ATAC-seq introduction and notes from Preissl et al 2018

ATAC = Assay for Transposable-Accessible Chromatin

Chromatin = chromosome DNA plus all proteins involved in its physical structure. DNA in the nucleus is "packaged" into a condensed structure. "Open chromatin" is less densely packaged, making it more accessible to proteins and other molecules (e.g. transcription factors, miRNAs, mutagens), and thus more likely to be transcribed or otherwise used.

ATAC-seq is a method to identify chromatin accessibility (regions of open chromatin).

Tn5 transposase = the enzyme used in ATAC-seq. It binds open chromatin and marks it by inserting known DNA sequences ("adapters").

Scientists use adapters as primer-binding sites to amplify DNA and make "libraries", specifically amplifying the Tn5-marked DNA regions of open chromatin.

Libraries are DNA in a test tube. Libraries are stuff, not yet data. To get data for analysis, the libraries are run through a sequencing machine.

ATAC-seq is cheaper and faster than ChIP-seq, another method that uses antibodies to pick up DNA-binding proteins and then identify the DNA sequences they are binding. ChIP-seq is useful for identifying the "binding motif" (short sequence pattern) that proteins bind, for example, hormone-binding transcription factors.

ATAC-seq identifies "active loci", regions of DNA that are open for use by the cell.

Why do ATAC-seq? Why is it useful to know regions of open chromatin?

• Useful to combine with RNA-sequencing data
• Helps identify if epigenetics is involved in your phenotype of interest (e.g. higher expression of InterestingGene1 in GroupA, compared to GroupB).
  
  

scATAC-seq = single cell ATAC-seq: sample processing involves dissociating tissue into single cells (or using material that already has single cells) and using each cell for ATAC-seq.

snATAC-seq = single nuclei ATAC-seq: sample processing involves isolating intact nuclei and using each nucleus for ATAC-seq.

Literature notes for myself

Preissl et al. 2018 "Single-nucleus analysis of accessible chromatin in developing mouse forebrain reveals cell-type-specific transcriptional regulation" Nature Neuroscience, 21, 432-439 (2018).

• Flash frozen tissue samples from 8 week old adult mice and from mouse embryos
• Frozen tissue was pulverized in liquid nitrogen with a mortar and pestle
• Two samples processed each timepoint
• Transcriptome generation to make distinct barcodes (prior to library prep)
• A & B oligos annealed to pMENTs oligos in separate reactions = A & B transposomes generated separately
• Mixed barcoded transposons with unloaded transposase Tn5 (from Illumina) and incubated 30 min, resulting in the Tn5 enzyme picking up the transposome DNA sequences
• Combinatorial barcoding (12xA, 8xB, for 96 barcode combinations)
• Sample processing for single nuclei ATAC-seq
• 5-20 mg of flash frozen tissue in 1.5ml lobind tubes
• Add 1 ml NPB buffer and incubate 15 min at 4C (reduced to 5 minutes for cell culture) to isolate nuclei
• 5% BSA (Sigma)
• 0.2% IGEPAL-CA630 (nonionic, non-denaturing detergent similar to NP40 and used to isolate membrane protein complexes) (Sigma)
• cOmplete protease inhibitor cocktail (Roche)
• 1 mM DTT (reducing agent, also used as an antioxidant) 
• All in PBS buffer
• Nuclei suspension was filtered through 30 um Cell-Tric (Sysmex)
• Centrifuged 5 min at 500 g 
• Nuclei pellet resuspended in 500 ul of 1.1x DMF buffer: 
• 36.3 mM Tris-acetate pH 7.8
• 72.6 mM potassium-acetate, 
• 11 mM Mg-acetate
• 17.6% DMF as a solvent (don't need PBS buffer)
• Nuclei were counted using a hemocytometer
• Concentrations adjusted to 500/uL
• 4500 nuclei were dispensed into each well in a 96-well plate
• Added 1 uL barcoded Tn5 transposome (0.25 uM) to each well
• This is the Tn5 enzyme loaded with the barcodes
• Mixed 5 times
• Incubated 60 min at 37C with 500 rpm shaking 
• Stop the reaction by adding 10 ul of 40 mM EDTA to each well
• EDTA sequesters Mg2+ ions which DNA-binding enzymes require for binding DNA, so it stops Tn5 enzyme
• Incubate 15 min at 37C with 500 rpm shaking
• Added 20 ul sort buffer to each well
• 2% BSA
• 2 mM EDTA
• All in PBS
• Combine all wells
• Filter through 30 um CellTric (Sysmex) into a FACS tube
• Add 3 uM Draq7 (Cell Signaling) - this is a DNA staining dye usually used as a viability dye, but here used to stain the nuclei for sorting. The 10X Genomics protocols use 7AAD dye.
• Sort 25 nuclei per well, in four 96-well plates (total 384 wells) with 18.5 ul EB
• 50 pM Primer i7
• 200 ng BSA (Sigma)
• Spin plates (at ???)
• Add 2 ul of 0.2% SDS
• Incubate 7 minutes at 55C with 500 rpm shaking
• Stop the reaction by adding 2.5 ul of 10% Triton-X "to quench the SDS"
• How does this work? The Triton-X displaces the SDS?
• Add 2 ul of 25 uM Primer i5
• Add 25 ul of NEBNext High-Fidelity 2x PCR Master Mix (NEB)
• PCR amplify samples for 11 cycles
• Combine all wells (volume is about 15.5 ml)
• See the paper for the rest of the protocol
• Sorted 20,000 single nuclei for ATAC-seq
• Data analysis in brief
• Alignment of pair-end sequencing reads to the genome with Bowtie2
• Alignment filtering: drop nonuniquely mapped (MAPQ<30) and improperly flagged (flag=1804) alignments
• Barcode error correction: expect barcodes with four 8-bp long indexes (i5, i7, p5, p7). Reads with >1 base mismatches for any index were dropped. Reads with only 1 mismatch were changed to their closest index.
  
• Reads separation based on barcode combinations - this separates out the single cell or single nuclei data
• Mark and drop PCR duplicates
• Mitochondrial reads were dropped
• Offset reads to adjust position of the Tn5 inserts
• Quality assessment for each cell: calculate accessible promoter coverage, number of reads, signal-over-noise ratio
• Cell selection - keep only cells that pass these thresholds
• >10% coverage of constitutively accessible promoters (per ENCODE annotations)
• >1000 reads
• reads in peak ratio greater than the estimation from corresponding bulk ATAC-seq levels
• Replicates separation
• Analyzed 15,767 single nuclei after data cleanup
• Mean read depth per nucleus = 9,275-18,397 reads
• Identified 20 cell types