Systematic functional characterization of antisense eRNA of protocadherin α composite enhancer.

Enhancers generate bidirectional noncoding enhancer RNAs (eRNAs) that may regulate gene expression. At present, the eRNA function remains enigmatic. Here, we report a 5' capped antisense eRNA PEARL (Pcdh eRNA associated with R-loop formation) that is transcribed from the protocadherin (Pcdh) α HS5-1 enhancer region. Through loss- and gain-of-function experiments with CRISPR/Cas9 DNA fragment editing, CRISPRi, and CRISPRa, as well as locked nucleic acid strategies, in conjunction with ChIRP, MeDIP, DRIP, QHR-4C, and HiChIP experiments, we found that PEARL regulates Pcdhα gene expression by forming local RNA-DNA duplexes (R-loops) in situ within the HS5-1 enhancer region to promote long-distance chromatin interactions between distal enhancers and target promoters. In particular, increased levels of eRNA PEARL via perturbing transcription elongation factor SPT6 lead to strengthened local three-dimensional chromatin organization within the Pcdh superTAD. These findings have important implications regarding molecular mechanisms by which the HS5-1 enhancer regulates stochastic Pcdhα promoter choice in single cells in the brain.

Enhancers are distal cis-acting elements that were originally found to stimulate gene expression in a location- and orientation-independent manner (Banerji et al. 1981). However, recent CRISPR inversion studies showed that enhancers are not orientation-independent in vivo—at least for those associated with CTCF (CCCTC-binding factor) sites (Guo et al. 2015; Lu et al. 2019). Detailed studies of specific gene loci such as the β-globin gene cluster have shed significant insights into our understanding of enhancer function (Collis et al. 1990; Tuan et al. 1992; Ashe et al. 1997). In addition, transcriptional enhancers can act in trans to regulate promoters on homologous chromosomes (Geyer et al. 1990). The enhancer transcription has recently been shown to be a general genome-wide phenomenon (Kim et al. 2010; Ørom et al. 2010), and transcribed noncoding RNAs from active enhancers are known as eRNAs (Kim et al. 2010; Hah et al. 2011; Mousavi et al. 2013; Andersson and Sandelin 2020). In particular, clusters of strong composite enhancers known as superenhancers (SEs) transcribe prevalent eRNAs and activate gene expression programs that often determine cell identities during development and in diseases such as cancers (Lovén et al. 2013; Whyte et al. 2013; Xiang et al. 2014; Pefanis et al. 2015; Henriques et al. 2018). Finally, recent studies suggest that transcription elongation factor SPT6 restricts eRNA transcription and R-loop formation in enhancers (Nojima et al. 2018). Despite widespread transcription and extensive investigations of eRNAs (Li and Fu 2019; Sartorelli and Lauberth 2020), their exact functions remain obscure.

The clustered protocadherin (Pcdh) genes are organized into three closely linked clusters of α, β, and γ. They are stochastically and monoallelically expressed in a cell-specific manner in the brain (Wu and Maniatis 1999; Esumi et al. 2005; Canzio et al. 2019; Jia et al. 2020). The encoded Pcdh proteins function as neural identity codes to specify tremendous numbers of neuronal connections (Canzio et al. 2019; Wu and Jia 2021). The human Pcdh α and γ, but not β, clusters have variable and constant genomic organization, similar to those of the immunoglobulin, T-cell receptor, and UDP-glucuronosyltransferase gene clusters (Wu and Maniatis 1999; Zhang et al. 2004). The variable region of the human Pcdhα gene cluster contains 13 highly similar alternate variable exons (α1–α13) and two C-type variable exons (αC1 and αC2), each of which is separately spliced to a single set of three downstream constant exons to generate diverse mRNAs (Wu and Maniatis 1999). Each Pcdhα alternate variable exon is preceded by a promoter that is flanked by two forward-oriented CTCF sites (Fig. 1A; Guo et al. 2012, 2015; Monahan et al. 2012; Canzio et al. 2019; Jia et al. 2020).

Figure 1.

Protocadherin HS5-1 enhancer produces a prominent antisense eRNA PEARL. (A) Schematic representation of the human Pcdhα gene cluster. The alternative and C-type isoforms are represented by blue and yellow boxes, respectively. HS7 and HS5-1 enhancers are indicated with black arrows. The HS5-1 eRNA transcription start site (TSS) was mapped to the HS5-1 enhancer region between the two CTCF-binding sites (CBSa and CBSb). (B) Shown are total RNA-seq, RNAPII, H3K4me3, H3K27ac, CTCF, and Rad21 ChIP-seq in the human Pcdhα cluster. Total RNA-seq shows transcribed sense (red) and antisense (blue) transcripts. (C) Detailed features of the HS5-1 region are shown. The HS5-1 enhancer transcribes weak sense and strong antisense transcripts. H3K4me3 and H3K27ac mark active promoters and enhancers, respectively. CTCF and Rad21 have two binding sites in the HS5-1 region. (D) Schematic of the 5′RACE experiment. (E) Agarose gels show HS5-1 eRNA products. (Lane 1) HS5-1 eRNA 5′RACE PCR products. Note that the smear below the prominent band PEARL represents several alternative TSSs (Supplemental Fig. S1B). (Lanes 2–4) Negative controls. (M) 1.5-kb DNA ladder. (F) Mouse cortical total RNA-seq. (G) Magnification of the mouse HS5-1 enhancer region. The mouse Pcdhα HS5-1 enhancer transcribes weak sense and strong antisense transcripts that are indicated in red and blue, respectively. (H,I) Quantitative RT-PCR analyses of the bidirectional eRNAs in HEC-1-B cells and the mouse cerebral cortex. (M) 1.5-kb DNA marker, (E) eRNA products, (NC) negative control. See also Supplemental Figure S1.

The Pcdhα cluster is regulated by a downstream superenhancer composed of two composite enhancers of HS7 and HS5-1 (hypersensitive sites 5–1) (Ribich et al. 2006). In particular, the HS5-1 enhancer, flanked by two reverse-oriented CTCF sites, is located at ∼30 kb downstream from the last constant exon (Fig. 1A; Guo et al. 2012, 2015; Monahan et al. 2012; Canzio et al. 2019; Jia et al. 2020). CTCF/cohesin-mediated active “loop extrusion” results in long-distance “double-clamping” chromatin interactions between the two forward–reverse pairs of convergent CTCF sites and brings the HS5-1 enhancer in close spatial contact with its target variable promoters to determine the Pcdhα promoter choice (Guo et al. 2012, 2015; Canzio et al. 2019; Jia et al. 2020). Specifically, antisense transcription of lncRNA from specific variable antisense promoters, which leads to DNA demethylation by TET enzymes and subsequent recruitment of CTCF proteins, is the key determinant of the Pcdhα promoter choice (Canzio et al. 2019). However, the mechanism by which long-distance chromatin interactions between the HS5-1 enhancer and its target promoters regulate Pcdhα promoter choice is not fully understood. Here, we report that an antisense eRNA transcribed from the HS5-1 enhancer is required for Pcdhα promoter activity through forming local R-loops and modifying higher-order 3D chromatin structures.

Results

Pcdhα HS5-1 antisense eRNA revealed by RACE experiments

Using the model cell line of HEC-1-B (Guo et al. 2015), we first performed strand-specific total RNA-seq experiments, which remove the abundant ribosomal RNAs (Ameur et al. 2011), and found prominent antisense transcripts from the HS5-1 enhancer region (Fig. 1B,C). These transcripts map to a position enriched of RNAPII (RNA polymerase II), H3K4me3 (histone 3 lysine 4 trimethylation), and H3K27ac (histone 3 lysine 27 acetylation), which are located between the two reverse-oriented CTCF sites (Fig. 1C). These hallmarks of active enhancers are consistent with recent findings of rich depositions of H3K4me3 at strong enhancers (Henriques et al. 2018).

To map the exact transcription start site (TSS), we carried out cap-dependent 5′ RACE (rapid amplification of cDNA ends) experiments and found a major 804-nt 5′ capped eRNA molecule that we named PEARL (Pcdh eRNA associated with R-loop formation), although several minor 5′ capped transcripts with different TSSs were also detected, suggesting considerable heterogeneity in its start site positions (Fig. 1D,E; Supplemental Tables S1, S2; Supplemental Fig. S1A,B). Despite repeated attempts of 3′RACE, we were unable to determine the exact transcription termination site (TTS). Total RNA-seq with mouse cortical tissues also revealed a prominent HS5-1 antisense transcript (Fig. 1F,G). In addition, sequence analyses showed that this HS5-1 eRNA transcript contains no conserved ORF (open reading frame), suggesting that it is noncoding. Moreover, in both human HEC-1-B cells and mouse cortical brain tissues, total RNA-seq experiments revealed weaker but detectable HS5-1 sense transcripts (Fig. 1C,G). To confirm the bidirectional eRNA transcription, we performed quantitative RT-PCR experiments using either a sense or antisense primer as a reverse transcriptional primer and found that both can produce cDNA (Fig. 1H,I), suggesting that the HS5-1 enhancer transcription is bidirectional. In addition, we used oligo d(T) as the reverse transcriptional primer and found that it can produce cDNA, suggesting that the HS5-1 eRNAs are polyadenylated (Fig. 1H,I). Finally, we quantified eRNA expression levels in model cell lines of HEC-1-B, SK-N-SH, HepG2, and HEK293T and found that HS5-1 eRNA expression levels correlate with those of Pcdhα expression (Supplemental Fig. S1C,D).

HS5-1 eRNA PEARL TSS deletion affects Pcdhα expression

To investigate the potential function of the eRNA PEARL in regulating Pcdhα gene expression, we specifically deleted the TSS region by CRISPR DNA fragment editing (Li et al. 2015; Shou et al. 2018) to avoid the perturbation of the two CBS (CTCF-binding site) elements (Fig. 2A,B), which are known to be essential for Pcdhα gene regulation (Guo et al. 2012; Jia et al. 2020). We isolated two independent homozygous single-cell CRISPR deletion clones (Supplemental Fig. S2). RNA-seq experiments revealed a significant decrease of Pcdh α6, α12, and αC1 expression levels upon TSS deletion (Fig. 2C); note that PcdhαC2 contains no CBS (Guo et al. 2012, 2015) and is not regulated by HS5-1 (Ribich et al. 2006). Next, we examined histone modifications of the HS5-1 enhancer and Pcdhα alternative promoters. We examined enrichment levels of H3K4me3 (Fig. 2D) by ChIP-seq with a specific antibody and found that the TSS deletion results in a significant decrease of H3K4me3 occupancy at the α6, α12, and αC1 promoters as well as at the HS5-1 enhancer region (Fig. 2E–H), consistent with the reduction of gene expression of α6, α12, and αC1 (Fig. 2C). We also performed ChIP-seq experiments with a specific antibody against RNAPII or H3K27ac (Fig. 2I,J) and found that deletion of the eRNA PEARL TSS results in a significant reduction in the occupancy of RNAPII and H3K27ac in the HS5-1 enhancer region (Fig. 2K,L), suggesting that the HS5-1 enhancer activity is impaired upon eRNA PEARL TSS deletion.

Figure 2.

Deletion of the HS5-1 eRNA PEARL TSS affects Pcdhα expression and chromatin landscape. (A) Schematic representation of the HS5-1 eRNA PEARL TSS deletion using the CRISPR/Cas9 system with dual sgRNAs. (B) Shown is the TSS deletion region relative to H3K4me3 and RNAPII ChIP-seq signals at the HS5-1 enhancer region. (C) RNA-seq of wild type (WT) and two eRNA TSS deletion single-cell CRISPR clones (ΔTSS clones 1 and 2) in the Pcdhα cluster (n = 3). (D) H3K4me3 ChIP-seq of WT and ΔTSS in the Pcdhα cluster. (E–H) Magnification of H3K4me3 ChIP-seq at the Pcdh α6, α12, and αC1 variable promoters as well as at the HS5-1 enhancer. (I,J) RNAPII and H3K27ac ChIP-seq of WT and ΔTSS in the Pcdhα cluster. (K,L) Magnification of RNAPII and H3K27ac ChIP-seq at the HS5-1 enhancer. (M) CTCF ChIP-seq of WT and ΔTSS in the Pcdhα cluster. (N–Q) Magnification of CTCF ChIP-seq at the Pcdh α6, α12, and αC1 variable promoters as well as at the HS5-1 enhancer. (R–T) QHR-4C chromatin interaction profiles of WT and ΔTSS with the HS5-1, Pcdhα6, or Pcdhα12 as a viewpoint (n = 2). Data are mean ± SD. (*) P < 0.05, (**) P < 0.01, (***) P < 0.001; unpaired Student's t-test. See also Supplemental Figure S2.

We next carried out ChIP-seq experiments with a specific antibody against CTCF and found that CTCF occupancy is significantly reduced at the Pcdh α6, α12, and αC1 promoters as well as the two CTCF sites flanking the HS5-1 enhancer (Fig. 2M–Q). Quantitative high-resolution chromosome conformation capture copy (QHR-4C) (Jia et al. 2020) with HS5-1 (Fig. 2R) as a viewpoint revealed a significant decrease of long-distance chromatin interactions between the HS5-1 enhancer and the Pcdhα target genes upon eRNA PEARL TSS deletion. Finally, QHR-4C with the alternative α6 or α12 promoter as a viewpoint (Fig. 2S,T) confirmed the decreased long-distance chromatin interactions between the distal enhancer and its Pcdhα target genes (Fig. 2S,T). These data suggest that the eRNA PEARL TSS may play a role in maintaining enhancer activity, mediating chromatin looping, and/or regulating Pcdhα gene expression.

PEARL is required for Pcdhα gene expression via chromatin looping

We next perturbed the eRNA PEARL transcription by inhibiting its initiation, blocking its elongation, or engineering its premature termination by CRISPR epigenetic and genetic methods. CRISPR interference (CRISPRi) could program a KRAB (Krüppel-associated box repressor domain)-fused dCas9 (catalytically dead Cas9) to specific genomic sites to interfere with eRNA transcription without perturbing primary DNA sequences. To this end, dCas9-KRAB was programed by sgRNAs (single-guide RNAs) to a region ranging from 361 to 557 bp upstream of the eRNA PEARL TSS to inhibit its transcription initiation (CRISPRi_i) (Fig. 3A) or to a location of 66 bp downstream from the TSS to block the RNAPII elongation (CRISPRi_e) (Fig. 3B). There were significant decreases of eRNA expression levels in both cases (Supplemental Fig. S3A). In addition, we inserted a pAS (polyadenylation signal) by homologous recombination to cause a premature termination of eRNA transcription (Fig. 3C; Supplemental S3B,C). Two homozygous single-cell CRISPR clones were obtained by screening CRISPR insertion clones and their genotypes were confirmed by PCR and Sanger sequencing (Supplemental Fig. S3D,E).

Figure 3.

HS5-1 eRNA PEARL affects Pcdhα gene expression. (A,B) Schematic of CRISPR interference (CRISPRi) for perturbation of the eRNA PEARL transcription initiation (CRISPRi_i) (A) or elongation (CRISPRi_e) (B). (C) Schematic representation of premature termination of the HS5-1 eRNA transcription by CRISPR DNA fragment insertion of 447-bp polyadenylation signal (pAS) sequences by homologous recombination. The sgRNA targeting site is indicated by a vertical black arrow and the donor sequences are indicated by the dotted line. (D–F) Shown are RNA-seq data after interfering with the HS5-1 eRNA PEARL transcription initiation, elongation, or termination in the Pcdhα cluster (n = 3), respectively. (G) H3K4me3 ChIP-seq patterns of CRISPRi_i, CRISPRi_e, and pAS insertion at the Pcdh α6, α12, and αC1 variable promoters as well as the HS5-1 enhancer in the Pcdhα cluster. (H,I) RNAPII and H3K27ac ChIP-seq of CRISPRi_i, CRISPRi_e, and pAS insertion. (J,K) Magnification of the HS5-1 enhancer region. (L,M) CTCF and Rad21 ChIP-seq of CRISPRi_i, CRISPRi_e, and pAS insertion in the Pcdhα cluster. (N–P) QHR-4C interaction profiles with the HS5-1, Pcdhα6, or Pcdhα12 as a viewpoint (n = 2). See also Supplemental Figure S3.

RNA-seq experiments revealed a significant decrease of expression levels of Pcdh α6, α12, and αC1 in all three methods of perturbing eRNA transcription (Fig. 3D–F). We then examined H3K4me3 levels in the Pcdhα cluster after interfering with PEARL transcription and found that H3K4me3 enrichments are decreased at the α6, α12, and αC1 promoters as well as at the HS5-1 enhancer (Fig. 3G), consistent with the RNA-seq results (Fig. 3D–F). In addition, RNAPII occupancy and H3K27ac enrichment at the HS5-1 enhancer were also reduced (Fig. 3H,I). We noted that, compared with perturbing eRNA PEARL transcription initiation, the RNAPII occupancy and H3K27ac enrichment in the HS5-1 enhancer region were almost abolished upon blocking eRNA elongation or inserting a premature termination signal (Fig. 3J,K). Next, we designed two sgRNAs to program dCas9 to the locations further downstream from the eRNA PEARL TSS (Supplemental Fig. S3F): one at 1158 bp downstream from the TSS and the other at 1959 bp downstream from the TSS. We found that the former reduced the expression levels of both PEARL and Pcdhα as well as the HS5-1 enhancer activity, but the latter had no effect, presumably because it is located after HS5-1 eRNA termination (Supplemental Fig. S3G–I). These data suggested that the eRNA PEARL is required for Pcdhα gene expression.

Previous studies have shown that eRNAs play a role in gene activation by mediating the formation of chromatin loops (Lai et al. 2013; Li et al. 2013; Hsieh et al. 2014; Xiang et al. 2014). In the Pcdhα cluster, the stochastic expression of Pcdhα depends on the long-distance chromatin interactions between distal enhancer and target variable promoters mediated by CTCF and its associated cohesin complex (Guo et al. 2012; Canzio et al. 2019; Jia et al. 2020). Therefore, we asked whether the eRNA PEARL plays a role in CTCF/cohesin-mediated chromatin looping between the distal enhancer and its target variable promoters. To this end, we first measured the enrichments of CTCF and Rad21 by ChIP-seq and found that they were reduced at the HS5-1 enhancer region (Fig. 3L,M). In addition, they were also reduced at the Pcdh α6, α12, and αC1 variable promoters (Fig. 3L,M). We then performed QHR-4C and found that there is a decrease of long-distance chromatin interactions between the HS5-1 enhancer and its target variable promoters (Fig. 3N–P). Although these CRISPR experiments have caveats for interfering with the HS5-1 enhancer function, all in all, these data suggest that the eRNA PEARL may be necessary for chromatin interactions between the Pcdhα HS5-1 enhancer and its target variable promoters.

Locally transcribed but not globally overexpressed PEARL regulates Pcdhα expression

To understand the function of eRNA PEARL, we activated eRNA transcription in cis locally by using the CRISPR activation (CRISPRa) system with a dCas9-VP160 protein programed to a region upstream of the eRNA TSS (Fig. 4A). We first confirmed the eRNA CRISPR activation by quantitative RT-PCR (Fig. 4B). We then performed RNA-seq and found that eRNA PEARL transcriptional activation results in a significant increase of expression levels of the Pcdh α6, α12, and αC1 genes (Fig. 4C). Consistently, there is a significant increase of long-distance chromatin interactions between the HS5-1 enhancer and its target variable promoters (Fig. 4D–F).

Figure 4.

Locally transcribed PEARL regulates Pcdhα expression. (A) Schematic of CRISPR activation (CRISPRa) with sgRNAs ranging from 361 to 557 bp upstream of the HS5-1 eRNA TSS. (B) The HS5-1 eRNA expression after activating the HS5-1 eRNA transcription (n = 3). (C) Shown is RNA-seq after activating the HS5-1 eRNA transcription in the Pcdhα cluster (n = 3). (D–F) QHR-4C chromatin interaction profiles of WT and CRISPRa with the HS5-1, Pcdhα6, or Pcdhα12 as a viewpoint (n = 2). (G) Overexpression of the eRNA PEARL in HEC-1-B cells (n = 3). (OE) Overexpression. (H) Shown are the Pcdhα expression levels after eRNA overexpression (n = 3). (I) RNA-seq for members of the Pcdhα cluster after overexpressing the eRNA PEARL (n = 3). (J–L) QHR-4C chromatin interaction profiles of WT and eRNA overexpression with the HS5-1, Pcdhα6, or Pcdhα12 as a viewpoint (n = 2). (M,N) Overexpression of the HS5-1 eRNA PEARL cannot rescue Pcdhα expression in the ΔTSS clone (n = 3). (O) RNA-seq shows no rescue of the ΔTSS clone (n = 3). (P–R) QHR-4C chromatin interaction profiles of WT and eRNA overexpression in the ΔTSS clone with the HS5-1, Pcdhα6, or Pcdhα12 as a viewpoint (n = 2).

We next overexpressed the eRNA PEARL in trans globally by a U6 promoter and found, surprisingly, that its overexpression has no effect on Pcdhα expression (Fig. 4G–I). In addition, its overexpression also has no effect on long-distance chromatin interactions between the HS5-1 enhancer and its target variable promoters (Fig. 4J–L). Finally, we overexpressed the eRNA PEARL in the TSS deletion CRISPR clones and found that it rescues neither Pcdhα gene expression (Fig. 4M–O) nor long-distance chromatin looping (Fig. 4P–R). Collectively, we concluded that locally transcribed, but not globally overexpressed, eRNA PEARL or its transcriptional process per se regulates Pcdhα gene expression.

Local PEARL transcripts function in cis within hypomethylated HS5-1 enhancers

To investigate why locally transcribed, but not globally overexpressed, PEARL regulates Pcdhα gene expression, we performed chromatin isolation by RNA purification followed by sequencing (ChIRP-seq) experiments (Chu et al. 2011). We synthesized 24 specific 3′-biotin DNA probes directed against the eRNA PEARL transcripts and divided them into odd and even pools to capture with streptavidin magnetic beads (Fig. 5A). We first confirmed that both odd and even probes enrich the eRNA PEARL transcripts (Fig. 5B). We found that there exist unique signals in the HS5-1 enhancer region with both odd and even probe pools, suggesting that the eRNA PEARL is located in the HS5-1 enhancer region (Fig. 5C,D). As a positive control, the noncoding RNA NEAT1 is specifically enriched in the MALAT1 locus in trans (Supplemental Fig. S4A; West et al. 2014). We also performed ChIRP followed by SDS-PAGE with silver staining and did not find any prominent protein in comparison with the known protein PSF association of NEAT1 (Supplemental Fig. S4B,C; West et al. 2014; Jiang et al. 2017a).

Figure 5.

The eRNA PEARL forms local R-loops in the HS5-1 enhancer. (A) Schematic representation of odd and even probe pools of the ChIRP assay. (B) Enrichment of the eRNA PEARL after the ChIRP experiments with odd or even probes compared with the control of the housekeeping gene GAPDH (n = 2). (C) The HS5-1 eRNA PEARL ChIRP-seq in HEC-1-B cells. (D) Magnification showing the PEARL occupancy. (E–H) The MeDIP-seq used a specific antibody against 5mC to detect DNA methylation in HEC-1-B cells (E,F) or mouse cortical tissues (G,H). (I) Schematic of DRIP experiment using the S9.6 antibody to capture R-loops. (J) The R-loop enrichment in the HS5-1 enhancer region by DRIP-qPCR (n = 2). IgG was used as a negative control. See also Supplemental Figure S4.

We next performed methylated DNA immunoprecipitation and sequencing experiments (MeDIP-seq) and found that, in contrast to hypermethylation in the Pcdhα variable exons (Supplemental Fig. S4D,E), the eRNA promoter within the HS5-1 enhancer is hypomethylated in both human HEC-1-B cells and mouse cortical brain tissues (Fig. 5E–H). Given that the R-loop structure is often formed by local RNAs in the hypomethylated region (Ginno et al. 2012; Arab et al. 2019; Niehrs and Luke 2020), these MeDIP-seq and ChIRP-seq data suggest that local eRNA PEARL transcripts function in cis within the hypomethylated HS5-1 enhancer region.

PEARL transcripts form local R-loops in situ in the enhancer region

R-loops or DNA–RNA hybrids are special three-stranded nucleic acid structures that form locally in vivo to perform physiological or pathological functions (Skourti-Stathaki and Proudfoot 2014; Crossley et al. 2019; García-Muse and Aguilera 2019; Niehrs and Luke 2020), but whether eRNA in the distal enhancer region regulates activation of target promoters via local R-loop formation in situ is not known. To this end, we performed DRIP (DNA–RNA immunoprecipitation) assays with the S9.6 antibody, which specifically recognizes DNA–RNA hybrids (Fig. 5I; Boguslawski et al. 1986; Ginno et al. 2012; Tan-Wong et al. 2019). We found that the eRNA PEARL is significantly enriched in the S9.6 immunoprecipitate, suggesting that the eRNA PEARL forms R-loops in the enhancer region (Fig. 5J).

LNA-mediated knockdown of PEARL

We next used locked nucleic acid antisense oligonucleotides (LNAs) to specifically block and/or degrade (by endogenous RNase H) eRNA PEARL transcripts (Fig. 6A; Bennett and Swayze 2010; Li et al. 2013). RNA-seq experiments showed that there are significant decreases of expression levels of Pcdhα genes (Fig. 6B). In addition, DRIP-seq with the S9.6 antibody showed that the HS5-1 R-loop is almost abolished upon LNA treatments (Fig. 6C) compared with no effect on the control NEAT1 R-loop (Supplemental Fig. S4F). Finally, QHR-4C experiments showed that there is a reduction of long-distance chromatin interactions between the HS5-1 enhancer and its target promoters (Fig. 6D–F). These data suggest that eRNA PEARL transcripts form local R-loops in the HS5-1 enhancer and affect Pcdhα gene expression through long-distance chromatin interactions.

Figure 6.

Local R-loop formation of eRNA PEARL participates in higher-order chromatin organization. (A) Shown are the expression profiles of the HS5-1 eRNA after antisense LNA-mediated silencing with two functional LNAs (n = 3). (B) RNA-seq after LNA perturbation (n = 3). (C) The R-loop enrichment in the HS5-1 enhancer by DRIP-seq. (D–F) QHR-4C interaction profiles with Pcdhα6, Pcdhα12, or the HS5-1 as a viewpoint (n = 2). (G) Schematic of the RNase H1 construct with no mitochondrial localization signal. (H) Western blot of the overexpressed RNase H1. (OE) Overexpression. (I) Quantitative analyses of the HS5-1 eRNA PEARL expression after overexpressing RNase H1 (n = 3). (J) RNA-seq of mock and RNase H1 overexpression (n = 3). (K–M) QHR-4C chromatin interaction profiles of WT and RNase H1 overexpression with the Pcdhα6, Pcdhα12, or the HS5-1 as a viewpoint (n = 2). See also Supplemental Figure S4.

Knockdown of R-loops by RNase H1 overexpression

RNase H1 enzyme hydrolyzes the RNA strand of RNA/DNA hybrids (Stein and Hausen 1969; Skourti-Stathaki and Proudfoot 2014; Nojima et al. 2018; Tan-Wong et al. 2019). We constructed a plasmid overexpressing V5-tagged RNase H1 lacking mitochondrial localization signal (Fig. 6G) and confirmed its overexpression by Western blot with a specific antibody against the V5 tag (Fig. 6H). We found that RNase H1 overexpression results in a significant decrease of the eRNA PEARL levels, suggesting that the eRNA PEARL forms DNA–RNA hybrids in vivo (Fig. 6I). In addition, RNase H1 overexpression leads to a significant decrease of expression levels of Pcdh α6, α12, and αC1 (Fig. 6J). Finally, the long-distance chromatin interactions between the HS5-1 enhancer and its target variable promoters are also decreased upon RNase H1 overexpression (Fig. 6K–M). In conjunction with the ChIRP-seq data (Fig. 5), this suggests that the eRNA PEARL forms local RNA/DNA hybrids that may participate in long-distance chromatin interactions.

SPT6 knockdown leads to increased intradomain interactions within the Pcdh superTAD

Recent studies found that transcription elongation factor SPT6 (suppressor of Ty 6) restricts eRNA transcription and R-loop formation in enhancer regions (Nojima et al. 2018). Consistent with this, we found that SPT6 depletion by sequence-specific antisense LNA (Fig. 7A) results in a significant increase of eRNA PEARL expression (Fig. 7B). Interestingly, SPT6 depletion also leads to increased long-distance chromatin interactions between the HS5-1 enhancer and its target promoters (Fig. 7C–E), suggesting that the eRNA PEARL may participate in higher-order chromatin organization.

Figure 7.

Enhancing the eRNA PEARL by perturbing SPT6 alters higher-order chromatin organization. (A,B) The SPT6 and eRNA PEARL expression levels after LNA-mediated SPT6 depletion (n = 3). (C–E) QHR-4C interaction profiles with the HS5-1, Pcdhα6, or Pcdhα12 as a viewpoint after SPT6 depletion (n = 2). (F–H) The interaction heat map of the Pcdh clusters from HiChIP experiments at the 10-kb resolution. (I–K) Catalytic-dead RNase H1(D210N)-centric higher-order chromatin organization within the Pcdh superTAD after perturbing PEARL or SPT6. (L) A model for local R-loop formation of eRNA PEARL regulating Pcdhα gene expression through long-distance chromatin looping. The eRNA PEARL expression and R-loop formation maintained by transcription elongation factor SPT6 facilitate proper long-distance chromatin contacts via CTCF/cohesin “loop extrusion.” See also Supplemental Figure S5.

To further explore 3D chromatin organization mediated by PEARL, we constructed a stable cell line expressing a catalytically dead RNase H1 mutant (D210N) (Supplemental Fig. S5A), which lacks the RNA hydrolyzing catalytic activity but still binds RNA/DNA hybrids. This RNase H1 mutant can bind R-loops but cannot resolve them (Chen et al. 2017). We then performed HiChIP experiments, which are similar to ChIA-PET or in situ HiC and detect protein-centric genome architecture (Mumbach et al. 2016), with an RNase H1-specific antibody. The overall heat map of HiChIP revealed RNase H1(D210N)-centric (presumably R-loop-mediated) chromatin conformation genome-wide (Supplemental Fig. S5B–D). Interestingly, PEARL depletion results in decreased chromatin interactions in the Pcdh clusters, whereas SPT6 depletion results in increased chromatin interactions (Fig. 7F–H).

Previous studies showed that the three Pcdh clusters are organized into one superTAD with two subTADs (Guo et al. 2015; Jiang et al. 2017b). Our HiChIP data showed that chromatin interactions are abundant within the Pcdh superTAD (Fig. 7I). Importantly, PEARL depletion by LNAs leads to significant decreases of intradomain interactions; in contrast, SPT6 depletion results in significant increases of local chromatin interactions within the Pcdh superTAD (Fig. 7J,K). Given that SPT6 ensures proper eRNA transcription and R-loop formation (Nojima et al. 2018), these data suggest that eRNA PEARL and the formed R-loops participate in the Pcdh higher-order chromatin organization.

Discussion

Giving the ongoing debate about whether eRNA transcripts are functional (Kaikkonen and Adelman 2018), our data support that eRNA PEARL is an important regulator for Pcdhα gene expression. Owing to the fact that eRNA is transcribed from enhancer DNA sequences, CRISPR genetic methods (deletion or insertion) also perturb the DNA sequences of the enhancer region. To minimize the effects of CRISPR genetic editing, we next used a CRISPR epigenetic method (CRISPR interference or CRISPR activation) to study the function of eRNA transcription and transcripts. Although it remains challenging to distinguish eRNA transcription and transcripts, we attempted to use various strategies to explore the function of eRNA PEARL. In particular, antisense LNAs caused specific degradation of local eRNA PEARL transcripts by endogenous RNase H without altering DNA sequences or the spacing between CTCF sites. These data support that the local eRNA PEARL transcripts are functional and are not merely by-products of the HS5-1 enhancer. All in all, the combined data from various experiments suggest that eRNA PEARL regulates Pcdhα expression via R-loop formation and promotes long-distance chromatin interactions between distal enhancers and target promoters.

The ∼60 clustered Pcdh genes encode large numbers of cadherin-like cell adhesion proteins that function as neural identity tags in individual cells in the brain (Canzio et al. 2019; Wu and Jia 2021). The enormous cell surface repertoires for neuronal identities are achieved by combinatorial expression of ∼15 members of the clustered Pcdh genes (two alternate α isoforms, four β isoforms, four alternate γ isoforms, and five C-type Pcdhs) (Guo et al. 2015; Canzio et al. 2019; Jia et al. 2020). Whereas expression of members of the Pcdhβγ clusters is regulated by a superenhancer downstream from the Pcdhγ cluster, members of the Pcdhα cluster are regulated by a superenhancer composed of HS5-1 and HS7. We show here, in mouse and cellular models, that HS5-1 transcribes a prominent antisense eRNA (Fig. 1; Supplemental Fig. S1) that forms R-loops locally in the HS5-1 enhancer region (Figs. 5, 6) and is required for the regulation of distal target promoters through modifying higher-order chromatin architecture (Figs. 6, 7). We recently found that REST/NRSF binds to the location between the PEARL TSS and CBSb and that releasing REST/NRSF initiates the derepression of Pcdhα expression (Tang et al. 2021), probably via PEARL transcription because of close location. Our data are consistent with a model that the R-loop formed by eRNA PEARL in the HS5-1 enhancer region participates in 3D chromatin organization to activate the chosen promoters of Pcdh α6, α12, and αC1 genes (Figs. 6, 7). It is tempting to speculate that R-loop formation at the HS5-1 enhancer might stall the “loop extrusion” of the CTCF/cohesin complex, thus bringing the distal enhancer in close contact with variable promoters. The stochastically chosen Pcdhα genes, in conjunction with balanced expression of members of the Pcdh β and γ clusters, determine individual neuron identity in the brain.

Materials and methods

Total RNA-seq

Total RNA-seq, which removes the abundant ribosomal RNAs, was performed as previously described (Ameur et al. 2011) with modifications. For more detailed procedures, see the Supplemental Material.

QHR-4C

Quantitative high-resolution chromosome conformation capture copy (QHR-4C) was performed as recently described (Jia et al. 2020) with modifications. For more experimental details, see the Supplemental Material.