A somatic role for the histone methyltransferase Setdb1 in endogenous retrovirus silencing

Abstract

Subsets of endogenous retroviruses (ERVs) are derepressed in mouse embryonic stem cells (mESCs) deficient for Setdb1, which catalyzes histone H3 lysine 9 trimethylation (H3K9me3). Most of those ERVs, including IAPs, remain silent if Setdb1 is deleted in differentiated embryonic cells; however they are derepressed when deficient for Dnmt1, suggesting that Setdb1 is dispensable for ERV silencing in somatic cells. However, H3K9me3 enrichment on ERVs is maintained in differentiated cells and is mostly diminished in mouse embryonic fibroblasts (MEFs) lacking Setdb1. Here we find that distinctive sets of ERVs are reactivated in different types of Setdb1-deficient somatic cells, including the VL30-class of ERVs in MEFs, whose derepression is dependent on cell-type-specific transcription factors (TFs). These data suggest a more general role for Setdb1 in ERV silencing, which provides an additional layer of epigenetic silencing through the H3K9me3 modification.

Fig. 1

Different ERV families are derepressed by Setdb1 KO in different cell types. a Cell-type-specific ERV derepression in Setdb1 cKO cells. Expression of ERV families in Setdb1 cKO ESC (day 6 after treatment with 4OHT (KO) or no treatment (WT)), iMEF (day 5 after treatment with 4OHT (KO) or no treatment (WT)), and E14.5 forebrain cells from WT and Emx2-Cre:Setdb1 fl/fl mice (KO). For GMP and LSK cells, bone marrow cells from Rosa-CreERT:Setdb1 cKO mice were transplanted into irradiated recipient mice, GMP and LSK cells were then isolated after injection of 4OHT for 2 weeks (KO) or control injection (WT). Only ERVs derepressed (≧2 fold) in at least one of the analyzed cell types with Setdb1 KO are listed. Heatmap indicates the relative expression level of representative ERV families (the RPKM value). The ERVs derepressed (≧1.5 fold) in Setdb1 KO iMEFs are highlighted in red. b H3K9me3 intensity profiles on different ERV families in different cell types. NGS plots show the fold enrichment of H3K9me3 from −5 kb to 10 kb around genomic ERV elements in ESC, forebrain, iMEF, and GMP. We selected ERVs containing -int element with flanked LTRs (See Methods). Position 1 is 5′ start site of the -int element. Positions of LTRs and int for each ERV element are indicated below the plots. c Bar plots showing the loss of H3K9me3 in ERV families in Setdb1 KO (day 7 after 4OHT treatment: red bar) vs. WT (black bar) iMEF. The y axis indicates the fold enrichment of normalized ChIP read density relative to input. ERV names written in green are analyzed in Fig. 1b. d ChIP-qPCR of H3K9me3 in the Gapdh promoter region indicated ERVs and major satellite (Major S) loci of Setdb1 cKO ESCs and iMEFs. Values are means ± s.d. from independent experiments (n = 3). *0.005 < P < 0.05, **0.0005 < P < 0.005, Student’s t-test

Fig. 2

MMVL30 families are derepressed in Setdb1 KO iMEFs. a Schematic structures of VL30 families. Blue boxes indicate highly homologous sequences between MMVL30-int and RLTR6-int. High magnification of LTR regions of U3 classes. Color boxes indicate highly homologous sequence among LTRs. White arrowhead and arrow indicate unique insertion and mutation in class II, respectively. Black arrowhead indicates RARE sites. b Expression of VL30 U3 classes in Setdb1 KO iMEFs. Boxplots show that expression of VL30 U3 classes in Setdb1 KO iMEFs (day 5 after 4OHT treatment (KO) or no treatment (WT)). RNA-seq reads overlapping each VL30 loci are counted and normalized using the length of the locus and million mapped reads (RPKM). U3 I (71 loci), U3 II (189 loci), U3 III (66 loci), and U3 IV (26 loci). **0.0005 < P < 0.005, ***P < 0.0005, Student’s t-test. c NGS plots of H3K9me3 ChIP-seq data at genomic VL30 loci (RLTR6_Mm - MMVL30-int) of ESCs, E14.5 forebrain cells and iMEFs. MMVL30-int with RLTR6_Mm U3 I (36), with U3 III (40) and with U3 IV (10). The number inside of parenthesis indicates the number of individual each element analyzed. LTRs are aligned and MMVL30-int starts from position 1. H3K9me3 intensity profiles are shown with dark blue line (U3 I), light blue one (U3 III), and yellow one (U3 IV). Positions of LTRs and int are indicated below the plots. d NGS plots of H3K9me3 ChIP-seq data at genomic “RLTR6_Mm - MMVL30-int” and “RLTR6_Mm - RLTR6-int” loci in Setdb1 cKO iMEFs (day7 after 4OHT treatment (KO) and no treatment (WT)). H3K9me3 fold enrichments (log2) are shown with dark blue line (WT) and light blue one (KO)

Fig. 3

VL30 activation requires cell-type-specific TFs. a, b RT-qPCR of the VL30 U3 classes in RNA from Setdb1 cKO iMEFs (a) or ESCs (33#6,) (b), untreated or treated with 1 μM all-trans retinoic acid (atRA), with or without 4OHT (n = 3 biological replicates). For ESC data (n = 2 biological replicates), we normalized each expression to VL30 U3 class I LTR expression of control iMEF. Error bars represent s.d. *0.005 < P < 0.05, **0.0005 < P < 0.005, ***P < 0.0005, Student’s t-test. c Upper panel; the derepression induced by Setdb1 depletion in each VL30 U3 class I element (totally 71 elements) is shown. 22 loci on the right side have typical PBS-pro near the 3′ end of 5′ LTR. Selected loci for parallel alignment shown in d are denoted by H, M, and L symbols based on their derepression (H: high assigned to the locus where RPKM^KO (RPKM of Setdb1 KO) is >100, M: mild assigned to 10 ≤ RPKM^KO < 100, and L: low assigned to RPKM^KO < 10, respectively). Lower panel: H3K9me3 ChIP-seq read counts on 5′ LTR at each VL30 U3 class I locus in Setdb1 cKO iMEFs (no treatment (WT: blue bar) and day 7 after treatment of 4OHT (KO: light blue bar)) are shown. d Alignment of U3 sequences of VL30 U3 class I. The red box indicates the consensus sequence of Elk family binding sites. The blue and green boxes indicate the consensus sequences of AP-1 and Ets, respectively. High, mild, and low indicate in Fig. 3c upper panel. Entire alignment of VL30 class I shown in Fig. 3c is in Supplementary Data 3. We defined the copies for RPKM^KO < 10 as silenced. e Inhibition of the MAPK pathway precludes the transcriptional activation of VL30 in Setdb1 KO iMEFs. RT-qPCR analysis of VL30 U3 class I. 4OHT treatment was applied for the first 4 days and cells were harvested at day 6. The MEK inhibitor PD0325901 (PD; 1 µM) was added for the last 24 h (n = 2 biological replicates). Error bars represent s.d. *0.005 < P < 0.05, **0.0005 < P < 0.005, Student’s t-test

Fig. 4

Dnmt1 and Setdb1 are required for the silencing of distinct sets of ERVs. a Expression of ERV families by the loss of Dnmt1, Setdb1, and simultaneous depletion of Dnmt1 and Setdb1 in iMEFs and ESCs. To obtain mRNA-seq data of iMEFs, Setdb1 cKO iMEFs were transfected with siRNA against Dnmt1 or control siRNA in combination with or without Setdb1 depletion by the 4OHT treatment, then mRNA was isolated. RNA-seq data from Dnmt1 cKO or double cKO of Dnmt1 and Setdb1 ESCs (6 days after treatment of 4OHT or without treatment) were reanalyzed. Heatmap indicates the magnitude of the RPKM value. b Setdb1 cKO iMEFs were transfected with siRNA against Dnmt1 either alone or in combination with the loss of Setdb1. RT-qPCR of VL30 U3 class I RLTR6, IAPEz, and RLTR4 was performed. Values represent mean expression relative to VL30 U3 I control (n = 2 biological replicates). Error bars represent s.d. *0.005 < P < 0.05, **0.0005 < P < 0.005, ***P < 0.0005, Student’s t-test. c The derepression induced by Dnmt1 KD, Setdb1 KO, and simultaneous depletion of Dnmt1 and Setdb1 in iMEFs in each VL30 U3 class I element (totally 71 elements) is shown. The number of RNA-seq reads overlapping with a locus was divided by the length of the locus and normalized by million mapped reads (RPKM) as shown in Fig. 3c top

Fig. 5

Distinct requirement of Setdb1, Trim28, and Zfp809 for VL30 silencing. a Boxplots indicate silencing effects on VL30 U3 I by Setdb1 and Zfp809. 71 VL30 U3 I (MMVL30-int with RLTR6_Mm U3 I) are divided into two groups according to with (22 loci) or without (49 loci) PBS-pro sequences. RNA-seq data of Setdb1 cKO iMEFs (no treatment (WT) or 4OHT 5d (KO)), Zfp809 KO MEFs derived from mutant embryos (KO) and wild type (WT) are shown. *0.005 < P < 0.05, **0.0005 < P < 0.005, Student’s t-test b Expression of ERV families in Trim28 KO MEFs. RNA-seq data for Trim28 WT and KO MEFs was reanalyzed. Heatmap indicates the magnitude of the RPKM value. For comparison, Setdb1 cKO iMEF data (Fig. 1a) was shown above. c Trim28 KO only marginally lead to VL30 derepression. iMEFs were transfected with a CRISPR-gRNA vector against Setdb1 or Trim28. Two days later, transfected cells (tRFP-positive) were sorted using FACS Aria. Four days later, cells were harvested, and RT-qPCR of VL30 U3 class I was performed. (Trim28; n = 3 biological replicates Setdb1; n = 3 technical replicates) Error bars represent s.d. **0.0005 < P < 0.005, ***P < 0.0005, Student’s t-test. Efficiencies of CRISPR KO were examined by immunoblotting with anti-Setdb1 or anti-Trim28 antibodies. d Reintroduced Setdb1 represses VL30 in Setdb1 KO iMEFs. Setdb1 was stably expressed in Setdb1 long-term-cultured KO iMEFs. RT-qPCR of VL30 U3 class I was performed. (n = 2 biological replicates). Error bars represent s.d. **0.0005 < P < 0.005, ***P < 0.0005, Student’s t-test. Expression level of Setdb1 was examined by immunoblotting

Fig. 6

Model of Setdb1 function in ERV silencing. a Setdb1 deposits H3K9me3 at ERV loci generally in differentiated cells, not only in ESCs. b In Setdb1 KO cells H3K9me3 is reduced at ERV loci. However, TFs are required for each ERVs to get depressed. Without accelerator (TFs) ERVs are not expressed even without brake (H3K9me3). c For example, in Setdb1 KO iMEFs derepression of VL30 class I require Elk, Ets, and AP-1 TFs. VL30 class II is expressed only in the presence of RA in Setdb1 KO iMEFs. This ERV derepression might cause IFN pathway activation