mRNAs. We performed FTO iCLIP-seq experiments, with four biological replicates, using inducible Flp-In T-REx HEK293 cells expressing FLAG-tagged FTO (Additional file 1: Fig. S2F; Additional file 5: Table S4). Consistent with previous studies [33, 37], we found that most FTO CITS peaks (~80%) fall within introns (Fig. 2A, left), whereas, after normalizing to the lengths of various transcript regions, ~28%, ~34%, and ~28% of FTO peaks map within 5’UTRs, 3’UTRs, and CDS, respectively (Fig. 2A, right). To examine whether ZBTB48 crosslink sites coincide with FTO binding sites, we plotted the ZBTB48 CITS around the FTO RNA-binding positions. Importantly, ZBTB48 iCLIP signal was significantly enriched in the vicinity of FTO-binding sites (Fig. 2B, p ≤ 0.001, Wilcoxon test; Additional file 1: Fig. S3A, left). Furthermore, ZBTB48 signal was also enriched around FTO peaks when previously published CLIP-seq data was used to map FTO RNA-binding sites (Additional file 1: Fig. S3B) [37].

Fig. 2

RNA-binding sites of ZBTB48 and FTO coincide on target transcripts. A Left, FTO iCLIP-seq CITS distribution. Absolute numbers of peaks are used. Right, Bar chart showing the distribution of FTO iCLIP-seq peaks per Kb along mRNA. B Metagene profile showing ZBTB48 iCLIP-seq signal density around either FTO RNA-binding sites or random sites. Quantification of signal densities was performed for +/-100bp around the center (∗∗∗p ≤ 0.001, Wilcoxon (Mann-Whitney) test). Shaded area: SEM. C FTO iCLIP-seq signal density around either ZBTB48 CITS or random sites (∗∗∗p ≤ 0.001, Wilcoxon test). D Standardized metaplot profiles showing the binding density (iCLIP/size-matched input) of FTO in either siZBTB48 or siNT cells. E Bar graph showing FTO RIP-qPCR experiments after knocking down ZBTB48 (biological replicates (n) = 4, student’s t test, ∗∗p ≤ 0.01, ∗p ≤ 0.05, error bars: SEM). F ZBTB48 iCLIP-seq signal density around either m6A sites or random sites. G FTO signal density around m6A (left) or m6Am (right) sites in cells treated with either siZBTB48 or siNT. Note: For F and G, quantification of signal densities was performed for +/-100bp around the center; ∗∗∗p ≤ 0.001, Wilcoxon test. H ZBTB48 CLIP dot blot showing the enrichment of m6A/m6Am modified RNA. Top panel: anti-m6A, bottom panel: Methylene blue. I Bar plot showing the enrichment of m6A. RIP was performed using either GFP-ZBTB48 or GFP-alone cells, and RNA was analyzed using an m6A ELISA kit. Synthetic RNA containing m6A was used as a positive control (error bars: SEM, n =3, student’s t test, ∗∗p ≤ 0.01). J m6A dot blot using polyadenylated RNA purified from either GFP-alone, or ZBTB48-overexpressing, or FTO-overexpressing cells (top panel: anti-m6A, bottom panel: Methylene blue, error bars: SEM, ∗p ≤ 0.05; student’s t-test; n = 3). See also Additional file 1: Fig. S3, S4, Additional file 5: Table S4

Consistently, when examining the RNA-binding profile of FTO around ZBTB48 crosslink sites, we found that FTO signal was also significantly enriched around ZBTB48 RNA-binding sites (Fig. 2C), and this enrichment was observed for 5’UTR, CDS, and 3’UTR regions (Additional file 1: Fig. S3C, p ≤ 0.001, Wilcoxon test). Consistent with their overlapping RNA-binding sites, the majority of the identified target transcripts were also shared between the two proteins (Additional file 1: Fig. S3A, right). To further examine the specificity of FTO enrichment around ZBTB48 sites, we then utilized our previously reported iCLIP-seq data for SP1 [47]. Consistently, SP1, which we have previously shown to function in alternative polyadenylation [47], was not significantly enriched around ZBTB48 sites (Additional file 1: Fig. S3D). Moreover, FTO peaks were depleted around ZBTB48 DNA-binding sites (Additional file 1: Fig. S3E), perhaps because transcription factor-binding regions in promoters or elsewhere are not transcribed, indicating that the ZBTB48-FTO interaction may not normally occur in the context of the DNA-related functions of ZBTB48. We conclude that the ZBTB48 and FTO RNA-binding sites are coincident across the transcriptome, consistent with our observation of a direct or indirect physical association between these factors. Collectively these observations suggest possible co-regulatory potential for these proteins on their common target mRNAs.

ZBTB48 modulates FTO RNA-binding

To address their possible functional interaction, we assessed whether ZBTB48 affects the RNA-binding activity of FTO by performing FTO iCLIP-seq in HEK293 cells, in 2 biological replicates, after knockdown (KD) with a ZBTB48 siRNA or non-targeting siRNAs (siNT) (Additional file 1: Fig. S4A). Western blotting analysis using anti-ZBTB48 antibody confirmed that the ZBTB48 protein levels were specifically depleted in siZBTB48-treated cells (Additional file 1: Fig. S4B). The expression of FTO remained unaffected upon ZBTB48 KD at the levels of both mRNA and protein (Additional file 1: Fig. S4B). Moreover, the subcellular localization of FTO did not change in HEK293 cells following ZBTB48 KD (Additional file 1: Fig. S4C). In siNT treated cells, FTO’s RNA-binding density was particularly enriched in the 5’UTRs (Fig. 2D), consistent with previous studies [33, 37] and results described above. In contrast, the RNA-binding density of FTO was substantially reduced all along target transcripts in siZBTB48 cells, and especially in the 5’UTR, even though the number of uniquely mapping reads was comparable across both conditions (Fig. 2D, Additional file 1: Fig. S4D; Additional file 6: Table S5). These results suggested that ZBTB48 recruits FTO to target transcripts in cells.

Since RNA-binding sites of FTO often coincide with those of ZBTB48, we also examined whether depleting ZBTB48 affects FTO binding to these sites. When we plotted FTO iCLIP-seq signal around ZBTB48-binding sites, we observed a reduction in the RNA-binding density of FTO in siZBTB48 cells in comparison with control cells (Additional file 1: Fig. S4E, p ≤ 0.001, Wilcoxon test). Furthermore, FTO RIP-qPCR experiments in cells treated with a different siRNA against ZBTB48 (siZBTB48 # 2) (Additional file 1: Fig. S4A and S4F) showed a significant reduction in FTO signal for the two FTO targets we examined in ZBTB48-depleted cells (Fig. 2E, p ≤ 0.01, student’s t-test). In contrast, RNA-binding by ZBTB48 in FTO depleted cells remained unaffected for the examined target transcripts (Additional file 1: Fig. S4G). We conclude that ZBTB48 helps recruit FTO to target mRNAs in cells.

ZBTB48 affects FTO RNA-binding around m6A/m6Am sites

We next utilized published m6A-iCLIP-seq (miCLIP-seq) data [28] to ask whether ZBTB48 affects binding of FTO around m6A/m6Am sites in cells. FTO showed an enrichment around m6A and m6Am sites, as assessed using both our FTO iCLIP-seq and previously reported FTO CLIP-seq data (Additional file 1: Fig. S5A-C) [37]. Consistently, ZBTB48 also exhibited strong enrichments around m6A and m6Am sites (Fig. 2F, Additional file 1: Fig. S5D). In contrast to FTO and ZBTB48, however, none of the other RNA-binding proteins we examined, including U2AF1, PTBP1, MKRN2 [48], and SP1 [47], showed any enrichment around m6A sites (Additional file 1: Fig. S5E). We then utilized our FTO iCLIP-seq data generated in siZBTB48 or siNT cells and observed that FTO binding around m6A/m6Am sites was reduced in ZBTB48-depleted cells in comparison with the control cells (Fig. 2G, p ≤ 0.001, Wilcoxon test). These results suggested that ZBTB48 might reduce cellular m6A/m6Am levels by recruiting FTO around m6A/m6Am sites on target mRNAs.

ZBTB48 binds m6A/m6Am-containing RNAs and regulates cellular m6A/m6Am via FTO

Since the above results indicated that ZBTB48 binds near m6A/m6Am sites and co-ordinates the targeting of FTO to at least a subset of methylated sites, we examined whether ZBTB48 preferentially associates with m6A/m6Am-modified RNAs. We first carried out additional validations on a previously characterized anti-m6A antibody through a dot-blot assay using total RNA derived from cells treated either with siMETTL3 or control siRNAs (Additional file 1: Fig. S5F and S5G) [14, 15]. Note that the anti-m6A antibody does not distinguish between internal m6A and 5’ cap-associated m6Am. We then used this antibody to probe RNA isolated in ZBTB48 CLIP experiments. Consistently, GFP-ZBTB48 immunoprecipitated RNAs with far more m6A/m6Am-modifications compared to GFP-alone (Fig. 2H). YY1, another C2H2-ZFP that is known to bind RNA [49], was used as an additional specificity control in these experiments and failed to show any appreciable m6A/m6Am signal (Fig. 2H). To further examine the ability of ZBTB48 to target m6A/m6Am-modified RNAs, we performed GFP-ZBTB48 RIP experiments with anti-GFP followed by ELISA-based assays for m6A. Consistently, we observed that, in comparison with the GFP control, RNA immunoprecipitated with GFP-ZBTB48 was m6A/m6Am modified (Fig. 2I). These results support the conclusion that ZBTB48 binds near m6A/m6Am residues on its target RNAs.

Since ZBTB48 targeted methylated transcripts and modulated the mRNA-binding activity of FTO, we examined if ZBTB48 influences total m6A/m6Am levels in mRNA. We isolated PolyA-enriched mRNAs from cells overexpressing GFP-ZBTB48, FTO, or GFP (Additional file 1: Fig. S6A) and subjected these mRNAs to m6A/m6Am dot-blot assays. As expected, FTO overexpression resulted in a significant reduction in global m6A/m6Am levels in mRNAs (Fig. 2J). Notably, in comparison with control cells expressing GFP, m6A/m6Am levels in cellular mRNAs were also significantly reduced upon GFP-ZBTB48 overexpression (Fig. 2J, p ≤ 0.05, student’s t-test). These results are consistent with the idea that ZBTB48 co-ordinates FTO-binding to target RNAs, thus indirectly regulating cellular m6A/m6Am levels.

ZBTB48 modulates FTO-mediated transcriptome-wide demethylation of mRNAs

To further assess the potential impact of ZBTB48 on cellular m6A/m6Am at a transcriptome level, we carried out miCLIP-seq experiments, in two biological replicates, using HEK293 cells overexpressing either GFP-ZBTB48, or FTO, or GFP-alone (Fig. 3A, Additional file 1: Fig. S6B; Additional file 7, 8, 9: Table S6, S7, S8). Since ZBTB48 and FTO both appear to target pre-mRNAs, we isolated total cellular RNA in these experiments, instead of using poly-A enriched processed mRNAs. The majority of identified CITS (FDR ≤ 0.01) fell within protein-coding genes with a strong enrichment near the stop codons and 3’UTR, as well as a smaller enrichment in the 5’UTR, and the DRACH sequence was found as the most significantly enriched motif in each dataset (Additional file 1: Fig. S6B), consistent with previous studies [28].

Fig. 3

ZBTB48 affects cellular m6A/m6Am levels. A Autoradiographs of 32P-labeled immunopurified (IP) anti-m6A-RNA complexes. Red boxes indicate the excised areas. B standardized metagene plots showing miCLIP-seq signal density for RNA purified from cells overexpressing GFP-alone, or GFP-ZBTB48, or GFP-FTO. C Boxplot representation of read counts for DRACH motif containing m6A sites across different samples, as indicated (Mann–Whitney test, ∗∗∗p ≤ 0.001; outliers not shown in the plot). D Box plot showing differential methylation analysis using miCLIP data generated from either GFP-ZBTB48-overexpressing or GFP-FTO-overexpressing cells versus GFP control cells. Analysis was performed with the DESeq2 package, which accounts for the read depth in each sample, using reads corresponding to internal m6A (left) and terminal m6Am (right) sites. Outliers are not shown in the box plots. Dotted lines indicate the base line signal (i.e., m6A or m6Am in GFP cells). Bottom bar plots indicate the mean values for the indicated groups along with p-values (ANOVA, post hoc Tukey HSD). E Genome browser snapshots of JUN and FOXD1 gene tracks show the location and coverage of miCLIP in GFP- or GFP-ZBTB48- or FTO-overexpressing cells. Locations of the putative m6A sites are highlighted. F Bar plots showing the m6A levels for JUN and FOXO3 transcripts estimated using m6A-RIP-qPCR on RNA purified from either ZBTB48-overexpressing or GFP-only cells (left) or from either FTO-overexpressing or GFP-only cells (right). Data are represented as % input. Note, for qPCRs: biological replicates n = 5, student’s t-test, ∗∗∗p ≤ 0.001, ∗∗p ≤ 0.01, ∗p ≤ 0.05, n.s.: non-significant, error bars denote SEM. See also Additional file 1: Fig. S5, S6; Additional file 6, 7, 8, 9: Table S5, S6, S7,S8

By comparing their signal densities, we found that, on average, m6A/m6Am intensities were substantially reduced in FTO- or ZBTB48-overexpression samples when compared with GFP controls (Fig. 3B). We next considered only DRACH-motif-containing peaks as the high-confidence m6A sites for further analyses (Additional file 1: Fig. S6C). Moreover, we also identified potential m6Am sites (see Methods for details) (Additional file 1: Fig. S6D). Consistently, we observed a significant reduction in both DRACH-motif-containing m6A sites in the 3’UTR, CDS, and 3’UTR, as well as m6Am, upon FTO- or ZBTB48-overexpression, in comparison with control samples (Fig. 3C and Additional file 1: Fig. S6E and S6F, p ≤ 0.001, Mann–Whitney). Differential analysis revealed that both FTO- and ZBTB48-overexpression samples had substantially reduced RNA methylation levels relative to control GFP samples (Fig. 3D-E, Additional file 1: Fig. S6G and S7A), suggesting a global reduction in transcriptome methylation levels. When we overlapped the m6A-containing genes, ~2200 genes that were present after GFP-overexpression appeared to lack a defined m6A CITS peak in FTO- or ZBTB48-overexpression samples (Additional file 1: Fig. S7B). We validated this change in m6A levels for two targets, JUN and FOXO3, using m6A-RIP-qPCR experiments in cells overexpressing either ZBTB48, or FTO, or GFP. Consistently, m6A levels were significantly reduced for both examined transcripts in ZBTB48- or FTO-overexpression samples in comparison with control GFP cells (Fig. 3F, p ≤ 0.001, t-test).

We initially utilized ZBTB48- and FTO-overexpressing cells in our experiments for consistency, because elevated FTO levels have previously been shown to significantly decrease cellular m6A levels [32]. Although the expression of known Mettl3/14-complex subunits and the m6Am writer enzyme PCIF1, as well as YTH-family proteins, was not significantly changed (see below), we noted that the overexpression of ZBTB48 elevated FTO expression levels (Additional file 1: Fig. S7C). The observed increase in FTO expression does not appear to have been caused directly by the transcription factor-related functions of ZBTB48, since ZBTB48 did not target the FTO promoter in our ChIP-seq data (Additional file 4: Table S3). ZBTB48 and FTO mRNA expression levels were comparable in wildtype cells (Additional file 1: Fig. S7D), suggesting that the observed increase in FTO levels upon ZBTB48 overexpression could simply reflect a feedback mechanism in which cells sense the need for additional FTO protein. However, this elevation of FTO could at least partly explain why overexpression of ZBTB48 causes reductions in the levels of RNA methylation.

Because of this issue, we also examined m6A/m6Am levels in siZBTB48- or siFTO-treated cells. Importantly, ZBTB48 knockdown did not significantly alter the expression level of FTO (Additional file 1: Fig. S4B), and vice versa (Additional file 1: Fig. S7E). Moreover, the expression of METTL3 and PCIF1 also remained unaffected in siZBTB48 or siFTO cells (Additional file 1: Fig. S7E). When we analyzed m6A-to-A and m6Am-to-A ratios in total RNA using quantification by LC-MS/MS, knockdown of ZBTB48 led to a significant increase in cellular m6A and m6Am levels in comparison with control knockdown samples (Fig. 4A, p ≤ 0.05; t-test), and consistent with previous reports [32], FTO depletion also resulted in an increase in both m6A and m6Am. Conversely, LC-MS/MS confirmed that overexpression of ZBTB48 or FTO decreased m6A and m6Am in purified polyadenylated mRNAs as well as in total RNA samples (Additional file 1: Fig. S7F, S7G), consistent with the findings in Fig. 2J. We validated these findings for two individual targets, JUN and FOXO3 by performing m6A-RIP followed by qRT-PCR experiments after knocking down ZBTB48 or FTO using different siRNAs (siZBTB48 # 2 and siFTO # 2) (Fig. 4B, p ≤ 0.001; t-test).

Fig. 4

ZBTB48 knockdown increases cellular m6A/m6Am. A LC-MS/MS shows that knockdown of ZBTB48 or FTO leads to an increase in m6A/A (left) and m6Am/A (right) in total RNA in comparison with the controls (i.e., siNT). Error bars show SEM (∗p ≤ 0.05; student’s t-test). B Bar plot showing the m6A levels for JUN and FOXO3 transcripts estimated using m6A-RIP-qPCR on RNA purified from either siZBTB48- or siNT-treated cells (left) or from either siFTO- or siNT-treated cells (right). Data are represented as % input. C-D Bar plots showing the m6A levels for JUN and FOXO3 transcripts estimated using m6A-RIP-qPCR on RNA purified from cells treated in the indicated ways. Data are represented as % input. Note, qPCRs in B-D: biological replicates n = 5, student’s t-test, ∗∗∗p ≤ 0.001, ∗∗p ≤ 0.01, ∗p ≤ 0.05, n.s.: non-significant, error bars denote SEM. See also Additional file 1: Fig. S6, S7.

To investigate whether ZBTB48-mediated modulation of m6A requires FTO, m6A-RIP-qPCR experiments were per