Essays on Science and SocietyGenomics, Proteomics, and Systems Biology

Three-dimensional genome structure of a single cell

See allHide authors and affiliations

Science  22 Nov 2019:
Vol. 366, Issue 6468, pp. 964-965
DOI: 10.1126/science.aaz7774

Since the 1880s, scientists such as Carl Rabl have been looking at cell nuclei under a microscope and speculating about their three-dimensional (3D) structure. We now know that each nucleus in our body carries 6 billion base pairs (bp) of DNA, which would be 2 m long if fully stretched. The linear sequence of this DNA was determined by the Human Genome Project in 2003; however, its 3D structure remains elusive.

Embedded Image

In recent decades, scientists began to uncover how the 2-m-long DNA folds into the 10-µm cell nucleus, taking advantage of next-generation DNA sequencing. This method, termed chromosome conformation capture [3C (1) or Hi-C (2)], provides a genome-wide view of DNA folding by averaging a large ensemble of cells. However, ensemble averaging cannot capture the relative positioning of different chromosomes and their interactions, because each chromosome, which occupies a contiguous territory in a cell nucleus, only contacts a few other chromosomes (3). Single-cell 3C has been achieved (4) but has limited spatial resolution.

New technology enables a holistic view of cell type–specific three-dimensional genome organizationGRAPHIC: N. DESAI/SCIENCE, FROM L. TAN

High-Precision Structures with Allelic Resolution

In Hi-C, a cell nucleus is chemically fixed and its genome cut into small pieces and then ligated on the basis of 3D proximity. This procedure creates numerous artificial linkages—termed chromosome “contacts”—between DNA loci that are distant on the linear genome but nearby in the 3D space.

Working in the laboratory of Xiaoliang Sunney Xie at Harvard University, I improved the resolution of single-cell 3C by inventing new whole-genome amplification methods. Through a highly uniform, transposon-based (5) amplification method termed multiplex end-tagging amplification (META) (6) and other improvements, I obtained an average of ∼1 million contacts per cell—five times as many as before. (High numbers of contacts are necessary for acquiring high-resolution 3D structures.)

The 23 chromosomes inherited from one's mother and the 23 from one's father differ by less than 0.1%, making it difficult to distinguish the parent of origin for each observed contact. Previous studies (7) have focused on haploid mouse cell lines, which are not applicable to diploid cells. I developed an algorithm to infer the parent of origin of each contact. The algorithm aggregates the minute (0.1%) parental difference from many neighboring contacts to accurately infer their common parent of origin. With these biochemical and computational advances—together termed diploid chromosome conformation capture (Dip-C)—we can now look at almost any cell in the human body (see the figure, left).

Diploid Human and Mouse Genomes in 3D

We achieved a resolution of 20,000 bp, or 100 nm, in the first 3D structure of a diploid human genome (6). That is, we simultaneously localized ∼300,000 DNA loci in each cell, without using a microscope. This capability is far beyond state-of-the-art optical imaging of chromosomes, which can achieve comparable resolution but can only measure ∼100 loci per cell, meaning it cannot resolve the whole genome.

Our structures provide a holistic 3D view of the human genome with unprecedented details. Zooming in from the whole nucleus to individual chromatin loops, we have visualized the fractal organization of DNA that was originally theorized from ensemble-averaged Hi-C measurements (2). In imprinted genes and female X chromosomes, where two alleles differ dramatically in gene expression, we observed clear structural differences as well. This highlights an intimate relationship between structure and function.

In each tissue, different types of cells perform different functions. We found that knowing the 3D genome structure alone allowed investigators to distinguish between the cell types in the absence of any other information. This “structure typing” not only puts Dip-C on the growing list of single-cell omics tools—such as transcriptomics (8) and epigenomics—for charting a complete cellular atlas of our body, it also provides a structural basis for the diverse functions of human cell types.

Embedded Image

An electron micrograph depicts human chromosomes and nucleus.


Structural Basis for Smell

To demonstrate the power of Dip-C, I turned to the mouse olfactory system. In the nose, neurons sense odors by expressing olfactory receptors (ORs) (9). ORs constitute the largest gene family in the mouse genome, consisting of ∼1100 G protein–coupled receptor (GPCR) genes from different chromosomes. Each neuron, however, only expresses a single OR, silencing all ∼1100 others (10) after a brief developmental period of multi-OR expression (11, 12). This “one neuron–one receptor” rule ensures that each neuron senses its own subset of odors and projects to the brain to form a map (13) of smell (see the figure, center).

Previous studies (14) have proposed interchromosomal interaction as a mechanism to achieve one neuron–one receptor expression. In this model, OR genes interact to ensure mutual silencing, while their enhancers jointly activate the one OR that will be expressed. However, older methods of fluorescence imaging cannot resolve all ORs at once.

Using Dip-C, I mapped all ∼1100 OR genes and their ∼60 enhancers in single cells throughout mouse development (15). What I learned is that, in a normal cell, where ORs are not expressed, most OR genes reside on the surface of the nucleus, and ORs from different chromosomes barely interact (see the figure, right). Only olfactory neurons bring together OR genes to form large, multichromosome aggregates at the nuclear center, which presumably silence all but one OR. Each cell harbors a few such aggregates, the largest of which summons nearly a dozen enhancers from different chromosomes. The formation of this enhancer hub coincides with the developmental transition in OR expression (11, 12). This distinct structure is the basis for one neuron–one receptor expression.

The importance of 3D genome organization goes beyond olfaction. It may enhance night vision by inverting euchromatin and heterochromatin to form a microlens (16), the 3D structure of which has also been determined by Dip-C (15). Chromosome conformation also helps our immune system perform V(D)J recombination, coordinates the formation of neural circuits through protocadherin genes, and holds therapeutics value for cancer epigenomics and genome editing. The ability to measure 3D genomes of single cells, then, will have far-reaching impacts on fundamental biology as well as on human health.

Embedded Image


Longzhi Tan

Longzhi Tan received his undergraduate degree from Massachusetts Institute of Technology and a Ph.D. from Harvard University. He started his postdoctoral fellowship at Stanford University in 2019. His research combines single-cell genomics and optogenetics to study the 3D chromatin basis of neurodevelopment and behaviors and to develop new tools for chromatin biology.

Embedded Image


Barbara C. Klump

Barbara Klump received her master's degree from Heidelberg University, Germany, and a Ph.D. from the University of St Andrews, UK. After completing a postdoctoral fellowship at the University of St Andrews, she moved to the Max Planck Institute of Animal Behavior in Radolfzell, Germany, where she is currently a postdoc in the Cognitive and Cultural Ecology Lab. Her research explores how a species's ecology shapes its behavioral repertoire.

Embedded Image


Humsa Venkatesh

Humsa Venkatesh received her undergraduate degree from the University of California, Berkeley, and her Ph.D. from Stanford University. She is currently completing her postdoctoral fellowship at Stanford University. Her research combines principles of neuroscience and cancer biology to understand the electrical components of cancer pathophysiology and harness these malignant dependencies for therapeutic intervention.

Embedded Image


Zibo Chen

Zibo Chen received his undergraduate degree from the National University of Singapore and his Ph.D. in biochemistry from the University of Washington. He is currently a postdoctoral scholar at the California Institute of Technology, where he is programming mammalian cells using proteins designed from scratch.

References and Notes

Acknowledgments: I thank my thesis adviser, X. S. Xie, for being a fantastic mentor, a visionary scientist, and a wonderful human being. I am also grateful to all members of the Xie laboratory and to my collaborators.
View Abstract

Stay Connected to Science

Navigate This Article