hnRNP K coordinates transcriptional silencing by SETDB1 in embryonic stem cells

Abstract

Retrotransposition of endogenous retroviruses (ERVs) poses a substantial threat to genome stability. Transcriptional silencing of a subset of these parasitic elements in early mouse embryonic and germ cell development is dependent upon the lysine methyltransferase SETDB1, which deposits H3K9 trimethylation (H3K9me3) and the co-repressor KAP1, which binds SETDB1 when SUMOylated. Here we identified the transcription co-factor hnRNP K as a novel binding partner of the SETDB1/KAP1 complex in mouse embryonic stem cells (mESCs) and show that hnRNP K is required for ERV silencing. RNAi-mediated knockdown of hnRNP K led to depletion of H3K9me3 at ERVs, concomitant with de-repression of proviral reporter constructs and specific ERV subfamilies, as well as a cohort of germline-specific genes directly targeted by SETDB1. While hnRNP K recruitment to ERVs is dependent upon KAP1, SETDB1 binding at these elements requires hnRNP K. Furthermore, an intact SUMO conjugation pathway is necessary for SETDB1 recruitment to proviral chromatin and depletion of hnRNP K resulted in reduced SUMOylation at ERVs. Taken together, these findings reveal a novel regulatory hierarchy governing SETDB1 recruitment and in turn, transcriptional silencing in mESCs.

Figure 1. hnRNP K is associated with SETDB1 and KAP1 in mESCs.

(A) IP scheme to identify SUMO-dependent binding partners of SETDB1 and silver stained gel showing protein content of the indicated fractions. The nuclear extract input (NE), negative control IP (IgG), SETDB1 IP (α-SETDB1), and the flow-through (FT), 0.25 M KCl and 0.5 M KCl fractions from the anionic column are shown. (B) Western blot of SETDB1 and SENP1 in the indicated fractions and IP. (C) Silver stained gel of FLAG-SETDB1 IP and western blot of FLAG-SETDB1, KAP1 and hnRNP K in immunopurified FLAG-SETDB1 complexes isolated from tamoxifen-induced Setdb1 KO mESCs stably expressing 3XFLAG-Setdb1 (KO+FLAG-Setdb1) or negative control uninduced Setdb1 conditional KO cells (cKO). Complexes were immunoprecipitated with FLAG antibodies and specifically eluted with 3XFLAG peptide. Asterisk marks a non-specific band. Cropped band at bottom of hnRNP K blot is IgG heavy chain (~55 kDa). (D) Co-IP assay of endogenous KAP1 and hnRNP K with SETDB1 from mESC nuclear extracts in the absence or presence of 10 mM SENP inhibitor (NEM). Note that hnRNP K is detected in mESCs as two bands at ~65 kDa and ~60 kDa, the larger of which represents full-length hnRNP K and the smaller is a splicing isoform hnRNP J, also produced from the Hnrnpk gene [77]. The hnRNP K isoform is associated with SETDB1. ‘NE’ represents ~10% of nuclear extract input and ‘IgG’ is the negative control IP. (E) Co-IP assay of endogenous SETDB1 with KAP1 or hnRNP K from mESC nuclear extracts in the presence or absence of NEM, as in (D).

Figure 2. hnRNP K directly interacts with unmodified KAP1 in a region containing the RBCC domain.

(A) GST pulldown assays using purified GST, GST-SUMO2 or GST-hnRNP K as baits with recombinant FLAG-SETDB1 or Ubc9 as prey proteins. ‘IN’ represents ~20% of pulldown prey protein input. (B) Western blot analysis of in vitro SUMOylation reactions performed on GST-tagged RanGAP1 C-terminal fragment (residues 419–587), full-length wt GST-KAP1 or GST-p53. + or—indicates presence or absence of SUMO1 in the reaction. (C) GST pulldown assays using SUMOylated or unmodified GST-tagged baits from (B) with purified recombinant FLAG-tagged SETDB1 as prey protein. ‘IN’ represents 10% of input SETDB1 prey protein. (D) GST pulldown assay as in (C) except using purified recombinant 6X-His-tagged hnRNP K as prey protein. (E) Schematic of wt KAP1 domain structure and KAP1 mutants used in GST pulldown assays. Unmodified GST-tagged wt KAP1, deletion mutants KAP1^PxVxL and KAP1^PB or GST alone were used as baits to pull down purified recombinant 6X-His-tagged hnRNP K. ‘IN’ represents 15% of the input hnRNP K protein.

Figure 3. siRNA knockdown of hnRNP K results in de-repression of proviral reporters and ERVs.

(A) Schematics of the MSCV (PBS^Gln)-GFP retroviral vector in the Setdb1^lox/-mESC line and the IAP LTR-PBS GFP reporter in the HA36 wt mESC line. (B) qRT-PCR analysis of Setdb1 and Hnrnpk mRNA in Setdb1^lox/- 33#6 MSCV-GFP cells transfected with control (siCtrl), Setdb1 or Hnrnpk siRNAs at 24 h post-transfection. Data are mean expression level relative to Gapdh mRNA and normalized to siCtrl. Data are means (+ s.d.) of three technical replicates. (C) Flow cytometry analysis on untransfected (-) cells and cells transfected with indicated siRNAs at 72 h post-transfection. Data represent mean percentage of SSEA1⁺, GFP⁺ (double-positive) cells from three biological replicates of 10,000 propidium iodide-negative (PI^-) cells per sample. Error bars are s.d. (D) qRT-PCR analysis of Setdb1, Hnrnpk and Mcaf1 mRNA in J1 wt or Dnmt3a^-/-; Dmnt3b^-/-; Dnmt1^-/-(Dnmt TKO) cells transfected with indicated siRNAs at 24 h post-transfection. Data are mean expression level (+ s.d.) normalized to expression levels in control siRNA transfected J1 wt mESCs, relative to level of Gapdh mRNA. (E) qRT-PCR analysis as in (D) of except for intact class I and II ERVs in J1 wt or Dnmt TKO mESCs transfected with the indicated siRNAs at 96 h post-transfection. *p < 0.01, **p < 0.001, Student’s two-tailed T-test, relative to siCtrl.

Figure 4. hnRNP K plays a role in repression of SETDB1 target genes.

(A) RNA-seq 2D scatterplots of gene expression (reads per kilobase per million mapped reads; log(RPKM) from two biological replicates of TT2 cells transfected with control or Hnrnpk siRNAs at 72 h post-transfection. Genes upregulated ≥2-fold (‘UP’) in hnRNP K KD cells are labelled in red, whereas genes downregulated ≥50% (‘DN’) are shown in blue. Data points correspond to n = 22,138 ENSEMBL annotated genes. (B) qRT-PCR analysis of selected lineage-specific genes identified as upregulated from the RNA-seq of hnRNP K KD cells. Fkbp6, Dazl, Mael, Taf7l are direct SETDB1/H3K9me3 target genes. Data are relative mean expression level (+ s.d.) for three technical replicates. (C) UCSC genome browser screenshot including tracks for SETDB1 ChIP-seq in wt [28] and H3K9me3 ChIP-seq in wt and Setdb1 KO mESCs and RNA-seq for wt and Setdb1 KO mESCs [17] along with total coverage tracks for hnRNP K KD replicates RNA-seq at the Dazl gene. Numbers on the right indicate y-axis scale. (D) Venn diagram showing overlap of genes upregulated in hnRNP K KD cells (264) and genes upregulated in SETDB1 KO cells that are either bound by SETDB1 and/or marked by SETDB1-dependent H3K9me3 (134) according to [17]. p = 8.7×10^-30, Fisher’s exact test (n = 22,138 ENSEMBL-annotated genes). (E) Native ChIP for H3K9me3 at the core promoters/TSS of the indicated genes on TT2 cells transfected with control, Setdb1 or Hnrnpk siRNAs at 72 h post-transfection. Data are mean enrichment relative to input from three technical replicates, error bars are s.d. *p < 0.05, **p < 0.005, Student’s two-tailed T-test, relative to siCtrl.

Figure 5. hnRNP K is bound at ERVs and is required for H3K9me3 deposition at proviral chromatin.

(A) Schematic of intact ERV structure and 5’LTR-internal sequence amplified (primers shown as arrows). (B) Crosslinked ChIP of hnRNP K in TT2 cells transfected with control or Hnrnpk siRNAs at 24 h post-transfection. Egr1 core promoter and TSS (-50 to +50) was amplified as a negative control locus. Data shown are mean enrichment levels relative to input material from three technical replicates, error bars show s.d. (C) Native ChIP for H3K9me3 at the indicated ERV 5’LTR-internal regions in TT2 wt mESCs transfected with control, Setdb1, or Hnrnpk siRNAs at 72 h post-transfection with the same primers as in (A). The Myc core promoter and TSS (-50 to +50) was amplified as a negative control. (D) Schematic of the MSCV (PBS^Gln)-GFP vector with black bars showing the positions of ChIP amplicons for the MSCV 5’LTR-PBS region and Gfp internal sequence. (E) Native ChIP for H3K9me3 as in (C) except on siRNA KD of Hnrnpkin unsorted (GFP⁺ and GFP^-) MSCV-GFP cells at 24 h post-transfection. (F) Native ChIP from the same KD cells as in (E) except for H4K20me3. *p < 0.001, **p < 0.0001, Student’s two-tailed T-test, relative to siCtrl.

Figure 6. hnRNP K is required for SETDB1 but not KAP1 recruitment to ERVs.

(A) Western blot analysis of SETDB1, KAP1 and hnRNP K in TT2 cells transfected with control, Setdb1 or Hnrnpk siRNAs at 24 h post-transfection. GAPDH was detected as a loading control. (B) Crosslinked ChIP of SETDB1 in the same cells as in (B) except at 72 h post-transfection. All qPCR data are mean enrichment relative to input of three technical replicates, error bars are s.d. The Ifna5 core promoter and TSS (-50 to +50) was amplified as a negative control locus. (C) Western blot analysis of KAP1 and hnRNP K in TT2 cells transfected with control, Kap1 or Hnrnpk siRNAs at 24 h post-transfection. (D) Crosslinked ChIP of KAP1 in the same cells as in (C) except at 72 h post-transfection. (E) Crosslinked ChIP as in (B) and (D) except of hnRNP K in TT2 cells transfected with control or Kap1 siRNAs at 72 h post-transfection. *p < 0.001, **p < 0.0001, Student’s two-tailed T-test relative to siCtrl.

Figure 7. SUMOylation on proviral chromatin is required for SETDB1 recruitment and is compromised upon hnRNP K knockdown.

(A) Illustration of the SUMO conjugation pathway including the activities of SUMO E1 activating heterodimer enzyme Aos1/Uba2 and SUMO E2 conjugating enzyme Ubc9. Anacardic acid was used to inhibit E1 activity while siRNAs were used to deplete Ubc9 transcripts to disrupt SUMO conjugation in the MSCV-GFP cell line. (B) Western blot analysis of KAP1, GAPDH, pan Histone H3 acetylation or H3K9me3 in MSCV-GFP cells incubated in varying concentrations of anacardic acid for 18 h prior to harvest. A mono-SUMOylated KAP1 band is detected at ~130 kDa (arrow). (C) Flow cytometry for GFP⁺ cells in the MSCV-GFP line alone (-), vehicle (DMSO) treated, or treated with 40 or 100 μM anacardic acid for 18 h. Data are mean of three biological replicates for 10,000 PI- cells per sample, error bars are s.d. (D) Flow cytometry data as in (B) except on cells untransfected (-) or transfected with control or Ubc9 siRNAs at 48 h post-transfection. Inset, qRT-PCR analysis of Ubc9 expression in cells transfected with control or Ubc9 siRNAs at 24 h post-transfection. Data are mean relative expression levels determined from three technical replicates, normalized to the level of β-actin (Actb) transcripts, error bars are s.d. (E) Crosslinked ChIP of SETDB1 on unsorted (GFP⁺ and GFP^-) MSCV-GFP cells transfected with control or Ubc9 siRNAs at 48 h post-transfection. Data are mean enrichment as a percent of input chromatin from three technical replicates, error bars are s.d. (F) Crosslinked ChIP as in (E) except for SUMO1 on TT2 cells transfected with control or Hnrnpk siRNAs at 24 h post-transfection. *p < 0.01, **p < 0.001, Student’s two-tailed T-test relative to vehicle or siCtrl.

Figure 8. Model for SETDB1/KAP1-mediated proviral silencing pathway.

(A) In wt mESCs, KRAB-ZFPs recruit KAP1 subunits and unmodified KAP1 recruits hnRNP K. SUMOylation of KAP1 and/or other proteins on chromatin is promoted by hnRNP K, resulting in recruitment of SETDB1/MCAF1 and deposition of H3K9me3. (B) In hnRNP K-deficient cells, the SUMOylation of KAP1 on chromatin is compromised, leading to loss of SETDB1 recruitment, loss of H3K9me3 and induction of proviral expression. (C) In MCAF1-deficient cells, SETDB1 recruitment is maintained but H3K9me3 is no longer deposited efficiently, leading to proviral de-repression.