Global identification and functional characterization of Z‐DNA in rice

Summary

Z‐DNA is a left‐handed double helix form of DNA that is believed to be involved in various DNA transactions. However, comprehensive investigations aimed at global profiling of Z‐DNA landscapes are still missing in both humans and plants. We here report the development of two techniques: anti‐Z‐DNA antibody‐based immunoprecipitation followed by sequencing (ZIP‐seq), and cleavage under targets and tagmentation (CUT&TAG) for characterizing Z‐DNA in nipponbare rice (Oryza sativa L., Japonica). We found that Z‐DNA‐IP⁺ (Z‐DNA recognized by the antibody) exhibits distinct genomic features as compared to Z‐DNA‐IP⁻ (Z‐DNA not recognized by the antibody). The concomitant presence of G‐quadruplexes (G4s) and i‐motifs (iMs) may promote Z‐DNA formation. DNA modifications such as DNA‐6mA/‐4acC generally disfavours Z‐DNA formation, while modifications like DNA‐5mC (CHH) and 8‐oxodG promote it, highlighting the distinct roles of DNA base modifications in modulating Z‐DNA formation. Importantly, Z‐DNA located at transcription start sites (TSSs) enhances gene expression, whereas Z‐DNA in genic regions represses it, underscoring its dual roles in regulating the expression of genes involved in fundamental biological functions and responses to salt stress. Furthermore, Z‐DNA may play a role in transcriptional initiation and termination rather than in transcriptional elongation. Finally, the presence of Z‐DNA in promoters is correlated with the coevolution of overlapping genes, thereby regulating gene domestication. Consequently, our study represents as a pivotal point and a solid foundation for reliably launching genome‐wide investigations of Z‐DNA, thereby advancing the understanding of Z‐DNA biology in both plants and non‐plant systems.

Introduction

Z‐DNA is a noncanonical left‐handed double‐helical nucleic acid structure, exhibiting a zig‐zag configuration as described by Wang et al. (1979). It was proposed in 1970 according to the unusual double‐helical DNA structure formed from the d(I‐C) polymer (Mitsui et al., 1970). It was subsequently discovered in a DNA fragment at an atomic resolution, which contains self‐complementary dG‐dC sequences such as hexamer d(CG)₃ or tetramer d(CG)₂, through X‐ray diffraction analysis conducted in the late 1970s (Crawford et al., 1980; Wang et al., 1979). Physically, Z‐DNA has remarkedly different helical parameters, such as diameter, number of residues per turn, pitch height, base tilt, rise per residue and rotation per residue, which markedly differ from those of B‐DNA and A‐DNA (Krall et al., 2023). These differences represent significant conformational variations among DNA structures.

It has been documented that formation and stability of Z‐DNA can be influenced by a variety of both endogenous and exogenous factors (Krall et al., 2023; Rich et al., 1984). These factors include intrinsic DNA sequences, chemical modifications of bases such as methylation and bromination, as well as mechanical stresses such as negative supercoiling (Peck et al., 1982; Rahmouni and Wells, 1989), torsional stress (Kim et al., 2021; Wittig et al., 1989) and DNA bending (Yi et al., 2022). Additionally, the activities of Z‐DNA‐binding proteins (Bartas et al., 2022; Lafer et al., 1985) and environmental conditions, including the presence of cations (e.g. Ca²⁺, Co⁴⁺, Ru⁴⁺), variations in salt concentration (e.g. NaCl, MgCl₂), polyamines (such as spermidine and spermine) and osmolytes like alcohols (methanol, ethanol or propanol) and polyols (sucrose or stachyose) (Behe and Felsenfeld, 1981; Krall et al., 2023; Subramani et al., 2019; Thomas et al., 1991) also play significant roles in modulating Z‐DNA stability. For instance, Z‐DNA is preferentially formed from alternating purine‐pyrimidine nucleotide sequences, with a propensity for formation in the following order: d(GC)_n > d(TG)_n > d(TA)_n (Ho et al., 1986; McLean et al., 1986; Rich et al., 1984; Rich and Zhang, 2003). Methylation (Behe and Felsenfeld, 1981) and bromination (Chevrier et al., 1986; Moller et al., 1984; Ross et al., 1989) at the C5 position of cytosine (C), as well as methylation or bromination at the C8 position of guanosine (G) (Moller et al., 1984; Ross et al., 1989; Xu et al., 2003) have been shown to promote the formation of Z‐DNA. Additionally, polyamines such as spermine and spermidine, along with high salt concentrations, such as 4 m NaCl (Behe and Felsenfeld, 1981), also contribute to this process. The bending force of DNA facilitates the transition from B‐DNA to Z‐DNA, thereby enabling the formation of Z‐DNA under physiological salt conditions (Yi et al., 2022).

Z‐DNA can be transiently stabilized by the consumption of additional energy or with the assistance of binding proteins (Bartas et al., 2022; Lee et al., 2010). Consequently, its genomic role was initially overlooked and remained a topic of debate. Increasing evidence indicates that negative supercoiling (Haniford and Pulleyblank, 1983b; Peck et al., 1982; Peck and Wang, 1983), and Z‐DNA binding proteins containing a conserved Z‐DNA binding domain Zα, like ADAR1 (double‐stranded RNA adenosine deaminase) (Herbert et al., 1995), DLM1 (Schwartz et al., 2001), E3L (Brandt and Jacobs, 2001) and ZBP1 (Ha et al., 2008), can stabilize Z‐DNA, leading to reconsideration of its potential biological implications. Currently, Z‐DNA has been reported to function in various cellular processes, including the recruitment of Z‐DNA binding proteins and other trans‐factors such as MeCP2 (Bayele et al., 2007; Ravichandran et al., 2019), transcriptional regulation (Cerna et al., 2004; Del Mundo et al., 2017; Ferl and Paul, 1992; Liu et al., 2001; Liu and Wang, 1987; Oh et al., 2002; Wolfl et al., 1995, 1996), genetic recombination (Blaho and Wells, 1989; Kobori et al., 1986; Wahls et al., 1990), genome instability (Duardo et al., 2023; McKinney et al., 2020), restriction of nucleosome formation (Garner and Felsenfeld, 1987), bacterial biofilm formation (Buzzo et al., 2021), nucleosome positioning (Mulholland et al., 2012; Wong et al., 2007), chromatin remodelling (Liu et al., 2006), DNA damage (Kha et al., 2010; Luokkamaki et al., 1993; Ngan et al., 1989), structural changes of chromatin or chromosomes (Javadekar and Raghavan, 2015; Wang et al., 2006), and human diseases (Herbert, 2019; Karki et al., 2021; Kuriakose and Kanneganti, 2018; Ravichandran et al., 2019; Vongsutilers and Gannett, 2018; Wang and Vasquez, 2007, 2023). Z‐DNA containing d(CG)₉ can reprogram the positioning of local nucleosomes, thereby creating a favourable chromatin environment that promotes the transcription of nearby genes in yeast (Wong et al., 2007). Long terminal repeat (LTR) retrotransposon elements derived Z‐DNA can serve as alternative promoters for modulating the transcription of genes in humans, such as GSDML, APOC1 and LY6K, which are associated with human diseases (Lee et al., 2022). However, further intensive studies are required to achieve a better understanding of Z‐DNA functions in the genome, particularly in plants.

Z‐DNA can be identified in vitro through a range of biophysical and biochemical techniques (Rich et al., 1984). The biophysical techniques encompass circular dichroism (CD) (Subramani and Kim, 2023), infrared spectroscopy (Zhang and Huang, 2023), nuclear magnetic resonance (NMR) (Choi et al., 2023), Raman spectroscopy (Wartell et al., 1982), fluorescence resonance energy transfer (FRET) (Jares‐Erijman and Jovin, 1996), and single‐molecule FRET (Jung et al., 2023), hydrogen‐deuterium exchange (Hartmann et al., 1982), isothermal titration calorimetry (ITC) and differential scanning calorimetry (DSC) (Ferreira and Sheardy, 2006), and X‐ray diffraction analysis (Crawford et al., 1980; Wang et al., 1979). The biochemical methods include gel electrophoresis (Haniford and Pulleyblank, 1983a; Singleton et al., 1982), sedimentation (Peck et al., 1982), nitrocellulose filtration (Kuhnlein et al., 1980), chiral molecule binding (Barton et al., 1984), nuclease sensitivity assays (Kilpatrick et al., 1983), chemical probe‐based detection (Ferl and Paul, 1992; Johnston and Rich, 1985), enzymatic probe detection (Wohlrab and Wells, 1987) and anti‐Z‐DNA antibody‐based detection (Rich et al., 1984; Spencer et al., 2021). These methods are low‐throughput, which restrict their applications on a genome‐wide scale. In stark contrast, only a few high‐throughput methodologies have been applied for the global profiling of Z‐DNA, including the first global view of Z‐DNA using Zaa‐based ChIP‐seq (Shin et al., 2016) and computer based prediction such as DeepZ in the human genome (Beknazarov et al., 2020; Beknazarov and Poptsova, 2023) and Z‐Hunt in the genomes of humans, Arabidopsis and rice (Ho et al., 1986; Schroth et al., 1992; Zhou et al., 2009). Therefore, increased investment is necessary to develop more robust high‐throughput methodologies for the global characterization of Z‐DNA.

Poly‐ and monoclonal antibodies have been developed to specifically recognize Z‐DNA structures both in vitro and in vivo (Lafer et al., 1981; Moller et al., 1982), marking a significant advancement in Z‐DNA biology. However, comprehensive antibody‐based global profiling of Z‐DNA remains absent in both humans and plants. To address this gap, we developed for the first time the Z‐DNA antibody (Z22)‐based immunoprecipitation (IP)‐seq and CUT&TAG methodologies to profile Z‐DNA across the rice genome. We subsequently integrated them with multiple omics data, including RNA‐seq, histone ChIP‐seq, BS‐seq, IP‐seq for G4/iMs, DNA6‐mA and 4acC and OxiDIP‐seq, to achieve a comprehensive characterization of Z‐DNA‐related genomic and epigenomic features. This integrated approach provides insights into the potential biological implications of Z‐DNA in rice.

Results

Global identification of Z‐DNA in the rice genome using ZIP‐seq

To identify Z‐DNA within the rice genome, we initially performed an anti‐Z‐DNA (Z22)‐based dot blotting assay using both reconstituted genomic DNA and synthetic oligonucleotides containing (positive control for the formation of Z‐DNA) or lacking (negative control for the lack of Z‐DNA formation) putative Z‐DNA‐forming sequences (PZFSs), both of which have already been characterized before (Lang et al., 1982; Schroth et al., 1992). Immuno‐signal can be clearly detected in the positive control and genomic DNA, but not in the negative control (Figure S1a). We also reconstructed a DNA G‐quadruplex (G4) and i‐motif (iM) sequence, which was confirmed by using G4P‐ and iMab‐based dot blotting assays, respectively (Figure S1b), for Z22‐based dot blotting assays. We found that immuno‐signal is undetectable in both G4 and iM sequence (Figure S1c). These results confirm the specificity of Z22 for recognizing Z‐DNA. We then developed a Z‐DNA‐IP‐seq (hereafter termed ZIP‐seq) approach for the global identification of Z‐DNA in nipponbare rice (Oryza sativa L., Japonica). The main procedures include: purification and fragmentation of genomic DNA followed by Z‐DNA reconstitution, incubation with Z22 followed by protein G capture, and recovery of Z22‐enriched Z‐DNA fragments for library preparation and sequencing (Figure S2). We generated two highly correlated biological replicates (Rep) for immunoprecipitation (IP) (r = 0.9) along with input and IgG as negative controls (Figure S3, Table S1). We identified 18 537/25349 and 21 987/25979 Z‐DNA peaks corresponding to Rep I and II, respectively, relative to IgG and input (Figure S4), leading to 12 189 common peaks for downstream analyses (Figure 1a). To validate the accuracy of these peaks, we associated them with PZFSs and observed an overlapping rate as 71%, leading to 8654 Z‐DNA (PZFS)‐IP⁺ (representing PZFSs isolated with Z22), 66 408 Z‐DNA (PZFS)‐IP⁻(representing PZFSs not isolated with Z22) and 3535 Z‐DNA (nonPZFS)‐IP⁺ (representing Z22‐enriched DNA devoid of typical PZFSs) (Figure 1b). As anticipated, the normalized read counts of Z‐DNA(PZFS)‐IP⁺ are remarkably enriched from the start to the end point of PZFSs as compared to Z‐DNA (PZFS)‐IP⁻ (Figure S5a). As illustrated in Figure 1c, the normalized ZIP‐seq read counts are predominantly distributed at around the transcription start sites (TSSs), and a small subpeak observed at the downstream of the transcription terminate sites (TTSs). This distribution is similar to the distribution profile of PZFSs (Figure S5b). Also, ZIP‐qPCR result shows that three positive loci corresponding to Z‐DNA peaks are more enriched as compared to a negative locus without Z‐DNA peaks (Figure 1d). After calculating the (GC)_n and (AT)_n content within the Z‐DNA peaks, we found that Z‐DNA (PZFS)‐IP⁺ exhibits a higher number of (GC)_n repeats but a lower number of (AT)_n repeats in comparison to Z‐DNA (nonPZFS)‐IP⁺ (Figure 1e, left). To confirm the folding potentials of Z‐DNA from both repeats, we studied synthetic (GC)_n and (AT)_n oligonucleotides using circular dichroism (CD). Our findings reveal that (GC)_n exhibits lower negative ellipticities at 295 nm compared to (AT)_n repeats. Additionally, the ellipticities of (GC)_n repeats were found to be anti‐correlated with copy numbers (Figure 1e, right), which is consistent with previous research indicating that (GC)_n has a higher propensity for Z‐DNA formation than (AT)_n (Wang et al., 1979). Oligos from Z‐DNA (nonPZFS)‐IP⁺ (sample 1 and 2 in Figure 1e, right) exhibit significantly lower negative ellipticities at 295 nm as compared to other oligos from Z‐DNA (PZFS)‐IP⁺ (Figure 1e, right, partially enlarged view). A representative Integrative Genomics Viewer (IGV) analysis spanning 23 kb on the Chr. 3 illustrates the biological reproducibility of Z‐DNA peaks between two biological replicates (Figure 1f).

Figure 1.

Global identification of Z‐DNA in the rice genome using ZIP‐seq. (a) Overlaps of Z‐DNA peaks between biological replicates. (b) Overlapping ratio between Z‐DNA‐IP⁺ and putative Z‐DNA forming sequences (PZFSs). (c) Distributions of Z‐DNA‐IP⁺ within genic regions. Normalized read counts of Z‐DNA‐IP⁺ were plotted across ±2.0 kb from the transcription start sites (TSSs) to the transcription terminate sites (TTSs) of genes. (d) ZIP‐qPCR for validation of 3 Z‐DNA peaks, ZIP‐1 and 2 for Z‐DNA (PZFS)‐IP⁺ and ZIP‐3 for Z‐DNA (nonPZFS)‐IP⁺. (e) CD assays of randomly selected 7 DNA oligoes corresponding to Z‐DNA‐IP⁺, 1 and 2 were from Z‐DNA (nonPZFS)‐IP⁺. (f) A representative IGV across 23 kb from the Chr.3 illustrating biological reproducibility of Z‐DNA‐IP⁺. Significance test was determined using t‐test **: P < 0.01.

Therefore, all above results represent that ZIP‐seq is a reliable method for characterizing Z‐DNA across the rice genome.

Genomic features of Z‐DNA

We next examined genomic distributions of Z‐DNA within the seven functionally annotated subgenomic regions, including 5′UTRs, promoters, exons, introns, 3′UTRs, downstream and intergenic regions. Z‐DNA (PZFS)‐IP⁺ is more enriched in exons, 5′UTRs and distal intergenic regions, but less enriched in introns, 3′ UTRs and promoters as compared to Z‐DNA (PZFS)‐IP⁻ (Figure 2a). Overall, the length of Z‐DNA (PZFS)‐IP⁺ is slightly shorter than that of Z‐DNA (PZFS)‐IP⁻ (Figure S6). We then calculated the GC content across ±1.0 kb from the start to the end point of PZFSs associated with Z‐DNA (PZFS)‐IP^+/−, and found that Z‐DNA (PZFS)‐IP⁺ exhibits a higher GC content than Z‐DNA (PZFS)‐IP⁻ (Figure 2b). After assessing the GC and AT skew, Z‐DNA(PZFS)‐IP^+/− does not show a clear GC or AT skew but displays a clear symmetric oscillation within the peak region. This observation suggests the potential presence of alternating purine‐pyrimidine nucleotides, such as GCGC or ATAT, within the PZFS regions (Figure 2c). As demonstrated in Figure 2d, Z‐DNA (PZFS)‐IP⁺ displays a higher frequency of SS (G/C), but a reduced occurrence of WW (A/T) dinucleotide sequences near the centre of PZFSs. Additionally, there is an approximately 200 bp positional shift in the summit of SS or WW when compared to Z‐DNA (PZFS)‐IP⁻.

Figure 2.

Genomic distributions and sequence features of Z‐DNA. (a) Genomic distributions of Z‐DNA‐IP⁺ across the rice genome. The whole genome was partitioned into 7 functionally annotated subdomains, including promoters, 5′UTRs, exons, introns, 3′UTRs, downstream and intergenic regions. Promoters refer to the 1 kb upstream of the TSSs. Downstream regions refer to 3 kb downstream of the TTSs. Distal intergenic regions refer to the regions away from the 1 kb upstream of the TSSs of one gene and 3 kb downstream of the TTSs of the nearest neighbouring gene. (b) GC content of Z‐DNA‐IP^+/− and random control. (c) GC (left) and AT (right) skew of Z‐DNA‐IP^+/− and random control. (d) Distributions of WW (A/T) or SS (G/C) nucleotides across ±1.0 bp around the centre of Z‐DNA‐IP^+/−. (e) Distributions of dinucleotides across ±500 bp around the centre of Z‐DNA‐IP⁺ (left), Z‐DNA‐IP⁻ (right). (f) Enrichment of motifs for potential binding of TFs associated with Z‐DNA‐IP⁺ (left) and Z‐DNA‐IP⁻ (right). Significance test was determined using t‐test **: P < 0.01.

We also conducted an analysis of the frequency of dinucleotide sequences within a ± 500 bp region surrounding the centre of Z‐DNA (PZFS)‐IP^+/−, and observed nuanced variations in dinucleotide frequency between Z‐DNA (PZFS)‐IP⁺ and Z‐DNA (PZFS)‐IP⁻. Specifically, Z‐DNA (PZFS)‐IP⁺ exhibits increased prevalence of CG/GC, while a decreased frequency of CC/GG and AA/TT within the ±100 bp vicinity of the centre for Z‐DNA (PZFS)‐IP⁺ in comparison to Z‐DNA (PZFS)‐IP⁻ (Figure 2e). We conducted similar analyses on alternation of purine/pyrimidine sequences within ±500 bp of the centre of Z‐DNA (PZFS)‐IP⁺ and found a greater enrichment of C/G^+/−, but a lesser enrichment of A/T^+/− within ±150 bp of the centre (Figure S7). We also plotted normalized PZFS‐IP^+/− read counts across ±2.0 kb from the centre of d(GC)_n, d(TG)_n and d(TA)_n repeats. Our analysis reveals that d(GC)_n and d(TA)_n display the highest and lowest PZFS‐IP⁺ read intensity, respectively (Figure S8), which is consistent with previous findings (Wang and Vasquez, 2007). To investigate whether Z‐DNA can serve as a platform for the recruitment of trans‐factors, we performed de novo motif identification using sequences from Z‐DNA (PZFS)‐IP^+/− and discovered that AP2/EREBP and BZR TF putative binding sites are common for Z‐DNA (PZFS)‐IP^+/−. In contrast, bZIP, Trihelix and TCP TF‐sites are remarkably enriched in Z‐DNA (PZFS)‐IP⁺, while REM, MYB and NLP sites are more abundant in Z‐DNA (PZFS)‐IP⁻ (Figure 2f). We also investigated Z‐DNA‐IP⁺ but without overlapping typical PZFSs, referred to as Z‐DNA (nonPZFS)‐IP⁺. Our analysis reveals that these sequences have lower GC content (Figure S9a) and fewer SS dinucleotide sequences but a higher prevalence of WW dinucleotide sequences (Figure S9b), and a similar trend of dinucleotide frequencies as compared to Z‐DNA (PZFS)‐IP⁺ (Figure S9c). In contrast, Z‐DNA (nonPZFS)‐IP⁺ exhibits a clear GC skew and a similar AT skew to Z‐DNA (PZFS)‐IP⁺ (Figure S10). These findings indicate that Z‐DNA (PZFS)‐IP⁺ loci possess unique intrinsic DNA sequences and genomic characteristics as compared to Z‐DNA (PZFS)‐IP⁻.

Z‐DNA distribution correlates with gene transcriptional activity

Z‐DNA has been shown to either promote or repress the transcription of a subset of genes (Ray et al., 2011; Wang and Vasquez, 2007). To evaluate the impact of Z‐DNA on gene transcription on a genome‐wide scale, we compared the expression levels of genes overlapping Z‐DNA (PZFS)‐IP^+/− or without PZFSs. Our analysis reveals that genes with Z‐DNA (PZFS)‐IP⁺ have the highest expression levels (Figure S11a), suggesting that Z‐DNA may facilitate gene transcription. Furthermore, genes containing one or two PZFSs are more expressed than those with more than 2 PZFSs (Figure S11b). We also compared the expression levels of genes with Z‐DNA (PZFS)‐IP^+/− located in different subgenomic regions. Genes with Z‐DNA (PZFS)‐IP⁺ in exons and downstream regions exhibit higher or lower expression levels, respectively, as compared to these with the corresponding Z‐DNA (PZFS)‐IP⁻(Figure S11c). We subsequently plotted the normalized Z‐DNA (PZFS)‐IP⁺ read counts across ±2.0 kb from the TSSs to the TTSs of genes with different expression levels (high, medium and low, as measured by FPKM). Our analysis reveals that the read intensity of Z‐DNA (PZFS)‐IP⁺ exhibits a direct correspondence with transcription levels of genes at around the TSSs, while a negative association with gene transcription levels occurs within the gene bodies. This finding indicates its dual (active and repressive) roles of Z‐DNA (PZFS)‐IP⁺ in the regulation of gene expression (Figure 3a).

Figure 3.

Correlation of Z‐DNA with gene transcription. (a) Curve plot showing an association between normalized read density of Z‐DNA‐IP⁺ and genes with different expression levels (FPKM values, high, medium, low). (b) Curve plot showing distributions of normalized read counts from Z‐DNA IP‐seq and RNA Pol II ChIP‐seq across ±2.0 kb from the TSS to the TTS of overlapping genes. (c) Curve plot showing distributions of normalized read counts from Z‐DNA IP‐seq and RNA Pol II ChIP‐seq located at 5′UTRs at around ±2.0 kb of the centre. (d) CUT&RUN‐qPCR assay for detecting changes of Z‐DNA abundance in stress‐responsive genes under salt conditions. Significance test was determined using t‐test. **: P < 0.01. (e) CUT&RUN‐qPCR assay for detecting changes of Z‐DNA abundance in sodium‐potassium transporter genes (OsHKT) under salt conditions. Significance test was determined using t‐test. *: 0.01 < P < 0.05; **: 0.01 > P > 0.001; ***: P < 0.001. (f) Expression levels of genes with in vivo Z‐DNA, without Z‐DNA and random control. (g) Curve plot showing an association between normalized read density of in vivo Z‐DNA and genes with different expression levels (FPKM values, high, medium, low). (h) Carbon dots (CDs) mediated transient transformation validated the enhancer activity of four Z‐DNA regions. Significance test was determined using Wilcoxon test, *: 0.05 > P > 0.01; **: 0.01 > P > 0.001; ***: P < 0.001.

Subsequently, we then plotted the normalized read counts from ZIP‐seq and RNA polymerase II (RNAPII, Pol II) ChIP‐seq (GSE142570) (Zhao et al., 2020) across ±2.0 kb from the TSSs to the TTSs of genes. This analysis shows that RNAPII reads exhibits a near co‐localization with Z‐DNA at the downstream of the TSSs, and a pronounced peak prior to the subpeak of Z‐DNA at the TTSs, (Figure 3b). This suggests that Z‐DNA may play a role in both transcriptional initiation and termination rather than transcriptional elongation. To further investigate this, we observed that Z‐DNA exhibits an overall anti‐correlation with the enrichment levels of RNAPII in all genomic regions and TEs (Figure S12). In contrast, Z‐DNA located at 5′UTRs exhibits a positive correlation with enrichment levels of RNAPII (Figure 3c). These findings support both the promotional and repressive roles of Z‐DNA in the regulation of gene transcription, which is dependent on the specific genomic region. To further investigate the biology of Z‐DNA, we constructed regulatory networks for OsBZR1‐4 and AP2/EREBP, which mediate brassinosteroid signalling in rice (Bai et al., 2007), containing motifs shared within Z‐DNA‐IP^+/−, and bZIP overrepresented in Z‐DNA‐IP⁺ as illustrated in Figure 2f (Figure S13, Table S2). We found that OsBZR1‐4 target genes have GO terms associated with the regulation of metabolic processes, gene expression and biological processes (Figure S14a), by contrast, bZIP target genes have GO terms related to membrane and kinase activities (Figure S14b). Also, changes of Z‐DNA abundance are positively correlated with the expression of several stress‐responsive genes under salt conditions, including OsERF096, OsDREBIE, OsSERF1 and OsJAZ9 (Figure 3d). Conversely, Z‐DNA negatively regulated the expression of three sodium‐potassium transporter genes, OsHKT (Figure 3e). These findings suggest potential roles of Z‐DNA in stress responses, particularly in relation to salt stress.

To interrogate whether Z‐DNA in vivo has similar functions in modulating gene transcription, we generated two well‐correlated biological replicates of Z22‐based CUT&TAG data sets (Figure S15), resulting in 3214 in vivo Z‐DNA peaks from the merged data. As expected, genes overlapping with in vivo Z‐DNA have higher expression levels than those without Z‐DNA and the random control (Figure 3f). Additionally, consistent with Z‐DNA (PZFS)‐IP⁺, in vivo Z‐DNA located at around the TSSs or within gene bodies show a positive or negative correlation with gene expression levels, respectively (Figure 3g). To further evaluate whether Z‐DNA loci can function as enhancers to modulate gene expression, we randomly selected four Z‐DNA loci for enhancer validation. Our findings indicate that each locus can indeed act as an enhancer to significantly increase expression of the reporter gene‐GFP as compared to the negative control with mini35S (Figure 3h).

Collectively, these results indicate that Z‐DNA exhibits a genomic region‐dependent role in the regulation of gene expression. Briefly, promoter Z‐DNA positively correlates with gene expression while gene body Z‐DNA anti‐correlates with gene expression.

Epigenomic features of Z‐DNA

Z‐DNA is flanked by two B‐to‐Z (or B‐Z) junctions with an unwinding helix, implying the presence of single‐stranded DNA (ssDNA) near Z‐DNA sites (Subramani et al., 2019; Yi et al., 2022). To evaluate whether Z‐DNA is partly associated with other non‐B DNA structures that may arise from ssDNA, we plotted normalized read counts of DNA G‐quadruplexes (G4s), and i‐motifs (iMs) across ±2.0 kb from the start to the end point of PZFSs within Z‐DNA (PZFS)‐IP^+/−. We found that PZFSs of Z‐DNA (PZFS)‐IP⁺ have overall higher enrichment of G4s and iMs as compared to Z‐DNA (PZFS)‐IP⁻ and random controls (Figure 4a,b). Interestingly, we observed that G4 and iM reads are more enriched in the flanking regions of PZFSs, which correspond to B‐Z junction regions (Figure 4a,b). To go a step further, we plotted the normalized read counts of Z‐DNA (PZFS)‐IP⁺ across ±2.0 kb from the start to the end point of PZFS within Z‐DNA (PZFS)‐IP⁺ overlapping with or without G4s or iMs in the B‐Z junction regions. We found that Z‐DNA (PZFS)‐IP⁺ with G4s or iMs has higher read abundance than the corresponding ones without G4s or iMs (Figure S16). This finding suggests that the presence of G4s or iMs may facilitate or promote the formation of Z‐DNA.

Figure 4.

Epigenomic features of Z‐DNA. (a) Curve plot showing distributions of normalized G4DP‐seq read counts across ±2.0 kb from the start to the end point of Z‐DNA‐IP^+/− and random control. (b) Curve plot showing distributions of normalized iM‐IP‐seq read counts across ±2.0 kb from the start to the end point of Z‐DNA‐IP^+/− and random control. (c) DNA methylation levels of CG (left), CHG (middle) and CHH (right) across ±2.0 kb from the start to the end point of Z‐DNA‐IP^+/− and random control (top). Dot blotting assays of Z‐DNA antibodies were conducted under control (CK) and high/low methylation treatments(left bottom panel), and ZIP‐qPCR analyses were performed on six Z‐DNA peaks corresponding to these treatments(right bottom panel). Significance test was determined using t‐test. *: 0.01 < P < 0.05; **: 0.01 > P > 0.001; ***: P < 0.001. (d) Normalized DNA 6mA IP‐seq read counts distributed across ±2.0 kb from the start to the end point of Z‐DNA‐IP^+/− and random control(left). Dot blotting assays of NIP (WT) and two Cas9‐edited transgenic lines with DNA 6mA modifier genes, LOC_Os01g16180 and LOC_Os10g31030 (right). (e) Normalized DNA 4acC IP‐seq read counts distributed across ±2.0 kb from the start to the end point of Z‐DNA‐IP^+/− and random control. (f) Curve plot showing distributions of normalized OxiDIP‐seq read counts across ±2.0 kb from the start to the end point of Z‐DNA‐IP^+/− and random control. (g) Curve plot showing distributions of normalized 8‐oxodG read counts across ±2.0 kb from the start to the end point of Z‐DNA‐IP⁺ with different occupancy (high, middle and low).

It has been reported that DNA methylation can promote the B‐Z transition, thereby facilitating the formation of Z‐DNA (Temiz et al., 2012; Zacharias et al., 1988, 1990). We thus assessed epigenetic features of Z‐DNA loci on a genome‐wide scale by measuring methylation levels of CG, CHG and CHH within ±2.0 kb from the start to the end point of Z‐DNA (PZFS)‐IP^+/− and random controls. We found that Z‐DNA (PZFS)‐IP⁺ regions have higher CG, CHG and CHH methylation levels than Z‐DNA (PZFS)‐IP⁻ regions (Figure 4c). Z‐DNA (PZFS)‐IP^+/− regions exhibit lower levels of methylation at CG and CHG sites, but slightly higher CHH methylation in Z‐DNA (PZFS)‐IP⁺. In contrast, the flanking regions of Z‐DNA (PZFS)‐IP⁺, corresponding to the B‐Z junctions, display higher levels of CHG and CHH methylation as compared to random controls (Figure 4c, upper panel). We conducted similar analyses for DNA 6mA and 4acC and found that the Z‐DNA (PZFS)‐IP^+/− regions exhibit higher levels of DNA‐6mA but lower levels of 4acC enrichment as compared to the random control. Additionally, the Z‐DNA (PZFS)‐IP⁺ regions show greater DNA‐6mA read abundance than the Z‐DNA (PZFS)‐IP⁻ regions (Figure 4d left, 4e). Furthermore, the flanking regions of Z‐DNA (PZFS)‐IP^+/−, corresponding to the B‐Z junctions, demonstrate the highest enrichment levels of both DNA‐6mA and 4acC (Figure 4d,e). In contrast, Z‐DNA (nonPZFS)‐IP⁺ displays lower enrichment levels of methylated CHH, DNA‐6mA and 4acC, as well as iMs and G4s, while exhibits higher levels of methylated CG and CHG as compared to Z‐DNA (PZFS)‐IP⁺ (Figure S17a,b). There is a strong association between DNA modifications and the formation of Z‐DNA. For instance, Z‐DNA (PZFS)‐IP⁺ overlapping DNA‐6mA or DNA‐4acC exhibits lower levels of read density enrichment than the corresponding ones without these modifications (Figure S18). Additionally, Z‐DNA (PZFS)‐IP⁺ shows a positive correlation with CG and CHH methylation levels (Figure S19a), while a negative correlation with enrichment levels of DNA‐6mA or ‐4acC (Figure S19b,c). These results suggest that Z‐DNA exhibits an overall negative association with the enrichment levels of both DNA‐6mA or DNA‐4acC, while showing a positive correspondence with CHH methylation levels. This highlights the distinct roles of DNA base chemical modifications in modulating Z‐DNA formation. Furthermore, we conducted ZIP‐qPCR for six Z‐DNA peaks using DNA with hyper/hypo‐DNA‐5mC (Figure 4c, left bottom panel) as previously described (Feng et al., 2024) and found that five peaks exhibit hypo‐DNA methylation, which promotes Z‐DNA formation while hyper‐DNA methylation disfavours Z‐DNA formation (Figure 4c, right bottom panel). Notably, only one peak displays an opposite trend regarding the association between hyper/hypo‐DNA methylation and Z‐DNA formation. Also, we conducted dot blotting assays by using DNA from NIP (WT) and two transgenic lines with Cas9‐edited loss‐of‐function mutation in two putative rice 6mA modifier genes, LOC_Os01g16180 and LOC_Os10g31030. Our results indicate that the two transgenic lines exhibit a global decrease in DNA‐6mA levels, accompanied by a global increase in Z‐DNA immune‐signal (Figure 4d right). This further suggests a negative association between DNA‐6mA and Z‐DNA formation on a genome‐wide scale (Figure S19b). Also, we associated Z‐DNA with 8‐Oxo‐7,8‐Dihydro‐2′‐Deoxyguanosine (8‐oxodG), which was identified by using OxiDIP‐seq (anti–8‐oxodG antibodies based immunoprecipitation coupled with high‐throughput sequencing) as previously described (Gorini et al., 2020) and found that overall Z‐DNA (PZFS)‐IP⁺ has a higher density of 8‐oxodG than Z‐DNA (PZFS)‐IP⁻, particularly around the B‐Z junction regions (Figure 4f). A similar trend was observed for Z‐DNA in vivo (Figure S20). Furthermore, 8‐oxodG exhibits an overall positive association with Z‐DNA formation, indicating that the presence of 8‐oxodG promotes Z‐DNA formation. In addition, we associated Z‐DNA(PZFS)‐IP^+/− with 6 histone marks, including 5 active marks H3K4/36me3, H3K9/27 ac, H4K12ac and 1 repressive mark H3K27me3. We found that genomic regions with Z‐DNA(PZFS)‐IP⁺ exhibit more enrichment of H3K4me3 and H3K36me3, but less enrichment of H3K27me3, and H3K9/27ac, H4K12ac as compared to genomic regions with Z‐DNA(PZFS)‐IP⁻ (Figure S21).

Taken together, all above results show that Z‐DNA possesses intrinsic epigenetic characteristics, such as its coexistence with quadruplex DNA structures and the enrichment of methylated CHH, DNA‐6mA/4acC and 8‐oxodG, as well as distinct histone marks (H3K4me3 and H3K36me3).

Potential involvement of Z‐DNA in gene coevolution and domestication

To evaluate the potential link between Z‐DNA and gene coevolution and domestication, we calculated the ka/ks values for genes associated with Z‐DNA in vivo, Z‐DNA (PZFS)‐IP^+/− and a random control. Our analysis reveals that genes with Z‐DNA‐IP⁺ and Z‐DNA in vivo display the lowest and the second lowest ka/ks value, respectively (Figure 5a, Figure S22). This finding suggests that both types of genes experience stronger purifying selection pressure as compared to genes with Z‐DNA‐IP⁻. Consistently, we observed that Z‐DNA‐IP⁺ (r = 0.78) or Z‐DNA in vivo (r = 0.81) in promoters indeed exhibits a higher correlation with overlapping genes than Z‐DNA‐IP⁻ (r = 0.68) (Figure 5b). After examining the putative selected genes (PSGs) characterized previously (Jing et al., 2023), we found that Z‐DNA‐IP⁺ (r = 0.81) in promoters exhibits a stronger correlation with overlapping PSGs than Z‐DNA‐IP⁻ (r = 0.65) (Figure 5c). Additionally, PSGs associated with Z‐DNA‐IP⁺ demonstrate a lower ka/ks ratio than those linked to Z‐DNA‐IP⁻ (Figure S23), suggesting that the former experiences a higher degree of negative selection pressure than the latter. These results indicate that Z‐DNA‐IP⁺ or Z‐DNA in vivo in promoters exhibits a more pronounced co‐evolutionary trend with overlapping genes than Z‐DNA‐IP⁻.

Figure 5.

Characterization of Z‐DNA in gene coevolution and domestication. (a) ka/ks values calculated for genes with Z‐DNA in vivo, Z‐DNA (PZFS)‐IP^+/− and random control. (b) Scatter plot showing correlation of Z‐DNA in vivo, Z‐DNA (PZFS)‐IP^+/− in promoters with overlapping genes. (c) Scatter plot showing correlation of Z‐DNA (PZFS)‐IP^+/− in promoters with overlapping PSG genes. (d) ka/ks values calculated for genes associated with Z‐DNA (PZFS)‐IP⁺ with or without 8‐oxodG modifications and random control. (e) Scatter plot showing correlation of Z‐DNA (PZFS)‐IP^+/− with or without 8‐oxodG modifications in promoters with overlapping genes. Significance test was determined using Wilcoxon test, ***: P < 0.001.

We aimed to interrogate whether Z‐DNA with or without 8‐oxodG exhibits distinct associations with gene coevolution and domestication. We also calculated the ka/ks value for genes overlapping with Z‐DNA with or without 8‐oxodG, and observed that genes containing Z‐DNA coexisting with 8‐oxodG have a lower ka/ks value than genes associated with Z‐DNA but lack of 8‐oxodG (Figure 5d). This finding suggests that genes with Z‐DNA coexisting with 8‐oxodG experience stronger negative selection pressure. Also, we conducted correlation analyses for promoter Z‐DNA with or without 8‐oxodG with the corresponding genes and found that Z‐DNA without 8‐oxodG exhibits a higher correlation with overlapping genes than Z‐DNA with 8‐oxodG (Figure 5e). This indicates that the presence of 8‐oxodG in Z‐DNA tends to promote gene domestication but does not facilitate gene coevolution.

Collectively, these results show that Z‐DNA may have significant implications for the regulation of gene domestication by influencing the coevolution of genes.

Discussion

We here for the first time report the application of Z22‐based IP‐seq and CUT&TAG techniques for the characterization of in vitro and in vivo Z‐DNA in the rice genome. It has been documented that R‐loop profiling is methodology‐dependent, including CUT&TAG with 2XHBD (the N‐terminal hybrid‐binding domain of RNase H1), R‐ChIP, MapR and DRIP‐seq, across the human genome (Wang et al., 2021). This is largely caused by the different binding specificity and efficiency to DNA–RNA hybrids (R‐loops) among the S9.6 antibody, engineered RNaseH1 devoid of the catalytic activity and engineered 2XHBD. Our results show that Z22‐based IP‐seq and CUT&TAG can produce a similar genomic profiling of Z‐DNA using the same antibody but with different experimental strategies. This suggests that the intrinsic nature of non‐B DNA recognition sensors may paly determinant roles in depicting global non‐B DNA landscapes across the genome.

This research could represent significant advancement in our understanding of Z‐DNA biology in plants and may potentially be extended to non‐plant systems, such as humans. Our findings indicate that PZFSs‐IP⁺ exhibits distinct sequence features, characterized by high G/C but low A/T content, as well as unique genomic distributions. Specifically, these features are more prevalent in exons, 5′UTRs and distal intergenic regions, while being less in promoters, introns and 3′UTRs, as compared to PZFSs‐IP⁻. Additionally, we found that d(CG)_n repeats facilitate the in vitro formation of Z‐DNA in rice, demonstrating a similar trend in Z‐DNA formation propensity as previously reported (Ho et al., 1986; McLean et al., 1986; Rich et al., 1984; Rich and Zhang, 2003). This phenomenon is partly attributed to the lower energy requirement for Z‐DNA folding from CG repeats as compared to other type of repeats (Peticolas et al., 1988). In addition, we showed that DNA modifications (such as 5mC, 6mA, 4acC and 8‐oxodG) and the presence of other non‐B DNA structures (such as G4s and iMs), in the B‐Z junction regions, may influence the formation and stability of Z‐DNA. These findings align well with previous literatures, which show that DNA 5mC and histone acetylation facilitate the B‐Z transition both in vitro and in vivo (Temiz et al., 2012; Zacharias et al., 1990; Zhang et al., 2016). Furthermore, it has been shown that DNA 5mC stabilizes Z‐DNA formation under physiological conditions in human cells (Lander et al., 2001) by modifying helix pitch, free energy and helical unwinding (Gruenbaum et al., 1982; Zacharias et al., 1988). DNA methylation is indeed recognized for reducing the energy required for conformational changes, thereby promoting the B‐Z conversion (Temiz et al., 2012). It is worth noting that qPCR results for DNA with changes in total DNA‐5mC or 6mA indicate that both DNA modifications exhibit positive and negative impacts on Z‐DNA formation. This suggests a complex relationship between these of DNA modifications and Z‐DNA formation, which may depend on the extent of original DNA methylation levels or changes in methylation. This trend is a similar to the trend for iMs we discussed before (Feng et al., 2024). The presence of 8‐methyl‐2′‐deoxyguanosine (m(8)G) within PZFSs can also promote the formation of Z‐DNA as well (Vongsutilers and Gannett, 2018; Wang et al., 2015; Xu et al., 2003). Additionally, the formation of vicinal G4s or iMs may facilitate negative supercoiling and/or promote helix unwinding at the B‐Z junctions (Peck and Wang, 1983; Wittig et al., 1991; Zacharias et al., 1988). Notably, approximately 29% of Z‐DNA sites are not associated with typical PZFS motifs, which we refer to as Z‐DNA (nonPZFS)‐IP⁺. The reduced potential for Z‐DNA folding was validated by CD spectroscopy and ZIP‐qPCR. This observation may stem from the fact that the sequences of Z‐DNA (PZFS)‐IP⁺ are richer in (GC)_n repeats but poorer in (TG)_n repeats. The (GC)_n repeats are more stable than(TG)_n repeats for Z‐DNA formation, as (GC)_n has a lower free energy for the conformational transition from B‐DNA to Z‐DNA as compared to (TG)_n (Pohl, 1967; Wang et al., 1979). Our study also provides evidence indicating that Z‐DNA (nonPZFS)‐IP⁺ exhibits lower potentials for Z‐DNA formation or stability as compared to Z‐DNA (PZFS)‐IP⁺ (Figure S24). Moreover, the core sequences of the Z‐DNA (PZFS)‐IP⁺ sequences consist of 1 bp‐sized alternating purine‐pyrimidine nucleotides (CG or AT), whereas those of Z‐DNA (nonPZFS)‐IP⁺ feature 5 bp‐sized arrangements (Figure S25). These results indicate that Z‐DNA (nonPZFS)‐IP⁺ sequences exhibit greater variability, which prevents them from perfectly aligning with the PZFS predictions.

In agreement with their distributions around the TSSs, Z‐DNA (PZFS)‐IP⁺ acts as an active genomic regulator, facilitating expression of overlapping genes. This is further supported by the positive association observed between Z‐DNA (PZFS)‐IP⁺ at TSSs and gene expression. Conversely, those located within gene bodies are anti‐correlated with gene expression, likely behaving as a roadblock to gene expression. This suggests that Z‐DNA may play either promotional or repressive roles in modulating the expression of genes with functions related to salt responses. We also demonstrated that Z‐DNA sites can act as hubs for the recruitment of TFs or other trans‐factors, thereby serving as cis‐elements like enhancers that regulate the expression of a subset of genes through intricate networks. Furthermore, it has been reported that Z‐DNA can act as both active and repressive regulatory elements involved in the regulation of gene transcription (Cerna et al., 2004; Wang and Vasquez, 2007). For instance, Z‐DNA can act as a negative regulatory element (NRE) for repression of human ADAM‐12 transcription (Ray et al., 2011), as it is located within gene bodies, thereby acting as a roadblock to the movement of RNA polymerase (Peck and Wang, 1985). This assertion is further substantiated here by the observed anti‐correlation between Z‐DNA read abundance and the enrichment levels of RNA PII reads. Of note, the roles of Z‐DNA in the regulation of gene transcription can also be mediated by other trans‐factors such as Z‐DNA binding protein (ZBP) as well as chromatin states, including chromatin remodelling, nucleosome replacement and histone acetylation (Garner and Felsenfeld, 1987; Jiao et al., 2022; Liu et al., 2006; Wong et al., 2007; Zhang et al., 2016). PZFSs have been found to be associated with active gene transcription in the human genome (Shin et al., 2016).

Z‐DNA has been suspected to serve as hot spots for oxidation due to its intrinsic G‐rich context. Additionally, 8‐oxodG containing Z‐DNA potentially acts as a transcriptional switch for modulating the transcription of a subset of genes (Fleming et al., 2019). However, the association of 8‐oxodG with Z‐DNA formation and subsequent biological implications remain largely uncharacterized on a genome‐wide scale. Here, we for the first time provide evidence showing that Z‐DNA‐IP⁺ exhibits a higher 8‐oxodG read density than Z‐DNA‐IP⁻. Furthermore, 8‐oxodG appears to play a promotional role in Z‐DNA formation. It has been reported that the presence of 8‐oxodG in a specific Z‐DNA site favours Z‐DNA formation in vitro (Fleming et al., 2019). Notably, our study highlights the roles of Z‐DNA in gene domestication and coevolution, further supporting previous reports that non‐B DNA structures serve as a driving force in genome evolution (Makova and Weissensteiner, 2023). In particular, the presence of 8‐oxodG can facilitate gene domestication. This phenomenon is partly attributed to 8‐oxodG's ability that destabilizes the genome and epigenome, which includes G‐C to T‐A base transversions, destabilization of double‐stranded DNA, DNA methylation and histone modifications (Gruber et al., 2018; Hahm et al., 2022; Plitta‐Michalak et al., 2022).

Conclusions

Our study for the first time conducts global profiling and functional characterization of Z‐DNA in rice by using Z22 antibody‐based ZIP‐seq and CUT&TAG techniques. The findings indicate that the formation and stability of Z‐DNA is determined by intrinsic DNA sequences alone or along with base modifications and/or other non‐B DNA structures. Importantly, Z‐DNA play a significant role in regulating gene expression and gene evolution. Therefore, our study represents as a pivotal turning point and a solid foundation for reliably launching Z‐DNA study on a genome‐wide scale, thereby advancing the understanding of Z‐DNA biology in both plants and non‐plant systems.

Materials and methods

Plant materials

Rice seeds of Nipponbare (Oryza sativa L., Japonica) and transgenic lines with CRISPR‐Cas9 edited two putative rice 6mA modifier genes, LOC_Os01g16180 and LOC_Os10g31030, were pre‐germinated at room temperature (RT) for 2–3 days. Uniformly germinated seeds were then transferred into soil to continue growing in a greenhouse at 28–30 °C with a 14 h/10 h light–dark cycle. Two‐week old seedlings above the ground were collected and quickly frozen in liquid nitrogen, then kept at −80 °C for storage.

Z‐DNA‐IP‐seq, Z‐DNA CUT&TAG and data analyses

For Z‐DNA‐IP‐seq, purified genomic DNA was subjected to fragmentation, achieving sizes ranging from 100 to 500 bp in sonication buffer (50 mM Tris–HCl, pH 8.0). A total of 15 μg of the fragmented DNA was subsequently diluted and denatured in Z‐DNA reconstitution buffer (20 mM sodium citrate buffer, pH 7.2, 4 m NaCl, 25% ethylene glycol, 250 μM Mn²⁺ and 0.7 m Mg²⁺) at 95 °C for 10 min, then slowly cooled down to RT overnight for DNA re‐association. The re‐associated DNA was then diluted with 400 μL of Z‐DNA‐IP buffer (20 mm sodium citrate, pH 7.2, 0.8 m NaCl, 25% ethylene glycol, 250 μm Mn²⁺ and 0.14 m Mg²⁺), and then incubated with 5 μg of rabbit anti‐Z‐DNA antibody [Z22] (IgG, kappa, Cat# Ab00783‐23.0, Absolute Antibody) overnight at 4 °C. The antibody‐bound DNA was subsequently incubated with 30 μL of washed magnetic protein G beads for an additional 4 h followed by washing three times. The antibody‐bound Z‐DNA was finally recovered using phenol:chloroform extraction and cold ethanol precipitation. Purified IPed Z‐DNA and control DNA from IgG and input were used for library preparation using Hieff NGS Ultima Pro DNA Library Prep Kit for Illumina (Yeasen, Cat.# 12197ES24, Shanghai), which were followed by sequencing on Illumina NovaSeq platform.

For Z‐DNA CUT&TAG, the experiment was conducted according to the instructions provided in the CUT&TAG kit manual (Cat#: TD904‐01, Vazyme). Briefly, cross‐linked rice nuclei were purified and incubated with ConA beads for capture of intact nuclei. The anti‐Z‐DNA antibody (Z22) was added to the cell (nucleus)‐bead complex for incubation at 4 °C overnight. The secondary antibody was introduced for an additional incubation with constant rotation at RT for 60 min. After thoroughly washing, the pA/G‐Tnp Pro was added for incubation with rotation at RT for an additional 1 h. After initiating fragmentation at 37 °C for 60 min followed by adding 2 μL of 10% SDS to stop the reaction by incubating at 55 °C for 10 min. The anti‐Z‐DNA antibody‐enriched Z‐DNA fragments were then recovered for library preparation and sequencing on the Illumina NovaSeq platform.

For data analysis, raw ZIP‐seq data were processed by using Fastp (version 0.21.0) (Chen et al., 2018) to eliminate adapter sequences. All clean reads from two biological replicates were aligned to the MSU v7.0 reference genome for Oryza sativa (http://rice.plantbiology.msu.edu/) by utilizing the Burrows‐Wheeler Aligner (BWA, version 0.7.17) (Li and Durbin, 2009) with default settings of its mem algorithm. Any PCR duplicates were completely removed by using the Picard tool. Reads with a mapping quality score lower than 10 were filtered out by using SAMtools. Spearman rank correlation coefficient was calculated by using the multiBamSummary and plotCorrelation functions from the deepTools suite (Ramirez et al., 2014), which indicates the reproducibility and consistency of biologically replicated data.

For Z‐DNA peak calling, ZIP‐seq data with aligned reads of a minimum length as 50 nucleotides (nts) were selected for peak calling, which was conducted by using MACS2 (version 2.1.1) (Zhang et al., 2008), with the parameters set as “callpeak ‐g 3.8e + 8 ‐p 1e‐4 ‐nomodel ‐f BAMPE”. Input and IgG data were included as background controls to ensure the specificity and accuracy of the peak calling processes.

Prediction of Z‐DNA regions

Z‐DNA regions across the rice genome were predicted using the enhanced software Z‐catcher2 (Li et al., 2009b) with default settings, in conjunction with the gta software (Cer et al., 2013) utilizing the following parameters: ‐skipAPR, ‐skipSTR, ‐skipDR, −skipMR, ‐skipIR and ‐skipGQ.

ZIP‐qPCR and Z‐DNA CUT&RUN‐qPCR assays

Anti‐Z‐DNA antibody‐enriched IP or CUT&RUN DNA was recovered for qPCR assays. The enrichment levels of both types of DNA were calculated using the $2^{\left(\mathit{\Delta \Delta }{C}_{\mathrm{t}}\right)}$ method and expressed as fold changes relative to the corresponding input. Each primer set was repeated three times in each qPCR. All qPCR primer sequences are listed in Table S3. Significance test was determined by using One‐way ANOVA analysis, with significance levels indicate as ***P < 0.001, **P < 0.01 and *P < 0.05.

Hypo/hyper‐DNA‐5mC and low DNA‐6mA related ZIP‐qPCR assays

Zebularine induced globally hypomethylated genomic DNA and M.CviP I‐treated genomic DNA, which exhibited a global elevation of DNA methylation, were prepared as previously described (Feng et al., 2024). Genomic DNA with a global reduction of DNA‐6mA was obtained from transgenic lines with CRISPR‐cas9 edited two putative rice 6mA modifier genes, LOC_Os01g16180 and LOC_Os10g31030 encoding a homologous MTA‐70 domain protein responsible for 6mA modification in A. thaliana. This DNA was utilized for the ZIP‐qPCR assay. All qPCR primers are listed in Table S3.

CD measurements

The synthesized forward and reverse strands of Z‐DNA oligonucleotides were annealed into double‐stranded (DS) DNA in HEPES buffer (10 mm HEPES, 50 mm NaCl, 10 mm MgCl₂, pH = 7.5). To reconstitute Z‐DNA, the dsDNA was denatured in Z‐DNA reconstitution buffer (10 mm HEPES, 4.5 m NaCl, pH = 7.5) at 95 °C for 10 min, after which the temperature was allowed to gradually cool to RT. For circular dichroism (CD) spectroscopy, 10 μm of each reconstituted Z‐DNA in Z‐DNA reconstitution buffer was scanned over a wavelength ranging from 220 to 320 nm, with a 1 nm bandwidth, a 0.5 s response time, and a 1 mm path length on a Chirascan Spectropolarimeter (Applied Photophysics, Surrey, UK). Dry purified nitrogen gas was employed to maintain a deoxygenation atmosphere. After subtracting the CD signal from the background solution, the final CD signal from the Z‐DNA oligos examined was used to indicate the folding potentials of Z‐DNA. All Z‐DNA oligo sequences are listed in Table S4.

Dot blotting assays

The intact purified genomic DNA, the synthesized DNA oligos with (positive control for the formation of Z‐DNA) or without (negative control for lack of Z‐DNA formation) PZFSs, and the synthesized DNA G‐quadruplex (G4) and i‐motif (iM) oligos were denatured and reannealed in a Z‐DNA reconstitution buffer (20 mm sodium citrate buffer, pH 7.2, 4 m NaCl, 25% ethylene glycol, 250 μm Mn²⁺ and 0.7 m Mg²⁺) at 95 °C for 8 min. The denatured and reannealed genomic DNA and oligos were then loaded onto Amersham Hybond‐N⁺ nylon membrane, which were pre‐blocked in 5% milk for 45 min at RT. Following pre‐blocking, the membrane was incubated overnight at 4 °C with the Z22, G4P, iMab, anti‐5mC and DNA‐6mA antibodies in IP buffer. Subsequently, the membrane was incubated with anti‐IgG (HRP) antibody for an additional 1.5 h. The procedures for immune‐signal development were the same as our previous procedures (Fang et al., 2019). Each blot was repeated at least two times. Reconstruction and immune‐detection of DNA G4 and iM were conducted as previously described (Feng et al., 2022; Ma et al., 2022; Zheng et al., 2020).

Carbon dots (CDs) mediated transient transformation for enhancer validation

10 ng/μL of plasmid DNA, specifically a modified pJIT163‐hGFP vector containing mini 35S with or without DNA fragments, were amplified from Z‐DNA peak for enhancer validation, which was mixed with CDs with 25× (200 mg/L) as previously described (Zhang et al., 2023b) and supplemented with 0.6% of Tween‐20 for transient transformation using 10‐day old rice leaves. The aforementioned CD solution was smeared on the surface of rice leaves using a paint brush. A total of 4 applications were made with intervals of 2–3 h between each application. The leaves were then allowed to grow for an additional 24 h in the growth chamber. After this period, CD‐treated leaves were collected for recording the GFP signal using a ZEISS confocal LSM 900. All primers used in this study are listed in Table S3.

Genomic features of Z‐DNA

ChIPseeker (Yu et al., 2015), an R package, was utilized to illustrate genomic distributions of Z‐DNA. The GC content and GC/AT skew were calculated using the following formula: (C+G)/(C+G+A+T) for GC content; ((G‐C)/(G+C)) for GC skew; ((A‐T)/(A+T)) for AT skew.

The density of dinucleotides within a 1‐kb region centred on the peaks was quantified using the ‘annotatePeaks.pl’ sub‐command of the HOMER software (Heinz et al., 2010), with the parameters set as follows: ‐hist 100 ‐di ‐size ‐500 500, while all other parameters remained as their default settings. The TBtools software (Chen et al., 2020) was employed to generate heatmaps for visualization of the resulting data.

Motif prediction

MEME‐ChIP (Machanick and Bailey, 2011) (http://memesuite.org/tools/meme‐chip) was utilized to identify motifs within the Z‐DNA peaks using specific parameters: a minimum motif width of 5 and a maximum width of 20. The Tomtom tool was used to screen the transcription factor (TF) database from Arabidopsis for matching potential TF‐binding sites corresponding to all identified motifs. The top five motifs with the most significant enrichment (E‐values) are presented in the text.

Read count normalization

The ±2.0 kb regions from the transcription start sites (TSSs) to the transcription termination sites (TTSs) of all genes associated with Z‐DNA were divided into 50 bp windows. The number of reads within each sliding window was first normalized by the window length (50 bp) and subsequently divided by the total number of uniquely mapped reads across the genome (Mb).

Construction of transcription factor‐centred regulatory network

The transcription factor‐centred regulatory network was constructed using the RiceTFtarget database (Zhang et al., 2023a) database (https://cbi.njau.edu.cn/RiceTFtarget/).

Analysis of OxiDIP‐seq data

The raw OxiDIP‐seq data were processed for removal of adapter and low‐quality bases using TrimGalore (version 0.4.3; Babraham Bioinformatics, Cambridge, Cambridgeshire, UK). Clean reads were aligned to the rice reference genome (MSU version 7.0) using Bowtie (Langmead et al., 2009). Additional duplicate reads were removed using SAMtools rmdup, version 1.4 (Li et al., 2009a). Peak calling for genomic sites with 8‐oxodG was performed using MACS2 (Zhang et al., 2008) with the following command and parameters: macs2 callpeak ‐‐keep‐dup 1 ‐g 3.8e+8 ‐B ‐‐extsize 200 ‐q 0.01 ‐‐nomodel ‐‐SPMR.

Gene coevolution and domestication analyses

In this study, multiple genome alignments and analyses were performed using the LAST software package (Kiełbasa et al., 2011), with NIP as the reference genome. The alignment results were further processed using last‐split and last‐postmask. The R package ggplot2 was used to plot scatter plots, and the cor.test function was applied to calculate the correlation R values and significance.

Funding

This work was supported by the National Key R&D Program of China (2023YFD1200800). The National Natural Science Foundation of China (U23A20179, 32201782, ZX2400789, 32070561 and U20A2030). Research on Key Technologies for Multi‐dimensional Precision Identification and Germplasm Creation of Salt alkali tolerant and Suitable Crops (2024BBF02002).

Author contributions

W.L.Z. conceived and designed the study. Z.X.H. analysed the data. Y.Y., Y.H.R. and Y.L.F. performed the experiments. A.A. helped with material preparation. A.F. and M.T. assisted with manuscript editing and data interpretation. W.L.Z. wrote the manuscript with contributions from all other authors.

Competing interests

The authors declare that they have no competing interests.

Supporting information

Figure S1 Verification of the anti‐Z‐DNA antibody (Z22) specificity using dot blotting assays.

Figure S2 The diagram showing main procedures of ZIP‐seq for global identification of Z‐DNA.

Figure S3 Correlation analyses of biologically replicated ZIP‐seq data sets.

Figure S4 Overlapping of Z‐DNA peaks between IP relative to IgG or input control in rep 1 or 2, respectively.

Figure S5 Distributions of normalized read counts of Z‐DNA‐IP^+/− across ±2.0 kb from the start to the end point of PZFSs (a) or from the TSSs to the TTSs of genes (b).

Figure S6 The length of Z‐DNA (PZFS)‐IP⁺ and Z‐DNA (PZFS)‐IP⁻.

Figure S7 Distributions of dinucleotides across ±500 bp around the centre of Z‐DNA‐IP⁺.

Figure S8 Distributions of normalized PZFS‐IP^+/− reads across ±2.0 kb of the centre of d(GC)_n, d(TG)_n and d(TA)_n repeats.

Figure S9 Sequence features of Z‐DNA (PZFS or nonPZFS)‐IP⁺.

Figure S10 GC (left) and AT (right) skew for Z‐DNA (PZFS or nonPZFS)‐IP⁺.

Figure S11 Impacts of Z‐DNA on gene expression.

Figure S12 Associations of Z‐DNA with read intensity of Pol II.

Figure S13 TF‐centred gene regulatory network for TFs shared within Z‐DNA‐IP^+/− (left) and overrepresented within Z‐DNA‐IP⁺ (right).

Figure S14 GO term enrichment analyses.

Figure S15 Identification of in vivo Z‐DNA by using CUT&TAG.

Figure S16 Distributions of normalized read counts from Z‐DNA–IP⁺ with or without G4s (left) or i‐motifs (right) across ±2.0 kb from the start and the end point of Z‐DNA.

Figure S17 Epigenetic features of Z‐DNA (PZFS/nonPZFS)–IP⁺.

Figure S18 Distributions of normalized read counts from Z‐DNA–IP⁺ with or without DNA‐6mA or 4acC across ±2.0 kb from the start to the end point of Z‐DNA.

Figure S19 Associations between DNA‐5mC (a), 6mA (b) and 4acC (c) modifications with PZFS‐IP⁺ read abundance.

Figure S20 Distributions of normalized OxiDIP‐seq read counts across ±2.0 kb from the start to the end point of Z‐DNA in vivo and random control.

Figure S21 The distribution of 6 histone marks (H3K4me3, H3K27me3, H3K36me3, H3K9ac, H4K12ac, H3K27ac) across ±2.0 kb from the start to the end point of Z‐DNA‐IP^+/− and random control regions.

Figure S22 Correlation analysis for random control and genes.

Figure S23 Comparison of ka/ks value for PSGs with Z‐DNA‐IP^+/− and random control.

Figure S24 Distributions of normalized read counts from Z‐DNA (PZFS/nonPZFS)–IP⁺ across ±2.0 kb from the start to the end point of Z‐DNA.

Figure S25 Frequency of G/C or A/T nucleotide distributed around the centre of (PZFS/nonPZFS)–IP⁺.

Table S1 Summary of ZIP‐seq data used in this study.

Table S2 Hub genes in transcription factor‐centred gene regulatory networks.

Table S3 Summary of primer sequences used in this study.

Table S4 Z‐DNA oligo sequences used in CD assays in this study.

Data availability statement

All genomic data produced in this study have been deposited in the NCBI Gene Expression Omnibus (GEO; http://www.ncbi.nlm.nih.gov/geo/) under accession number GSE252954, while it remains in private status:opsfkiesxlsxzkj.