Simultaneous profiling of multiple chromatin proteins in the same cells

Abstract

Methods derived from CUT&RUN and CUT&Tag enable genome-wide mapping of the localization of proteins on chromatin from as few as one cell. These and other mapping approaches focus on one protein at a time, preventing direct measurements of co-localization of different chromatin proteins in the same cells and requiring prioritization of targets where samples are limiting. Here, we describe multi-CUT&Tag, an adaptation of CUT&Tag that overcomes these hurdles by using antibody-specific barcodes to simultaneously map multiple proteins in the same cells. Highly specific multi-CUT&Tag maps of histone marks and RNA Polymerase II uncovered sites of co-localization in the same cells, active and repressed genes, and candidate cis-regulatory elements. Single-cell multi-CUT&Tag profiling facilitated identification of distinct cell types from a mixed population and characterization of cell-type-specific chromatin architecture. In sum, multi-CUT&Tag increases the information content per cell of epigenomic maps, facilitating direct analysis of the interplay of different chromatin proteins.

Keywords: H3K27ac; H3K27me3; RNA polymerase; chromatin; histones; profiling; single-cell; transcription.

Figure 1:. Simultaneous profiling of multiple chromatin proteins using multi-CUT&Tag.

A, Diagrammatic representation of multi-CUT&Tag workflow. B, Genomic landscape comparing single Ab CUT&Tag and triple Ab multi-CUT&Tag for H3K27me3, H3K27ac and RNAPII S2P. The shaded regions represent domains with high enrichment of the indicated proteins or histone modifications. C-E, Heatmaps depicting enrichment of reads from single Ab CUT&Tag and triple Ab multi-CUT&Tag over the genomic locations of peaks of published ChIP-seq data specific for H3K27me3 (+/−5kb; GSE123174) (C), H3K27ac (+/−2kb; GSE31039) (D) and RNAPII S2P (+/−2kb; GSE112113) (E). F-H, Average enrichment of reads from triple Ab multi-CUT&Tag over the peak locations (called from published ChIP-seq data) of H3K27me3 (F), H3K27ac (G) and RNAPII S2P (H). See also Figure S1, Figure S2, and Table S1.

Figure 2:. Specificity of multi-CUT&Tag profiles.

A, Correlation matrix of single and triple Ab multi-CUT&Tag maps. Pearson correlations were calculated using the normalized read counts surrounding total H3K27ac, H3K27me3, and RNAPII S2P peaks from multi-CUT&Tag libraries. B-E, Nascent transcription levels, as measured by 4SU-seq (GSE93538) for genes (B, D) or predicted ESC CREs (C, E) separated into quintiles based on read densities from multi-CUT&Tag maps of RNAPII S2P (B, C) or H3K27ac (D, E). Predicted CREs were taken from a previous study (Moorthy et al., 2016). Data re represented as boxplots with median (dark line), 25^th and 75^th percentiles (bottom and top of boxes, respectively), and a maximum of 1.5 times the interquartile range (whiskers) shown. F, Browser tracks of multi-CUT&Tag maps of H3K27ac, RNAPII S2P, and 4SU-seq depicting a region downstream of the Myc gene enriched for predicted cell type-specific enhancers. G-H, Detailed maps of two candidate CREs marked by H3K27ac and RNAPII S2P, that exhibit nascent transcription (4SU-seq) found at many enhancers. ENCODE candidate CREs (cCREs) from E14 ESCs are shown for reference; potential promoter-distal enhancers in this dataset are highlighted in yellow. See also Figure S1 and Figure S2.

Figure 3:. Co-localization of histone modifications and RNAPII directly detected by multi-CUT&Tag.

A, Schematic representation of homogenous and mixed reads derived from multi-CUT&Tag B-D, Genome browser tracks showing enrichment of homogenous reads (with barcodes for a single Ab on both ends) and mixed reads (one barcode specific to each Ab on either end) from dual Ab multi-CUT&Tag profiles of H3K27me3+H3K27ac (B), H3K27me3+RNAPII S2P (C), and H3K27ac+RNAPII S2P (D). Each track is normalized to the same number of reads to allow direct comparison of homogeneous and mixed read profiles E-G, Aggregate enrichment of homogenous reads for each Ab listed over the peaks called from mixed reads in dual Ab multi-CUT&Tag for H3K27me3+H3K27ac (E), H3K27me3+RNAPII S2P (F), and H3K27ac+RNAPII S2P (G). See also Figure S3.

Figure 4:. multi-CUT&Tag profiling in single cells.

A, Diagrammatic representation of scMulti-CUT&Tag approach. B-C, Chromatin landscapes showing comparing bulk CUT&Tag maps with scMulti-CUT&Tag maps, both in aggregate over all single cells and individual cells, at regions of enrichment of H3K27me3 (B) and H3K27ac (C). For H3K27me3, the cells with six or more cut sites within the region were shown and for H3K27ac, the cells with at least one cut site within the region were shown. For both Abs, cells were ordered by read coverage within the regions depicted. D, Violin plots depicting the normalized cut sites for H3K27me3 and H3K27ac within peaks specific for specified genomic regions across single cells. Peaks were called from aggregate single-cell datasets for each Ab, as described in STAR Methods. See also Figure S4.

Figure 5:. Unbiased clustering of integrated scMulti-CUT&Tag profiles uncovers cell type-specific differences in chromatin architecture.

A, Latent semantic indexing (LSI) plots derived from integrated H3K27ac and H3K27me3 scMulti-CUT&Tag profiles, revealing cell type-specific clusters. B-C, LSI projections (B) and pseudo-bulk analysis (C) of H3K27me3 and H3K27ac enrichment surrounding the TSC gene Cdx2 within each cell cluster. D-E, Violin plots showing mean enrichment of H3K27me3 (D) and H3K27ac (E) in ESCs (n=1750 cells) and TSCs (n=199 cells) near genes exhibiting cell type-specific enrichment of one or both chromatin marks, as indicated. Two-sided Wilcoxan Rank-Sum Tests were used to test for significance of differences between cell types. See also Figure S5.