Introduction

G-quadruplexes (G4s), non-canonical DNA secondary structures with quartets of Guanines bonded by Hoogsteen base pairing, are instrumental in regulating gene expression (Sengupta et al., 2020; Varshney et al., 2020). G4s were primarily observed to be able to regulate gene expression when present around transcription start sites (TSSs) (Huppert and Balasubramanian, 2007; Rawal et al., 2006; Verma et al., 2008). G4s can regulate gene expression by directly regulating recruitment of transcription factors and RNA polymerase or via alteration of DNA accessibility by modulating the epigenetic state of the gene promoters (Hussain et al., 2017; Kumar et al., 2011; Lago et al., 2021; Mukherjee et al., 2019; Saha et al., 2017; Sharma et al., 2021; Varshney et al., 2020). Recent studies have implicated the role of G4s in long-distance gene regulation (Robinson et al., 2021).

High-throughput chromosome conformation capture techniques reveal that specific regions of the human genome interact in three dimensions (3D) via chromatin looping and formation of topologically associated domains (TADs) (Bonev Boyan and Cavalli Giacomo, 2016; Denker and De Laat, 2016; Roy et al., 2018). Interestingly, recent computational studies observed enrichment of G4s in TAD boundaries along with higher enrichment of architectural proteins like CTCF and cohesin (Hou et al., 2019). Further, multiple studies noted the presence of G4s correlated with enhancer histone marks like H3K27Ac and H3K4Me1, and predominantly open chromatin regions (Calo and Wysocka, 2013; Hou et al., 2021; Shlyueva et al., 2014)

Although these studies implicate the role of G4s in long-range interactions and/or enhancer function, this was not directly tested. Here we asked if G4 structures might directly alter 3D chromatin, and affect long-range interactions including the epigenetic state of chromatin. To address this, we inserted an array of G4s into an isolated locus devoid of G4-forming sequences using CRISPR-Cas9 genome editing. To evaluate the specific function of G4s, a similar sequence of identical length but devoid of G4-forming capability was introduced. Using these pair of cell lines, we observed insertion of G4s specifically led to the recruitment of enhancer histone marks and increased expression of genes in a 10 Mb window. 3C and Hi-C results showed induced long-range interactions throughout the genome affecting topologically associated domains (TADs) that were specifically due to the incorporated G4s, and not found in case of the control insertion.

Results

Insertion of an array of G4s in an isolated locus

First, we sought to insert an array of G4s in a relatively isolated locus. We looked into Hi-C data from Rao et al., 2014 and identified a region that was markedly isolated with little or no interaction with its surrounding regions (as shown by snapshots of Hi-C interaction matrices obtained using the 3D genome browser (Wang et al., 2018) in Figure S1). In addition, this region was devoid of any G4s in the vicinity (no G4 forming motifs in a ±2.5 kb window).

Thereafter we artificially inserted an array of G4 forming sequences (275 bp long) at this region (chr12:79,872,423-79,872,424, hg19 genome assembly) using CRISPR-Cas9 genome editing (Figure 1A, S2). To study specific effects due to G4s, if any, a control sequence of identical length was inserted in HEK293T cells at the same locus where specific G/Cs necessary for G4 formation were substituted so that G4s are not formed by this sequence (Figure 1A, S2); we also ensured that the GC content was minimally affected by the substitutions (72.4% from 76.73%). Homozygous insertion was confirmed by PCR using primers adjacent to the insertion site followed by Sanger sequencing (Figure 1B, S3). The array of G4-forming sequences used for insertion was previously reported to form stable G4s in human cells (Lim et al., 2010; Monsen et al., 2020; Palumbo et al., 2009).

Figure 1:

Chromatin epigenetic landscape upon insertion of G4s

To understand how the introduction of G4s altered the local chromatin, chromatin immunoprecipitation (ChIP) of different chromatin-modifying histone marks was done followed by qRT-PCR using primers spanning the inserted locus. The PCR primers were designed such that none of the primers bind to any site of G/C alteration in the mutated control insert. We observed significant increase in H3K4Me1 and H3K27Ac enhancer marks in the G4-array when compared to the G4-mutated control (Figures 2A, B). However, there was no G4-specific change in the presence of chromatin compaction marks, H3K27Me3 and H3K9Me3 (Figures 2C, D). The G4-dependent recruitment of H3K4Me1 (associated with enhancers (Heintzman et al., 2009, 2007)) and H3K27Ac (associated with active enhancers and promoters (Creyghton, 2010; Heintzman et al., 2009, 2007)) indicated enhancer-like characteristics of the inserted G4s.

Figure 2:

The inserted G4-array acted as an enhancer

We next asked how the insertion of the G4-array influenced the expression of surrounding genes. To understand the distance-dependent gene regulatory impacts of the inserted G4-array, the mRNA expression of the nearest three genes and then some arbitrarily chosen genes further away up to 5 megabases (Mb) both up and downstream from the insertion site was quantified. Notably, the expression of four of the tested genes (PAWR, PPP1R12A, NAV3, and SLC6A15) increased in the G4-array insert compared to the mutated insert control cells (Figure 3A). Based on this enhanced expression, we further tested and observed a concomitant increase in the recruitment of Ser5 phosphorylated RNA Pol II in the surrounding gene promoters (Figure 3B). Next, we tested if chromosomal looping interactions between the insertion site and the gene promoters were involved in these long-distance effects by using chromosome conformation capture (3C). The 3C assay between the insertion locus and the gene promoters could only be performed till the NAV3 promoter 1.6 Mb away. Beyond this distance, there was not any significantly detectable PCR amplification of 3C interaction products. The 3C assays revealed that there was a G4-dependent increase in chromosomal looping interactions of the insertion locus with the gene promoters (Figure 3C). These results suggested that the insertion of the G4-array led to the formation of an enhancer.

Figure 3:

To understand the mechanism behind the enhancer-like property of the inserted G4-array we analyzed the recruitment of transcriptional coactivator p300 (Kalkhoven, 2004). There was an increase in recruitment of both p300 and as well as the more functionally active acetylated p300/CBP to the G4-array when compared against the mutated control (Figures 3D, E). Together, these results supported the enhancer-like function of the inserted G4-array.

LNA-mediated disruption of the inserted G4s reverses enhancer phenotype

To further establish that the enhancer effects upon the G4 array insertion are due to the formation of G4s, we wanted to see if some of the effects observed could be reversed upon disrupting the inserted structures. Specific Locked Nucleic Acid (LNA) probes were designed to target and disrupt the G4 using a similar approach as shown by others (Cadoni et al., 2021; Chowdhury et al., 2022; Kumar et al., 2008). Three probes were designed with stretches of mostly cytosines (Cs) as LNAs which would hybridize with stretches of guanines (Gs) in the G4-array insert important for the structure formation (Figure 4A; see Methods). We observed that there was a significant decrease in the expression of PPP1R12A and NAV3, two of the genes initially observed to have G4-dependent enhanced expression (Figure 3A), when the G4 array inserted cells were treated with the G4 targeting LNAs (Figure 4B). As expected, a decrease in the H3K4Me1 and H3K27Ac enhancer histone modifications was observed within the insert upon the LNAs treatment (Figures 4C, D). This indicated that the disruption of the inserted G4s can reverse some of the enhancer functions observed upon G4 insertion, further supporting the role of the G4 structure in the observed enhancer functions.

Figure 4:

Domain-wide increase in looping interactions by G4s

For in-depth analysis of the long-range changes in chromatin architecture upon G4 insertion, we performed genome-wide interaction by Hi-C. First, we compared all the Hi-C contacts originating within a ±10 kb window comprising the G4-array insert, or the G4-mutated control insert. Compared to the mutated control, the G4-inserted locus had more than twice as many genome-wide Hi-C interactions (6390 vs 3133) (Figures 5A, B, Supplementary Table 1). To rule out the possibility of artifacts due to the insertion we independently analyzed Hi-C data in HEK293T cells reported earlier (taken from GSE44267, Zuin et al., 2014). After normalizing for sequencing depth, the number of Hi-C contacts from the same window in HEK293T was relatively similar to the G4-mutated insert control (3968 and 3133 respectively, Figure 5C, Supplementary Table 1). Together, these showed that a significant number of new long-range interactions were induced throughout the genome due to the inserted G4s, but not from the inserted control sequence.

Figure 5:

For closer analysis, we focused on intrachromosomal Hi-C interaction matrices of the G4-array insert, or the mutated control insert. This was centered on the insertion locus on chromosome 12 (chr12:7,80,72,423-8,16,72,423; insertion site marked with arrows in Figures 6A, B). The number of Hi-C interactions in the G4-array insert was clearly enriched compared to the G4-mutated insert control, as expected from the global Hi-C contacts noted above. We noted that while the interactions from the G4-array insert were significantly more, the insertion per se did not affect the overall domain architecture, which was largely similar between G4 or G4-mutated inserts as clear from Figures 6A and B. Further, we asked if the domain architecture was retained from that seen in HEK293T cells (with no insertion): Comparison using reported HiC data for the same region from HEK293T cells showed this to be the case confirming that the chromatin domain architecture remained relatively unchanged on introducing the G-array or G4-mutated regions (Figure S4).

Figure 6:

To evaluate the effect of G4s in more detail, we plotted a Hi-C heatmap to show the enhanced or reduced (differential) contacts in the G4-array insert compared to the G4-mutated insert control cells (Figure 6C; relatively enriched/reduced contacts in the G4-array insert w.r.t. the G4-mutated insert plotted in red or blue, respectively; using Juicebox for analysis). This clearly showed that the G4-array induced significantly more Hi-C interactions; interestingly this was particularly evident in the downstream regions. For a closer analysis, we mapped the interaction frequency in a ±100 kb window centered on the insertion site. This clearly showed the difference in the number of interactions between the upstream regions vis-a-vis the region downstream of the insertion (Figure 6D).

To further confirm we used an independent HiC analysis method, HOMER (Hypergeometric Optimization of Motif EnRichment, Heinz et al., 2018) to compute the enhanced/reduced long-range interactions in the G4-array insert, compared to the control G4-mutated insert. Differential analysis using HOMER showed that the inserted locus induced significantly higher number of interactions in the case of G4-array insert relative to the control G4-mutated case (Figure 6E). When we plotted the significantly different chromosomal interactions with minimum 20 interaction reads, it was again clear that the number of interactions with the G4-array insertion region was significantly enhanced in the downstream region relative to the upstream (Figure 6F).

Together these show a clear role of G4s in inducing long-range interactions. A similar sequence devoid of G4-forming capability did not induce such interactions. Furthermore, the overall nature of the TAD was not disturbed, and largely consistent with what is noted in cells with no insertion. Overall, these support that the insertion of G4s induced long-range interactions with minimal organizational changes in the 3D chromatin domain, underlining the molecular role of G4s in the arrangement of 3D chromatin.

A second significant feature was notable at the insertion locus. The number of induced long-range interactions was more significant downstream of the insertion site, compared to the upstream region (Figures 6C, D, F). A close look at the Hi-C contact matrices indicated that the site of insertion was very close and downstream to the TAD boundary (Figures 6A-C). We reasoned that the G4-dependent long-range interactions were largely within the TAD, and limited in the upstream region due to the TAD boundary. This is clearly seen in Figure 6C, akin to an ‘architectural stripe’ displaying that the inserted G4 array had enhanced Hi-C interactions across the domain, thus prominently featured in the downstream regions.

Discussion

To directly test if G4s affect long-range chromatin organization we artificially inserted an array of G4s in the chromatin. Hi-C experiments clearly showed an enhanced number of cis-and trans-chromosomal long-range interactions emanating from the introduced G4s. This was G4-specific because a similar sequence devoid of G4-forming capability introduced at the same site did not result in enhanced interactions. Furthermore, interestingly, most new long-range interactions following G4 incorporation were downstream from the site of insertion. This is likely because the G4 insertion locus was proximal to the upstream TAD boundary thereby restricting most new interactions to the downstream regions within the TAD (Figures 5, 6).

The insertion of the G4 array led to enhanced expression of genes up to 5 Mb away compared to cells with the G4 mutated control insertion. Furthermore, there was enrichment in the H3K4Me1 and H3K27Ac enhancer histone marks, along with recruitment of transcriptional coactivator p300 and the more functionally active acetylated p300/CBP. This was clearly due to the introduction of G4s and not found upon the introduction of the G4-mutated control sequence. Moreover, the enhancer marks were reduced when the inserted G4s were specifically disrupted (Figure 4). Together, these directly support the function of G4s as enhancer elements and as factors that enhance long-range chromatin interactions. It is possible that such interactions are contextually dependent on the type of G4 structure, and/or the adjacent sequence context, and further studies will be necessary to elucidate these.

To ensure that the observed effects were from intracellular G4 formation we accounted for the following while designing the experiments. First, we introduced an array of G4s that were reported earlier to form stable G4s inside human cells; the sequence was from the hTERT promoter region with multiple arrayed G4s that have been well studied (Lim et al., 2010; Monsen et al., 2020; Palumbo et al., 2009). Second, we selected an insertion locus that was otherwise devoid of intrinsic G4s in a ±2.5 kb window. Third, the selected insertion locus was relatively sparse in long-range interactions. Fourth, we independently inserted a sequence of identical length (and similar GC%) which does not form G4s at the same locus (G4-mutated control). All results were compared to the G4-mutated insertion.

Existing literature shows promoter G4s are involved in regulating gene expression (Huppert and Balasubramanian, 2007; Rawal et al., 2006; Verma et al., 2008). Additionally, G4s have been reported to regulate chromatin epigenetics through both cytosine methylation and histone modifications (Halder et al., 2010; Mao et al., 2018; Sengupta et al., 2020). Previous studies by us further show that promoter G4s regulate gene expression by recruiting histone-modifying regulatory complexes (Hussain et al., 2017; Mukherjee et al., 2019; Saha et al., 2017; Sharma et al., 2021). Here we aimed to study how G4s affect the expression of genes far from their location, and if this was through G4-induced modifications in long-range 3D chromatin interactions.

Multiple studies have correlated the presence of G4s with long-range associations. CTCF, an architectural protein primarily involved in TAD boundary formation, was observed to bind to G4s and G4 stabilization was noted to enhance CTCF occupancy (Tikhonova et al., 2021). In addition, G4s were noted to be enriched in TAD boundaries and associated with the formation of chromatin loops (Hou et al., 2019). G4s were also found to coincide with open chromatin regions and H3K27Ac and H3K4Me1 ChIP-Seq peaks, which are markers for transcriptional enhancers (Calo and Wysocka, 2013; Hou et al., 2021; Shlyueva et al., 2014). Most of these regions were observed to overlap with annotated enhancers, and promoters regulated by such enhancers were enriched in G4s (Williams et al., 2020). Furthermore, it was proposed that inter-molecular G4 formation between distant stretches of Gs may lead to DNA looping (Hegyi, 2015; Liano et al., 2022). The YY1 transcription factor was found to bind to G4s and dimerization of G4-bound YY1 led to chromatin looping interactions and consequent regulation of target gene expression (Li et al., 2020). A recent study also showed that R-loop-dependent G4 formation led to stronger CTCF binding (Wulfridge et al., 2023). In addition, it was shown that G4s assist in RNA polymerase II-associated chromatin looping (Yuan et al., 2023). Our findings here provide a direct molecular basis for the above observations by demonstrating a causal role of G4s in inducing both long-range associations and enhancer function. Although from one particular G4-forming stretch, and at one specific genomic locus, this illustrates the function of G4s in 3D gene regulation. Together these shed new mechanistic light on how DNA secondary structure motifs directly control the state of 3D chromatin and thereby biological function.

Materials and Methods

Cell lines and Cell Culture Conditions

HEK 293T cells were cultured in Dulbecco’s Modified Eagle’s Medium-High Glucose (DMEM-HG) supplemented with 10% FBS and 1XAnti-Anti (Gibco).

Primary Antibodies

Histone H3 rabbit polyclonal (Abcam ab1791), H3K4Me1 rabbit polyclonal (Abcam ab8895), H3K27Ac rabbit polyclonal (Abcam ab4729), H3K4Me3 mouse monoclonal (Abcam ab1012), H3K27Me3 mouse monoclonal (Abcam ab6002), H3K9Me3 rabbit polyclonal (Abcam ab8848), p300 rabbit monoclonal (CST 54062), Ac-p300/CBP rabbit polyclonal (CST 4771), TRF2 rabbit polyclonal (Novus NB110-57130).

Genomic Insertions using CRISPR-Cas9 genome editing

For the genomic insertions CRISPR-Cas9 genome editing technique was used (Ran et al., 2013). For the G4 array insertion, 275 bp long hTERT promoter region was PCR amplified from HEK 293T genomic DNA. For the insertion of the mutated G4s, a synthetic DNA template was synthesized and cloned into pUC57 vector by Genscript Biotech Corp, where 12 Gs were substituted with Ts (see supplementary methods for detailed sequences). Both the G4 array and the G4 mutated insertion templates were PCR amplified using longer primers where the short homology arms were introduced as overhangs of the primer for the accurate insertion at the 79M locus via homologous recombination (see supplementary methods for primer sequences) (Paix et al., 2017). For cleavage at the 79M locus (chr12:79,478,643-79,478,644 (hg38)), the gRNA sequence, 5’-ACTATGTATGTACATCCAGG-3’, was cloned into the pX459 v2.0, a gift from Feng Zhang, that co-expresses cas9 protein and the gRNA. Guide RNAs (gRNAs) were designed using the CRISPOR tool (Haeussler et al., 2016). Once the gRNA cloned vector and the insertion donor templates were ready, they were transfected into HEK293T cells and the transfected cells were selected using puromycin, whose resistance gene was present in the pX459 vector. Then these selected cells were serially diluted to isolate clones originating from single cells. Many such clones were screened to detect cells with homozygous/heterozygous insertion of the G4 array or mutated G4 insert by performing locus-specific PCR. Either primers adjacent to the insertion site or cross primers, i.e., one primer within the insert and another from the adjacent region, were used to screen and identify insertions. While using adjacent primers, a shift in PCR product with an increase in amplicon size by 275 bp (size of the insert) indicated successful insertion (see supplementary methods for primer sequences).

ChIP (Chromatin Immunoprecipitation)

ChIP assays were performed as per the protocol previously reported in (Mukherjee et al., 2018). Immunoprecipitation was done using relevant primary antibodies. IgG was used for isotype control in all ChIP experiments. Three million cells were harvested and crosslinked with ∼1% formaldehyde for 10 min and lysed. Chromatin was sheared to an average size of ∼250-500 bp using Biorupter (Diagenode). 10% of sonicated fraction was processed as input using phenol– chloroform and ethanol precipitation. ChIP was performed using 3 μg of the respective antibody incubated overnight at 4°C. Immune complexes were collected using herring sperm DNA-saturated magnetic protein G Dynabeads and washed extensively using a series of low salt, high salt and LiCl Buffers. The Dynabeads were then resuspended in TE (Tris-EDTA pH 8.1) buffer and treated with proteinase K at 65° C for ∼5 hrs. Then, phenol-chloroform-isoamyl alcohol was utilized to extract DNA. Extracted DNA was precipitated by centrifugation after incubating overnight at −20 ° C with isopropanol, 0.3M sodium acetate and glycogen. The precipitated pellet was washed with freshly prepared 70% ethanol and resuspended in TE buffer. ChIP DNA was analyzed by qRT-PCR method (see supplementary methods for primer sequences).

Real-time PCR for Gene (mRNA) expression

Total RNA was isolated using TRIzol® Reagent (Invitrogen, Life Technologies) according to the manufacturer’s instructions. RNA was quantified and cDNA was synthesized using iScript cDNA Synthesis Kits. A relative transcript expression level for genes was measured by quantitative real-time PCR using a SYBR Green based method (see supplementary methods for primer sequences). Average fold change was calculated by the difference in threshold cycles (Ct) between test and control samples. GAPDH gene was used as internal control for normalizing the cDNA concentration of each sample.

Chromosome Conformation Capture (3C)

Chromosome Conformation Capture (3C) assay was done as per the protocol reported in (Cope and Fraser, 2009) with certain modifications. Briefly, about 5-6 million cells were crosslinked using 1% formaldehyde for 10 minutes and then lysed to isolate the nuclei. Nuclei were digested overnight by HindIII and then ligated in a diluted reaction so that intramolecular ligation is favored. After ligation, the reaction mixture was treated with proteinase K at 65° C to de-crosslink the DNA, followed by RNase A treatment. Then, phenol-chloroform-isoamyl alcohol was utilized to extract DNA. Extracted DNA was precipitated by centrifugation after incubating overnight at −80 ° C with 70% ethanol, 0.1M sodium acetate and glycogen. The precipitated pellet was washed with freshly prepared 70% ethanol and resuspended in TE buffer. 3C looping interactions were analyzed by TaqMan qRT-PCR method. For comparison, each interaction frequency was normalized to the interaction between exons 2 and 8 of the human α-actin (ACTA2)(Hadjur et al., 2009). See supplementary methods for primer sequences.

G4 disruption using LNA probes

Probes were designed to specifically bind to regions of genomic DNA containing G repeats which would form the G stems of the G4 structure. The probes containing LNA nucleotides should hybridize with the target with higher stability than the stability of the G4 structure thus destabilizing the G4. The probes used to target the G4 array insert were: 5’-C*CCGACCCCTCC*C-3’, 5’-C*CAGCCCCCTCC*G-3’, 5’-C*CCCTCCCCTTC*C-3’. Stretches of three or more Cs are shown in bold, LNA nucleotides within the probes are underlined, the ends of the probes were protected using phosphorothioate bonds, shown as *. Approximately 0.8 μg of LNA probes (all 3 mixed in equimolar amounts) were transfected per million cells. Cells were treated with the LNA probes for 108 hours by transfecting thrice with a gap of 36 hours in between. The schematic below shows the LNA probes designed to disrupt the inserted G4 structures along with the inserted G4 array sequence to show the specific sites of hybridization by the LNA probes.

Hi-C

Hi-C was performed using the Arima-HiC Kit as per the manufacturer’s protocol. After the proximally-ligated Hi-C templates were generated, sequencing libraries were prepared using NEBNext Ultra II DNA Library Prep Kit as per the Arima-HiC Kit’s protocol. The quality of the sequencing libraries was cross-checked using TapeStation (Agilent Technologies) and the KAPA Library Quantification Kit (Roche) before proceeding with sequencing using NovaSeq 6000 (Illumina).

Hi-C data analysis

Hi-C reads were mapped to the hg19 human genome and processed using default parameters using Juicer (https://github.com/ aidenlab/juicer). Hi-C count matrices were generated at 5kb, 10kb, 25kb, 50kb, 100kb, and 250kb using Juicer. Hi-C heatmap figures were rendered using Juicebox (https://github. com/aidenlab/Juicebox/wiki/Download). Hi-C contacts originating in the loci flanking the G4 insertion site were generated using bedtools (https://bedtools.readthedocs.io/en/latest/). The circos plots were rendered using Circos (http://circos.ca). To identify significant interaction the data was processed using homer (http://homer.ucsd.edu/homer/) using analyzeHiC function. The bins showing 2-fold enrichment in G4 WT over G4 Mut and vice-versa were retained for filtering contacts for representation on circos plots.

Data Availability

The sequencing data underlying this article are available in the NCBI Sequence Read Archive, scheduled to be released upon publication and would be accessible using the following link-https://www.ncbi.nlm.nih.gov/sra/?term=PRJNA1048044. The rest of the data are available in the article and its online supplementary material. Further inquiries can be directed to the corresponding author.

Funding

This work was supported by The Wellcome Trust DBT India Alliance (IA/S/18/2/504021).