Series on 3D Genome Organization
Part I: 3D Genome Folding - you need it, but it’s not enough
Bowl of ramen
The nucleus is kind of like a bowl of ramen. The noodles are the chromatin, the flakes
are the transcription factors, the broth is the nuclear solution, and the bowl is the nuclear
membrane. The broth is warm and salty, and contains all the minerals that the flakes and
noodles need to stay bouncy, elastic, and have just the right amount of chewiness (chromatin
structure is sensitive to temperature and ionic environment). You get that extra flavor and
texture hit from those little flakes sticking to the noodles (gene activity).
You could point out that this analogy falls out of "flavor" since the bowl is not permeable
to specific proteins and RNAs. Or, you could point out that the transcription factors are bound
to specific parts of DNA, and not randomly scattered like those yummy flake nuggets. But
where the analogy really falls flat, is that the entire bowl of noodles (chromatin) are not
randomly coiled up in the bowl (nucleus) at all. And therein lies the rub, and the difference
between a tasty bowl of noodles and every single cell in your body that is being used to chew,
taste, and digest those delicious strands of noodles.
The DNA in each of your cells’ nuclei, when stretched and lined up end to end, measures
2 meters long. And yet all of this DNA is packaged - along with trillions of molecules of protein -
into a nucleus microns in diameter. This packaging is not random, of course. It is extremely
precise and highly functional. Different cells fold their DNA differently in order to maximize the
function required of them. In an extreme example of functional genome folding, the silent
heterochromatin of photoreceptor cells in the eyes of nocturnal animals - which are more
densely packed - are pushed to the nuclear center, where they can act as collecting lenses to
maximize the amount of light received. In another example, the mammalian sperm replaces an
entire nucleus worth of normal packaging proteins (histones) with protamines. These are types
of variant histone proteins that allow for extra tight packaging, so that the sperm can take on a
more streamlined shape for efficient swimming - winner winner chicken dinner!
Proper DNA folding is not just important for unique cell functions, but it’s also essential for
all nuclear activities - from proper regulation of transcription to DNA repair and from
X-chromosome inactivation to DNA replication. Importantly, misfolding of DNA leads to
various pathologies. Happy mutations leading to changes in genome folding can even result
in the evolution of new sexual behaviors in moles!
Loop extrusion
For all its importance, the rules of exactly how DNA is folded so precisely and yet so
dynamically, and so deterministically and yet so readily reprogrammed, are not well
understood. It also remains unclear whether different organisms all along the tree of life, with
all their varying sizes and contents of genomes, fold their DNA into the nucleus under similar
or different rules to come to such similar and yet highly different functional conclusions.
Some key players have been identified. CCCTC-binding factor (CTCF) is a highly
conserved zinc-finger transcription factor found in all eukaryotes. Historically, it has defied
classification, as under different genomic contexts it can either activate or repress gene
transcription. When placed in between enhancers and promoters (bits of DNA necessary for
gene expression), it can actually block - or “insulate” - the enhancer’s ability to activate
These elements that bind to CTCF and other related proteins are called insulators. They
demarcate boundaries in the genome where genes can have their own nuclear environments,
controlled by the appropriate enhancers.
How do enhancers and promoters find each other? Cohesin, a multicomponent protein
complex, forms a ring-like structure that entraps DNA strands inside the cohesin ring. An
ATP-driven motor then proceeds to pull the entrapped DNA through the ring, in a process
termed “loop extrusion”.
This extrusion process occurs randomly and processively throughout the genome, and
eventually the appropriate enhancers and promoters, which can be sometimes hundreds of
kilobases or even megabases away, come into close proximity.
Enhancers bound by transcription factors, and promoters bound by core transcriptional
machinery, then can interact through a protein complex called mediator. At this point, all the
necessary protein components required for productive transcription are together, and it is
thought that the gene starts to be transcribed. Cohesin continues to extrude until it bumps into
CTCF, and due to the stability of CTCF binding at insulators, and possibly interactions between
CTCF and cohesin, cohesin stops extruding and falls off the DNA at some point. This then
prevents enhancers and promoters at different sides of insulators from ever meeting each
other, thus demarcating distinct chromatin environments in the nucleus.
But is this really the whole story? Some evidence suggests that DNA looping, which
brings enhancers and promoters into close proximity, precedes transcription by hours and
appears unchanged during early embryonic development. In fact, super-resolution microscopy
revealed that when genes become actively transcribed, they move further away from their
Even more troubling is the fact that while removing all CTCF and cohesin from a cell
leads to loss of proper chromatin folding, very little change in global gene expression pattern are observed. So why is the cell going to so much trouble to bring enhancers and promoters
close together? What function does this serve? In other words, what is the relationship
between transcription and 3D genome organization? Is proper DNA folding necessary and/or
sufficient for transcription? Put even more simply, it’s a chicken and egg question: does folding
come first, or is it transcription that precedes folding?
Chicken and egg
While these are some of the most important remaining questions in the field of
transcription, they have been historically difficult to answer. First, transcription occurs
sporadically, so while a population average of a cell type shows that the gene is being
expressed in large amounts, any given cell could either be expressing or not expressing the
gene. To be sure that the cell is expressing a given set of genes at an exact moment, and to
simultaneously measure the DNA folding in the cell, genome-wide, is beyond the limits of
contemporary life sciences. Technologies are available to dissect gene expression and
distances between loci in live cells, but these are only limited to a single locus at a time, and
the distance measurements do not give an exact idea of the various conformations the DNA
molecules-especially the intervening DNAs between two tagged loci- take. To study changes
in global 3D chromatin architecture and the associated changes in gene expression
simultaneously has therefore been nearly impossible. In our new paper in Nature Genetics,
we took advantage of a classic genetic trick from our friend, the yeast loving, banana
hopping, fruit fly, to unravel this hot bowl of messy ramen.
Fruit flies to the rescue
Fruit flies have been used as a model to study gene expression and chromosome biology
since the turn of the last century. As you might have learned in Biology 101, geneticists have
bombarded these tiny creatures from the farmers market with all kinds of mutational rays,
generated an array of mutants, and studied the resulting changes in the flies to great effect.
While some of these mutations are quite alarming(such as the well-known homeobox
transformations that cause legs to grow out of the head of the fly), another interesting series
of mutants were used in this study to illuminate the role of 3D genome organization in
transcriptional regulation.
Bone morphogenetic protein (BMP) signaling is critical throughout embryonic development.
In the humble fruit fly, tightly controlled BMP signals during the very first stages of
embryogenesis dictate the amount of certain transcription factors that localize to a nucleus.
A gradient of BMP signaling from the top (dorsal) to the bottom (ventral) axis of the embryo
dictate the amount of a crucial transcription factor, Dorsal (whose mammalian ortholog is NFᴋB),
that is transported into the nucleus. The amount of nuclear Dorsal controls the gene expression
patterns of the nucleus and ultimately the cell fate of that nucleus. Low levels of Dorsal guides
cells to form the dorsal ectoderm (outer layer), medium levels channel them to the
neuroectoderm (neuronal tissues), and high levels drive them to form the mesoderm
(muscles and other internal tissues). Importantly, the changes in gene expression and
associated changes in chromatin state (active vs repressed) at specific gene groups have
already been very well studied,
using mutants in the BMP signaling pathway that affect the concentration of nuclear Dorsal.
The key trick is that these mutants allowed for uniform levels of Dorsal from top to bottom.
For example, one mutant leads to uniformly high levels of Dorsal, driving all cells in these
embryos into the mesoderm fate, while another mutant leads to uniformly low levels of Dorsal,
prompting all cells to a dorsal ectoderm fate. Since cell cycles are extremely short in the early
fruit fly embryo, with some of these cell divisions occurring every 60 minutes, all gene
expression is achieved (with the control of specific enhancers regulating expression of specific
genes) in this brief time window. Therefore, the experimentalist has exquisite control over the
cell type and gene expression pattern. An added bonus is that fruit flies lay a ton of eggs, so
material is abundantly available. This makes genomic techniques that require large amounts of
starting material possible.
Mattias Mannervik at Stockholm University knew first hand how these mutant embryosUnexpected results... and an unexpected collaboration
Recall the more traditional and commonly agreed upon model of gene expression, where
cohesin-mediated loop extrusion brings together enhancers and promoters, triggering gene
activation. In contrast to this model, the authors found that the 3D folding of genes and
enhancers are independent of gene activity. In other words, regardless of whether a given
gene is on in one cell type or off in another, the regulatory DNA surrounding the gene is folded
in the exact same way! This result runs contrast to decades of previous literature, where looping
of enhancers to promoters have been postulated to result in gene activity.
Because of the stunning nature of the discovery, one might have brushed it off as some sort of artifactual discovery due to technical limitations. However, when Juanma presented their groups’ discovery at the Montpellier Institute of Molecular Genetics, he was stunned to discover that another group had used the same mutant embryos, analyzed chromatin architecture using a completely independent method, and came to the same conclusion.
Fate would have it that Mounia Lagha, a more recent alumna of the Levine lab back at
UC Berkeley, also became interested in the question at around the same time. Being an
alumna of the Levine lab, she too was very familiar with these mutant fruit fly lines. Her weapon
of choice was not the sequencer however, but instead the microscope. Using high-resolution
live imaging, she pioneered major discoveries in the Levine lab and as an independent
Principle Investigator. Instead of using Hi-C, their study used high resolution live imaging to
come to the same conclusionthat, regardless of whether genes were active or repressed, the chromatin surrounding the
enhancers and promoters of these genes are similarly positioned in 3D space.
News soon reached the rest of the two groups. After fevered discussions, the groups
decided to co-submit to Nature, where they promptly got editorially rejected and bumped down
to Nature Genetics. Anyone who publishes in the life sciences knows how this game of journal
chess plays out. Given the shocking nature of the results, and especially since the results were
negative, it would have been hard in the best of circumstances to get reviewed or accepted at
Cell, Science or Nature. The fact that the studies were done in fruit flies does not help either.
One could always argue that this was some weird fruit fly trickery. In any case, after the papers
went out for review at Nature Genetics, the groups posted their independent papers on
BioRxiv. And of course, as is customary in the age of social media, Juanma tweeted about the
Little did they know that this action would have major repercussions for a new postdoc across the pond, back at the lab of Mike Levine, now at Princeton. Having started my postdoctoral career a few months before, I was searching for a project that I could sink my teeth into. I came across these mutant fly lines, suggested to me by another postdoc in the lab, João Raimundo. I recognized that these lines could be a very good tool for addressing this key question in the field. When I was a graduate student in the lab of Oliver Rando, a fellow graduate student Hsieh Tsung Han developed a technology called Micro-C, which uses Micrococcal Nuclease (a non-specific DNase that is inhibited/impeded by nucleosomes/histone octamers bound to DNA) of DNA instead of restriction enzymes, to improve the limits of Hi-C approach from a few kilobases down to the resolution of a nucleosome (147bp). Having browsed various Hi-C maps previously made in fruit flies, and given the compactness of the fruit fly genome, I knew that the resolution limits of Hi-C was a major issue. If we wanted to draw strong conclusions about potentially subtle changes in 3D space between enhancers and promoters of genes that are active or repressed, we needed as much resolution as we could get.
I met Juanma at a Keystone Chromatin meeting in Whistler, Canada two years ago, when I was thinking about wrapping up my dissertation and finding a postdoc position somewhere. Again, fate would have it that the postdoc I worked closely with on my main project, Ana Boskovic (who now runs her own lab at EMBL Rome), was close friends with Juanma back when she was a graduate student and he was a postdoc at Maria-Elena Torres Padilla’s lab when she herself was at Max Planck (now she directs the Institute of Epigenetics and Stem Cells in Munich). Juanma, Ana and I would go on to conference together. We even after, after, after partied together.
When I saw Juanma’s tweet, I was completely stunned. Not by the results, but because I was in the middle of doing the exact same experiments, using Micro-C instead of Hi-C. I had just wrapped up analysis for one mutant, and was ramping up for experiments on another mutant, when the tweets hit me one night. I was cursing the science gods. I panicked. I was frustrated. Mind you this was in the middle of the COVID pandemic shutdown, and I was already frustrated with homeschooling a toddler, and an inability to work in the lab to push my fragile postdoc career in its infancy forward. Weirdly, I was relieved as well. My hunch was correct! The experiments were indeed good experiments - two other groups did it! They also found the same puzzling results that I found, which I had just presented a couple of weeks before in a lab meeting. Mike was shocked as well when he saw my results and had pushed me to do more follow up experiments. What would he say when he heard that we had been scooped, by two alumni of the lab no less?
Sometimes when your back is against the wall, a good idea emerges. Since I had nothing to lose, and given my acquaintance with Juanma, I decided to write him an email. Would they be interested in bringing my results into the paper to support their conclusions? I knew from reading their paper that Hi-C was indeed not good enough for them to draw clear cut conclusions. The fruit fly genome was simply too compact. To my confusion, Juanma did not write back. I told Mike, we forgot about it, and I pursued other experiments I had been working on in the background in earnest.
To my surprise, a month later, Liz - the first author of the paper - wrote me an email out of the blue. She asked me whether I had thought more about the collaboration and suggested setting up a Zoom meeting. I was confused. This email made it seem like I did not respond to their email. I rummaged through my Junk Mail, and there it was. Juanma did write me back and was very happy with my proposal. They actually needed my data. Turns out, the reviewers agreed with my own personal analysis of their paper. The reviewers wanted a higher resolution view of 3D chromatin architecture as well, and my Micro-C results were a perfect fit. Elation! I had an in and my efforts would not go to waste. Sometimes, a little vulnerability goes a long way, and a friendly hand across the pond doesn’t hurt either.
When I finally broached the idea with Mike, he was initially skeptical. This was expected; he would have liked us to go the competitive route and come out with our own paper, but I convinced him against this. We were behind Juanma’s group in terms of actual data acquisition, and they had already submitted the paper. Our paper would be rushed and incomplete, whilst joining forces will provide a more compelling study that would clarify the story for the entire field. In any case, he finally acquiesced.
We quickly set up the collaboration. They were awestruck by the resolution that Micro-C allowed. The results were clear as day - chromatin was folded in the same way, whether genes were on or off. Liz and Juanma worked my results into their manuscript, and we re-submitted together. Soon after, our paper was accepted. Turns out, Mounia’s group did not have such luck and were asked for experiments that would take many more months to complete, made especially difficult as it was in the middle of a COVID shutdown year. Eventually, Mounia’s paper got accepted as well. One big happy ex- and current Levine lab member reunion party!
So 3D chromatin architecture doesn't matter?
The results from these two papers, out now in Nature Genetics, have shocked the entire transcription and 3D genome organization field. People are perplexed. I’ve had my ear to the ground since the original BioRxiv papers came out, and the word is that the people are not happy. Traditionally, the field of 3D genome organization had deep roots in fruit fly research. More recently, it has been basically taken over by the mammalian cell culture crowd. The results from these two papers run contrary to the majority of mammalian results, with an extremely large group of scientists as followers of the mantra that enhancer and promoter looping equals gene activity. It must be a fruit fly thing, they say. It cannot be that enhancers and promoters being in close proximity does not result in gene activation! It simply cannot be! The papers are shaping up to be widely ignored by the mainstream mammalian chromatin crowd, or marked down as some fruit fly eboration. Of course, this has a lot to do with funding - so much money has been spent trying to understand this very question of the relationship between transcription and 3D genome organization (hundreds of millions of dollars), that the answer cannot be that it DOES NOT MATTER!
Or can it? A more shrewd, objective observer of the data would come out with a different conclusion altogether. As noted previously, much of the mammalian crowd have ignored transcriptional repression for decades. They have simply shrugged it off, filtered it into their mental Junk Mail.
What the studies actually show is that not only gene activation driven by enhancer-promoter proximity; repression is as well. Enhancers enhance gene activity when bound by transcriptional activators, but they repress gene activity when bound by transcriptional repressors. The genome goes through all the trouble of bringing the enhancers and promoters together, so that either activators or repressors can communicate their message to the promoter. In other words, the state of the gene, whether on or off, depends on the type of transcription factors that are present in that cell, in that specific nuclear environment, not by how close the enhancer and promoters are in 3D space. 3D genome organization is not sufficient for gene activation. From all previous evidence, it is essential and possibly necessary, but it’s not sufficient.
Explained another way, during embryogenesis, the genome folds in a way such that genes are “primed” for activation. When the appropriate extracellular signal is received, such as BMP signaling, cascading protein modification events (including possibly formation or dissolution of transcriptional condensates) lead to recruitment of activators and/or dispersion of repressors. The actual chromatin folding or architecture is already set up for this to occur in a timely manner: all the gene needs is to switch up the proteins in its regulatory milieu to actualize its expression program - a gene switch! In this way, a developing embryo can shuffle through numerous alliterations of self-perpetuating gene programs quickly, through the timely coordination of thousands of genes that have to be regulated and expressed to perfection.
So is the genome like a bowl of ramen? I think we’d all agree that it’s a little more complicated than that. The nucleus continues to hold its juicy secrets for us to discover. While science can feel like a nutritionless broth, occasionally the flavor profile can be just right. With time, as better technology becomes available, and as we think of more precise and insightful ways to ask the right questions, the ramen will keep delivering that warm and nourishing mouthful.
In part 2 of my series of 3D genome organization, I will discuss our upcoming paper on some of the rules of chromatin folding in the early embryo, and how it controls transcription dynamics.
Comments
Post a Comment