Skip to main content


New Tool Predicts Response to Immunotherapy in Lung Cancer Patients

Posted on by

A purple irregular cell is releasing purple particles. It is surrounded my smoother blue cells. National Institutes of Health.
Credit: XVIVO Scientific Animation, Wethersfield, CT

With just a blood sample from a patient, a promising technology has the potential to accurately diagnose non-small cell lung cancer (NSCLC), the most-common form of the disease, more than 90 percent of the time. The same technology can even predict from the same blood sample whether a patient will respond well to a targeted immunotherapy treatment.

This work is a good example of research supported by the NIH Common Fund. Many Common Fund programs support development of new tools that catalyze research across the full spectrum of biomedical science without focusing on a single disease or organ system.

The emerging NSCLC prediction technology was developed as part of our Extracellular RNA Communication Program. The program develops technologies to understand RNA circulating in the body, known as extracellular RNA (exRNA). These molecules can be easily accessed in bodily fluids such as blood, urine, and saliva, and they have enormous potential as biomarkers to better understand cancer and other diseases.

When the body’s immune system detects a developing tumor, it activates various immune cells that work together to kill the suspicious cells. But many tumors have found a way to evade the immune system by producing a protein called PD-L1.

Displayed on the surface of a cancer cell, PD-L1 can bind to a protein found on immune cells with the similar designation of PD-1. The binding of the two proteins keeps immune cells from killing tumor cells. One type of immunotherapy interferes with this binding process and can restore the natural ability of the immune system to kill the tumor cells.

However, tumors differ from person to person, and this form of cancer immunotherapy doesn’t work for everyone. People with higher levels of PD-L1 in their tumors generally have better response rates to immunotherapy, and that’s why oncologists test for the protein before attempting the treatment.

Because cancer cells within a tumor can vary greatly, a single biopsy taken at a single site in the tumor may miss cells with PD-L1. In fact, current prediction technologies using tissue biopsies correctly predict just 20 – 40 percent of NSCLC patients who will respond well to immunotherapy. This means some people receive immunotherapy who shouldn’t, while others don’t get it who might benefit.

To improve these predictions, a research team led by Eduardo Reátegui, The Ohio State University, Columbus, engineered a new technology to measure exRNA and proteins found within and on the surface of extracellular vesicles (EVs) [1]. EVs are tiny molecular containers released by cells. They carry RNA and proteins (including PD-L1) throughout the body and are known to play a role in communication between cells.

As the illustration above shows, EVs can be shed from tumors and then circulate in the bloodstream. That means their characteristics and internal cargo, including exRNA, can provide insight into the features of a tumor. But collecting EVs, breaking them open, and pooling their contents for assessment means that molecules occurring in small quantities (like PD-L1) can get lost in the mix. It also exposes delicate exRNA molecules to potential breakdown outside the protective EV.

The new technology solves these problems. It sorts and isolates individual EVs and measures both PD-1 and PD-L1 proteins, as well as exRNA that contains their genetic codes. This provides a more comprehensive picture of PD-L1 production within the tumor compared to a single biopsy sample. But also, measuring surface proteins and the contents of individual EVs makes this technique exquisitely sensitive.

By measuring proteins and the exRNA cargo from individual EVs, Reátegui and team found that the technology correctly predicted whether a patient had NSCLC 93.2 percent of the time. It also predicted immunotherapy response with an accuracy of 72.2 percent, far exceeding the current gold standard method.

The researchers are working on scaling up the technology, which would increase precision and allow for more simultaneous measurements. They are also working with the James Comprehensive Cancer Center at The Ohio State University to expand their testing. That includes validating the technology using banked clinical samples of blood and other bodily fluids from large groups of cancer patients. With continued development, this new technology could improve NSCLC treatment while, critically, lowering its cost.

The real power of the technology, though, lies in its flexibility. Its components can be swapped out to recognize any number of marker molecules for other diseases and conditions. That includes other cancers, neurodegenerative diseases, traumatic brain injury, viral diseases, and cardiovascular diseases. This broad applicability is an example of how Common Fund investments catalyze advances across the research spectrum that will help many people now and in the future.


[1] An immunogold single extracellular vesicular RNA and protein (AuSERP) biochip to predict responses to immunotherapy in non-small cell lung cancer patients. Nguyen LTH, Zhang J, Rima XY, Wang X, Kwak KJ, Okimoto T, Amann J, Yoon MJ, Shukuya T, Chiang CL, Walters N, Ma Y, Belcher D, Li H, Palmer AF, Carbone DP, Lee LJ, Reátegui E. J Extracell Vesicles. 11(9):e12258. doi: 10.1002/jev2.12258.


NIH Common Fund

Video: Unlocking the Mysteries of Extracellular RNA Communication (Common Fund)

Extracellular RNA Communication Program (ERCC) (Common Fund)

Upcoming Meeting: ERCC19 Research Meeting (May 1-2, 2023)

Eduardo Reátegui Group for Bioengineering Research (The Ohio State University College of Engineering, Columbus)

Note: Dr. Lawrence Tabak, who performs the duties of the NIH Director, has asked the heads of NIH’s Institutes, Centers, and Offices to contribute occasional guest posts to the blog to highlight some of the interesting science that they support and conduct. This is the 27th in the series of NIH guest posts that will run until a new permanent NIH director is in place.

Visualizing The Placenta, a Critical but Poorly Understood Organ

Posted on by

Ultrasound image of placenta
Ultrasound images showing fetal (left) and maternal (right) placental vasculature. Credit: Eastern Virginia Medical School and University of Texas Medical Branch

The placenta is the Rodney Dangerfield of organs; it gets no respect, no respect at all. This short-lived but critical organ supports pregnancy by bringing nutrients and oxygen to the fetus, removing waste, providing immune protection, and producing hormones to support fetal development.

It also influences the lifelong health of both mother and child. Problems with the placenta can lead to preeclampsia, gestational diabetes, poor fetal growth, preterm birth, and stillbirth. Although we were all connected to one, the placenta is the least understood, and least studied, of all human organs.

What we do know about the human placenta largely comes from studying it after delivery. But that’s like studying the heart after it’s stopped beating. It doesn’t help us predict complications in time to avert a crisis.

To fill these knowledge gaps, NIH’s Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD) developed the Human Placenta Project (HPP) to noninvasively study the placenta during pregnancy. Since 2014, this approximately $88 million collaborative research effort has been developing ultrasound, magnetic resonance imaging (MRI), and blood-based biomarker methods to study how the placenta functions in real time and in greater detail.

As illustrated in the image above, advanced ultrasound tools allowed HPP researchers at Eastern Virginia Medical School, Norfolk, and the University of Texas Medical Branch, Galveston, to gain a detailed look at the placenta’s intricate arrangement of blood vessels, or vasculature. By evaluating both fetal (left panel) and maternal (right panel) placental vasculature in 610 pregnant people starting at 13 weeks of gestation, the investigators aimed to identify early changes that predicted later complications.

They observed that such changes can start in the first trimester and affect both the vasculature and placental tissue. While further research is needed, these findings suggest that placental ultrasound monitoring can inform efforts to prevent and treat pregnancy complications.

Another HPP team led by Boston Children’s Hospital is developing an MRI strategy to monitor blood flow and oxygen transport through the placenta during pregnancy. Interpreting and visualizing MRI data of the placenta is challenging because of its variable shape, the tendency of muscles in the uterus to begin tightening or contracting well before labor [1], and other factors.

As shown in the video above, the researchers developed a way to account for the motion of the uterus and “freeze” the placenta to make it easier to study (left two panels of video) [2]. They also developed algorithms to better visualize the complex patterns of placental oxygen content during contractions (center panel) [3]. The scientists then carried out initial visualizations of blood flow through the placenta shortly after delivery (second panel from right) [4].

They now intend to map these MRI findings to the placenta itself after delivery (far right panel), which will allow them to explore how additional factors such as gene expression patterns and genetic variants contribute to placental function. Ultimately, they plan to apply these MRI techniques to monitor the placenta in real time during pregnancy and identify changes that indicate compromised function early enough to adjust maternal management as needed.

Other HPP efforts focus on identifying components in maternal blood that reflect the status of the placenta. For example, an HPP research team led by scientists at the University of California, Los Angeles, adapted non-invasive prenatal testing methods to analyze genetic material shed from the placenta into the maternal bloodstream. Their findings suggest that distinctive patterns in this genetic material detected early in pregnancy may indicate risk for later complications [5].

Another HPP team, led by investigators at Columbia University, New York, helped establish that extracellular RNAs (exRNAs) released by the placenta into maternal circulation reflect the placenta’s status at a cellular level beginning in the first trimester. To harness the potential of exRNA biomarkers, the investigators are optimizing methods to isolate, sequence, and analyze exRNAs in maternal blood.

These are just a few examples of the cutting-edge work being funded through the HPP, which complements NICHD’s longstanding investment in basic research to unravel the physiology of and real-time gene expression in the placenta. Unlocking the secrets of the placenta may one day help us to prevent and treat a range of common pregnancy complications, while also providing insights into other areas of science and medicine such as cardiovascular disease and aging. NICHD is committed to giving this important organ the respect it deserves.


[1] Placental MRI: Effect of maternal position and uterine contractions on placental BOLD MRI measurements. Abaci Turk E, Abulnaga SM, Luo J, Stout JN, Feldman H, Turk A, Gagoski B, Wald LL, Adalsteinsson E, Roberts DJ, Bibbo C, Robinson JN, Golland P, Grant PE, Barth, Jr WH. Placenta. 2020 Jun 1; 95: 69-77.

[2] Spatiotemporal alignment of in utero BOLD-MRI series. Turk EA, Luo J, Gagoski B, Pascau J, Bibbo C, Robinson JN, Grant PE, Adalsteinsson E, Golland P, Malpica N. J Magn Reson Imaging. 2017 Aug;46(2):403-412.

[3] Volumetric parameterization of the placenta to a flattened template. Abulnaga SM, Turk EA, Bessmeltsev M, Grant PE, Solomon J, Golland P. IEEE transactions on medical imaging. 2022 April;41(4):925-936.

[4] Placental MRI: development of an MRI compatible ex vivo system for whole placenta dual perfusion. Stout JN, Rouhani S, Turk EA, Ha CG, Luo J, Rich K, Wald LL, Adalsteinsson E, Barth, Jr WH, Grant PE, Roberts DJ. Placenta. 2020 Nov 1; 101: 4-12.

[5] Cell-free DNA methylation and transcriptomic signature prediction of pregnancies with adverse outcomes. Del Vecchio G, Li Q, Li W, Thamotharan S, Tosevska A, Morselli M, Sung K, Janzen C, Zhou X, Pellegrini M, Devaskar SU. Epigenetics. 2021 Jun;16(6):642-661. 


Human Placenta Project (Eunice Kennedy Shriver National Institute of Child Health and Human Development/NIH)

Preeclampsia (NICHD)

Understanding Gestational Diabetes (NICHD)

Preterm Labor and Birth (NICHD)

Stillbirth (NICHD)

Abuhamad Project Information (NIH RePORTER)

Grant Project Information (NIH RePORTER)

Devaskar Project Information (NIH RePORTER)

Williams Project Information (NIH RePORTER)

Note: Acting NIH Director Lawrence Tabak has asked the heads of NIH’s Institutes and Centers (ICs) to contribute occasional guest posts to the blog to highlight some of the interesting science that they support and conduct. This is the 10th in the series of NIH IC guest posts that will run until a new permanent NIH director is in place.

Millions of Single-Cell Analyses Yield Most Comprehensive Human Cell Atlas Yet

Posted on by

A field of playing cards showing different body tissues

There are 37 trillion or so cells in our bodies that work together to give us life. But it may surprise you that we still haven’t put a good number on how many distinct cell types there are within those trillions of cells.

That’s why in 2016, a team of researchers from around the globe launched a historic project called the Human Cell Atlas (HCA) consortium to identify and define the hundreds of presumed distinct cell types in our bodies. Knowing where each cell type resides in the body, and which genes each one turns on or off to create its own unique molecular identity, will revolutionize our studies of human biology and medicine across the board.

Since its launch, the HCA has progressed rapidly. In fact, it has already reached an important milestone with the recent publication in the journal Science of four studies that, together, comprise the first multi-tissue drafts of the human cell atlas. This draft, based on analyses of millions of cells, defines more than 500 different cell types in more than 30 human tissues. A second draft, with even finer definition, is already in the works.

Making the HCA possible are recent technological advances in RNA sequencing. RNA sequencing is a topic that’s been mentioned frequently on this blog in a range of research areas, from neuroscience to skin rashes. Researchers use it to detect and analyze all the messenger RNA (mRNA) molecules in a biological sample, in this case individual human cells from a wide range of tissues, organs, and individuals who voluntarily donated their tissues.

By quantifying these RNA messages, researchers can capture the thousands of genes that any given cell actively expresses at any one time. These precise gene expression profiles can be used to catalogue cells from throughout the body and understand the important similarities and differences among them.

In one of the published studies, funded in part by the NIH, a team co-led by Aviv Regev, a founding co-chair of the consortium at the Broad Institute of MIT and Harvard, Cambridge, MA, established a framework for multi-tissue human cell atlases [1]. (Regev is now on leave from the Broad Institute and MIT and has recently moved to Genentech Research and Early Development, South San Francisco, CA.)

Among its many advances, Regev’s team optimized single-cell RNA sequencing for use on cell nuclei isolated from frozen tissue. This technological advance paved the way for single-cell analyses of the vast numbers of samples that are stored in research collections and freezers all around the world.

Using their new pipeline, Regev and team built an atlas including more than 200,000 single-cell RNA sequence profiles from eight tissue types collected from 16 individuals. These samples were archived earlier by NIH’s Genotype-Tissue Expression (GTEx) project. The team’s data revealed unexpected differences among cell types but surprising similarities, too.

For example, they found that genetic profiles seen in muscle cells were also present in connective tissue cells in the lungs. Using novel machine learning approaches to help make sense of their data, they’ve linked the cells in their atlases with thousands of genetic diseases and traits to identify cell types and genetic profiles that may contribute to a wide range of human conditions.

By cross-referencing 6,000 genes previously implicated in causing specific genetic disorders with their single-cell genetic profiles, they identified new cell types that may play unexpected roles. For instance, they found some non-muscle cells that may play a role in muscular dystrophy, a group of conditions in which muscles progressively weaken. More research will be needed to make sense of these fascinating, but vital, discoveries.

The team also compared genes that are more active in specific cell types to genes with previously identified links to more complex conditions. Again, their data surprised them. They identified new cell types that may play a role in conditions such as heart disease and inflammatory bowel disease.

Two of the other papers, one of which was funded in part by NIH, explored the immune system, especially the similarities and differences among immune cells that reside in specific tissues, such as scavenging macrophages [2,3] This is a critical area of study. Most of our understanding of the immune system comes from immune cells that circulate in the bloodstream, not these resident macrophages and other immune cells.

These immune cell atlases, which are still first drafts, already provide an invaluable resource toward designing new treatments to bolster immune responses, such as vaccines and anti-cancer treatments. They also may have implications for understanding what goes wrong in various autoimmune conditions.

Scientists have been working for more than 150 years to characterize the trillions of cells in our bodies. Thanks to this timely effort and its advances in describing and cataloguing cell types, we now have a much better foundation for understanding these fundamental units of the human body.

But the latest data are just the tip of the iceberg, with vast flows of biological information from throughout the human body surely to be released in the years ahead. And while consortium members continue making history, their hard work to date is freely available to the scientific community to explore critical biological questions with far-reaching implications for human health and disease.


[1] Single-nucleus cross-tissue molecular reference maps toward understanding disease gene function. Eraslan G, Drokhlyansky E, Anand S, Fiskin E, Subramanian A, Segrè AV, Aguet F, Rozenblatt-Rosen O, Ardlie KG, Regev A, et al. Science. 2022 May 13;376(6594):eabl4290.

[2] Cross-tissue immune cell analysis reveals tissue-specific features in humans. Domínguez Conde C, Xu C, Jarvis LB, Rainbow DB, Farber DL, Saeb-Parsy K, Jones JL,Teichmann SA, et al. Science. 2022 May 13;376(6594):eabl5197.

[3] Mapping the developing human immune system across organs. Suo C, Dann E, Goh I, Jardine L, Marioni JC, Clatworthy MR, Haniffa M, Teichmann SA, et al. Science. 2022 May 12:eabo0510.


Ribonucleic acid (RNA) (National Human Genome Research Institute/NIH)

Studying Cells (National Institute of General Medical Sciences/NIH)

Human Cell Atlas

Regev Lab (Broad Institute of MIT and Harvard, Cambridge, MA)

NIH Support: Common Fund; National Cancer Institute; National Human Genome Research Institute; National Heart, Lung, and Blood Institute; National Institute on Drug Abuse; National Institute of Mental Health; National Institute on Aging; National Institute of Allergy and Infectious Diseases; National Institute of Neurological Disorders and Stroke; National Eye Institute

New Clues to Delta Variant’s Spread in Studies of Virus-Like Particles

Posted on by

About 70,000 people in the United States are diagnosed with COVID-19 each and every day. It’s clear that these new cases are being driven by the more-infectious Delta variant of SARS-CoV-2, the novel coronavirus that causes COVID-19. But why does the Delta variant spread more easily than other viral variants from one person to the next?

Now, an NIH-funded team has discovered at least part of Delta’s secret, and it’s not all attributable to those widely studied mutations in the spike protein that links up to human cells through the ACE2 receptor. It turns out that a specific mutation found within the N protein coding region of the Delta genome also enables the virus to pack more of its RNA code into the infected host cell. As a result, there is increased production of fully functional new viral particles, which can go on to infect someone else.

This finding, published in the journal Science [1], comes from the lab of Nobel laureate Jennifer Doudna at the Howard Hughes Medical Institute, the Gladstone Institutes, San Francisco, and the Innovative Genomics Institute at the University of California, Berkeley. Co-leading the team was Melanie Ott, Gladstone Institutes.

The Doudna and Ott teams have developed an exciting new tool to study variants of the coronavirus. It’s a lab construct called a virus-like particle (VLP). These specially made VLPs have all the structural proteins of SARS-CoV-2 (shown above), but they contain no genetic material. Consequently, they are non-infectious replicas of the real virus that can be studied safely in any lab. Scientists don’t have to reserve time in labs equipped with heightened levels of biosafety, as is required when working with whole virus.

The VLPs also allow researchers to explore changes found in the coronavirus’s other essential proteins, not just the spike protein on its surface. In fact, all of the SARS-CoV-2 variants of concern, as defined by the World Health Organization (WHO), carry at least one mutation within the same stretch of seven amino acids in a viral protein known as the nucleocapsid (N protein). This protein, which hasn’t been widely studied, is required for the virus to make more of itself. It is also involved in the virus’s ability to package and release infectious RNA.

In the Science paper, Doudna and colleagues took a closer look at the N protein. They did so by developing a special system that used VLPs to package and deliver viral RNA messages into human cells.

Here’s how it works: The VLPs include all four of SARS-CoV-2’s structural proteins, including the spike and N proteins. In addition, they contain the RNA sequence that allows the virus to recognize its genetic material within the cell, so that it can be packaged into the next generation of viral particles.

Though the particles look just like SARS-CoV-2 from the outside, they lack the vast majority of the viral genome on the inside. But they do have one other key component: a snippet of RNA that makes cells invaded by VLPs glow. In fact, the more RNA messages a VLP delivers, the brighter the cells will glow. It allowed the researchers to spot successful invasions, while also quantifying the amount of RNA a particular VLP packed into a cell.

The researchers then produced SARS-CoV-2 VLPs including four mutations that are universally found within the N proteins of more transmissible variants of concern. That’s when they discovered those variants produced and delivered 10 times more RNA messages into cells.

The increased RNA also fits with what has been observed in people infected with the Delta variant. They produce about 10 times more virus in their nose and throat compared to people infected with the older variants.

But did those findings match what happens in the real virus? To find out, the researchers and their colleagues tested the N protein mutation found in the Delta variant in a high-level biosafety lab. And, indeed, their studies showed that the mutated virus within infected human lung cells produced about 50 times more infectious virus compared to the original SARS-CoV-2 variant.

The findings suggest that the N protein could be an important new target for effective COVID-19 therapeutics, and that tracking newly emerging mutations in the N protein might also be important for identifying new viral variants of concern. This new system is a powerful tool, and one that can also be used for exploring how newly arising variants in the future might affect the course of this terrible pandemic.


[1] Rapid assessment of SARS-CoV-2 evolved variants using virus-like particles. Syed AM, Taha TY, Tabata T, Chen IP, Ciling A, Khalid MM, Sreekumar B, Chen PY, Hayashi JM, Soczek KM, Ott M, Doudna JA. Science. 2021 Nov 4:eabl6184.


COVID-19 Research (NIH)

Doudna Lab

NIH Support: National Institute of Allergy and Infectious Diseases

Artificial Intelligence Accurately Predicts RNA Structures, Too

Posted on by

A mechanical claw grabs molecular models
Credit: Camille L.L. Townshend

Researchers recently showed that a computer could “learn” from many examples of protein folding to predict the 3D structure of proteins with great speed and precision. Now a recent study in the journal Science shows that a computer also can predict the 3D shapes of RNA molecules [1]. This includes the mRNA that codes for proteins and the non-coding RNA that performs a range of cellular functions.

This work marks an important basic science advance. RNA therapeutics—from COVID-19 vaccines to cancer drugs—have already benefited millions of people and will help many more in the future. Now, the ability to predict RNA shapes quickly and accurately on a computer will help to accelerate understanding these critical molecules and expand their healthcare uses.

Like proteins, the shapes of single-stranded RNA molecules are important for their ability to function properly inside cells. Yet far less is known about these RNA structures and the rules that determine their precise shapes. The RNA elements (bases) can form internal hydrogen-bonded pairs, but the number of possible combinations of pairings is almost astronomical for any RNA molecule with more than a few dozen bases.

In hopes of moving the field forward, a team led by Stephan Eismann and Raphael Townshend in the lab of Ron Dror, Stanford University, Palo Alto, CA, looked to a machine learning approach known as deep learning. It is inspired by how our own brain’s neural networks process information, learning to focus on some details but not others.

In deep learning, computers look for patterns in data. As they begin to “see” complex relationships, some connections in the network are strengthened while others are weakened.

One of the things that makes deep learning so powerful is it doesn’t rely on any preconceived notions. It also can pick up on important features and patterns that humans can’t possibly detect. But, as successful as this approach has been in solving many different kinds of problems, it has primarily been applied to areas of biology, such as protein folding, in which lots of data were available for researchers to train the computers.

That’s not the case with RNA molecules. To work around this problem, Dror’s team designed a neural network they call ARES. (No, it’s not the Greek god of war. It’s short for Atomic Rotationally Equivariant Scorer.)

To start, the researchers trained ARES on just 18 small RNA molecules for which structures had been experimentally determined. They gave ARES these structural models specified only by their atomic structure and chemical elements.

The next test was to see if ARES could determine from this small training set the best structural model for RNA sequences it had never seen before. The researchers put it to the test with RNA molecules whose structures had been determined more recently.

ARES, however, doesn’t come up with the structures itself. Instead, the researchers give ARES a sequence and at least 1,500 possible 3D structures it might take, all generated using another computer program. Based on patterns in the training set, ARES scores each of the possible structures to find the one it predicts is closest to the actual structure. Remarkably, it does this without being provided any prior information about features important for determining RNA shapes, such as nucleotides, steric constraints, and hydrogen bonds.

It turns out that ARES consistently outperforms humans and all other previous methods to produce the best results. In fact, it outperformed at least nine other methods to come out on top in a community-wide RNA-puzzles contest. It also can make predictions about RNA molecules that are significantly larger and more complex than those upon which it was trained.

The success of ARES and this deep learning approach will help to elucidate RNA molecules with potentially important implications for health and disease. It’s another compelling example of how deep learning promises to solve many other problems in structural biology, chemistry, and the material sciences when—at the outset—very little is known.


[1] Geometric deep learning of RNA structure. Townshend RJL, Eismann S, Watkins AM, Rangan R, Karelina M, Das R, Dror RO. Science. 2021 Aug 27;373(6558):1047-1051.


Structural Biology (National Institute of General Medical Sciences/NIH)

The Structures of Life (National Institute of General Medical Sciences/NIH)

RNA Biology (NIH)

RNA Puzzles

Dror Lab (Stanford University, Palo Alto, CA)

NIH Support: National Cancer Institute; National Institute of General Medical Sciences

Next Page