Earlier this year, before I joined my current thesis lab, I had very few ideas of what the 3D genome constitutes of. I vague knew that it is responsible for long-range promoter-enhancer interactions, and that certain diseases, from neurodegenerative diseases to cancer, are caused by a miswiring of such long-range interactions. Ever since, I learned that the 3D genome is a very exciting field with rapid advances happening on both experimental and computational grounds. Even though there are new papers coming out frequently, we currently know very little how exactly these 3D structures regulate gene expression, how they are important for normal cellular growth, and how to manipulate them.

As part of my attempt to scope the field, I will write about the developments of 3D genome studies, the discovery of the hierarchical structure of genome folding, the multidisciplinary approaches to dissect these structures from sequencing, imaging, and biophysical standpoints, the computational advances in analyzing 3D genome contact maps, and attempts to either manipulate these 3D structures to facilitate new promoter-enhancer interactions, or predict its determinants using ML-based methods. Of course, the literature on 3D genome is ever expanding, so doing a thorough review of everything will be quite impossible. Instead, I’ll rely on my selection bias and present the works I think are pretty impressive and hopefully represent the points I’d like to convey.

??? Can we study 3D chromatin structures the same way we study protein structures ???

The answer is, unfortunately, no. The DNA is way way longer than a regular protein. This comes with a lot of complications. For proteins, we can isolate it from cells then run cryo-EM to determine its 3D structure. In DNA, we cannot do that (technically there are experimental techniques that utilize electron tomography to study 3D genome structures, but they are very complicated and much less used than some methods introduced below). Furthermore, the 3D shape of the DNA is very dynamic and context-dependent, meaning the same DNA strand can be folded very differently across different cell types, or across different stages (think of the cell cycle for example) of even the same single cell. Therefore, the 3D folding problem for chromatin is a completely different beast compared to the 3D folding problem for proteins.

So what can we do to study it? Luckily, the field has come up with many ingenious ways to measure these 3D structures. Roughly speaking, there are 2 main hierarchies of methods one can use to study 3D chromatin: sequencing assays and imaging techniques. The core idea of these techniques is to be able to somehow take a snapshot of the 3D chromatin inside the cell. Since the cell is very dynamic and things keep moving around, in the majority of the methods, we have to keep everything fixed in space and time before we can take its snapshot, otherwise everything will be a gray blur.

Genome profiling assays

The main motivation of these sequencing assays is to profile all of the pairwise-interacting loci in the chromatin, since these interactions reveal whether two loci, despite being far away on the linear distance of the genome, are actually in close proximity in 3D space. To do so, we need the following key steps: keep the cells fixed in time, extract and distinguish the interacting loci with non-interacting ones, and sequence those loci. We achieve that by first applying formaldehyde cross-linking to the cells, where it forms covalent bonds between molecules to fix everything in place. Now we enrich for the interacting loci by marking them with a specific molecular handle and pull this handle down to separate them from the non-interacting ones. To do so, first we use a restriction enzyme that recognizes a specific DNA sequence and cuts the DNA there. At the cut site, the DNA becomes sticky, so that we can stick our molecular handle called biotin in there. Then we utilize a cellular mechanism called non homologous end joining (or end ligation, NHEL) that, whenever the DNA strand is broken, seals the cut site with our biotin in there. The assay is optimized in a way that prioritizes ligation only at cross-linked sites. With biotin held firmly in place, we use a molecular grabber called streptavidin with a very high affinity to biotin and pull down those pairwise-interacting DNA fragments. Finally, with our sample consisting of purely interacting DNA fragments, we can use high-throughput sequencing to read these fragments.

The workflow above is the basis of Hi-C. Variants of it (in situ Hi-C, Micro-C, single-cell Hi-C, …) are fundamentally the same, except they improve on one or more of the Hi-C procedures:

• In in-situ (means in place) Hi-C, as the name suggests, they do cell lysis after all the prep steps. This reduces the spurious interactions between free-flowing DNA and nuclear DNA with mitochondrial DNA.

• In Micro-C, the restriction enzyme is replaced with an MNase, an endo-exonuclease that digests DNA chains unbounded by nucleosomes. The advantage of Micro-C is that, unlike restriction enzymes, MNase does not need to recognize a specific DNA sequence to cleave the DNA, enabling it to detect more finer-resolution pairwise DNA-DNA interactions.

• In single-cell Hi-C, the procedures are done in intact single nuclei. When doing sequencing, the researchers come up with clever ways to tell apart which cell a particular DNA read comes from by either using droplet-based or split-pool approaches.

After bioinformatic data analyses, what we obtain is a square contact matrix, where each element i, j in the matrix contains the normalized counts of reads mapped to both loci i and j. We do not obtain a 3D rendering of the DNA polymer in 3D space, like what we would expect when we look at 3D protein structures; rather, the 3D chromatin contact matrix roughly encodes the frequency each pair of locus interacts with each other. The higher the signal at each element (i, j) of the matrix, the more likely the 2 loci (i, j) interact with each other more frequently, the more likely they are in the same spatial vicinity in 3D. This contact matrix turns out to have many interesting structures with intimate relationships to the location of genes and cis-regulatory elements that researchers are still actively working on deciphering.

Imaging assays

Maybe you’re still unconvinced that the high signal in a contact matrix does not necessarily mean the 2 loci are close in 3D space. Apart from sequencing assays, the field also develops imaging techniques that facilitate the observation of genomic loci in the nucleus. With these technologies, the researchers can observe that the same genomic loci with high signal in the contact matrix are physically closer together. To do it, researchers use short nucleotide sequences called probes that are complementary to the loci of interest. These probes also attach a fluorescent mark that allows researchers to observe where they are in the nucleus, like a lighthouse in the middle of the dark sea. To introduce it to the cells, they also have to use formamide to keep everything in the nucleus fixed, then use heat to denature the DNA double strand, then expose the cells to the probes. Now, we can observe the cells under a super-resolution microscope. This whole technique is called DNA Fluorescence In-Situ Hybridization (DNA-FISH). When researchers design probes with DNA-FISH for 2 loci that, despite being very far away on the linear genome, are believed to stay close together in 3D space via chromatin folding, they observe under the microscope that the loci are physically close together, a physical evidence!

So far, our techniques rely on only taking a snapshot of the cell, because the experimental techniques kill the cells in the process. That means we can’t really observe how the 3D chromatin structures dynamically change over time, really limiting our ability to make conclusions of its functions and dynamics. That’s when live cell imaging comes to rescue. The idea is conceptually similar to DNA-FISH: we add fluorescent probes to the 2 loci we are interested in studying, and then we observe it under the microscope. However, unlike DNA-FISH, the experimental procedure of live cell imaging is designed that allows the cells to live while also having the same loci tagged with fluorescent markers. Now, we can observe these markers move inside the nucleus over time and chart out its spatial locations. The main limitation though is that these imaging techniques are only used to detect interactions between a certain number of loci at a time, missing out a vast majority of other pairwise interactions.

Ending remarks

Congrats on making it this far! In this blog, I introduce the very basics of 3D chromatin assays, both from the sequencing and the imaging perspectives, to show how it is a completely different world to study chromatin structures compared with protein structures. There are many other very cool techniques the field develops that I have not touched on, such as HiChIP for both 3D chromatin and DNA-protein binding profiling using immuno-precipitation, SPRITE and RD-SPRITE for joint multi-wise interactions between DNA and RNA, or other multimodal single-cell techniques, such as single-nucleus methyl 3C (snm3C) for joint profiling of 3D chromatin and DNA methylation in single nuclei, or other advanced imaging technologies, each with their specific advantages, drawbacks, and use cases.

As alluded to above, the 3D contact matrix is very fascinating in that it reveals 3D chromatin structures are inherently hierarchical. Just like if we start from outer space and keep zooming in, we will see the whole solar system with the Sun and planets, then the Earth with land and sea, then the land with mountains and cities, then cities with buildings, roads, …, the 3D contact matrix is the same: first the broad plaid patterns called compartments, then square blocks called domains, and finally high-intensity dots called loops. Each structure has been intensively studied, its formation investigated, its functions mapped. Yet, so many unknowns remain. Nature designs it so that the whole human DNA roughly 2 meters long is fit inside a nucleus of roughly 6 micrometers, an unimaginably difficult Origami task, while at the same time ensuring that genes are turning on or off in correct patterns for cells to grow. Therefore, the 3D chromatin is folded in structures that we have just begun to discover and understand.

This blog is born as my attempt to share my learning process through my long PhD journey. Therefore, the blog will be messy and may not capture the most comprehensive overview of a topic, as you’d find in peer-reviewed articles; however, it will be an authentic look into what I learn, what I think, and how it fits into my ever growing knowledge of this field. I plan to first write one blog every month, mostly summarizing what I have read and learnt that month, but also may include my PhD research progress or a very deep dive in a particular topic that I’ve long been interested in.

With that, stay tuned for my next post!