Chromatin immunoprecipitation with sequencing (ChIP-seq) with population cells or tissues yields ensemble epigenomic profiles that only represent the population average, which eliminates cell-to-cell epigenetic heterogeneity. Chromatin immunocleavage with sequencing (ChIC-seq) (Ku et al., 2019) that based on antibody-guided chromatin cleavage under targets is a practical alternative to ChIP-seq. Many single-cell ChIC-based methods, such as scChIC-seq (Ku et al., 2019), CUT&Tag (Kaya-Okur et al., 2019), CoBATCH (Wang et al., 2019), ACT-seq (Carter et al., 2019), and 10× scCUT&Tag (Bartosovic et al., 2021; Wu et al., 2021), have been developed and applied to study cell-type profiling, developmental trajectory, epigenetic heterogeneity, and transcriptional regulation in animals. However, due to the existence of cell walls, it is difficult to obtain single cells in plants. On the other hand, the existing scChIC methods require many single-cell barcoding procedures after tagmentation, which might lead to some DNA leakage and reduced mapping efficiency. Hence, a robust single-cell ChIC-seq method for plant epigenomic research is required emergently. In our previous study, we developed a rapid ChIC-based chromatin profiling protocol, nucleus CUT&Tag (nCUT&Tag) (Ouyang et al., 2021). Here, we combined the nCUT&Tag assay with 10× Single Cell ATAC, developing an easy-to-use single-nucleus CUT&Tag (snCUT&Tag) method in rice.

We reformulated the bulk-cell nCUT&Tag with the droplet-based single-cell barcoding technology to develop snCUT&Tag (Figure 1a). Briefly, ~10⁶ of fixed rice-seedling nuclei were collected and immuno-cleaved with protein G-Tn5 (Vazyme, Nanjing, China, cat no. TD901). Following Tn5-assisted in situ tagmentation, ~20 000 Tn5-indexed nuclei were loaded onto the 10× Genomics Single Cell ATAC microfluidics device, in which each nucleus is partitioned in a gel-bead-in-emulsion droplet and barcoded with unique oligonucleotides (Figure 1a). Finally, the chromatin features of every single nucleus could be captured by pooling and sequencing the barcoded DNA fragments.

We first profiled the H3K4me3 histone mark in seedlings from the Xian group rice (Minghui 63) using snCUT&Tag. We generated over 480 million read pairs, of which 98.7% were indexed with a unique barcode (Figure 1c). After mapping the reads to Minghui 63 reference genome and removal of duplicates, we obtained the H3K4me3 profiles of 3679 single nuclei, with a median of 401 fragments per nucleus (Figure 1c). For the 3679 aggregate single nuclei, we called 28 795 peaks (Figure 1b,c).

Next, we compared the aggregate snCUT&Tag data with bulk-cell nCUT&Tag (Ouyang et al., 2021) and eChIP-seq data (Zhao et al., 2020). The aggregate snCUT&Tag peaks exhibited a high degree of consistency with nCUT&Tag and eChIP-seq data (Figure 1b). Coinciding with the high consistency, global scatterplots displayed significant correlations among the snCUT&Tag, nCUT&Tag, and eChIP-seq libraries (Figure 1d). As demonstrated by few reads falling in non-peak regions (Figure 1b), the snCUT&Tag data exhibited a significantly high fraction of reads in peaks (FRiP) value (median FRiP = 0.923) (Figure 1c,e). snCUT&Tag displayed similar peak signal profiles to nCUT&Tag and eChIP-seq, with H3K4me3 peaks mainly enriching at transcription start sites (Figure 1f). These results indicated that snCUT&Tag is a robust method that can be used to profile single-nucleus chromatin features.

Since the epigenomic data can be used to predict chromatin interactions (Fulco et al., 2019; Ouyang et al., 2020), we examined the H3K4me3 snCUT&Tag data and predicted cell-type specific enhancer–promoter interactions according to the activity-by-contact model (Bartosovic et al., 2021; Fulco et al., 2019). Firstly, we performed dimensionality reduction and clustering according to Bartosovic et al. (2021). The results showed that rice seedlings were partitioned into 17 cell clusters, with significantly high enrichment-signal specificity (Figure 1g). Furthermore, we annotated cell types for the 17 clusters by investigating the marker peak-associated functional genes. For instance, a marker peak in cluster 6 showed specific H3K4me3 enrichment around a malate dehydrogenase-coding gene, which was reported as a marker gene of mesophyll cells in maize (Marand et al., 2021), implying that cluster 6 should be classified as mesophyll cells. Then, we predicted enhancer–promoter interactions. Our results demonstrated that the predicted chromatin interactions were well-overlapped with H3K4me3 ChIA-PET data, and many enhancer-centred chromatin interactions showed cell-type biases that only occurred in specific cell clusters (Figure 1h). Expectedly, ~51.5% (15 881 of 30 839) of predicted enhancers overlapped with ATAC-seq peaks (Figure 1i), showing significant enrichment of ATAC-seq signals around the centre of the predicted enhancers (Figure 1j,k). Our results indicate that snCUT&Tag can be applied to dissect epigenetic heterogeneity and predict cell type-specific enhancer–promoter interactions, which might facilitate the annotation of regulatory elements and reconstruction of 3D genome structures.

In conclusion, we developed a simple and practical single-nucleus ChIC-seq method (snCUT&Tag) for single-cell plant epigenomic studies. We demonstrated that snCUT&Tag could be used to dissect the cell lineage and epigenetic heterogeneity and reconstruct chromatin topology of each single nucleus. Since nCUT&Tag is applicable for studying bulk-cell chromatin features with nuclei isolated from various plant species and tissues (Ouyang et al., 2021), we believe that snCUT&Tag may have a broad spectrum of application in profiling single-nucleus histone marks for other types of plant materials as well. Expectedly, snCUT&Tag provides an avenue for plant single-cell epigenomic studies and may help understand the 3D genomic and epigenetic basis of transcription regulation in a single cell.