Placement: Promote to Homepage

Eight from MIT named 2021 Sloan Research Fellows

Awards honor, support young professors in the Media Lab and departments of Biology, Brain and Cognitive Sciences, Chemical Engineering, EECS, and Mathematics.

MIT News Office
February 19, 2021

The Alfred P. Sloan Foundation announced Feb. 16 that it has awarded Sloan Research Fellowships to eight MIT professors in the MIT Media Lab and in the departments of Biology, Brain and Cognitive Sciences, Chemical Engineering, Electrical Engineering and Computer Science, and Mathematics. The fellowships, which honor pre-tenure faculty members, will support their research with two-year, $75,000 awards.

“The Sloan Research Fellowship Program recognizes and rewards outstanding early-career faculty who have the potential to revolutionize their fields of study,” according to the Sloan Foundation.

Fadel Adib, associate professor and Doherty Chair in Ocean Utilization, directs the Signal Kinetics group at the MIT Media Lab. His group invents, builds, and deploys wireless and sensor technologies to address complex problems in society, industry, and ecology. His team’s work focuses on bringing wireless capabilities to extreme domains like the ocean and the human body and to enable new applications that are infeasible using today’s technologies. His research extends beyond communication and networking to enabling novel micro-sensing, powering, and perception tasks. These capabilities aim at helping address major societal challenges in health care, climate change, and automation.

“We are excited about continuing to push our technologies deeper into the oceans and the human body,” says Adib, whose team invented the world’s first net-zero power underwater communication technology and wireless systems that power and communicate with batteryless micro-implants inside the human body. He aims to use this funding to further his team’s efforts in underwater GPS, in-body sensing, and robotic automation. “This Sloan fellowship will allow my team to continue taking risks in pursuing high-impact projects to understand and address global challenges ranging from climate change to health care and automation.”

Joseph “Joey” Davis, the Whitehead Career Development Assistant Professor in Biology, investigates the massive molecular “machines” that carry out important cellular processes, such as protein synthesis and degradation. He uses cryo-electron microscopy (cryoEM) to visualize these molecules at near-atomic resolution as they are being assembled and changing shape while they work. In collaboration with Simons Professor of Mathematics Bonnie Berger, his team has developed a new computational tool. Called cryoDRGN, it leverages neural networks to extract molecular motions from cryoEM data and create 3D movies. The Sloan Foundation award will help Davis combine cryoDRGN with a related imaging technique — electron cryotomography — to observe molecular structures directly inside living cells. Using this powerful combination, he hopes to uncover how machines like the ribosome, which synthesizes proteins, assemble in their native cellular environment. Ultimately, he aims to identify new antibiotic targets in this assembly pathway.

“We hope that the combination of cryoDRGN and electron cryotomography will enable us to directly visualize how key molecular machines are assembled within the cell,” Davis says. “This information will be critical in truly understanding how nature builds these machines so rapidly and efficiently, and will help us understand what aspects of the assembly process fail when cells are mutated and as they age. I am incredibly grateful to the Sloan Foundation for their support of our work.”

Steven Flavell, the Lister Brothers Career Development Assistant Professor in Brain and Cognitive Sciences and The Picower Institute for Learning and Memory, said the Sloan Foundation’s award will help him conduct experiments to uncover how animal nervous systems generate internal states that represent needs and desires, such as hunger, and then produce behaviors, such as roaming around in search of food. His lab plans to use a multidisciplinary experimental approach in their studies, which employ the simple model of the C. elegans worm whose nervous system contains only 302 neurons. Though simple, the model has proven to produce important insights across many areas of biology.

“Over the course of each day, an animal’s nervous system may transition between a wide range of internal states that influence how sensory information is processed and how behaviors are generated,” Flavell says. “These states of arousal, motivation, and mood can persist for hours, play a central role in organizing human behavior, and are commonly disrupted in psychiatric disease. However, the fundamental neural mechanisms that generate these states remain poorly understood. We envision that these studies will ultimately reveal fundamental principles of neural circuit function that may generalize across animals.”

Heather Kulik, associate professor in chemical engineering, advances first-principles and machine learning computational chemistry to accelerate materials and catalyst discovery. Her group has developed the first machine learning models capable of predicting normally time-consuming quantum mechanical properties of transition metal complexes, rapidly uncovering design principles in weeks instead of lifetimes. Her group develops large-scale quantum mechanical modeling methods and applies them to reveal how enzymes work and how to take inspiration from nature to design next-generation catalysts.

“The award from the Sloan Foundation will enable my group to continue advancing computational materials and bio-inspired catalyst discovery,” Kulik says. “The flexible nature of the support ensures we can continue to push forward these interdisciplinary efforts at the boundaries between fields.”

Luquiao Liu, associate professor in the Department of Electrical Engineering and Computer Science, focuses on understanding and exploiting spin-related physics in solid-state material and devices. Most recently, Luqiao has been doing research on developing material and carrying out electrical measurement on charge-spin interactions to achieve electrically induced magnetic switching, and exploring new methods to realize quantum control over the transport of magnons and other quasiparticles, which could be useful in future hybrid quantum systems for information processing.

“This fellowship will strengthen our capabilities in identifying new material and physics mechanisms that can be used to achieve functions that are unique to spintronic systems, with the long-term goal of realizing efficient computing in the classical and quantum domains,” he says.

Karthish Manthiram, the Theodore T. Miller Career Development Chair and assistant professor in chemical engineering, studies the carbon footprint behind most chemicals and materials that we encounter every day — there is a carbon footprint associated even with the fabric of the clothes we wear, the food we eat, and the disinfectants we spray, he says. To find ways to synthesize these chemicals and materials in a sustainable manner that eliminates the carbon footprint, the Manthiram lab is pioneering the development of a paradigm in which carbon dioxide, dinitrogen, and water can be converted into a wide range of chemicals and materials using renewable electricity. In essence, this would mean that a device that breathes air, drinks water, and takes in solar photons could in principle someday make many of the chemicals that society relies on. The lab specifically looks for ways to facilitate the molecular-level dance through which chemical bonds are broken and formed, so that desired molecules can be made more selectively, efficiently, and at faster rates.

“The support of the Sloan Research Fellowship will allow my group to advance the decarbonization of the material world, through electrically driven synthesis of critical chemicals beginning with just carbon dioxide, dinitrogen, and water,” Manthiram says. “We will pursue new frontiers in synthesizing even more complex molecules starting with these ubiquitous feedstocks.”

Dor Minzer, assistant professor of mathematics, works in the fields of mathematics and theoretical computer science. His interests revolve around computational complexity theory, or — more explicitly — probabilistically checkable proofs, Boolean function analysis, and combinatorics. Minzer’s more recent research has utilized and extended some of the insights gained from the work on probabilistically checkable proofs in order to make progress on several open problems in the field of analysis of Boolean functions, such as the Fourier Entropy conjecture and the stability problem for the edge isoperimetric inequality, as well as to other problems in theoretical computer science.

“The Sloan fellowship will allow us to continue pursuing difficult and important challenges in theoretical computer science, whose solution is likely to have wide impact on the field,” Minzer says.

Lisa Piccirillo, assistant professor mathematics, specializes in the study of three- and four-dimensional spaces. She is broadly interested in low-dimensional topology and knot theory, and employs constructive techniques in four-manifolds. Her work in four-manifold topology has surprising applications to the study of mathematical knots. She received an inaugural 2021 Maryam Mirzakhani New Frontiers Prize, created in 2019 by the Breakthrough Foundation to recognize outstanding early-career women in mathematics, for “resolving the classic problem that the Conway knot is not smoothly sliced.” For all other small knots, “sliceness” is readily determined, but this particular knot had remained a mystery since John Conway presented it in the mid-1900s. After hearing about the problem at a conference, Piccirillo took only a week to formulate a proof.

In all, the Sloan Foundation awarded fellowships to 128 tenure-track, but not-yet-tenured, scholars in the United States and Canada this year.

Vander Heiden and Lourido receive promotions
February 18, 2021

Effective July 1, Matthew Vander Heiden and Sebastian Lourido will be promoted to Full Professor and Associate Professor (Without Tenure), respectively.

Vander Heiden is Associate Director of the Koch Institute for Integrative Cancer Research, a member of the MIT Center for Precision Cancer Medicine, a member of the Ludwig Center for Molecular Oncology, and a member of the Broad Institute. He joined the department in 2010 and earned tenure in 2017. His work focuses on the biochemical pathways cells use and how they are regulated to meet the metabolic requirements of cells in different physiological situations. His lab investigates the role of metabolism in cancer, particularly how metabolic pathways support cell proliferation. They aim to translate their understanding of cancer cell metabolism into novel cancer therapies. His promotion to full professor reflects his international standing in his field, his excellent and dedicated teaching, and his service to the department and the broader scientific community.

Lourido is a member of the Whitehead Institute and Latham Family Career Development Professor. He joined the department and Whitehead Institute in 2017. His lab is interested in the molecular events that enable apicomplexan parasites to remain widespread and deadly, infectious agents. They study many important human pathogens, including Toxoplasma gondii, to model features conserved throughout the phylum. They seek to expand our understanding of eukaryotic diversity and identify specific features that can be targeted to treat parasite infections.

Posted: 2.18.21
Catching key moments of cancer progression
Whitehead Institute
February 9, 2021

Important moments of cancer — mutation, tumor formation, metastasis —  are fleeting, easy-to-miss events. Even with modern medical technologies and methods, they often happen unobserved, and are only realized later when these cells spawn life-threatening conditions.

In recent years, however, new methods of tracking individual cells through time have allowed researchers to get closer to the origin of these events, and Whitehead Institute scientists are turning the power of these technologies to study cells involved in several different types of cancer. “With [these technologies], you can track down rare events in the past, identify all the offspring of an event, and see how they’ve changed their behavior,” said Whitehead Institute Member Jonathan Weissman.  “You can ask, when a cell picks up an oncogene, how does it mutate, and further evolve, and proliferate and metastasize?”

Read on to learn how three Whitehead Institute Members are using specially engineered mice, CRISPR-based technologies, and other methods to track cells at different stages of cancer development, pushing the boundaries of what we understand about how the disease starts and proliferates. From the initial beginnings of a tumor, sometimes in the darkness of a still-forming embryo, to a tumor’s growth and eventual metastasis to other sites in the body, Whitehead Institute scientists Rudolf Jaenisch, Jonathan Weissman, and Robert Weinberg study the pivotal points in a cancer’s growth and spread.
The birth of a tumor

For around 800 children each year in the United States, the seeds are sown during fetal development for a rare and unpredictable childhood cancer called neuroblastoma. To understand how the disease develops, scientists need to study what happens to these cancerous “seeds” as the embryo  matures. But they can hardly study a living human embryo, and fetal development is such a complex process that it is near impossible to simulate it in the lab.

To make matters more complex, the cancer grows and then may shrink unpredictably after the children are born and as they age. Sometimes the tumors disappear on their own; other times they grow uncontrollably.

In 2020, Institute Founding Member Rudolf Jaenisch’s lab introduced a new way of tracking the cells involved in the disease using his tried and true method for modeling such complex conditions: chimeric mice. A chimera is a conglomeration of two species — in this case, mostly mouse, with a few strategically placed human stem cells.

To create chimeras to study neuroblastoma, Jaenisch, who is also a Professor of Biology at Massachusetts Institute of Technology (MIT), along with collaborators at the Koch Institute for integrative Cancer Research at MIT and the Dana-Farber Cancer Institute, engineered human stem cells with glowing proteins to make them easy to see under the microscope. The cells contained a special genetic-switch that allowed the researchers to induce tumors by adding a certain chemical. These human stem cells were then induced to form a more specialized cell type – a neural crest stem cell.

Neural crest cells are a group of developing stem cells that go on to form the peripheral nervous system as well as other parts of the body such as the facial bones. It is neural crest cells that mutate into neuroblastoma tumors in humans, so the researchers hoped that by using these cells, they could create “human” tumors in mice. The researchers injected these human neural crest cells into mice so that they could readily incorporate with the host’s cells.

After the mice were born, the researchers were then able to take samples from the mice over the course of their development to see whether these implanted cells would form neural crest-derived tumors and, if so, what happened as they grew — something they would never have been able to do with neuroblastoma tumors in human infants and children.

The tumors in the chimeric mice pups developed in similar forms to human neuroblastomas — specifically, they formed characteristic rosette shapes — very similar to those seen in patient’s samples. With the help of Stefani Spranger, an assistant professor of biology at MIT and a Member of the Koch Institute for Integrative Cancer Research at MIT, they were able to track the cells’ interactions with the mice’s immune systems, and learn how the cancer “tricks” the immune cells into letting it stick around.

Now that they are able to model the formation of neuroblastoma tumors, the researchers hope to find a way to eliminate the tumors in the mice. “This is a model that will allow us to approach how to get rid of the tumors,” said Malkiel Cohen, a former postdoc in Jaenisch’s lab and first author of the paper, published in the journal Cell Stem Cell describing the work.

Cancer genealogy

A recent project from the lab of Whitehead Institute Member Jonathan Weissman focuses on another essential moment of cancer progression: metastasis. Metastasis happens when cancer spreads from a primary tumor to distant places in the body. Weissman, an expert in genome editing, created a CRISPR-based method to track the lineage of individual cancer cells in real time as they proliferate and metastasize.

Weissman and his collaborators, including graduate student Matthew Jones and then postdoctoral researcher Jeffrey Quinn, adapted the technology from a similar tool designed by Michelle Chan, a former postdoc in Weissman’s former lab at the University of California, San Francisco (UCSF) who is now an assistant professor at Princeton University. Chan designed a CRISPR mechanism to track the lineages of embryonic cells as they developed into specialized tissues.

“What Michelle Chan was able to do was uncover how tissues that look very similar to one another, actually come from disparate sources,” said Matthew Jones, a graduate student in Weissman’s lab and a co-first author on the paper describing the new method. “That has rapid implications about how tissues organize themselves, and we wanted to apply it to cancer.”

To create this system, the researchers engineered cancer cells with added genes: one for Cas9, the gene that codes for CRISPR’s “molecular scissors,” others for fluorescing proteins for microscopy, and a few sequences that would serve as targets for the CRISPR technology.

They then implanted thousands of these cells into the lungs of mice to simulate a tumor, using a model designed by Trever Bivona, a cancer biologist at UCSF. As the cells in the model tumors began to divide, the Cas9 protein began making small snips in the target sites in the cells’ DNA. When the cells fixed these snips, they patched in or deleted a few random nucleotides, leading to a unique “barcode” in each cell. Because these barcodes were added to each cell’s DNA, they were heritable and able to be passed on through generations of cells.

With the help of Nir Yosef, a computer scientist at the University of California, Berkeley, the researchers organized the data into “family trees” of cancer cells spanning multiple generations. By taking samples from different areas of the body, the researchers were able to resolve exactly when a cell jumped from where it started, in the lungs, to a distant tissue.

When they compared the trees, the researchers noticed that some cells were highly metastatic, jumping around multiple times over the course of the experiment, while others stayed put throughout. “We were excited to uncover some of these really rare, but consequential events that happened in the past that you would never be able to observe, and rarely be able to infer from a static snapshot,” Jones said.

By comparing highly metastatic and non-metastatic cells, they were able to identify metastasis-associated genes and answer questions about how the tumors were evolving and adapting. “It’s an entirely new way to look at the behavior and evolution of a tumor,” Weissman said. “We think it can be applied to many different problems in cancer biology.”

The next steps

A natural tumor begins with a single cell, mutated in a way that leads it to “go rogue,” so to speak. To mimic this in a model system, Dian Yang, another post-doc in the Weissman Lab, is collaborating with researchers in the lab of MIT Professor of Biology and Koch Institute Director Tyler Jacks’ to create a mouse model with the CRISPR lineage recording tool embedded in its DNA.

The model is based on a mouse model created by Jacks’ lab to study lung cancer. The lab has created genetically engineered mice that when left alone, live completely normal lives. But upon adding a trigger (an enzyme called “Cre”) — in Jacks’ and Yang’s case, they deliver the trigger to the lung using a virus — oncogenic signals are activated and lead to spontaneous tumors in the mice’s lungs.

Being able to “switch on” the mouse model in this way has a number of advantages.  “Each tumor will start from one single transformed cell, which we can then watch in its native environment as it evolves,” Yang said. “Then, we can look back later at how the cells metastasize.”

Adding the CRISPR system to Jacks’ existing cancer model will also allow researchers to study cancerous cells on a broader time scale. “Usually, when we harvest samples from the mice, it is like a snapshot, just one sample at one stage,” Yang said. “You can see what it looks like, you can analyze gene expression at the time of sampling, you can even take a time series, but you don’t know what happened in the past.”

In a sense, the lineage recording technology embedded in the genome of the mice now makes it possible to look back in time. “When you have a million cells in a tumor, you can use the lineage network of these cells to find out how they’re related, and connect the current state with the past evolutionary lineage history,” he said. “I think this will provide a new dimension of biological information for us to understand biology that is not just limited to cancer biology.”The making of metastasis

Weissman’s lab’s method for tracking the lineage of cancer cells can illuminate the nature of the cells that leave the primary tumor and scatter throughout the body. But what actually happens to these cells to cause them to metastasize?

That’s where Whitehead Institute Founding Member Robert Weinberg’s research comes in. Weinberg has been studying cancer for decades. His early work helped to answer the question of how cells that form a primary tumor become cancerous. Weinberg identified the first human oncogene, a gene that causes otherwise-normal cells to form tumors. This finding, combined with others, demonstrated that cancer is a disease driven by damaged genes, at least in its origins. Weinberg has since turned his attention to the question of how cancer cells acquire the ability to spread.

Around 90 percent of cancer deaths are caused not by the primary tumor, but by its metastasis. Based on previous work, there is no single genetic switch that can be flipped to equip a cancer cell for metastasis. Instead, cells must go through a series of changes over time. Most cancer cells fail to acquire all of the necessary traits, and so may, for example, spread to new tissues but rarely form tumors there. Weinberg’s lab tracks cancer cells to help fill in the “road map” that cancer cells follow on the way to metastasis, in the hopes that their insights can be used to prevent or treat metastatic cancers.

One important change that cancer cells undergo is called the epithelial to mesenchymal transition (EMT), a cellular process that causes the cancer cells to express different genes and go from being immobile to mobile and invasive. Cancer cells undergoing this transition to be able to spread are called “quasi-mesenchymal.” The ability to spread does not fully explain metastasis, however.

“There’s two aspects of the metastasis problem,” Weinberg said. “The first aspect is how cancer cells physically leave the primary tumor and get seeded into a distant tissue. In other words, the physical translocation of the cells. The second step represents the subsequent ability of the already seeded cells to figure out how to make a living in a distant tissue.”

In other words, how do those transplanted cells adapt to a new tissue environment, which might offer them only inhospitable conditions? “That represents the major unsolved problem of metastasis,” he said.

Weinberg hopes to study this more in the future; for now though, his lab has found that studying quasi-mesenchymal cells can serve another purpose. Anushka Dongre, a postdoc in Weinberg’s lab, found that these cells are resistant to a common type of cancer treatment, known as immune checkpoint therapy, and can even protect the other cancer cells around them from that therapy. If as little as ten percent of a tumor consists of quasi-mesenchymal cells, then the whole tumor may become resistant.

By using a tumor’s epithelial/mesenchymal profile, Dongre demonstrated that she could predict how likely a breast cancer tumor was to respond to a particular immune checkpoint therapy. This finding could help physicians match patients to the best treatment plan, by indicating ahead of time whether the treatment will work. She also identified a way to eliminate the quasi-mesenchymal cells’ protective effect by suppressing a certain enzyme that they employ to defend themselves.

Weinberg’s lab continues to study pivotal changes in the lives of cancer cells, such as the EMT, so that they can better understand metastasis and, they hope, help find effective treatments for patients with metastatic cancers.Tracking cells into the future

Scientists have been tracking cells for more than a century, and Whitehead Institute scientists will be tracking cancer cells for decades to come. In the coming years, Weinberg plans to continue to investigate the mysteries of metastasis. For Weissman’s part, he hopes to continue refining his CRISPR technique, with the end goal of eventually being able to predict the behavior of cancer cells. “We want to be able to measure where they are, where they’re going at any time, and then predict where they’re going to be in the future,” he said.

With new technologies and ever-expanding fields, there is limitless potential in these various methods. “That’s what is so exciting about the cell tracking field right now,” said Matt Jones. “It’s really pushing the boundaries on what we can capture from our measurements.”

Robert Weinberg receives 2021 Japan Prize

The award recognizes Weinberg’s pioneering achievements in the field of cancer biology.

Eva Frederick | Whitehead Institute
February 10, 2021

The Japan Prize Foundation has named MIT Professor Robert Weinberg as one of the recipients of its 2021 awards in the category of Medical Science and Medicinal Science, citing Weinberg’s contributions to the development of a multi-step model of how cancer begins and progresses, and the application of that model to improve cancer treatments and outcomes.

Weinberg, along with co-recipient Bert Vogelstein of the Johns Hopkins University School of Medicine, will receive his award in April at a presentation ceremony attended by the emperor and empress of Japan.  “Dr. Weinberg’s work has led to the identification of critical genes for cancer development that have subsequently been approved as therapeutic targets, resulting in thousands of lives being saved,” writes the Japan Prize Foundation in their news release.

“This award is a tribute to the brilliant scientists who have worked alongside me during my time at the Whitehead Institute,” says Weinberg, a Whitehead Institute founding member who is the Daniel K. Ludwig Professor for Cancer Research at MIT, as well as an extramural member of the David H. Koch Institute for Integrative Cancer Research at MIT.

In 1979, Weinberg and his lab discovered the first gene associated with tumor formation in humans, also known as an oncogene. In the decades since, he has devoted his career to studying not only the genetic basis of cancer, but also the ways in which cancerous cells spread and proliferate throughout the body. His work, along with Vogelstein’s, is credited with the development of new areas of cancer research, including the idea of targeted cancer therapies, and the broader field of precision medicine.

Weinberg joins a list of distinguished scientists from around the world who have received the prestigious Japan Prize, which is intended to express Japan’s gratitude to the international community. Each year, the Japan Prize Foundation selects two specialized fields of science and technology and solicits nominations from over a thousand scientists and engineers across Japan and abroad. This year, these scientists nominated 385 individuals, and three received a prize. In addition to Weinberg and Vogelstein, Martin A. Green, a professor at the University of New South Wales, was also honored this year, in the category of Resources, Energy, Environment, and Social Infrastructure.

“Weinberg’s work on oncogenes and tumor suppressor genes in cancer research has helped create the paradigm of cancer progression as we know it today, and has led the field of cancer biology in new and fruitful directions,” says Whitehead Institute director and MIT professor of biology Ruth Lehmann. “His research has laid the foundation for the development of new treatments that are improving the lives of cancer patients around the world.”

Lehmann receives Morgan Medal for scientific achievements
Whitehead Institute
January 27, 2021

Whitehead Institute director and Member Ruth Lehmann is the 2021 recipient of the Genetics Society of America’s (GSA) Thomas Hunt Morgan Medal for lifetime contributions to the field of genetics. The GSA is an international community of more than 5,000 scientists; and the Morgan Medal is one of the most prestigious awards for career achievement in the field of genetics.

The Morgan Medal was bestowed  in recognition for Lehmann’s groundbreaking work revealing the unique biology of the specialized cells that give rise to egg and sperm. Known as germ cells, they are the only cells in the body with the power to build a new organism and transmit this potential to future generations. Lehmann’s research using the fruit fly Drosophila melanogaster has uncovered the molecular mechanisms by which germ cells are distinguished from the other cells of the body and how they migrate into position during development of the embryo. These and other highly influential discoveries from the Lehmann lab have spawned insights into many aspects of animal development, cell migration, cell signaling, RNA regulation, genome integrity, and mitochondrial inheritance.

This is the second consecutive year that a Whitehead Institute Member has received the Thomas Hunt Morgan Medal. It was awarded in 2020 to Whitehead Founding Member and former director Gerald R. Fink.

The Davis and Berger labs combined cryo-electron microscopy and machine learning to visualize molecules in 3D.

February 4, 2021
Machine-learning model helps determine protein structures

New technique reveals many possible conformations that a protein may take.

Anne Trafton | MIT News Office
February 4, 2021

Cryo-electron microscopy (cryo-EM) allows scientists to produce high-resolution, three-dimensional images of tiny molecules such as proteins. This technique works best for imaging proteins that exist in only one conformation, but MIT researchers have now developed a machine-learning algorithm that helps them identify multiple possible structures that a protein can take.

Unlike AI techniques that aim to predict protein structure from sequence data alone, protein structure can also be experimentally determined using cryo-EM, which produces hundreds of thousands, or even millions, of two-dimensional images of protein samples frozen in a thin layer of ice. Computer algorithms then piece together these images, taken from different angles, into a three-dimensional representation of the protein in a process termed reconstruction.

In a Nature Methods paper, the MIT researchers report a new AI-based software for reconstructing multiple structures and motions of the imaged protein — a major goal in the protein science community. Instead of using the traditional representation of protein structure as electron-scattering intensities on a 3D lattice, which is impractical for modeling multiple structures, the researchers introduced a new neural network architecture that can efficiently generate the full ensemble of structures in a single model.

“With the broad representation power of neural networks, we can extract structural information from noisy images and visualize detailed movements of macromolecular machines,” says Ellen Zhong, an MIT graduate student and the lead author of the paper.

With their software, they discovered protein motions from imaging datasets where only a single static 3D structure was originally identified. They also visualized large-scale flexible motions of the spliceosome — a protein complex that coordinates the splicing of the protein coding sequences of transcribed RNA.

“Our idea was to try to use machine-learning techniques to better capture the underlying structural heterogeneity, and to allow us to inspect the variety of structural states that are present in a sample,” says Joseph Davis, the Whitehead Career Development Assistant Professor in MIT’s Department of Biology.

Davis and Bonnie Berger, the Simons Professor of Mathematics at MIT and head of the Computation and Biology group at the Computer Science and Artificial Intelligence Laboratory, are the senior authors of the study, which appears today in Nature Methods. MIT postdoc Tristan Bepler is also an author of the paper.

Visualizing a multistep process

The researchers demonstrated the utility of their new approach by analyzing structures that form during the process of assembling ribosomes — the cell organelles responsible for reading messenger RNA and translating it into proteins. Davis began studying the structure of ribosomes while a postdoc at the Scripps Research Institute. Ribosomes have two major subunits, each of which contains many individual proteins that are assembled in a multistep process.

To study the steps of ribosome assembly in detail, Davis stalled the process at different points and then took electron microscope images of the resulting structures. At some points, blocking assembly resulted in accumulation of just a single structure, suggesting that there is only one way for that step to occur. However, blocking other points resulted in many different structures, suggesting that the assembly could occur in a variety of ways.

Because some of these experiments generated so many different protein structures, traditional cryo-EM reconstruction tools did not work well to determine what those structures were.

“In general, it’s an extremely challenging problem to try to figure out how many states you have when you have a mixture of particles,” Davis says.

After starting his lab at MIT in 2017, he teamed up with Berger to use machine learning to develop a model that can use the two-dimensional images produced by cryo-EM to generate all of the three-dimensional structures found in the original sample.

In the new Nature Methods study, the researchers demonstrated the power of the technique by using it to identify a new ribosomal state that hadn’t been seen before. Previous studies had suggested that as a ribosome is assembled, large structural elements, which are akin to the foundation for a building, form first. Only after this foundation is formed are the “active sites” of the ribosome, which read messenger RNA and synthesize proteins, added to the structure.

In the new study, however, the researchers found that in a very small subset of ribosomes, about 1 percent, a structure that is normally added at the end actually appears before assembly of the foundation. To account for that, Davis hypothesizes that it might be too energetically expensive for cells to ensure that every single ribosome is assembled in the correct order.

“The cells are likely evolved to find a balance between what they can tolerate, which is maybe a small percentage of these types of potentially deleterious structures, and what it would cost to completely remove them from the assembly pathway,” he says.

Viral proteins

The researchers are now using this technique to study the coronavirus spike protein, which is the viral protein that binds to receptors on human cells and allows them to enter cells. The receptor binding domain (RBD) of the spike protein has three subunits, each of which can point either up or down.

“For me, watching the pandemic unfold over the past year has emphasized how important front-line antiviral drugs will be in battling similar viruses, which are likely to emerge in the future. As we start to think about how one might develop small molecule compounds to force all of the RBDs into the ‘down’ state so that they can’t interact with human cells, understanding exactly what the ‘up’ state looks like and how much conformational flexibility there is will be informative for drug design. We hope our new technique can reveal these sorts of structural details,” Davis says.

The research was funded by the National Science Foundation Graduate Research Fellowship Program, the National Institutes of Health, and the MIT Jameel Clinic for Machine Learning and Health. This work was supported by MIT Satori computation cluster hosted at the MGHPCC.