4D Nucleome Network Wiki

AGENDA

CUT and RUN Presentation by Steven Henikoff

Features of the CUT and RUM method:

• Limit digestion of targeted fragments
• Can profile insoluble proteins.
• Give high resolution TF mapping and prefers one surface of DNA duplex.
• Compared to ChiP-gives better resolution.
• Maps 3D contact sites
• Do Not use crosslinking
• Can enrich ChiA-PET data bc we are having high resolution
• Utilize low cell number (100 cells)
• Profile cell differentiation.
• Profile flash-frozen tumors

Discussion

Do you detect enrichment of the reads on open chromatin, such as promoters?

We have looked at H3K27me3, which is not expected to be enrich on active sites, so we don’t see any enrichments on atac-seq enrich sites. Is more complicated when looking at transcription factors because is often expected to be unlocalized at accessible sites, however for stage specific transcription factor, for example sox2 in pluripotent specific genes and foxa2 in endoderm, they typically localize at accessible sites, but with specificity border targets.

Over digestion (by doing time series) is also recommended to get accessible sites. If the background is high due to particular antibodies short times are recommended.

Shall we do multiple time points or it is optional?

Is optional. There is data that looks very good with 3 min digestion, however you can not get full recovery, but if the cell numbers are small, the digestion is efficient with short time periods. For example, we have data for yeast that we carried out for one second, so there is a wide range of digestion. Perhaps, some factors are more accessible than others and require shorter times points. For example, for H3K27me3, 30 min is enough and for H3K27ac with few minutes is better. This probably depend on the accessibility. \It's always good idea to do different time points when you starting out until find out the one that is ideal. Binding sites are recovered at different points, is there a way of combining data from different digestions time series to have more uniform data sets? With the respect to the time series, we would like to think is that we are looking at quantitative data because we are having similar results for different time points, even at what seems to be saturation where the whole population has been cleaved out. That itself is very useful because we can do some inferences about occupancy, which i won't trust if were not actually getting similarity in short time point, however we prefer to combine all the times point and consider them technical replicates, because they are so similar. For the data that is submitted to the portal, there are different data sets that are digested 1 min or 3 min. DCIC is very strict in terms of what we defined as replicates in experimental parameters that recall those not replicates in different conditions, but in this particular experiment, shall we do an exception and group different experiments? Is this will be a better representation on the portal? Yes, as i user i prefer have more and deeper data to get better resolution and if it pases Quality control will say yes and combine the time points. How do you validate reproducibility between two replicates whether is biological or else? I use correlation metrics, which is the simplest way to do it. Basically we do a lot pass filter to remove all the ceros. Cut and Run has a lot advantages over ChiP-SEQ, however both methods share the use of antibodies and antibodies specificity becomes an issue for QC. Do you agree that the same antibody QC standard recommended for ChIP-seq will apply also for cut and run? Steven Henikoff: I would agree, however we don't know what the standard is. Bing Ren: ENCODE has recommended two types of essays: western blot and knockdown experiments to reduce the protein or intensity. They also recommend certain fraction, perhaps 50% of the signal from the western should belong to the antibody. \These are minimal expectation for antibody specificity. ChiA-PET might have the same issue on the specificity of the antibodies.

Have you tried to do CUT and RUN in the absence of CTCF in the cells?

NO.

Note: Bing Ren has engineered mouse embryonic stem cell lines that express auxin-inducible-degron (AID) fused with CTCF, which can be used to try CUT&RUN without CTCF.

Have you look footprints in the direct binding regions vs indirect binding regions and seen specific transcription factors that can be identified in the direct binding sites?

We did, but we didn't see anything that looks different. For example, we thought that maybe we will see that the indirect sites may come up later than the direct sites in the time course, however we saw the opposite. The direct sites where CTCF is bound is perhaps because is blocked by the factor itself and might not be accessible as the contact sites. It's a little tricky to came up with an answer with that question.