4D Nucleome Network Wiki

4DN OMICS Minutes, May 14, 2018

\ 1.- Overview of TSA-Seq method (Andrew Belmont’s research group). \The information about this work is in the following pre-print -bioRxiv preprint: https://www.biorxiv.org/content/early/2018/04/25/307892

\ Tyramide Signal Amplification (TSA) -Seq method principles: Is a genomic proximity mapping method that can measure chromosome distance at any chromosome loci to particular nuclear compartment and transform that information into sequencing readout. This is the only currently genomic method available that can measure distance. \

Features of the TSA-Seq method:

• Turn genomic distance from nuclear speckles into sequencing readings
• Nuclear speckles also called Interchromosomal Granule Clusters (IGCs) do not contain chromatin.
• Works for insoluble structures
• Does not require direct interaction
• Limited labeling bias
• High purification efficiency (less bias)
• Easier quality control for antibodies

\ Current \ChIP and DamID methods do not work very well on nuclear speckles, because they rely on direct crosslinking between Protein-DNA or require direct contact for the enzyme to modify the DNA, however, TSA-seq method can map organization and do not rely on direct interaction.

\ Q1: Although ChIP won't be able to review interactions between nuclear speckles and chromatin, DamID method might have a chance because the enzyme appear to harvest a range of regions across large distances. \Would DamID be able to reveal the distance between chromatin regions?

A: Using EM by serial sectioning, the chromatin is ~200 nm distance from the granule core, so we think that it won’t work, but \recently Dr. Bass developed a new modified DamID method where he can visualize the product of the Dam Methylase and is working for nucleoli and lamin, but doesn't get a DNA methylation signal using a speckle target, so DamID doesn't work. \

Another issue is the ‘closeness’, so we need to know how close is close. Currently, there is no clear definition whether 0.1 micron is close or 0.5 is close etc. \so we needed also a method that directly measure distances to better assess functional significance at certain distances. TSA-seq method is able to do this. \

2.- Technical details:

\ Two check steps or quality controls in the protocol: 1) Microscopy; 2) Dot blot Input DNA in some TSA-Seq experiments are not flat and would be filtered. \

Q2: Is the TSA-Seq value comparable between different biological samples? For example, one in ES cells and the other one in GM12878.

\ A: Y-values are expected to be different in different cells, and only the distribution counts, therefore, those values should not be directly comparable. \

A: We do have done experiments and we do see changes at the dynamic range and some of that is based on contrast on a given staining, but some are probably based on cell shaped and if we want to map distance and compare distance to speckles we have to do distance calibration which require microscopy and certain production marks followed by TSA score to predict the distance.

\ Q3: How well this method actually works converting the track to a distance? Assuming that you did microscopy experiments at the single locus in a particular cell type and then you want to extrapolate to the rest to the cell type. Is this desirable if it could meant to work? \

A: \Is partially base on target, we are approximately the collection of speckles as a difference from point source and it seem to work very well. In the case of 562 the residual is very small. This has been also using it as a minimization problem and and can get better fit to different distance map but in the case for lamina is different and we need to talk about deconvolution. Depending of the shape of the nucleus you can got to convolve the signal from the whole lamina to predict the signal to given spot in the nucleus, however we haven’t done this yet and this will be the next step, which we will use a deconvolution approach, which will require some assumption about the nuclear shape for things that are more continuous. \

Q4: How variable are nuclear speckles to their genomic neighborhood since you are able to correlate the distance measure from imaging to the pool of raw data and since there is a huge amount of synchronization between cells? \Also, this will define the limit of how well you will deconvulate the distance or infer the distance from the enrichment value. \

A:This mean how heterogenous the population is. Basically, the distance that you get is an average and from those medians and distance points the distance distribution became very broad and we are able to predict the mean distance quite well. In terms of near to speckles is more deterministic that anything that we have seen before in the sense that if you look at the probes from A-C, which protect from large peaks, the medians from those are 0.1- 0.2 microns and near 100% of the loci were half a micron. These things are positioned more reproducibly than things positioned to the nuclear lamina or other compartments.

\ There is a different distribution of genomic distance from speckle peaks to lamina based on \from these large spackle peaks, which directly interact with the sparkles vs smaller ones and now comparing a few cell types looks like that most of these interaction regions are very close in another cell types as well.

\ 3.- Data processing: * FASTQ files * View quality of reads * Trimming reads * Mapping * Remove duplicates * Alignment quality \

Calculation of TSA score represent enrichment

Input signal is not flat, variation is also taken in consideration

\ Erez (min 37) Q4: What is the best way to integrate TSA-seq from different conditions using the same antibody? It appears that by using different antibodies the compactness of peaks and the noise profile changes. \

A: Conditions 1 and 2 tend to overlap because we obtained similar results and if we have those two different conditions we can minimize the residual to get the distance mapping. For condition 3 at much higher resolution for things that are very close and things about certain distance are invisible. For condition 1 and 2 we are getting the mean of the distribution, but if we integrate condition 3 we might get an idea of the shape distribution. In condition 3, you are looking at the distribution of things that are close to the speckle and you may be able to distinguish two different situations. In one situation you have all fraction of cells with the locus very close to the speckle and a lot of cells were far away, while in another cell population you might have a continuous distribution. In theory we could do this, but it may require a modeling approach.

\ Q5: In terms of data processing, such as \trimming and removing duplicates, is there significant different in the analysis pipeline between all the protocols (e.g. Repli-seq, DamID and TSA-Seq)?

A: We have not coordinated with the other labs to compare the protocols

\ Comments: It would be definitely better if the protocols between different technologies are comparable. Also, Fish validation would help to have some testing data at the low distance to test for the exponential decay better. \

Q6: How can you tell if different chromatin are interacting with different speckles? Is it possible to tell?

A: Currently we assume the ones in a peak are interacting with the same speckle

\ Comments: We may be able to tell if we combine TSA-Seq results with Hi-C results Note: This is being explored in the joint analysis WG \

The next step would be to produce a protocol-like document for TSA-Seq. The next meeting would be cancelled due to memorial day.