Article Figures & Data
Figures
- Figure 1. The MLL3/4 complexes associate with PROSER1, a novel proline and serine rich protein, the DNA demethylase TET2 and the glycosyltransferase OGT.
(A) FLAG-UTX immunoprecipitation (IP) followed by mass spectrometry identifies all known subunits of the MLL3/4 complexes along with PROSER1, a novel serine and proline rich protein, the DNA demethylase TET2 and the glycosyltransferase OGT. SC, spectral counts; TP, peptide counts; abundance = SC × 50 (kD)/protein size (kD). (B) Western blot of FLAG-UTX IP from HEK293 cells confirming interaction of UTX with PROSER1, TET2 and OGT. UTX interacts with the H3K4 methyltransferases MLL3 and MLL4, RBBP5 a core component of the MLL3/4 complexes, PROSER1, TET2, and OGT. HEK293 cells with a FLAG-tag expressing plasmid were used as an IP control. Nuclear extracts were used as input. Actin was used as a loading control for the inputs. (C) Glycerol gradient sedimentation after FLAG-UTX IP reveals co-fractionation of PROSER1 and TET2 with components of the MLL3/4 complexes (red box).
- Figure 2. PROSER1 interacts with members of the TET family of DNA demethylases and OGT.
(A) FLAG-PROSER1 IP followed by mass spectrometry identifies the glycosyltransferase OGT and all three TET family proteins. (B) Western blot of FLAG IP from FLAG-HA-NeonGreen-PROSER1 (FHNG-PROSER1) knock-in HEK293 cells confirming interaction of PROSER1 with OGT, TET1, TET2, and UTX. Wild-type (WT) HEK293 cells were used as an IP control. Nuclear extracts were used as input. Actin was used as a loading control for the inputs. The asterisk indicates the IgG heavy chain. (C) Top: Domain structure and constructs of mouse TET2. Blue: cysteine-rich dioxygenase (CD) domain. TET2 FL, full length construct. TET2 N500, N-terminal 500 amino acids (aa). TET2 N1041, N-terminal 1,041 aa. TET2 CD, TET2 cysteine–rich dioxygenase domain. Bottom: HEK293 cells were transiently transfected with a series of FLAG-tagged mouse TET2 constructs as shown on the top. WB of FLAG IPs from total cell lysates showing interaction of PROSER1 and OGT with TET2 FL and TET2 CD. Total cell lysates were used as inputs. Actin was used as a loading control for the inputs. The asterisk indicates the IgG heavy chain. (D) Top: Domain structure and constructs of the human TET2 cysteine–rich dioxygenase (CD) domain. Light blue, cysteine-rich domain. Red, double-stranded β-helix (DSBH1) domain 1. Olive, low-complexity insert (LCI) region. Purple, double-stranded β-helix (DSBH2) domain 2. Bottom: HEK293 cells were transiently transfected with a series of FLAG-tagged human TET2 constructs as shown on the top. WB of FLAG IPs from total cell lysates depicting interaction of PROSER1 and OGT with TET2 DSHB2. Total cell lysates were used as inputs. Actin was used as a loading control for the inputs. The asterisk indicates the IgG heavy chain. TET2 Cys-rich, TET2 LCI, and TET2 DSHB2 could not be detected in the inputs.
- Figure S1. PROSER1 interacts with components of the DNA demethylation machinery and the MLL3/4 complexes.
(A) Western blot (WB) of FLAG-PROSER1 IP from HEK293 cells confirming interaction of PROSER1 with OGT, TET2, and UTX. HEK293 cells with a FLAG-tag expressing plasmid were used as an IP control. Nuclear extracts were used as input. Actin was used as a loading control for the inputs. (B) Schematic of the FLAG-HA-NeonGreen-PROSER1 protein expressed in the FLAG-HA-NeonGreen-PROSER1 (FHNG-PROSER1) knock-in HEK293 cell line. (C) FHNG-PROSER1 cells were transiently transfected with a series of FLAG-tagged mouse TET2 constructs as shown in Fig 2C. WB of HA IP from total cell lysates showing interaction of TET2 FL and TET2 CD with FHNG-PROSER1 (all marked by arrows). Actin was used as a loading control for the inputs. The asterisks indicate the IgG heavy chain (upper) or an unspecific band (lower). For the FLAG WB, a longer exposure and a shorter exposure are shown. (D) CRISPR/Cas9 KO strategy to create PROSER1 KO clones in HEK293 cells.
- Figure 3. PROSER1 mediates TET2 O-GlcNAcylation and stability by promoting the interaction between OGT and TET2.
(A) Top: Domain structure and constructs of the human TET2 double-stranded β-helix 2 (DSBH2) domain. TET2 DSHB2-N: N-terminal 93 aa of DSHB2. TET2 DSHB2-C: C-terminal 66 aa of DSHB2. Bottom: HEK293 cells were transiently transfected with the FLAG-tagged constructs as shown on the top. Western blot of FLAG IPs showing interaction of PROSER1 and OGT with TET2 DSHB2-C. Total cell lysates were used as inputs. Actin was used as a loading control for the inputs. The asterisks indicate an unspecific band (upper panel) or the IgG heavy chain (lower panel). TET2 DSHB2-N and TET2 DSHB2-C could not be detected in the inputs. Failure to detect FLAG-TET2 DSBH2-C after FLAG IP might be due to its positive charge at the pH of the sample buffer. (B) WT and PROSER1 KO HEK293 cells were transiently transfected with FLAG-tagged TET2 DSHB2. WB of FLAG IPs depicts decreased interaction of OGT with TET2 DSHB2 in the absence of PROSER1. Total cell lysates were used as inputs. Actin was used as a loading control for the inputs. The asterisks indicate unspecific bands (middle panel) or the IgG heavy chain (lower panel). TET2 DSHB2 could not be detected in the inputs. (C) TET2 IP from nuclear extracts of WT and PROSER1 KO HEK293 cells. Approximately equal amounts of TET2 were IPed from WT and PROSER1 KO cells. A significant decrease in OGT binding to TET2 and a strong reduction of TET2 O-GlcNAcylation is observed in PROSER1 KO versus WT HEK293 cells. Nuclear extracts were used as inputs. Actin was used as a loading control for the inputs. The asterisks indicate unspecific bands, the arrow O-GlcNAcylated TET2. (D) WB for the indicated proteins from nuclear extracts of WT and PROSER1 KO HEK293 cells. Actin was used as a loading control. (E) FLAG IP from total cell lysates of FHNG-PROSER1 HEK293 cells treated with control or OGT siRNA. PROSER1 protein levels are decreased upon OGT knockdown. IPed PROSER1 from control and OGT siRNA-treated cells was adjusted to equal amounts to show that OGT catalyzes O-GlycNAcylation of PROSER1. Total cell lysates were used as inputs. Actin was used as a loading control for the inputs. The asterisk indicates the IgG heavy chain. (F) Model of PROSER1 function. PROSER1 mediates the recruitment of OGT to TET2 to facilitate O-GlcNAcylation of TET2 and PROSER1. In the absence of PROSER1, OGT association with TET2 is impaired resulting in decreased TET2 and PROSER1 O-GlcNAcylation and protein stability.
- Figure S2. TET1 and TET2 mRNA levels are unchanged in PROSER1 KO compared with WT cells.
Genome browser tracks showing the RNA-seq profiles of the TET1 and TET2 genes in WT and PROSER1 KO HEK293 cells. Two replicates are displayed for each genotype.
- Figure 4. PROSER1 regulates the chromatin association of TET1/2 to mediate UTX/H3K4me1–dependent enhancer activity.
(A) Heatmaps centered on 18,017 reproducible PROSER1 binding sites obtained from three different PROSER1 antibodies in WT HEK293 cells. Occupancy of PROSER1, UTX, OGT, TET1, TET2, H3K4me1, H3K4me2, H3K4me3, and H3K27ac is displayed. (B) Venn diagram depicting the overlap between reproducible PROSER1, UTX, and TET2 peaks in WT HEK293 cells. (C) Heatmaps centered on 4,421 UTX/H3K4me1-down-regulated regions in PROSER1 KO compared with WT cells. Occupancy of PROSER1, UTX, OGT, TET1, TET2, H3K4me1, H3K4me2, H3K4me3, and H3K27ac is displayed. (D) Genome browser tracks depicting the ChIP-seq profiles of PROSER1, UTX, OGT, TET1, TET2, H3K4me1, H3K4me2, H3K4me3, and H3K27ac in WT and PROSER1 KO HEK293 cells at the DDB2 promoter (left) and SOX2 enhancer (right). (E) Genome browser tracks showing the RNA-seq profiles of the DDB2 and SOX2 genes in WT and PROSER1 KO HEK293 cells. Two replicates are displayed for each genotype. (F) Violin plots showing transcriptional down-regulation of the 400 genes associated with the 4,421 regions displaying lower enrichment of UTX and H3K4me1 in PROSER1 KO compared with WT cells. Expression values were standardized for each gene across the samples. The median interquartile range and the 1.5× interquartile range are shown for each violin plot.
- Figure S3. PROSER1 co-localizes genome-wide with UTX, TET1, TET2, and OGT.
(A, B, C, D) Heat maps centered on (A) 109,378 UTX peaks; (B) 69,815 TET1 peaks; (C) 16,982 TET2 peaks and (D) 23,386 OGT peaks in WT HEK293 cells. Occupancy of PROSER1, UTX, OGT, TET1, TET2, H3K4me1, H3K4me2, H3K4me3, and H3K27ac is displayed. (E) Genomic distribution of all possible 1 kb genomic regions (control) and identified reproducible PROSER1, UTX, TET1, TET2, and OGT peaks in WT HEK293 cells. TES, transcription end site. (F) Venn diagram showing the overlap between down-regulated UTX and down-regulated H3K4me1 peaks in PROSER1 KO versus WT HEK293 cells (both FC ≥ 2). (G) Genomic distribution of all 4,421 UTX/H3K4me1 down-regulated peaks.
- Figure 5. PROSER1 regulates DNA demethylation on UTX/H3K4me1-dependent enhancers and CpG islands.
(A) Dot blot analysis of genomic 5mC, 5hmC, 5fC, and 5caC in WT and PROSER1 KO HEK293 cells. 5mC is unchanged and 5hmC, 5fC, and 5caC are decreased in PROSER1 KO versus WT cells. Methylene Blue staining was performed to ensure that equal amounts of DNA were used from WT and PROSER1 KO cells. (B) Heat maps displaying whole genome bisulfite sequencing (WGBS) and hMeDIP-seq results centered on 4,421 UTX/H3K4me1-down-regulated regions in PROSER1 KO versus WT cells. (C) Heat maps showing WGBS and hMeDIP-seq results centered on 128 hypermethylated CpG islands in PROSER1 KO compared with WT cells. (D) Heat maps centered on hypermethylated DMR500 regions in PROSER1 KO compared with WT cells overlapping with at least one PROSER1, UTX, or H3K4me1 peak in WT cells. Occupancy of PROSER1, UTX, OGT, TET1, TET2, H3K4me1, H3K4me2, H3K4me3, and H3K27ac and WGBS results are shown.
- Figure S4. PROSER1 regulates DNA demethylation at CpG islands and differentially methylated regions.
(A) Genome-wide differential methylation analysis (methylation change ≥ 25%; FDR ≤ 0.05) of differentially methylated cytosines, CpG islands (CGIs), CGI shores, promoters (Promoter), differentially methylated regions of 100 bp (DMR100), 500 bp (DMR500), and 1,000 bp (DMR1000) size. (B) Genomic distribution of all hypomethylated and hypermethylated differentially methylated cytosines identified in PROSER1 KO compared with WT HEK293 cells. TES, transcription end site. (C) Heat maps centered on hypermethylated CGIs in PROSER1 KO compared with WT HEK293 cells. Occupancy of PROSER1, UTX, OGT, TET1, TET2, H3K4me1, H3K4me2, H3K4me3, and H3K27ac and whole genome bisulfite sequencing results are shown. (D) Heatmaps centered on hypermethylated DMR1000 regions in PROSER1 KO compared with WT HEK293 cells overlapping with at least one PROSER1, UTX, or H3K4me1 peak in WT cells. Occupancy of PROSER1, UTX, OGT, TET1, TET2, H3K4me1, H3K4me2, H3K4me3, and H3K27ac and whole genome bisulfite sequencing results are shown.
- Figure S5. PROSER1 regulates gene expression associated with DNA demethylation at CpG islands and differentially methylated regions and controls 5hmC levels genome-wide.
(A, B, C) Gene set enrichment analysis analysis showing transcriptional down-regulation of genes associated with (A) hypermethylated CGIs, or (B) DMR500, and (C) DMR1000 regions overlapping with at least one PROSER1, UTX, or H3K4me1 peak in PROSER1 KO compared with WT HEK293 cells. (D) Venn diagram depicting the overlap of identified 5hmC peaks from hMeDIP-seq data in WT and PROSER1 KO HEK293 cells.
- Figure 6. Loss of PROSER1 function causes DNA hypermethylation encroachment at CpG islands.
(A) Representative genome browser tracks depicting the ChIP-seq profiles of PROSER1, UTX, H3K4me1, TET1, TET2, and whole genome bisulfite sequencing results at CGIs with DNA hypermethylation encroachment in PROSER1 KO versus WT cells. Examples for 5′ encroachment (MGA), bidirectional encroachment (IGFBP7), and 3′ encroachment (HOXC6) in WT and PROSER1 KO cells are displayed. (B) Heat map showing the whole genome bisulfite sequencing results of 483 identified CGIs with DNA hypermethylation encroachment in PROSER1 KO compared with WT cells. Clustering was performed based on the pattern of DNA hypermethylation encroachment: 5′ end encroachment (orange), bidirectional encroachment (gray), and 3′ encroachment (green). The color scale shows the DNA methylation range difference between PROSER1 KO and WT cells. Missing values (Not-a-Number; NaN) are depicted in white. (C) Average per-bin DNA methylation level of the CGIs with DNA hypermethylation encroachment described in Fig 6B categorized into groups of 5′ end encroachment (orange), bidirectional encroachment (gray), and 3′ end encroachment (green).
- Figure S6. PROSER1 controls DNA demethylation at CpG islands and transcription of associated target genes.
(A) Genome browser tracks showing the RNA-seq profiles of the MGA, IGFBP7, and HOXC6 genes in WT and PROSER1 KO HEK293 cells. Two replicates are displayed for each genotype. (B) Heat maps showing hMeDIP-seq results of 483 identified CGIs with DNA hypermethylation encroachment in PROSER1 KO compared with WT HEK293 cells. Heat maps were grouped based on 5′ end encroachment (left), bidirectional encroachment (middle), and 3′ end encroachment (right). The color scale depicts the 5hmC methylation range.
- Figure 7. Model of PROSER1 function at enhancers and CpG islands and broader implication for PROSER1 as a general mediator of OGT activity in chromatin signaling.
(A) Model of PROSER1 function at enhancers and CpG islands. PROSER1 mediates OGT interaction with and O-GlcNAcylation of TET2 to control TET2 stabilization at enhancers and CGIs that are controlled by the activity of the MLL3/4 complexes. TET2 is involved in the recruitment of the MLL3/4 complexes via their complex-specific subunit UTX. In the absence of PROSER1 TET2 stabilization at enhancers and CGIs is impaired resulting in reduced recruitment of the MLL3/4 complexes. TF, transcription factor. (B) Western blot of IP with an O-GlcNAc–specific antibody from WT and PROSER1 KO HEK293 cells confirming strongly reduced O-GlcNAcylation of TET2 in PROSER1 KO compared with WT cells. Nuclear extracts were used as input. Actin was used as a loading control for the inputs. (C) IP with an O-GlcNAc–specific antibody followed by mass spectrometry identifies TET1 and OGT as differentially O-GlcNAcylated in WT versus PROSER1 KO HEK293 cells. Many components of the H3K4 methyltransferase and demethylase family also display reduced O-GlcNAcylation in PROSER1 KO compared with WT cells. Of note, QSER1, a protein just recently described as a factor that protects DNA methylation valleys from de novo methylation also showed reduced O-GlcNAcylation in PROSER1 KO versus WT cells.
- Figure S7. Exploring alternative mechanisms of PROSER1-mediated changes in DNA methylation.
(A) Western blot for members of the DNA methyltransferase family from nuclear extracts of WT and PROSER1 KO HEK293 cells. Actin was used as a loading control. (B) DNA methylation rank plots showing regions associated with TET1 down-regulation, TET2 down-regulation, TET1 up-regulation, TET2 up-regulation, UTX/H3K4me1 down-regulation, and PROSER1-occupied regions in PROSER1 KO versus WT HEK293 cells (all FC ≥ 2). Regions with CpGs covered by at least 10 whole genome bisulfite sequencing reads (please see “whole genome bisulfite sequencing data processing” in the “Materials and Methods” section for further details) were ranked from hyper- to hypomethylated on the x-axis.
- Figure S8. Simplified example of CGI separation into five segments.
Segments with no CpGs were assigned a “Not-a-Number” (NaN) value. Please see “Identification of CGI encroachment events” in the “Materials and Methods” section for further details.
Supplementary Materials
Table S1. CpG islands with bidirectional or monodirectional DNA hypermethylation encroachment in PROSER1 KO versus WT cells. [LSA-2021-01228_TableS1.xlsx]
Table S2 Primers for molecular cloning.
In this Issue
Subjects
Related Articles
Cited By...
- No citing articles found.