Genome-wide measurements of protein-dna interaction by chromatin immunoprecipitation

Transcription

1 Genome-wide measurements of protein-dna interaction by chromatin immunoprecipitation D. Puthier. laboratoire INSERM, Aix-Marseille Université, TAGC/INSERM U928, Parc Scientifique de Luminy case 928

2 Outline Chromatine structure Chip-Seq approach Method Expected output Sequencing technologies Technical considerations Data analysis Quality control Mapping Peak detection Practical session (Galaxy)

3 Chromatin constraints Each diploid cell contains about 2 meters of DNA High level of compaction required Accessibility required Replication Transcription DNA repair Specific machinery required

4 Chromatin has highly complex structure with several levels of organization Genetics: A Conceptual Approach, 2nd ed.

5 Beads on a string Figure 4: Chromatin fibers purified from chicken erythrocytes. Each nucleosome (~12-15 nm) is well resolved, along with the linker DNA between the nucleosomes. Given the resolution, other components, if present, such as a transcribing RNA polymerase or transcription factor complexes, should be resolvable

6 Histones and nucleosomes Histones Small proteins (11-22 kda) Highly conserved Basic (Arginine et Lysine) N-terminal tails subject to post translational modification Nucleosome Octamers of histone (H2A,H2B,H3,H4) x 2 146bp DNA

7 Nucleosome structure

8 Histone post translational modification Lysine acetylation Lysine methylation Arginine methylation Serine phosphorylation Threonine phosphorylation ADP-ribosylation Ubiquitylation Sumoylation...

9 Some alternative modifications

10 The Brno nomenclature The nomenclature set out here was devised following the first meeting of the Epigenome Network of Excellence (NoE), at the Mendel Abbey in Brno, Czech Republic. For this reason, it can be referred to as the Brno nomenclature.

11 Chromatine immuno-precipitation (ChIP)

12 Expected output Strand-dependent bimodality in tag density

13 A nice example (NFKB) clear strand assymetry probably single binding events

14 More complexe situation...

15 Sequencing technologies

16 Sequencing by ligation (SOLiD) We have a flow cell (basically a microscope slide that can be serially exposed to any liquids desired) whose surface is coated with thousands of beads each containing a single genomic DNA species, with unique adapters on either end. Each microbead can be considered a separate sequencing reaction which is monitored simultaneously via sequential digital imaging. Up to this point all next-gen sequencing technologies are very similar, this is where ABI/SOLiD diverges dramatically (see figure 4)

17 Sequencing by ligation (SOLiD)

18 Sequencing by ligation (SOLiD)

19 Sequencing by ligation (SOLiD) According to colour code table four different dinucleotides may correspond to the same colour. This ambiguity is resolved using primer n-1 round The first ligation of this primer round analyses dinucleotide with one known nucleotide

20 ChIP-Seq: technical considerations Quality of antibodies: one of the most important factors ('ChIP grade') High sensitivity Fivefold enrichment by ChIP-PCR at several positive-control regions High specificity The specificity of an antibody can be directly addressed by immunoblot analysis (knockdown by RNA-mediated interference or genetic knockout) Polyclonal antibodies may be prefered Offer the flexibility of the recognition of multiple epitopes Cell Number Typically (e.g, RNA polymerase II/histone modifications) (less-abundant proteins)

21 ChIP-Seq: technical considerations Open chromatin regions are easier to shear Higher background signals Two solutions Isotype control antibodies Immunoprecipitate much less DNA than specific antibodies Overamplification of particular genomic regions during the library construction step (PCR) Input Non-ChIP genomic DNA Better control

22 ChIP-Seq: technical considerations Replicates To ensure reliability of the data No consensus on the correct number of replicates ( 2?) Use of a different antibody wherever possible Chromatin fragmentation Sonication in SDS-containing buffers may improve fragmentation May be appropriate for transcription factors tightly bound to DNA May result in loss of signal for proteins not directly bound to DNA

23 ChIP-Seq: technical considerations Size selection Melt agarose gel slices at room temperature in the supplied buffer Heating to 50 C decreased the representation of A/T-rich sequences A/T rich sequences may denaturate and may not reanneal Higher affinity of spin columns for double stranded DNA GC rich sequences enriched Amplification Overamplification can typically be avoided by decreasing the number of PCR cycles

24 Quality control for high throughput sequence data FastQC

25 Common procedures for ChIP-seq data analysis

26 Read mapping a Work well for Sanger and 454 reads, allowing gaps and clipping. b Paired end mapping. c Make use of base quality in alignment.dbwa trims the primer base and the first color for a color read. e Long-read alignment implemented in the BWA-SW module. fmaq only does gapped alignment for Illumina paired-end reads. g Free executable for non-profit projects only.

27 Mapping Reads mapped to multiple sites ('multi-reads') are usually discarded during 'normal' analysis. Peaks in highly repetitive regions are overlooked CSEM (ChIP-Seq multi-read allocation using E-M algorithm) Assign multi-reads based on region coverage. Number of accepted mismatches is typically 0-2

28 Peak-calling algorithms Depends on peak shapes Sharp MACS (Model-based Analysis of ChIP-Seq) SPP/WTD (Window Tag Density) SPP/MTC (Mirror Tag Correlation) CisGenome, Broad CCAT (control-based ChIP-Seq analysis tool) SICER (Spatial clustering approach for the identification of ChIP-enriched regions) Sharp and Broad ZINBA (Zero-Inflated Negative Binomial Algorithm)

29 Peak-Calling: WTD Window Tag Density (SPP package) pd= positive downstream pd pu= positive upstream pu nd = negative downstream nd nu nu = negative upstream

30 Peak-Calling: MTC Mirror Tag Correlation (SPP package) Strand cross-correlation profile

31 MACS: pre-processing For experiments with a control MACS linearly scales the total control tag count to be the same as the total ChIP tag count. MACS removes duplicate tags if more than expected based on sequence depth

32 MACS: computing 'd' 1. Given a sonication size (bandwidth) MACS slides 2bandwidth windows across the genome to find regions with tags more than mfold enriched relative to a random tag genome distribution 2. Randomly samples 1,000 of these high-quality peaks Separates Watson and Crick tags. Aligns them relative to window center Distance between the modes of the Watson and Crick peaks in the alignment is defined as 'd'

33 MACS: computing 'd' 3. Shift all the tags by d/2 toward the 3' ends to the most likely proteindna interaction sites.

34 MACS: Peak detection Slides 2d windows across the genome to find candidate peaks with a significant tag enrichment The location with the highest fragment pileup ('summit'), is predicted as the binding location.

35 Histone Methylation near Transcription Start Sites

36 MERCI