Correlated expression dynamics of XACT and T113.3

To understand whether the transcription of XACT and T113.3 is controlled by a shared regulatory network, we first explored the patterns and dynamics of their expression in hESCs and upon their differentiation. We observed, using RNA-fluorescence in situ hybridization (FISH), that XACT and T113.3 are co-expressed from the active X chromosome (Xa) in both male (H1) and female (H9) hESCs before differentiation, with a minority of cells expressing either transcript individually (Fig. 2a and Supplementary Fig. ^2A). Their expression dynamics is also strongly correlated across a 10-day differentiation time course of male and female hESCs (Fig. 2b and Supplementary Fig. ^2B), with both genes being turned off from day 3 onwards.

Open in a separate window

Fig. 2

XACT and T113 are co-regulated in pluripotent contexts.

a Analysis of XACT (green) and T113.3 (magenta) expression by RNA-FISH in male H1 hESCs. Percentage of cells co-expressing XACT and T113.3 from the same locus is indicated. Quantification of the different patterns of expression is represented on the right (n=145). The white scale bar represents 5µm. b RT-qPCR analysis of XACT and T113.3 expression during a 10-day undirected differentiation of H1 hESCs. The bar charts correspond to the average of three independent differentiation experiments. Error bars indicate the standard deviation (SD). The correlation coefficient for the expression dynamics of the two genes is shown (r=0.94, Pearson correlation). c Plot of T113.3 vs. XACT expression levels (log2 reads per kilobase per million mapped reads [RPKM]) in early (E3, E4, and early E5) and late stage (E5, E6, and E7) female (upper panels) and male (lower panels) pre-implantation embryos, with corresponding Spearman’s correlation score and p-value. Dataset from ref. ³³. d Hi-C heatmap of the LHFPL1/LRCH2 locus in H1 hESCs and in HFF ⁴³ visualized with Juicebox⁷⁰. Red and green arrows indicate the orientation of CTCF binding sites. Blue arrows and dotted lines show the boundaries of the domains in H1 hESCs.

We further explored the transcriptional dynamics of XACT and T113.3 in human pre-implantation embryos using public single-cell RNA sequencing (RNA-seq) datasets^33,40–42. Expression from both loci can be detected concomitantly between the four- and the eight-cell stages, likely following the major wave of zygotic gene activation (Supplementary Fig. ^2C). Furthermore, the expression detected from the XACT and T113.3 loci is correlated at early (E3 to early E5) and late (E5 to E7) stages of development in both female and male embryos (Fig. 2c). A weaker transcriptional correlation between XIST and T113.3 is observed in early stages of pre-implantation development, but is mostly lost at later stages (Supplementary Fig. ^2D). These observations confirm that the XACT and T113.3 loci display similar transcriptional dynamics in different pluripotent contexts, in hESCs and in early embryos, suggesting that they may be co-regulated.

To corroborate this hypothesis, we analyzed Hi-C datasets⁴³ from H1 hESCs and from human foreskin fibroblasts (HFF), in which XACT and T113.3 are expressed and silenced, respectively. In hESCs, the 5′-region of XACT is embedded within a local compartment that includes T113.3 and expands into the upstream LRCH2 gene, whereas in fibroblasts this local compartmentalization is absent (Fig. 2d). This indicates that the regulatory sequences of XACT and T113.3 belong to the same TAD in cells where both genes are expressed.

Blocking T113.3 transcription impairs XACT expression

To test whether the expression of XACT or of T113.3 are interdependent, we performed CRISPR interference (CRISPRi), using a catalytically dead Cas9 coupled with a repressive KRAB domain (dCas9-KRAB), to induce local heterochromatinization of the TSS regions of XACT and T113.3⁴⁴. Given the similarity of XACT/T113.3 expression dynamics in female and male pluripotent contexts, we performed all our functional assays in male H1 hESCs. We first generated H1 hESCs stably expressing the CRISPRi machinery together with gRNAs targeting the promoter of either XACT or T113.3 (Fig. 3a, top scheme and Supplementary Fig. ^3A). Reverse transcription–quantitative PCR (RT-qPCR) analysis showed the efficiency of the CRISPRi system, with the levels of XACT and T113.3 RNAs being strongly reduced when their respective promoter is targeted (Fig. 3a). Although interfering with XACT expression does not affect the expression levels of T113.3, targeting the T113.3 promoter induces a significant decrease in XACT lncRNA levels (Fig. 3a). As targeted cells maintain a normal morphology and stable expression of pluripotency markers (Supplementary Fig. ^3B), this effect is unlikely due to an alteration of the pluripotent state that could have resulted from the transcriptional interference of T113.3. Moreover, accumulation of the H3K9me3 mark induced by T113.3 CRISPRi spreads only locally, up to 6kb away from the targeted site, but does not reach the XACT promoter (Fig. 3b), whereas the distribution of H3K4me3 and H3K27ac changes only slightly across the loci (Supplementary Fig. ^3C). Altogether, this suggests that the T113.3 locus participates in the regulation of XACT expression, either through the T113.3 transcript, the act of transcription, or the chromatin landscape around the LTR7 forming the T113.3 promoter.

Open in a separate window

Fig. 3

Transcription of T113.3 regulates the expression of XACT in hESCs.

a The top scheme depicts the CRISPRi strategy used to target the XACT and T113.3 promoters, with the positions of the two sgRNAs (colored arrows) and primer pairs used to measure the transcript levels of each gene (black arrows). Quantification of the global RNA levels of XACT and T113.3 by RT-qPCR upon XACT CRISPRi (left bar charts) and T113.3 CRISPRi (right bar charts). Error bars represent the SD of at least four different passages of hESCs infected with guides targeting either XACT or T113.3. b ChIP-qPCR analysis of H3K9me3 enrichment across the XACT/T113.3 locus upon XACT CRISPRi (green squares) or T113.3 CRISPRi (red circles) in hESCs. The positions analyzed are indicated in the scheme with arrows. ChIP-qPCR data for untransfected cells (gray inverted triangles) and cells transfected with a vector without a sgRNA (black triangles) is also shown. Right graph represents the relative enrichment of H3K9me3 in control positions (SOX2 TSS and hXIC19). Error bars represent the SD of three passages of each of two independent hESC lines infected with different guides targeting either XACT or T113.3 (n=6). *P-values<0.05, **P-values<0.01, ***P-values<0.001, and ****P-values<0.0001. Statistical significances were determined using a one-way ANOVA with sg_empty.

T113.3 is dispensable for XACT expression

The LTR7/HERVH class of retrotransposons has been linked to the transcriptional control of human pluripotency^15,17. We thus hypothesized that the LTR7/HERVH defining the T113.3 gene could link the expression of the XACT/T113.3 locus to the pluripotency network. To test this hypothesis, we developed parallel strategies to independently assess the involvement of the T113.3 RNA or of the T113.3 locus in the regulation of XACT.

First, we knocked down (KD) the T113.3 RNA using LNA-gapmers (locked nucleic acid, antisense oligonucleotides) targeting different intronic or exonic regions (Fig. 4a, upper panel). As a control, we designed LNA-gapmers for different regions of the XACT transcript. H1 hESCs transfected with XACT- or T113.3-specific LNA-gapmers show a significant reduction of their respective target (Fig. 4a), without any morphological alteration or changes in the expression of pluripotency or differentiation markers (Supplementary Fig. ^4A). However, interfering with the levels of either T113.3 or XACT RNAs does not lead to any changes on the RNA levels of the other transcript (Fig. 4a), indicating that the T113.3 transcript is dispensable for the expression of XACT in hESCs (and vice versa).

Open in a separate window

Fig. 4

Neither T113.3 gene nor T113.3 RNA regulate XACT expression.

a Scheme of XACT and T113.3 loci showing the position of the LNA GapmeRs used for XACT (green arrows) and T113.3 (red arrows) KD, and the position of the primers used for the RT-qPCR (black arrows). RT-qPCR analysis of T113.3 and XACT expression in cells transfected with LNA GapmeRs targeting T113.3 (top panels) or XACT (bottom panels) (minimum of six replicates per condition). b Scheme of the XACT and T113.3 locus showing the guide positions (red arrows) used for the deletion of the T113.3 gene and the position of the primers used for RT-qPCR (black arrows). RT-qPCR analysis of T113.3 and XACT expression in H1 WT, T113.3 KO, and INV hESC clones. n=3. c RT-qPCR analysis of XACT expression during a 10-day undirected differentiation of H1 T113.3 WT, KO, and INV hESC lines. n≥3. Error bars represent the SD. Statistical significances were determined using a one-way ANOVA (all tests compared to LNA_SCR). *P-values<0.05, **P-values<0.01, ***P-values<0.001, and ****P-values<0.0001.

We next tested the role of the T113.3 LTR7/HERVH in the regulation of XACT, by using CRISPR/Cas9 to delete a 25kb region encompassing the entire T113.3 gene (including the whole LTR7/HERVH element at its 5′-end) in H1 hESCs (Supplementary Fig. ^4bB). We screened individual colonies by PCR for the presence of a deletion or an inversion of the T113.3 gene and selected three independent clones of each genotype (wild type (WT), deleted, and inverted; Supplementary Fig. ^4B) for further analysis. All clones, independently of their genotype, showed comparable expression levels of X-linked (AMOT, HTR2C, ATRX), pluripotency (OCT4), and differentiation marker genes (NODAL) (Supplementary Fig. ^4C). As expected, T113.3 RNA could not be detected in clones carrying a deletion of the locus; in addition, the inversion of the T113.3 gene lead to the complete abrogation of T113.3 transcription (Fig. 4b). However, none of these mutations affected the expression levels of XACT in hESCs (Fig. 4b). In addition, the expression dynamics of XACT during undirected differentiation of the T113.3 mutant hESCs (Fig. 4c) and the global kinetics of the differentiation were unaffected (Supplementary Fig. ^4D). These results clearly show that the LTR7/HERVH defining the T113.3 gene is dispensable for the expression of XACT in hESCs and for its silencing dynamics during cellular differentiation.

Identification of a pluripotent-specific enhancer

The analysis of H3K27ac distribution across the locus, a mark of active enhancers, pointed to a potential regulatory region approximately 2kb upstream of the T113.3 TSS (Supplementary Fig. ^3C). Examination of ATAC-seq and chromatin immunoprecipitation sequencing (ChIP-seq) datasets from H1 hESCs⁴⁵ confirmed the existence of a broad acetylation domain, which is not found in fetal fibroblasts (Fig. 5a and Supplementary Fig. ^5A), suggesting that this region may serve as a pluripotent-specific enhancer. Further inspection of ChIP-seq profiles revealed that this domain encompasses a CTCF peak and binding sites for several transcription factors, including pluripotency factors OCT4, SOX2, and NANOG.

Open in a separate window

Fig. 5

A common enhancer wires the XACT/T113.3 locus to the pluripotency network.

a ATAC-seq (black) and ChIP-seq data for H3K4me3, H3K27ac (blue), CTCF (brown), OCT4, SOX2, and NANOG (orange) enrichment over the T113.3 TSS region in H1 hESCs. Publicly available data were obtained from the ENCODE project⁵⁷ and CISTROME DB⁴⁵. Called peaks for CTCF, OCT4, SOX2, and NANOG are shown as black rectangles. b Scheme of the T113.3 locus showing the guide positions (black arrows) used for targeting either the CTCF (top panel) or TFB (bottom panel) sites upstream of the T113.3 gene. Quantitative RT-PCR analysis of XACT (left) and T113.3 (right) expression in CTCF- (top panel) or TFB- (bottom panel) targeted hESCs. The bar charts correspond to the average of at least two independent clones. c KD efficiency of SOX2, OCT4, and NANOG using an siRNA approach was assessed by western blotting. H1 hESCs were used as a non-transfected control cell line (no_si). siSCR was used as a non-targeting siRNA. Tubulin and H3.3 were used as loading control. d Quantification of XACT (top bar chart) and T113.3 (bottom bar chart) steady-state RNA levels by RT-qPCR in cells transfected with siRNAs targeting SOX2, OCT4, NANOG, or a combination of siRNAs targeting both OCT4 and SOX2. Error bars indicate the SD. Statistical significances were determined using a one-way ANOVA (all conditions compared with WT). *P-values<0.05, **P-values<0.01, ***P-values<0.001, and ****P-values<0.0001.

To test the enhancer activity of this domain and identify which element(s) are involved in the regulation of XACT and T113.3, we used CRISPR/Cas9 to delete either the CTCF or the transcription factor-binding (TFB) sites (Supplementary Fig. ^5B). We selected at least two independent clones of non-mutated, deleted, or inverted genotypes (Supplementary Fig. ^5B). All these clones retain normal expression levels of pluripotent markers (NANOG, OCT4), indicating that none of these mutations significantly impact pluripotency (Supplementary Fig. ^5C). Deletion or inversion of the CTCF site did not affect steady-state levels of T113.3 or XACT transcripts compared with WT H1 ESCs (Fig. 5b). In contrast, deletion of the TFB site completely abolished T113.3 transcription and drastically reduced the levels of XACT RNA (Fig. 5b). Interestingly, inversion of the TFB site did not perturb the expression of neither XACT nor T113.3, further supporting the role of this genomic element as an enhancer, acting independently of its orientation. Thus, we conclude that a pluripotent-specific proximal enhancer, upstream of the T113.3 gene, is essential for the regulation of the XACT/T113.3 loci in hESCs.

To confirm the role of the pluripotency transcription factors OCT4, SOX2, and NANOG in the activity of this enhancer, we used small interfering RNAs (siRNAs) to KD their expression, either individually or in combination. Western blotting analysis revealed that the simultaneous targeting of SOX2 and OCT4 resulted in the depletion of all three master regulators of the pluripotency network (Fig. 5c), consistent with the fact that OCT4 positively regulates NANOG⁴⁶. In this condition specifically, we observed a significant reduction of T113.3 and XACT transcript levels (Fig. 5d and Supplementary Fig. ^5D), further supporting the dependence of the locus towards master regulators of the pluripotency circuitry. We excluded an indirect effect linked to KD-induced differentiation, as upregulation of differentiation markers is observed in all conditions where OCT4 is targeted, whereas the effect on the XACT/T113.3 loci is seen only when the three factors are simultaneously depleted (Supplementary Fig. ^5D). Although we cannot exclude that these transcription factors play a more complex role regulating XACT and T113.3, namely by directly acting at their TSS, the combined results of the siRNA KD and LTR48B deletion suggests that the enhancer upstream of T113.3 connects the transcriptional regulation of the locus to the pluripotency network in hESCs.

Knockdown using siRNA and LNA GapmeRs

All transfections were performed using Lipofectamine RNAiMax (ThermoFisher Scientific), according to the manufacturer’s recommendations. LNA GapmeRs used in this study were designed using the Exiqon online tool, with sequences spanning different regions of the XACT and T113.3 transcripts. For XACT and T113.3 KDs, cells were reverse transfected with LNA GapmeRs (Exiqon, QIAGEN) at a final concentration of 50 or 100nM (with similar results). A scrambled LNA GapmeR and a LNA GapmeR targeting MALAT1, labeled with fluorescein amidite, were used as negative and transfection controls, respectively. Cells were collected at 24 or 48h post transfection (with a stable efficiency of KD for both time points).

For KD of pluripotency factors, cells were transfected with 50nM of siRNAs specific for POU5F1 (Ambion s10872), NANOG (Amibon s36650), or SOX2 (Dharmacon ON-TARGETplus SMARTpool L-011778–00–0005). siRNAs for MALAT1 (Ambion 4390843) and a scrambled siRNA (Ambion 4455877) were used as positive and negative controls, respectively. When co-transfecting multiple siRNAs, 25nM of each siRNA was used. After transfection of 1×10⁶ cells, these were plated in one well of of a six-well plate previously coated with Matrigel (Corning), with 10µm of Y-2763 (Merck Millipore). Transfected cells were collected 72h after transfection. Sequences of siRNAs are provided in Supplementary Table ¹

RNA-FISH on hESCs

Cells grown on 12mm Matrigel-coated coverslips (Corning) were fixed 24 to 48h later in a 4% paraformaldehyde/phosphate-buffered saline solution (Electron Microscopy Science) for 12min at room temperature. Cells were then permeabilized for 5min in ice-cold cytoskeletal buffer (NaCl 100mM, sucrose 300mM, MgCl₂ 3mM, PIPES 10mM) supplemented with 0.5% Triton, 1mM EGTA, and 2mM Vanadyl Ribonucleoside complex (VRC, New England Biolabs), washed three times with 70% ice-cold ethanol, and kept at −20°C.

Fluorescent probes were obtained by nick-translation, with either spectrum-green or spectrum-orange dUTPs (Abbott Molecular), as previously described³⁵. The following probes were used in this study: XACT (RP11–35D3, BACPAC), XIST (a 10kb fragment corresponding to XIST exon 1, gift from Dr C. Brown, University of British Columbia), and T113.3 (WI2–767I20, BACPAC) (Supplementary Table ²).

For hybridization, ~100ng of labeled probes (enough for three 12mm coverslips) were precipitated with 10µg of sheared salmon sperm DNA (ThermoFisher Scientific) and 5µg of human Cot-1 DNA (ThermoFisher Scientific), and then denatured in deionized formamide (Merck, Sigma-Aldrich) for 7min at 75°C. Denatured probes were then competed with the human Cot-1 DNA by incubating 30min at 37°C. Coverslips were dehydrated in sequential washes of 80%, 90%, and 100% ethanol just prior to overnight incubation with the probes at 37°C in 50% formamide/50% hybridization buffer (4× SSC (saline-sodium citrate), 20% dextran sulfate, 2mg/ml bovine serum albumin, and 2mM VRC). After three 50% formadehyde/2× SSC washes and three 2× SSC washes at 42°C for 5min, coverslips were mounted in Vectashield containing 4′,6-diamidino-2-phenylindole (DAPI; Vector Laboratories).

Microscopy and image analysis

All images were acquired with a fluorescence DMI-6000 inverted microscope with a motorized stage (Leica) using a HCX PL APO ×100 oil objective and a CCD Camera HQ2 (Roper Scientifics) using the Metamorph software (version 7.04, Roper Scientifics). Approximately 40 optical Z-sections were collected at 0.5µm steps across the nucleus for each wavelengths (DAPI [360nm, 470nm], fluorescein isothiocyanate [470nm, 525nm], and Cy3 [550nm, 570nm]). Stacks were processed using ImageJ. Throughout the manuscript, the three-dimensional FISH experiments are represented as a two-dimensional projection of the stacks (maximum intensity projection).

RNA extraction and RT-qPCR

Total RNAs were extracted from cells using TRIzol (ThermoFisher Scientific) and treated using the DNA free kit (ThermoFisher Scientific) for 1h at 37°C, to remove genomic DNA contamination. Five hundred nanograms of total RNAs were reverse-transcribed using SuperScript IV reverse transcriptase (ThermoFisher Scientific) and random primers (Promega). cDNA levels were measured using real-time qPCR with the Power SYBR Green PCR master mix (ThermoFisher Scientific) on a ViiA-7 real-time thermal cycler (Applied Biosystems). All samples were analyzed in technical duplicates. Normalization was performed using the reference gene GAPDH following the ΔCT or the ΔΔCT method. Sequences of RT-qPCR primers are provided in Supplementary Table ¹.

Characterization of the T113.3 transcript

For the reconstruction of the T113.3 transcript, bam and bigWig files (GSM758573) were downloaded from the ENCODE website⁵⁷. Transcripts were assembled using Scallop v.0.10.4³⁷ with default parameters. bigWig and corresponding Scallop reconstruction were visualized using the IGV genome browser. The T113.3 transcript was cloned from cDNA of H1 hESCs. PCR was performed using Platinum Taq DNA polymerase (ThermoFisher Scientific), using primers complementary to the first and last exons of T113.3 (primer sequences are provided in Supplementary Table ¹). To allow the amplification of a bigger number of potential transcripts, an elongation time of 2 min 30 sec per cycle, was used. The PCR products were resolved on a 2% agarose gel, purified using a NucleoSpin Gel and PCR clean-up kit (Macherey-Nagel), cloned using the TOPO TA Cloning Kit (ThermoFisher Scientific) and sequenced.

CRISPR/Cas9 deletions

Single guide RNA (sgRNA) sequences flanking each target region were obtained using the web-based tool CRISPOR (http://crispor.tefor.net/) and can be found in the ^{Supplementary Information}. For each targeted region, upstream and downstream sgRNAs were cloned under an U6 promoter into the pSpCas9(BB)−2A-GFP (Addgene #48138) and the pSpCas9(BB)-2A-mCherry (generated in house, by replacing the green fluorescent protein (GFP) with a mCherry reporter using the NEBuilding HiFi DNA Assembly Cloning Kit [New England Biolabs]). All sgRNAs sequences can be found in the Supplementary Table ¹ and vectors are listed in Supplementary Table ⁴.

Using the Amaxa 4D-NucleofectorTM system (Lonza), 1×10⁶ H1 hESCs were transfected with 2.5µg of both plasmids (to a total of 5µg). Cells were sorted by fluorescence-activated cell sorting (INFLUX 500-BD BioSciences) 48h after transfection. Double-positive cells (100, 200 or 400; GFP+/mCherry+) were plated into a matrigel or Laminin-coated 6cm petri dish in mTeSR supplemented with 1× CloneR (Stemcell Technologies). Individual colonies were manually picked into 96 wells ~10 days after transfection. Deletions and inversion events were screened by PCR (primer sequences can be found in Supplementary Table ¹).

CRISPRi, lentiviral production, cell infection

sgRNAs targeting either XACT or T113.3 promoters were designed using the web-based tool CRISPOR (http://crispor.tefor.net/) and cloned into the pLKO5.sgRNA.EFS.tGFP vector (Addgene #57823). sgRNA sequences can be found in Supplementary Table ¹. Lentiviral particles were produced by transient transfection of HEK293T cells (ATCC® CRL-3216™) with the packaging plasmids pMD2.G (Addgene #12259) and psPAX2 (Addgene #12260), together with a lentiviral dCas9/KRAB-mCherry construct using calcium phosphate. Lentiviral particles carrying sgRNA-expressing vectors were also obtained from HEK293T cells using the same method. The culture media was collected 48h after transfection of HEK293T and lentiviral particles were concentrated by ultracentrifugation. For each construct, the lentiviral titer was determined by infection of HEK293T cells and FACS analysis.

For CRISPRi, one million H1 hESCs were infected in suspension with lentiviral particles containing the dCas9-mCherry-KRAB (multiplicity of infection (MOI) of 10) and sorted by FACS (INFLUX 500-BD BioSciences) every week for 2–3 weeks, to enrich for mCherry-expressing cells (expressing dCas9-mCherry-KRAB). This cell line was then infected with lentiviral particles containing single sgRNA constructs (MOI of 10) targeting the TSS of either XACT or T113.3, and sorted as above, to enrich for GFP- and mCherry-expressing cells (co-expressing dCas9-mCherry-KRAB and sgRNAs). Cells infected with a lentiviral particle containing an empty sgRNA-expressing vector were used as control.

Chromatin immunoprecipitation

ChIP experiments were performed in H1 hESCs stably expressing the CRISPRi constructs, as previously described⁵⁸. Briefly, cells were crosslinked with 1% formaldehyde for 10min, then quenched with 0.125mM glycine for 5min. Cells were then incubated 30min in swelling buffer (5mM PIPES pH 8, 85mM KCl, 0.5% NP-40). Chromatin was extracted and sonicated in TSE150 buffer (0.1% SDS, 1% Triton, 2mM EDTA, 20mM Tris-HCl pH 8, 150mM NaCl, and protease inhibitors) with a Bioruptor Sonication System (Diagenode, UCD-200). Five micrograms of sonicated chromatin per sample was incubated overnight with 2–5μg of antibody/protein A magnetic beads complexes (2µl per IP of anti-H3K4me3 and anti-H3K27Ac; 2µg per IP of anti-H3K9me3). The beads were subsequently washed in TSE150, TSE500 (20mM Tris-HCl pH 8, 2mM EDTA, 0.1% SDS, 1% Triton X-100, 500mM NaCl), washing buffer (10mM Tris-HCl pH 8, 1mM EDTA, 250mM LiCl, 0.5% NP-40, 0.5% Na-deoxycholate) and finally twice in TE buffer. Elution from the beads was done in TE/1% SDS. Samples were then reverse-crosslinked at 65°C overnight. DNA was finally purified with phenol–chloroform and resuspended in water. The samples were analyzed by qPCR. Primer sequences are provided in Supplementary Table ¹ and the list of antibodies in Supplementary Table ³.

Protein extraction and western blotting

Total proteins were extracted using RIPA buffer (50mM Tris-HCl pH 8, 1mM EDTA, 0.5mM EGTA, 1% Triton X-100, 0.5% sodium deoxycholate, 0.1% SDS, 150mM NaCl, and protease inhibitors) at 4°C for 30min. Extracts were sonicated on a Bioruptor Sonication System (Diagenode, UCD-200) for 10min (30s ON, 30s OFF). Proteins were quantified using the BCA protein assay kit (ThermoFisher Scientific). Total protein extracts were resuspended in 5× Laemmli buffer at a final concentration of 500µg/mL. Ten micrograms of denatured proteins were loaded into a 4–12% gradient polyacrylamide gel (ThermoFisher Scientific) and transferred onto Invitrolon polyvinylidene difluoride membranes (ThermoFisher Scientific). The membranes were blocked for 1h with 5% milk in TBST (10mM Tris pH 8.0, 150mM NaCl, 0.5% Tween 20) and incubated overnight at 4°C with the following antibodies: anti-NANOG (Abcam ab21624, 1:200), anti-SOX2 (Abcam ab97959, 1:1000), anti-OCT4 antibody (Abcam ab181557, 1:1000), anti-TUBULIN (Sigma-Aldrich T9026, 1:10 000), and anti-H3.3 (Merck Millipore 09–838, 1:1000) (Supplementary Table ³). A peroxidase-conjugated antibody was used to reveal the proteins of interest with the Pierce ECL Western blotting substrate (ThermoFisher Scientific). Uncropped scans of blots are provided in the Source Data file

XACT/T113.3 sequence conservation and synteny

XACT and T113.3 sequences were selected from the hg38 human genome assembly for comparison with orthologous sequences from various primate species. These orthologous sequences were identified using BLASTN⁵⁹ and confirmed with liftOver⁶⁰ on panTro5 (chimpanzee), gorGor4 (gorilla), nomLeu3 (gibbon), rheMac8 (rhesus macaque), and calJac3 (marmoset) genome assemblies. Synteny conservation for XACT, T113.3, and neighboring protein-coding genes (LHFPL1, AMOT, HTR2C and LRCH2) was analyzed for the above sequences.

Sequence identities between hg38 and other primates was determined using the MAFFT multiple alignment tool⁶¹ with default parameters. Each primate sequence was independently aligned to hg38 orthologous sequence (using the above mentioned genome assemblies, with the exception of gorilla, in which we used gorGor5). Identity was calculated counting the number of nucleotide matches divided by multiple sequence alignment length. Due to incomplete assemblies of multiple primate species, nucleotides annotated as “N” and neighboring sequencing gaps were discarded and not considered for identity calculation. For the protein-coding genes, only the exons were taken into account to calculate the sequence conservation.

Pluripotency factor-binding sites analysis

To detect pluripotency transcription factors binding regions, Rsat matrix-scan⁶² was performed using the XACT/T113.3 enhancer hg38 sequences and the following Position-Specific Scoring Matrices (PSSMs): the transfac M01307 for OCT4, M01123 for NANOG, and the jaspar MA0143.1 for SOX2. The most probable sequences (lowest p-values) were then selected as putative binding sites. The same method was used for each primate enhancer sequences. P-values were calculated using Rsat matrix-scan tool.

LTR48B human consensus sequences and logo were extracted from the Dfam website⁶³. A multiple alignment with MAFFT was performed to compare the XACT/T113.3 enhancer sequence with the LTR48B consensus sequence. P-values were calculated as described above. In parallel, a Hidden Markov Model file was extracted from Dfam website. This file was converted to a PSSM using hmmlogo from hmmer tool (hmmer.org). From this matrix, the sequences corresponding to the identified binding sites in the XACT/T113.3 enhancer were selected. Consensus and reference matrices were compared using RSAT compare-matrices.

For ChIP-seq analysis, OCT4 and NANOG binding profiles on LTR48B and LTR7 were analyzed using available ChIP-seq datasets: SRR6177948 for OCT4, SRR6177944 for NANOG, and SRR6177932 for IgG⁴⁸. LTR7 and LTR48B sequences in the human genome were selected using the RepeatMasker database⁶⁴. ChIP-seq reads were aligned with hg38 using bowtie2 aligner⁶⁵ using the following parameters: “bowtie2 -p<threads>–end-to-end–no-mixed–no-discordant–minins 100–maxins 1000 -x hg38”. This assigns reads with multiple hits of similar mapping quality to one of those locations randomly. To select only uniquely mapped reads, we used a previously reported custom script⁶⁶. SAM files were sorted, compressed and indexed using samtools sort and samtools index⁶⁷. PCR duplicates were then removed using MarkDuplicates from the Picard tool (http://broadinstitute.github.io/picard/). bigWigs were obtained with bamCompare from deepTools⁶⁸, which calculates the log2 ratio between the ChIP-seq and a ChIP-seq IgG control. A matrix for the relative-coverage of reads in the bigWig was then calculated for all the LTRs belonging to the LTR48B and LTR7 subfamilies using computeMatrix (deepTools). This matrix was finally plotted as a heatmap with plotHeatmap or as a binding profile around subfamilies of LTRs with plotProfile (deepTools).

To calculate the number of uniquely mapped reads aligning to different classes of LTRs, we calculated the number of reads that aligned to each individual LTR using samtools bedcov. For each individual LTR, a reads per kilobase per million mapped reads value was calculated. A control set of sequences was extracted taking 1000 random sequences from hg38 of a length of 390nt (average length of LTR48B and LTR7 elements in the genome). ggplot2 and ggpubr⁶⁹ were used to create boxplots and determine statistical values (t-test). Peak calling was performed with MACS2 with default parameters, using IgG ChIP-seq as a control sample. To calculate the number of peaks on LTR7 and LTR48B, we used bedtools intersect to cross the LTRs in the genome to the called peaks.

We are grateful to members of the Rougeulle lab and members of the Epigenetics and Cell Fate Department for helpful discussion and evaluation of the work leading to this publication. We thank Anne-Valérie Gendrel, Céline Morey, Déborah Bourc’his, Sophie Polo, Miguel Branco, and Andrea Cerase for critical reading of the manuscript and valuable feedback. We also thank the Microscopy and Vectorology Platforms at the Epigenetic and Cell Fate Department, and the ImagoSeine cell-sorting facilities at the Institut Jacques Monod for access to instruments and technical advice. This study was supported by the LabEx “Who Am I?” (ANR-11-LABX-0071) and the Université de Paris IdEx (ANR-18-IDEX-0001) funded by the French Government through its “Investments for the Future” program. This work was also supported by the European Commission Network of Excellence EpiGeneSys (HEALTH-F4–2010–257082), Agence Nationale pour la Recherche (ANR-14-CE10–0017–01), and Ligue contre le cancer. M.C. has received financial support from the European Commission Network of Excellence EpiGeneSys (HEALTH-F4–2010–257082), Agence Nationale pour la Recherche (ANR-14-CE10–0017), and Labex Who Am I (ANR-11-LABX-0071). M.M. was supported by La Ligue contre le cancer. L.E.C. is supported by a fellowship from the French Ministry of Education and Research. O.R. was supported by a fellowship from the French Ministry of Education and Research, from the French Medical Research Foundation (FRM), and from the Labex Who Am I (ANR-11-LABX-0071).