Mapping the Microworld to Reveal Biology's Hidden Networks

From connecting cultures to charting continents, maps allow us to understand the world through its countless relationships.

This practice of creating and using maps is transforming other fields, too — helping shed light on the “microworld” of biomedicine. Advancements in imaging, sequencing, and data science are allowing scientists to observe the links joining molecular and cellular activity in intricate detail. St. Jude researchers are meticulously connecting all manners of biological function and dysfunction, ranging from cell migration to neurological disease development. These maps uncover biology’s hidden networks and form dynamic resources to address core scientific questions systematically.

Data science clarifies the interactions behind cell migration

How cells move from one location to another (cell migration) influences many processes in the body, including how immune cells travel to an infection, how the brain develops, and how wounds are repaired. Cell migration can also be exploited by disease, such as when cancer metastasizes.

Chemical signals secreted by cells, called chemokines, are vital to this process. Chemokines function as molecular breadcrumb trails, allowing one set of cells to recruit another set of cells to carry out important tasks. For example, cells release chemokines at wound sites to recruit immune cells to fight infections. Proteins on the surface of the migrating cell, called G protein-coupled receptors (GPCRs), recognize and bind the chemokine and transmit the signal to the cell interior.

However, some chemokines can bind to multiple different GPCRs, and some GPCRs can also recognize multiple chemokines. That such an interconnected web orchestrates the migration of cells raises a key question: How do individual GPCRs recognize specific chemokines, and vice versa?

To answer these questions, a team led by senior co-corresponding author M. Madan Babu, PhD, FRS, senior vice president of data science, Center of Excellence for Data-Driven Discovery director, and Department of Structural Biology member, and first and co-corresponding author Andrew Kleist, MD, PhD, lead scientist in the Babu group, analyzed protein sequence and structural information to map the interfaces between GPCRs and chemokines.

In a study published in Cell, the researchers used this data science framework to reveal key features of chemokine and GPCR sequences and structures that determine how they find each other. They discovered that chemokines and GPCRs use distinct regions, comprising structured elements and flexible, disordered elements. Some regions are common to many members of that protein family, while others are unique. These regions combine to identify each chemokine or GPCR, like an encryption key.

To ensure that the right chemokine binds the right GPCR, both proteins must match their unique encryption key to that of their binding partner. A match leads to a signal from the GPCR, instructing the cell to follow that breadcrumb trail, in the form of a chemokine gradient, to its point of origin.

“We found that cells have an elegant system that uses structure and disorder together to control cell migration,” said Babu. “With that understanding, we can now rationally introduce small changes in a chemokine’s structure to ultimately alter cell migration in desired ways.”

As proof of concept, the scientists changed chemokine-GPCR binding preferences in T cells to dial back a signal that normally stops their movement and alter its expected cell migration. The framework to assist with the rational design of chemokines and receptors is available online. It represents a first step toward rationally manipulating therapeutic cell movement, such as in CAR T–cell therapy.

We found that cells have an elegant system that uses structure and disorder together to control cell migration. We can now rationally introduce small changes in a chemokine’s structure to ultimately alter cell migration in desired ways.

M. Madan Babu, PhD, FRS

Department of Structural Biology

Brain map connects neurons to the spinal cord

The chemokine-GPCR relationship acts as a built-in security check to ensure the correct signal reaches the correct cell recipient to direct that cell to a new location. However, this relationship is just one way that cells communicate across relatively large distances with high specificity to regulate fundamental processes.

For example, within the nervous system, commands from the brain’s motor control systems travel along dense networks of axons to the spinal cord and on to recipient motor neurons, guiding body movement. In the middle of this interaction is a third group of neurons called interneurons, which act as gatekeepers and signal processors to ensure information is appropriately distributed and encoded in the spinal cord.

How interneurons specifically fit into this neuronal network has been poorly understood, so Jay Bikoff, PhD, Department of Developmental Neurobiology, created an interactive map that shows precisely how different brain regions connect with interneurons in the spinal cord. Published in Neuron, the researchers focused on V1 interneurons, a diverse group of inhibitory interneurons that restrict motor neuron signaling.

“Defining the cellular targets of descending motor systems is fundamental to understanding neural control of movement and behavior,” said corresponding author Bikoff. “We need to know how the brain is communicating these signals.”

To map the relevant circuits, the researchers used a genetically modified version of the rabies virus that is missing a key glycoprotein from its surface, preventing it from spreading between neurons. By reintroducing this glycoprotein to a specific population of interneurons, the virus could make a single jump across synapses before becoming stuck again. The researchers used a fluorescent tag to track the virus, allowing them to pinpoint which regions of the brain connect to the interneurons.

The three-dimensional brain map enabled the team to identify 26 distinct brain structures that directly connect V1 interneurons with the brain’s motor control systems. An accompanying web resource allows researchers to make accurate predictions about the network that connects different brain structures to the spinal cord. For example, they found that signals from the brain connect to spinal interneurons in patterned ways, with some pathways favored over others.

“We understand what some of the identified brain regions do from a behavioral perspective, but we can now make hypotheses about how these effects are mediated and what the role of the V1 interneurons might be,” explained Bikoff. “It will be very useful for the field as a hypothesis-generating tool.”

We understand what some of the identified brain regions do from a behavioral perspective, but we can now make hypotheses about how these effects are mediated and what the role of the V1 interneurons might be. It will be very useful for the field as a hypothesis-generating tool.

Jay Bikoff, PhD

Department of Developmental Neurobiology

Proteomics connects function with dysfunction in neurodegenerative disease

Mapping biological systems, whether chemokine-GPCR interactions, interneuron signaling, or other processes, reveals the complex, interconnected networks that coordinate functions across the body. By offering maps of various biological systems as evergreen resources for the wider community, scientists can engage with these networks on their own terms to generate independent hypotheses. Such was the expectation when Junmin Peng, PhD, Departments of Structural Biology and Developmental Neurobiology, launched a proteomics paradigm shift, starting with an investigation into protein turnover in mice.

The timeline for protein lifespan in cells varies greatly, with some cellular processes requiring longer-lived proteins. Turnover is a tightly regulated but poorly understood part of this process, with dips or surpluses in protein levels directly contributing to conditions such as Alzheimer’s disease.

In collaboration with the Center for Proteomics and Metabolomics, the team measured protein abundance and turnover rates across eight tissues and nine brain regions in mice to comprehensively chart information for over 11,000 unique proteins. The resulting map, called Turnover-PPT, published in Cell, offers detailed insights into protein regulation mechanisms that support tissue diversity.

By analyzing correlations between RNA levels, protein abundance, and protein turnover, the researchers determined that when proteins work together, they often turn over together — and in a tissue-specific manner. This revealed a level of coordination that had not been appreciated before.

Additionally, the team demonstrated how protein turnover helps directly regulate many processes, including transcription and translation. They also teased out the importance of phosphorylation in the turnover of many proteins, including controlling the stability of neurodegenerative disease-related proteins such as Tau and α-synuclein.

Scientist's hand working with a laboratory machine.

“Currently, we are analyzing protein turnover at the bulk level, but this approach could be extended to study protein turnover changes under different stress conditions or disease states, even at the level of individual cells,” co-corresponding author Peng said. “This work is the first large-scale analysis of its kind and sets the foundation and tone for future research in this area.”

Peng further demonstrated the power of large-scale proteomic mapping in another collaborative study on Alzheimer’s disease, also published in Cell. In this study, researchers formed regulatory network maps and a pioneering list of protein subnetworks that drive Alzheimer’s disease. This was achieved by combining proteomics with matched genetic, clinical, and pathological data from 198 human brain tissue samples.

Among the top drivers identified was the protein AHNAK, which helps form the glia-neuron network of the brain and is significantly associated with cognitive function. The researchers confirmed that AHNAK levels were higher in Alzheimer’s disease models. Probing deeper, they revealed that AHNAK was at the center of subnetworks containing proteins most often linked to Alzheimer’s disease. When they reduced AHNAK levels in disease models with compromised neuron signaling, neuronal activity was restored, and neurodegeneration could be stalled. As added validation, the knock-on effects to connected proteins aligned with predicted links from the network map.

“This map provides valuable subnetwork information that can be used with biomarkers to stratify patients and predict who will benefit from which treatment,” co-corresponding author Peng said. “We hope this is just the beginning, and that these identified drivers can be studied further in disease pathogenesis.”

St. Jude spatial omics sets the standard for the field

Network maps such as those developed by Peng act as guides to inspire new lines of inquiry and drive therapeutic ingenuity. In fact, researchers across St. Jude are taking the lead to ensure future generations of scientists are set up with robust, validated, and dynamic guides. Jasmine Plummer, PhD, Center for Spatial Omics director, Departments of Developmental Neurobiology and Cell & Molecular Biology, has positioned herself, and the Center, as both trusted experts in cellular identity and innovators of technology at the forefront of high-resolution tissue analysis.

The unmatched detail of spatial transcriptomics has allowed researchers to understand cellular microenvironments such as tumors and developing brains like never before. However, the large time, money, and resource investment in generating high-quality spatial maps of tissue omics profiles means ensuring a return on that investment is vital. Published in Nature Biotechnology, co-first and co-corresponding author Plummer and a global coalition formed the Spatial Touchstone project to fill this need within the spatial transcriptomics community and improve access to quality control.

The resulting collated dataset, which includes samples from multiple diseased and healthy tissues measured across two platforms, provides critical insight into how different tissues should look in a sample through a paired software tool, Spatial QM. An accompanying user-friendly application called the Spatial Touchstone Portal (STP) allows users to screen preliminary samples against the dataset.

The final components are the Spatial Touchstone Standard Operating Procedures (STSOP), which democratize a range of protocols from tissue preparation to data acquisition. “Every individual lab is different, but we wanted a set of protocols where people can feel confident in their methods and the expected outcomes,” Plummer said. “We analyze many samples using this strategy and are leaders in the field, so if a researcher’s sample keeps failing, they should consider using our protocols.”

For Plummer, providing the most cost-efficient and robust platform for spatial mapping of cells does not stop at data validation or sample preparation guidelines. Key techniques such as single-cell RNA sequencing remain prohibitively expensive, especially for those without access to the specialized equipment or computational infrastructure it typically requires. However, almost all research institutions have easy access to microscopes.

Considering this, a team co-led by Plummer created a method that combines single-cell RNA analysis with microscopy. The technique called Single-Cell Transcriptomics Analysis and Multimodal Profiling through Imaging (STAMP), published in Cell, can look at millions of single cells for a fraction of the cost of existing approaches.

“We’ve created a technique that gives us an advantage in the numbers game of single-cell analysis,” said co-corresponding author Plummer. “STAMP is an order of magnitude more cost-effective and allows us to profile a million cells simultaneously, compared to the tens of thousands typical of current methods, making it far more scalable.”

The researchers separated cells from tissues until they were individual, unconnected cells before fixing, or “stamping,” them onto microscopy slides. The scientists then added fluorescent molecules, which light up when bound to specific RNA sequences. The researchers found they could characterize many immune cells simultaneously and discriminate between the developmental stages of induced pluripotent stem cells at a 47-fold cost reduction compared to conventional approaches.

“STAMP gives us the best of both worlds in single-cell analysis: quantitative gene expression data and the ability to visually examine the cells under a microscope,” Plummer said. “We hope that these features, combined with its accessibility and cost-effectiveness, enable others to discover new biology and clinical opportunities in the future.”

STAMP is an order of magnitude more cost-effective and allows us to profile a million cells simultaneously, compared to the tens of thousands typical of current methods, making it far more scalable.

Jasmine Plummer, PhD

Departments of Developmental Neurobiology and Cell & Molecular Biology

Single-cell data processing tool uncovers hidden drivers

Single-cell RNA sequencing is quickly becoming a standard technology for mapping cellular identity, and approaches such as STAMP have broadened access to it as a resource. Careful consideration must also be given to downstream data processing, as many current methods connect cells based on linear correlations in gene expression profiles. While useful at a broad scale, such methods often lack the sensitivity needed to resolve subtle yet critical differences between cells, like trying to navigate through a cave with only a map of the surface.

With this objective in mind, Jiyang Yu, PhD, Department of Computational Biology interim chair, developed the single-cell Mutual Information-based Network Engineering Ranger (scMINER), a computational framework designed to infer biologically meaningful gene networks from single-cell RNA sequencing data. Published in Nature Communications, scMINER uncovers key “hidden drivers” that help differentiate cells, but that may not be apparent from RNA expression levels alone.

The researchers pre-trained a model using a large single-cell RNA sequencing dataset to learn general patterns of gene–gene relationships across many cellular contexts. They then fine-tuned the model using a dataset of interest, such as drug-resistant versus drug-sensitive tumor cells, to incorporate context-specific information. scMINER then generates single-cell activity profiles, which summarize the expression of neighboring genes, predicting protein activity in a way that RNA expression alone cannot capture.

“This is the power of our hidden driver approach — it captures factors correlated with protein activity beyond transcriptional changes,” Yu stated. “scMINER infers cluster-specific hidden drivers from single-cell RNA-sequencing data, ascribes missing information, and identifies underlying drivers for each cell type across different conditions: developmental, immunological, and more.”

Mapping out the future of biomedical research

The nature of biological function and dysfunction is only truly seen when viewed in the context of knowing what healthy networks look like and how they are subsequently disrupted in disease. Mapping these interactions, which drive molecular and cellular activity, provides immediate insight into pressing biomedical questions but also allows scientists to address questions yet to be asked.

The resources generated by St. Jude scientists present biomedical science holistically and at scales never achieved before; the tools developed to untangle the networks within position current and future scientists at the precipice of scientific discovery. While these developments are immediately impactful, the true legacy of mapping the microworld is yet to come.