henhaobio
Sign in

Nature CommunicationsGenomicsPeer-reviewed journal article

A rapid and robust method for single cell chromatin accessibility profiling

Xi Xian CheniD, Ricardo J. MiragaiaiD, Kedar Nath NatarajaniD, Sarah Amalia TeichmanniD

PDFDOIPreprintCodeData
In-depth full-text digest available · Jump to digest ↓

Abstract

The assay for transposase-accessible chromatin using sequencing (ATAC-seq) is widely used to identify regulatory regions throughout the genome. However, very few studies have been performed at the single cell level (scATAC-seq) due to technical challenges. Here we developed a simple and robust plate-based scATAC-seq method, combining upfront bulk Tn5 tagging with single-nuclei sorting. We demonstrated that our method worked robustly across various systems, including fresh and cryopreserved cells from primary tissues. By profiling over 3,000 splenocytes, we identify distinct immune cell types and reveal cell type-specific regulatory regions and related transcription factors.

In-depth summary

Claude Opus 5.5 · xhigh effort · 8 Nov 2026 · from the full text of version 2 (11 Oct 2018) · CC BY 4.0

Upfront bulk Tn5 tagging then single-nucleus sorting into plates gives scATAC-seq libraries with higher complexity than Fluidigm C1, resolving 12 clusters among 3,385 mouse splenocytes.

Earlier single-cell ATAC-seq methods tagment each cell separately in costly devices (Fluidigm C1, Takara ICELL8) or need large amounts of custom-modified Tn5 for combinatorial indexing. Tagging nuclei in bulk before FACS lets any lab with standard equipment build plate libraries in a few hours, and it allows marker-based sorting of tagged subpopulations.

Key results

  1. ShownIn mouse spleen, the method gave more than 3,000 high-quality single-cell profiles. With no visible batch effect between the two animals, they grouped into B, T and NK subtypes plus small granulocyte, dendritic cell and macrophage clusters.

    3,385 of 3,648 splenocytes (two C57BL/6Jax mice) passed QC (>10,000 reads, >90% mapping); medians: 643,734 reads/cell, 2.1% mitochondrial reads, library size 31,808.5; 78,048 peaks (q<0.01); 12 spectral clusters matched to ImmGen bulk ATAC-seq; ~120,000 reads/cell captures most unique fragments; a 55-cell cluster aggregate resembles bulk. · Abstract; Bioinformatics analysis; Figure legends

  2. ShownFACS after bulk tagging isolates single nuclei with few doublets. On matched cell lines, plate libraries had higher complexity and signal-to-noise than Fluidigm C1 libraries.

    HEK293T/NIH3T3 mix (one 384-well plate): of 307 passing wells, 303 were singlets (167 human, 136 mouse; human:mouse read ratio >10 or <0.1) and 4 were doublets. K562 and mESC run side by side: plate libraries had higher Picard library size and FRiP and comparable or lower mitochondrial fraction than C1; values shown only graphically. · Abstract; Species mixing experiments; scATAC-seq using Fluidigm C1; Figure legends

  3. ShownThe method works on cryopreserved primary cells processed right after thawing, giving libraries with high signal-to-noise from both human and mouse tissues.

    Thawed human and mouse skin fibroblasts (hSF, mSF) and mouse cardiac progenitors (E8.5, E9.5): median unique-fragment library sizes 52,747 (mSF) to 104,608.5 (mCPC_E8.5); mitochondrial DNA <13% in all; median FRiP 0.50 (mSF) to 0.60 (hSF); clear nucleosomal insert-size banding. · Abstract; Cell isolation; Figure legends

  4. ShownSurface antibody staining survives Tn5 tagging. A subpopulation can therefore be sorted after bulk tagging (TagSort), and its data match those from cells purified before tagging (SortTag).

    TagSort (anti-CD4-PE staining before tagging, PE+/DAPI+ nuclei sorted) vs SortTag (CD4 T cells purified with a Miltenyi depletion kit before tagging): median library sizes 30,953 vs 25,830; signal around peaks correlated at Pearson r=0.96; PE signal persisted after tagmentation. · Abstract; Plate scATAC-seq on CD4+ T cells (TagSort vs…; Figure legends

  5. ShownA logistic regression classifier yields cluster-specific marker peaks, mostly intronic or distal intergenic, that recover known lineage genes and transcription factor motifs. Separately, browser tracks show Cd83 accessibility differing between the two NK subsets.

    Top 500 marker peaks were taken per cluster (500 × 12 = 6,000); genes near the top 50 included Cd4 (cluster 3), Cd8a/Cd8b1 (cluster 6) and Cd9 (cluster 4). HOMER found ETS and ETS-IRF motifs in many clusters, T-box motifs in NK and memory CD8 T cells, POU motifs in marginal zone B cells; several Cd83 peaks in CD27+ but not CD27- NK cells. · Abstract; Bioinformatics analysis; Figure legends

Methods, data and reproducibility

Pools of 5,000 to 50,000 cells are tagmented with Nextera Tn5 in a digitonin/DMF buffer. DAPI-stained nuclei are then FACS-sorted into 96/384-well plates holding SDS/proteinase K lysis buffer, quenched with Tween-20 and indexed by qPCR-guided PCR (18 cycles). Binarised peak counts went through TF-IDF LSI, spectral clustering, logistic regression and HOMER.

  • HEK293T and NIH3T3 cells (25,000 each) mixed for a species-mixing doublet test · human and mouse · n = 307 passing wells (one 384-well plate) · E-MTAB-6714
  • K562 and mouse embryonic stem cells profiled side by side with the plate method and Fluidigm C1 · human (K562) and mouse (mESC) · E-MTAB-6714
  • Cryopreserved skin fibroblasts (human, mouse) and mouse cardiac progenitor cells at E8.5 and E9.5 · human and mouse · n = four tissues · E-MTAB-6714
  • Splenocytes after red blood cell lysis from two C57BL/6Jax mice (two 96-well and nine 384-well plates) · mouse · n = 3,648 profiled; 3,385 passed QC · E-MTAB-6714
  • Splenic CD4 T cells profiled by TagSort (sort after tagging) and SortTag (purify before tagging) · mouse · E-MTAB-6714
  • ImmGen bulk ATAC-seq of immune cell types, used as reference for comparison and cluster annotation · mouse (Immunological Genome Project) · PRJNA392905
  • Public Fluidigm C1 scATAC-seq data reprocessed with the same pipeline · PRJNA274006; PRJNA299657
Methods & tools
Nextera Tn5 transposase (Illumina FC-121-1030), FACS single-nucleus sorting (DAPI), Fluidigm C1 (comparison), snakemake, cutadapt, hisat2 (mm10; concatenated human-mouse genome for species mix), samtools, Picard MarkDuplicates, MACS2, bedTools (coverageBed, mergeBed), HOMER (annotatePeaks.pl, motif enrichment), Scikit-learn (TF-IDF LSI, spectral clustering, logistic regression)
Software & environment
snakemake pipeline with exact tools and flags in the repository Snakefile; analyses in Jupyter notebooks using Scikit-learn; reads aligned to UCSC mm10. No software version numbers are given in the text.
Reproducibility gaps
The detailed wet-lab steps are in a Supplementary Protocol outside the provided text. Tool versions are not listed in the manuscript (deferred to the GitHub Snakefile). The C1 comparison metrics are reported only in figures.
Limitations (4)
  • The pre-tagging antibody sorting strategy was tested only with CD4, an abundant marker on a splenocyte subpopulation; other surface markers and tissues would each need separate validation. · stated by the authors
  • Access to rare cell populations through post-tagging sorting is limited by how frequent the rare type is and by how many cells can be tagged upfront. · stated by the authors
  • Cluster identities rest on correlation with ImmGen bulk ATAC-seq and nearest-gene marker peaks, without orthogonal protein or RNA validation. The Cd83 permissive state in CD27+ NK cells is inferred from accessibility alone. · noted in this summary
  • The Fluidigm C1 comparison covers two cell lines, reports metrics only graphically and used a modified C1 protocol (digitonin replacing NP-40). The proposed explanation, more efficient bulk tagging, was not directly tested. · noted in this summary

Read this if you want a device-free, plate-based scATAC-seq protocol that tags cells with Tn5 in bulk before FACS, or if you need chromatin accessibility profiles of antibody-sorted subpopulations from a tissue.

Citations
1,147
FWCI
90.86
Field percentile
top 1%

Reproducibility ledger

Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.

Sign in to file a reproducibility report.

Related papers

Similarity Map
SelectedSame fieldCross-field
THISMezger '18Yang '24Baker '18Mimitou '20Seufert '23Zhang '24
75% semantic matchGenomics
Anja Mezger, Sandy L. Klemm, Ishminder K. Mann+7 more·Nature Communications·Sep 3, 2018· 149 citations