Research has found the ANS interacting with tumors in a variety of organs, including the brain, prostate, breast, and stomach, as well as osteosarcoma, which affects bone cells. Sympathetic and parasympathetic nerve fibers infiltrate the tumor microenvironment (TME), and they have direct impacts on tumor growth (tumorigenesis) and spread (metastasis). Depending on the type of signaling involved, these interactions can stimulate the cancer cells or inhibit them.

Interaction Between Tumors and the ANS

Tumor growth is driven both directly and indirectly by the sympathetic nervous system. The most well understood direct route is catecholamine signaling, especially via norepinephrine, which stimulates tumorigenesis and metastasis in prostate, breast, and colon cancers. In 2019, a team from Roswell Park Comprehensive Cancer Center found that catecholamines activate receptors on myeloid-derived suppressor cells (MDSCs), which then release factors that promote the growth of blood vessels. The TME benefits from increased vasculature as well as peripheral nerve fiber density. Much of this evidence is based on correlation–larger tumors and aggressive cancers tend to have more vasculature and innervation–but recent findings are starting to explain the complex causality and relationships within the TME.

Indirectly, the sympathetic nervous system promotes tumorigenesis by repressing immune function, the body’s primary natural mechanism for keeping cancer cells in check. A group at Columbia University learned in 2016 that a neural circuit–involving vagally modulated memory T cells–controls the number of MDSCs produced by the spleen, which has a downstream effect on colorectal cancer. Osteosarcoma is another area where neuroimmune interactions affect the development of the tumor microenvironment. Researchers from Ningbo No.6 Hospital have found that nerve growth factor (NGF) and brain-derived neurotrophic factor (BDNF) act as promoters of tumor growth in the TME. They suggest several pathways to interrupt this process while reducing associated pain and immunosuppression side effects.

The interaction between the ANS and tumors in the stomach operates slightly differently and relies heavily on the parasympathetic nervous system, which has a more direct action on the organ. Parasympathetic nerves release acetylcholine for several functions, including muscle contraction, and gastric cancer cells proliferate in the presence of acetylcholine. This effect has been demonstrated by three methods: vagotomy, nerve-blocking botulinum, and gene editing. An increased density of sympathetic nerves in the stomach also correlates to poor outcomes for gastric cancer patients, but the precise signaling mechanisms between the ANS and these tumor cells are not well understood.

Harnessing the ANS for Cancer Treatments

Various attempts have been made to control tumorigenesis via ANS pathways, primarily in mice but sometimes in human patients as well. Vagal neurectomy in mice has reduced the incidence of tumors in the gastric tract. Repressing sympathetic nerve signaling has had similar effects on prostate and colorectal cancers. Some of these harness direct mechanisms via the reduction of nerve density in the TME, and others work indirectly via neuroimmune pathways. In yet other cases, interventions haven’t affected tumor growth, but they have controlled a common side effect of many cancer chemotherapies: an increase in pain, which is another function of the ANS.

Significant work remains in order to understand the specific ways in which sympathetic and parasympathetic nerves interact with cancer cells, but what we already know has led to treatments that harness these neural circuits to affect the tumor microenvironment.

Furthering Research into Cancer and the ANS

Since “cancer neuroscience” is a relatively new field, and given its multidisciplinary nature–involving neurobiology, immunology, cell biology, oncology, and genetics–it is ripe for investment and collaboration, as pointed out in a oft-cited paper by Michelle Monje and others in 2020. The ANS plays a critical role in this nexus, and its potential for therapies that simultaneously minimize side effects should elevate its priority in future research.

The symptoms of cancer and its treatments overlap in numerous ways with other autonomic disorders. They include fatigue, brain fog, homeostatic imbalance, and remissions followed by relapses, to name a few. Discoveries that illuminate the neural mechanisms of these symptoms could have applications to cancer patients, which is why it’s essential to foster greater communication between physicians, immunologists and cellular neuroscientists who study the autonomic nervous system. The more we learn about the ANS, the more it becomes clear how essential its function (and dysfunction) is in every aspect of human health.

This framework has been extraordinarily productive. It led to the discovery of hormones such as leptin, ghrelin, CCK, and GLP-1, and to the identification of hypothalamic circuits that regulate energy balance. It also shaped how obesity has been conceptualized for decades, as a disorder of dysregulated intake relative to metabolic demand.

Yet this model largely focuses on hormones and brain centers that regulate energy balance. It says less about how the gut senses a meal and conveys information to the brain. Emerging evidence suggests that the vagus nerve is not simply a conduit for fullness signals, but a sensory interface that carries multiple dimensions of feeding including volume, nutrient identity, and internal state. The gut routes different aspects of a meal through specialized neural circuits that influence both physiology and cognition, including motivation, reinforcement, and memory.

Mapping the Sensory Interface

How did we learn that the vagus carries different kinds of information from the gut?

The answer emerged gradually, through a sequence of anatomical, molecular and functional studies. The gastrointestinal tract is densely innervated by vagal sensory terminals whose cell bodies sit just outside the brain. These neurons extend specialized endings into the stomach and small intestine, where they detect both mechanical stretch and chemical signals generated during digestion. Long before molecular tools were available, anatomical studies by Powley, Berthoud, and others identified distinct vagal terminal structures, including intraganglionic laminar endings (IGLEs) embedded within the muscle layers of the gut wall and mucosal endings that innervate the intestinal villi. These morphologically distinct structures suggested that vagal neurons might be specialized to detect different features of a meal.

A major advance came in 2016, when the Liberles lab demonstrated this anatomical diversity corresponded to genetically distinct sensory neurons. Neurons expressing GLP1R preferentially formed IGLE terminals consistent with stretch detection, whereas neurons expressing GPR65 innervated the intestinal mucosa, consistent with chemosensory roles. This work linked molecular identity to anatomy and established that the vagus nerve transmits at least two fundamentally different types of information about how much has been eaten and what has been eaten.

Single-cell RNA sequencing soon refined this picture further. The Knight laboratory identified additional vagal sensory clusters and generated mouse lines that allowed their terminals and functions to be mapped directly. Within the mechanosensory class, two IGLE-forming populations were identified: the previously described GLP1R-expressing neurons and a newly characterized Oxtr-expressing population. Optogenetic stimulation of either population reduced food intake and promoted meal termination, providing causal evidence that stretch-sensitive vagal pathways signal fullness to the brain. In contrast, stimulation of chemosensory populations did not acutely suppress feeding, suggesting that these neurons participate in different aspects of eating behavior.

A Gut-Reward Circuit

If these chemosensory neurons do not terminate a meal, how do they contribute to eating behavior?

In 2018, we activated right-sided vagal sensory neurons innervating the upper gastrointestinal tract. This engaged a defined pathway from the intestine through two brainstem hubs to dopamine-producing neurons in the midbrain. Using optogenetic stimulation to mimic the intestinal input to the brainstem hubs, we observed increased dopamine neuron activity and behavioral reinforcement.

Animals learned to self-stimulate this gut-vagal pathway. In real-time place preference experiments, activation of the circuit caused animals to spend more time in the stimulation-paired chamber. In flavor-nutrient conditioning paradigms, pairing circuit activation with a neutral flavor was sufficient to induce learned preference, whereas, inhibiting the pathway during nutrient infusion prevented reinforcement learning. Importantly, these effects were independent of taste. Nutrients delivered directly into the intestine, bypassing oral sensory input, were sufficient to drive reinforcement. The gut itself could signal reward!

Parallel Nutrient Pathways and Motivational Drive

Since gut circuitry engages brain reward regions, we wondered which specific intestinal signals might provide motivation to eat. Could fats and sugars, the defining components of many obesogenic foods, recruit this brain reward pathway?

Using in vivo calcium imaging during controlled nutrient infusion into the gut, we were surprised to find that fats and sugars do not converge onto a single chemosensory population. Instead, they activate distinct subsets of vagal sensory neurons. These fat-responsive and sugar-responsive populations are anatomically separable and project through parallel pathways that drive dopamine release in the midbrain

To determine whether these populations are functionally distinct, we used a mouse genetic tool for activity-dependent labeling (FosTRAP mice) to selectively target neurons activated by fat or sugar. Deletion of fat-responsive neurons abolished reinforcement learning for flavors paired with intestinal fat infusion but did not impair sugar-driven learning. Conversely, deletion of sugar-responsive neurons disrupted sugar reinforcement while leaving fat reinforcement intact. These results indicate that fat and sugar are encoded by separate vagal circuits, each capable of independently driving reinforcement learning for its associated nutrient..

We next asked whether activating these circuits is sufficient to promote motivation. Optogenetic stimulation of either fat-responsive or sugar-responsive pathways was sufficient to support self-stimulation, demonstrating that activation of each circuit is intrinsically reinforcing. In other words, the gut contains parallel channels that can independently generate motivational drive for specific macronutrients.

If separate circuits independently promote reinforcement, we reasoned that their combined activation might amplify motivational drive. To test this, we gave mice access to solutions containing fat alone, sugar alone, or a combination of fat and sugar, carefully matched for total caloric content. Despite equivalent calories, mice consumed nearly twice as much of the combined solution compared to either nutrient alone. Consistent with this behavioral effect, the fat-sugar combination produced greater dopamine release than a single macronutrient. When fat and sugar are consumed together, the gut engages multiple reinforcement circuits at once, producing a stronger motivational signal than either nutrient alone.

In nature, fat-rich and sugar-rich foods are typically distinct. Thus, separate reinforcement systems may have evolved to ensure motivation to seek out scarce energy sources. However, in modern food environments, processed foods frequently combine both macronutrients within the same food. Simultaneous recruitment of parallel gut-reward circuits may therefore increase motivational drive and vulnerability to overconsumption.

Importantly, similar findings have been reported in humans. Dana Small’s lab showed that foods combining fat and sugar elicit greater activity in reward-related brain regions than foods containing either macronutrient alone, and participants are willing to pay more for these combined foods. This convergence across species suggests that parallel nutrient reward circuits are evolutionarily conserved.

From Motivation to Memory

How do these reinforcement signals shape food choice over time?

Reinforcement increases the motivational value of a food in the moment. For those effects to shape future intake, however, they must be encoded and stored. Eating behavior is shaped not only by current physiological need, but also by remembered outcomes of prior meals.

The hippocampus has long been recognized for its role in contextual and spatial learning, and prior studies have suggested that it influences food seeking behavior. Until recently, however, its direct role in controlling food intake remained unclear.

We asked whether gut-derived nutrient signals engage hippocampal circuits to influence spatial memory and eating. Using activity-dependent labeling and circuit-specific manipulations, we found that nutrient-responsive vagal pathways activate defined hippocampal ensembles. Deleting these hippocampal neurons impaired spatial memory for nutrient location and selectively reduced intake of that specific nutrient without affecting consumption of others. Conversely, stimulation of the same circuit enhanced memory recall and increased intake of the associated nutrient.

These experiments demonstrate that hippocampal circuits are not merely involved in seeking food but are necessary for regulating nutrient-specific intake. Importantly, disruption of this pathway altered feeding behavior in models of obesity, indicating that gut-to-hippocampus signaling contributes to the control of body weight. These findings extend the function of the gut-brain axis beyond satiety and reinforcement to the spatial and contextual consequences of eating, thus shaping future consumption patterns in a nutrient-specific manner.

When the System Breaks Down

What happens when the flow of information from the gut to the brain is disrupted?

Obesity is often described as a disorder of excess intake relative to expenditure. Within the framework of a layered gut-brain architecture, however, it may also reflect impaired sensory communication. If stretch-sensitive neurons fail to accurately convey fullness, meal termination may be delayed. If nutrient-responsive pathways are altered, reinforcement learning may no longer reflect true metabolic benefit. If gut-derived signals fail to properly engage memory circuits, the ability to use prior physiological experience to guide intake may be compromised. In addition, if these signals are normally partitioned across parallel pathways, disruption may not simply increase or decrease appetite uniformly. It may alter the balance between physiological regulation and motivational drive, or between internal state and remembered outcome.

The remarkable success of GLP-1 receptor agonists underscores the therapeutic power of targeting gut-brain pathways. These drugs effectively reduce appetite and promote weight loss by engaging central satiety circuits while also reducing the motivational pull of food. In doing so, they may compensate for impaired peripheral signaling and restore control over food intake. However, weight regain after discontinuation indicates that while GLP-1 receptor agonists effectively dampen appetite, they do not fundamentally repair the neural circuits that integrate satiety, reinforcement, and memory. Future strategies may therefore require not only suppressing hunger signals, but also restoring accurate sensory communication between the gut and brain.

Vagal Afferent Signaling in Eating and Beyond

By understanding obesity as a disorder of sensory integration, rather than solely of willpower or caloric excess, we shift attention toward the neural circuits that interpret information from the intestine. Restoring or recalibrating these pathways may complement pharmacological advances and improve long-term outcomes.

Much of my scientific work has focused on how the vagus nerve carries information about the body’s internal state to the brain. What has become increasingly clear is that these signals do more than regulate hunger or fullness. They shape how the brain assigns value to food, how it stores memories of what we have eaten, and how internal state influences future behavior. This broader view of vagal signaling may have implications not only for obesity and metabolic disease, but also for conditions in which motivation, learning, and cognition are altered.

After the Pioneer plaques, Carl Sagan took an interest in neuroscience. During this time, he popularized the notion of a “reptile brain” lurking deep beneath our wrinkled neocortex in his 1977 book, The Dragons of Eden. The original form of this theory was strong and simplistic, which has given it cultural staying power. It’s also prompted some fierce ongoing criticism from professional neuroscientists, e.g., a recent whole book chapter.

Counting the many colors of the lower brain

In a twist of fate, recent neuroscience findings are showing the lower brain to be a surprisingly “different beast” than the cortical areas above. And the difference, in a sense, is again one of “hyperfine structure”, i.e., a level of subdivision unknown to previous science. This dawned on me upon seeing this late 2023 social media post from the Allen Institute, where they reported the lower (‘ventral’) part of the brain was discovered to have highly numerous distinguishable cell types:

This post was part of a social media thread announcing the Institute’s major part in a flagship achievement from the first decade of the US BRAIN Initiative: a striking ten-paper collection in the journal Nature entitled the “Brain Initiative Cell Census Network (BICCN) 2.0”. Of these ten, four were “whole brain cell-type atlases”, including the Allen Institute’s publication, two from Harvard, and one from its neighbor the Broad Institute. A summary of the BICCN 2.0 results explained that such atlases combine two technologies which advanced greatly during the prior decade: spatial transcriptomics and single cell RNA sequencing.

Spatial transcriptomics makes brain atlases “4-dimensional” by considering the expression of genes in brain cells (i.e., presence of mRNA), in addition to the three-dimensions of their spatial location. Hence the colorful images comprising the public Allen Brain Cell (ABC) atlas.

While spatial transcriptomics can explore hundreds of marker genes, single cell RNA sequencing (aka scRNA-seq) can probe much deeper into the “molecular identity” of an individual cell, in some cases sequencing for activity of all known genes (“whole transcriptome”). To produce these milestone atlases, data of both types were computationally combined at large scale (1-10 million cells). Given the high spatial and molecular resolution, the largest atlases are conveyed with statistical representations, such as UMAP plots, to concisely represent the categories of cells, alternately called “cell types” or “transcriptional clusters” across the papers.

The BICCN 2.0 landmark achievement was the scope and scale of these cell atlases, for the first time covering a whole mammalian (mouse) brain. This is not only impressive as a technical achievement, but important scientifically: it enabled unbiased comparison between brain areas for the first time, which revealed a large number of cell categories in the lower “ventral” brain regions. This finding was also a win for reproducible science, with these same top-level scientific results being found by different teams using differing technologies, from the newer MERFISH (Allen team) to scale-up of the earlier Slide-Seq (Broad team).

The authoring teams were clearly surprised by these results, using phrases such as “astonishing diversity” and “remarkably high” while describing the number of distinguishable brain cell categories in the ventral brain. The two studies with the broadest category coverage both found over 5000 distinctive brain cell categories by space and gene expression profile. Nearly all categories were neuronal, with relatively fewer non-neuronal and immune brain cell categories, and with the majority (over 3000) of distinctive categories falling in the hypothalamus, midbrain, or hindbrain.

Ventral is an anatomical term for the “belly” side of an organism, making the term neatly universal across vertebrate brains, from fish to quadrupeds (like mice) to upright bipeds like us. As humans, we can refer to it more colloquially as the “lower brain”, whereas rodent and fish scientists would likely refer to it as the “rear brain”.

There’s a whiff of “history rhyming” seeing the rich color-coding of the recent brain atlases, since several of the historically known anatomical nuclei of the lower brain have been described based on their literal color, e.g., the red nucleus and locus coeruleus (“blue spot” in Latin) are long-known nuclei of the midbrain and pons, respectively. But it’s not a “history repeat”, as the recent brain atlases are as much statistical as spatial. The second C in BICCN stands for “census” and these publications give us a first-pass count of what may be independent functional pathways carried out by specialized brain cells, the majority of which live in the lower brain.

From the oldest brain structures to new frontiers of medicine

Brain cell types are a recipe for wiring up independent neural circuits, as described in a recent article in . My background is electrical engineering, and these thousands of tightly packed independent circuits make me think of the multi-chip modules powering recent mobile computers. Such metaphors are appealing to geeks (like us!), but what interests most are medical implications and applications. Early translational science stories in our Newsletter show that cell-type specific functions can not only be experimentally dissected but sometimes directly related to medical states and symptoms. For instance, fainting control is a function pinpointed not merely to the anatomical region area postrema (AP) but to specific cell types within AP; and the length of “post-syncope” downtime was likewise mapped to specific cell types within the hypothalamus.

These brain atlas results were reported in mice, which can raise doubts about their relevance to human medicine, so it’s important to note three hopeful things. First, the same core finding (high brain cell type density of the lower brain) has been replicated in human brains. Second, as the Allen Institute’s post noted, the ventral brain is more “ancient” and less “divergent” than the dorsal part, meaning it’s largely the same across species over evolutionary time. This suggests that lower brain findings in mice (and other mammals) are more likely to translate to humans. Finally, the lower brain is the one connected–by the spinal cord, vagus nerve, and other cranial nerves–to the body! While that’s hardly new, the capacity of neuroscience to map physiological functions (and dysfunctions) within the lower brain’s “hyperfine structure” is unlike ever before.

Carl Sagan had a vivid imagination, conceiving that Pioneer plaques depicting our knowledge of “hyperfine structure” might convey a message to extraterrestrials and that the lower brain might be a “reptile brain” within. How might we (more realistically!) imagine that pioneering neuroscience studies can start translating to “moonshot medicine” of the 21st century? Medical education is historically known for using mnemonics, and I’d like to imagine that medical students will soon remember that the lower brain spans “from hypothalamus to hindbrain”; that it’s rich with mappable physiological functions (and dysfunctions); and that physicians will increasingly partner with neuroscience colleagues to develop new insights and pave the way for new treatments. What medicine calls deep brain stimulation (DBS) today scratches the surface of the lower brain, with targets such as pallidus and the subthalamic nucleus1 that just cross the line into the increased cell type density of the lower brain. Could upcoming treatments, surgically or otherwise, venture into the deeper and more “hyperfine structures” of the lower brain? Might they target physiological functions and dysfunctions more specifically? I can dare to imagine.

For most of medical history, the nervous and immune systems were viewed as largely separate entities operating independently. The nervous system controls all our thoughts, movements, and sensations, while the immune system defends against infection and injury. While these remain fundamentally true, recent work in the emerging field of neuroimmune interactions shows that the two systems engage in constant bidirectional communication. At our core, we’ve long known that the immune and nervous systems are intertwined in important ways. When we experience a case of food poisoning, we may become severely nauseous and develop a long-lasting aversion to the specific food, intentionally avoiding it in the future. This learned behavior occurs because bacterial toxins activate gut immune cells, which then send sensory signals to the brain, thereby encoding this aversive episode within neural networks.

Even a simple sneeze or allergic reaction involves rich and intricate neuroimmune communication. Allergens, such as dander, trigger immune cells in your airways to release histamine and activate specific sensory neurons that then initiate the sneeze reflex. So we know that the intricate, wide-ranging networks of the nervous system communicate with the many compartments and sentinel cells of the immune system. We also know that this neuroimmune communication is fundamental to maintaining health and homeostasis. When it’s disrupted, it contributes to a myriad of chronic inflammatory diseases and disorders of the autonomic nervous system.

But how does the nervous system encode specific information about immune signals? Does it simply register “inflammation present” or “inflammation absent,” like a binary signal, or does it capture something more nuanced and specific? Finding answers to these fundamental questions may inform important future therapeutic approaches for treating chronic inflammation and certain types of dysautonomia.

The Neural Code of Inflammatory Cytokines

Early work in the field demonstrated that neuroimmune communication can occur through slow humoral pathways, such as via blood and hormones. However, recent work from our lab has demonstrated the existence of more direct, immediate neural pathways that provide rich, real-time immune information to the brain. Specifically, we show that sensory neurons of the vagus nerve (i.e. vagal sensory neurons) encode the primary protein messengers of immune signaling – inflammatory cytokines – as distinct patterns of neural activity.

The vagus nerve (cranial nerve X) is part of our peripheral nervous system, the vast network of nerves and ganglia that lies outside of the brain and spinal cord. As the longest cranial nerve in the body, the vagus nerve connects to all our internal organs and serves as a major conduit for body-brain signaling. These vagal sensory neurons are the critical “first responders” in the body-brain axis, transmitting visceral signals about inflammation and other interoceptive signals to the lower brainstem. To monitor the activity of these neurons, we used a a microscopy technique called calcium imaging to record genetically encoded calcium signals while cytokines were administered. This allowed us to observe, in real time, how individual vagal sensory neurons responded to these immune signals:

We found that individual sensory neurons exhibit distinct, real-time responses to specific inflammatory cytokines. Some neurons respond selectively to individual cytokines, while others encode multiple cytokines but maintain unique, cytokine-specific activity signatures. Moreover, these cytokine-specific neural activity patterns were maintained when the cytokines were administered at different sites in the body, suggesting that there may be a preserved neural code for inflammation that informs the brain about what type of inflammation is occurring, while the neural circuit wiring within the peripheral nervous system provides information about where the inflammation is happening.

Our work was complemented by findings from other laboratories that identified some of the downstream brainstem circuits in the nucleus of the solitary tract (NTS) that receive and process these vagal immune signals, acting as a rheostat for inflammatory responses. Together, these findings show that lower brain regions tightly modulate the peripheral immune response and that the precise encoding of inflammation as neural signals occurs in the periphery.

The implications of this precise neural encoding became even more apparent when we examined states of chronic inflammation. Using the dextran sulfate sodium (DSS)-induced colitis model in mice, a widely used model for inflammatory bowel disease (IBD), we observed profound changes in how the vagal sensory neurons signaled. The inflammatory state dramatically increased the number of active neurons but decreased the amplitude of their signals, suggesting that baseline "noise" in their signals to the brain was increased. Chronic inflammation also decreased the responsiveness and precision of cytokine-specific neural representations, potentially blurring the distinct neural signatures of inflammatory cytokines and making them more challenging for subsequent neural stations (e.g. brainstem, thalamus, cortex) to interpret. Importantly, these findings show that chronic inflammation fundamentally alters the excitability and responsiveness of vagal sensory neurons, potentially disrupting the fidelity of body-brain neuroimmune communication.

Studies from other labs have shown that specific cortical brain regions encode episodes of inflammation, offering the promise that we may someday be able to target these regions for therapeutic benefit.

Bioelectronic Medicine: Targeting Neural Circuits to Treat Disease

As chronic inflammatory disorders are on the rise globally, new therapies beyond pharmaceuticals and biologics that are broadly immunosuppressive may be needed to help restore neuroimmune homeostasis. The potential to treat disorders such as chronic inflammation with devices that target the nervous system underpins the emerging field of bioelectronic medicine. While implantable devices such as cardiac pacemakers and deep brain stimulators have been in use for many decades, the idea that a device can be used to treat an immune disorder is unconventional. However, this idea has become a reality for patients living with moderate-to-severe rheumatoid arthritis (RA), an autoimmune disease.

In July 2025, the FDA approved a vagus nerve stimulation (VNS) implant for patients with RA who do not receive adequate relief from traditional drugs or biologics. This first-of-its-kind neuroimmune modulation therapy was based on decades of laboratory work with preclinical models to map the neural circuits and molecular mechanisms that regulate inflammation. But the FDA approval of VNS for RA is just the beginning. Every week, there are important scientific advances that lead us closer to the reality of device-based medicine, including bioelectronic sensors to detect inflammation in real-time and wireless, flexible implantable stimulators. It is our hope that a wide range of novel tools will enable the next generation of bioelectronic therapies to treat inflammation beyond RA, including autonomic dysfunction, sepsis, Long COVID, and other chronic conditions. In the meantime, we and others will continue to study the rich neuroimmune communications that govern health and disease.

Remember the flu? Well, it turns out that a specific subset of vagal sensory neurons can also protect against influenza virus infection by interacting with immune cells in the lungs, highlighting the potential of neuroimmune modulation to help control viral infections. It’s just another example of how closely intertwined these two systems are. So the next time you experience a fever or body aches from a virus, consider that your nervous system is having a conversation with your immune system, and that this complex dialogue is one that we are just now beginning to understand. For the millions of people living with chronic inflammatory conditions, decoding this neuroimmune communication offers hope that stimulating the body’s own regulatory circuits may one day restore the balance disrupted by disease.

Figure References

1. Huerta, Tomás S., et al. "Neural representation of cytokines by vagal sensory neurons." Nature Communications 16.1 (2025): 3840.

How bacteria burrow: biofilms

The main culprit of this infection was Pseudomonas aeruginosa, a common bacterium found in the environment. A defining feature of P. aeruginosa is its ability to form a biofilm, a sticky, protective matrix that helps the bacteria evade immune responses and resist antibiotics. To investigate our research question, biofilm specialist Dr. Joe Harrison collaborated with our lab and provided two important P. aeruginosa strains: one that could not form a biofilm due to mutations in polymers PEL and PSL, used to build the protective exopolysaccharide layer (EPS^-), and a normal, biofilm-producing strain (EPS⁺).

From initial immune suspicions to new neuro-immune trail

The beginnings of this work were conducted by Dr. Elise Granton, an anesthesiology physician-scientist who was a PhD student at the time. Initial observations indicated that both strains caused vast differences in sickness when administered into the airways of mice. Like humans, mice with EPS⁺ showed minimal sickness despite harboring millions of bacteria in their alveolar space. Interestingly, the EPS^- strain caused significant sickness responses and hypothermia. Naturally, coming from an immunology lab, Dr. Granton set out to identify which immune cells were recruited and what the inflammatory cytokine profile looked like in these mice. To our surprise, there were no differences in infiltration by neutrophils, an inflammatory white blood cell that acts as a first responder to infection in the lungs. In addition, inflammatory cytokines could not explain the difference in the observed phenotypes either. At this point, we had ruled out the usual suspects. The immune system alone couldn’t explain what we were seeing.

After consulting with neuroimmune expert Dr. Isaac Chiu (Harvard Medical School), it became apparent that sensory neurons, which are often associated with pain sensation, may have a role in detecting infection and inducing sickness responses. These communicators of pain and noxious stimuli highly innervate the lung airspace. This intrigued us and has been a pivotal step in our lab to investigate neuroimmune interactions in the lungs. First, we identified that chemical ablation of TRPV1⁺ nociceptors resulted in reduced sickness in the EPS^- mice. Next, we found that Toll-like receptor 4 (TLR4), an important immune receptor for detecting lipopolysaccharide (LPS), was also required for sickness responses. With Dr. Granton leaving for medical school, this is where I, a PhD student in the lab at the time, came in to help. Confirming Dr. Granton’s findings, I found that breaking down this EPS⁺ biofilm with PEL/PSL enzymes induced similar sickness to the EPS^- infected mice; thus, we identified that biofilms mask LPS signature and prevent detection. Together, this information prompted us to generate a novel TRPV1-TLR4 knockout mouse, which when tested with EPS^-, also failed to mount a sickness response.

Following the neuro-immune trail to the brain

These exciting findings started to answer our initial question: Pseudomonas aeruginosa biofilms continue to cause alveolar damage by masking their LPS signature from nociceptors in the lungs. Yet, a looming question overshadowed our findings: “Where do those neurons go?” Nociceptors innervating the lungs signal afferently through two routes, the dorsal root ganglia and the vagal nodose ganglia. This is where Dr. Christophe Altier (UCalgary) and his post-doc at the time, Dr. Manon Defaye, now an independent investigator (Laval University), shared their expertise. Through sophisticated adeno-associated virus (AAV) Cre-lox systems, Dr. Defaye injected TRPV1-specific Cre-producing AAV2 vectors into Tlr4^flox/flox nodose ganglia, resulting in localized conditional neuronal TLR4 knockouts that showed reduced sickness during EPS^- infection.

Now that we knew these vagal signals were transmitted through the nodose ganglion, we asked, “Where do these neurons dispatch their messages to, and who answers the call to elicit such profound sickness?” Thankfully, UCalgary investigators Drs. Deborah Kurrasch and Jaideep Bains heeded our calls on this exponentially growing, multidisciplinary project. Dr. Paris Moazen, a post-doc in the Kurrasch lab, helped identify that corticotropin-releasing hormone neurons were highly activated in the paraventricular nuclei of the hypothalamus (hPVN), the center for sickness response. We then blocked these neurons with a CRH receptor antagonist, which silenced sickness during EPS^- infection. Dr. Tamás Füzesi, a post-doc from the Bains lab, utilized this information and bilaterally injected AAV-expressing hM4D(Gi), a chemogenetic receptor (called a DREADD) that let us temporarily slow down neuron activity in the presence of an inert compound, stereotactically into the hPVN of CRH^reporter (Crh^Cre) mice. Shutting off these CRH neurons effectively blocked EPS^--mediated sickness. These powerful neuroscience tools were mind-blowing 🤯 to us immunologists at the time! Using them has truly transformed how we think about and investigate neuroimmune interactions.

Sensing and signaling sickness, in the body and society

Our final question in this complex cascade of events was “Can TRPV1-TLR4 neurons perceive LPS-producing bacteria beyond P. aeruginosa?” E. coli produces LPS and often causes acute pneumonia in a planktonic form; therefore, we tested whether a virulent strain is also detected by TRPV1-TLR4 neurons. Indeed, these neurons were responsible for detecting E. coli, and neuronal TLR4 knockouts mitigated sickness. Together, our study highlights that our lungs do more than just breathe – they sense, signal, and respond when something is wrong. Biofilms circumvent this detection and prevent neural circuits that drive protective sickness responses from being alerted.

Sickness is often regarded as a negative consequence of infection. During the COVID-19 pandemic, healthy individuals became wary of those who were sick and exhibited avoidance behaviors. In turn, those who were sick would begin to isolate to prevent the spread of infection and seek medical attention if required. Therefore, appropriate sickness responses may be a positive consequence of infection. However, patients with chronic biofilm infections often become indolent or desensitized to their sickness, allowing the infection to propagate and cause tissue damage in unsuspecting hosts. Therefore, appropriate sickness responses, paradoxically, seem to represent a beneficial host adaptation that serves as an indicator of individual and community health status.

Towards tailored therapeutics: balancing the brain’s bacterial responses

Striking the balance between beneficial and deleterious host responses and sickness during biofilm infection is tricky. Whereas appropriate inflammation can help the immune system combat invading pathogens and initiate vital wound healing and anti-inflammatory processes, inappropriate inflammation can result in cytokine storms that often cause multiple organ failure. The answer to this conundrum may involve the brain, which is a multifaceted center interconnected by complex neuroimmune circuits. Our findings at this neuroimmune intersection support the notion that unmasking LPS from biofilms may initiate some of the beneficial inflammatory pathways. In contrast, promising developments in vagal nerve stimulation therapy may be the answer for overzealous immune responses by utilizing neuroimmune modulation that dampens inflammation. Combining enzymes that break down biofilms and implantable vagal stimulation devices that are FDA approved may enable physicians like Dr. Yipp to strike this balance in the ICU.

Since our initial findings, others have demonstrated the importance of neuronal protection during influenza infection by altering immune cell landscapes. Beyond infection, we have begun to untangle nociceptor regulation of immune cell dynamics. Recently, we have discovered that nociceptors are protective during pulmonary fibrosis by modulating macrophage release of a neuropeptide that drives TGF-β-induced Siglec-F⁺ neutrophil inflammation. These exciting findings are beginning to reshape how we approach inflammation and healing within the lungs. We believe the next steps in neuroimmune therapy would be to harness the power of neuroimmune modulation on immune cell landscapes, raising the possibility of tailoring immune responses to specific inflammatory insults.

Following “recovery” from acute COVID-19, some patients develop a complex and debilitating chronic illness, often called Long COVID. The persisting symptoms include post-exertional malaise—fatigue and other symptoms made worse by physical or cognitive exertion—as well as unrefreshing sleep, cognitive impairment (problems with thinking, memory, etc.), and orthostatic intolerance (worsening symptoms when upright).

This illness resembles a similar disorder—myalgic encephalomyelitis/chronic fatigue syndrome (ME/CFS)—that often develops following what appears at first to be just another acute “viral” infection. But unlike most such illnesses, this “flu-like” illness does not end. Patients often describe it as “a flu that never goes away.”

Long COVID and ME/CFS involve hundreds of millions of people, globally, and cause hundreds of billions of dollars in medical care costs and lost workforce productivity. Recent discoveries in neuroscience suggest underlying mechanisms that may explain the symptoms.

Post-acute infection syndromes

Neither Long COVID nor ME/CFS is unique. They are similar to illnesses that develop following other viral, bacterial and protozoal infections such as West Nile virus infection, post-treatment Lyme disease syndrome, and post-giardiasis syndrome. Recently, it has been proposed that all of these illnesses are examples of a category called post-acute infection syndromes (PAIS)12.

Underlying pathophysiology

Initially, it was unclear if any underlying biological abnormalities accompanied the symptoms of these illnesses. This led some physicians to dismiss patients seeking care for their symptoms. However, it now is clear that Long COVID and ME/CFS share not only similar symptoms but also a large number of similar abnormalities involving the central (CNS) and autonomic (ANS) nervous systems, the immune system, energy metabolism, redox imbalance, the gut microbiome, and the vascular system3.

From pathophysiology to symptoms

Understanding how this complex, multi-system underlying pathophysiology leads to the symptoms of the illness should lead to effective treatments. Clearly, some of the documented underlying abnormalities would seem capable of directly causing the symptoms through their effects on the brain. For example: 1) dysautonomia could reduce the flow of blood through the brain’s microcirculation; 2) endothelial vascular dysfunction could do the same; 3) autoantibodies targeting epitopes in the central and autonomic nervous system could plausibly affect cognition and autonomic function; 4) the impaired ability to generate adenosine triphosphate (ATP) from glucose, fatty acids, amino acids and oxygen which has been documented could alter neuronal function.

Sickness symptoms and behavior

As a physician-scientist caring for people with ME/CFS, and studying the illness over the past 40 years, I was struck by the number of people who described the illness as “a flu that never goes away.” That raised two questions: 1) Why do we feel the way we do when we get the flu, and 2) Why, in some people, do these flu-like symptoms persist?

Pursuing the first question led me to discover the large literature on “sickness behavior”. It’s been known for decades that following experimentally-induced infection, humans and most animals studied (including fruit flies and worms) exhibit “sickness behavior”, including reduced activity of all sorts and reduced appetite. In humans, these behavioral changes clearly are in response to symptoms - we feel exhausted and we ache so we don’t move very much. We experience “brain fog” so we avoid cognitive challenges. We feel sleepy so we sleep. We have a poor appetite, so we don’t eat and digest. One can assume the same is true in other animals, even if they cannot articulate the symptoms. (If a parrot were to chirp that it was “tired and sleepy”, you wouldn’t believe it.)

At that time, 40 years ago, animal studies suggested that inflammation somehow triggered the behavior. Cytokines were being identified as molecular orchestrators of the inflammatory response, and recombinant DNA techniques were producing cytokines for use as therapeutics. Investigators reported that the infusions of cytokines were followed promptly by “flu-like” symptoms; then, when the infusion stopped, the symptoms promptly resolved. This supported the notion that sickness symptoms in humans were generated by molecules of inflammation.

Why has such “sickness behavior” following an infection been preserved by evolution throughout the animal kingdom? A plausible reason is this: sickness behaviors involve reductions in activities that consume ATP: movement, thinking, digesting, etc. By reducing energy-consuming behaviors, more ATP remains available to fight the infection and heal the injury caused by infection.

How would evolution have preserved sickness symptoms? In the past five years, in rodents, several neural circuits based in the hypothalamus and brainstem have been identified that appear to mediate sickness behavior. These neural circuits are triggered by molecules of inflammation, primarily cytokines and eicosanoids.

So, while specific infectious agents may be capable of generating certain symptoms more prominently than other infectious agents—such as the loss of taste and smell that can occur with SARS-CoV-2 infection in people with COVID-19—all infectious agents appear capable of generating a similar and stereotyped group of symptoms.

What triggers the neuroinflammation that then triggers the symptoms?

Neuroinflammation can result from infection of, or injury to, the CNS. In the past 20 years, it has become clear that inflammation outside of the brain, such as in the gut or lungs, can also trigger neuroinflammation—via blood-borne signals that penetrate the blood-brain barrier and retrograde neural signals that travel up the vagus nerve. There is considerable evidence in ME/CFS of smoldering low-grade inflammation in the gut caused by dysbiosis of the gut microbiota. In Long COVID, the immune response sometimes fails to fully eliminate the nucleic acids and antigens of SARS-CoV-2: they provoke an ongoing inflammatory response.

Summing it up

As shown in Figure 1, and summarized in detail in a recent publication4, many different underlying pathophysiological abnormalities can plausibly lead to some of the symptoms of Long COVID and ME/CFS. In addition, we argue that invocation of the sickness symptoms response may also be an important cause of the symptoms. Sickness symptoms and behavior in rodents appear to be generated by specific, dedicated, recently-discovered neural circuits. These circuits, and the neuroinflammation that activates them, suggest that targeting some aspect of neuroinflammation therefore may prove effective in reducing the symptoms.

About the Author

Snippet of Figure 1 from Neuronal wiring diagram of an adult brain | Nature, licensed under CC-BY. Representation of the computed segmentation of all neurons and ‘wires’ (axons and dendrites) for a whole adult fly brain.

The acronym Ansyme begins with ‘advanced’ neuroscience and this FlyWire work clearly hits the ‘advanced’ mark! This landmark dataset mapped all the (microscopic) neurons and the (more microscopic) connections between them in a brain. It was funded by the BRAIN Initiative, which is co-led by NINDS and whose own director likewise featured it as a 2024 highlight. The “A” in BRAIN also stands for advanced, and the Initiative was launched with an initial focus on pushing technological boundaries, i.e., what neuroscience can investigate. FlyWire demonstrates that fully.

Ansyme seeks to foster ‘systemic medicine’ applications of neuroscience, namely for humans, and someone outside this BRAINy club might wonder how relevant a highly detailed anatomical map of the fly brain may be to such ends. As if on cue, a mammalian connectome was just revealed last month from another multi-institutional team called MICrONS. This one was yet more technologically advanced, spanning a cubic millimeter of mouse visual cortex, and can be considered the crowning achievement in connectomics for the BRAIN Initiative’s first decade. For comparison, the adult fly brain spans about a quarter of a millimeter along its longest dimension, but its connectome has something else going for it: completeness.

One of our early articles showed that achieving whole-brain completeness with another neuroscience technology, spatial transcriptomics, was statistically transformative. It allowed an unbiased counting of cell types across brain regions that produced a major discovery: much higher-than-anticipated cell type diversity in the lower brain. While ‘more cell types’ is an abstract statement, its potential has been made more concrete by our early posts all highlighting research showing specific cell types of the lower nervous system can be mapped to specific functions, some of which are medical symptoms and states.

Could the completeness of a connectome, fly or otherwise, give us more gains in the realm of systemic medicine? I believe the answer is “yes”, judging from a tandem publication to the FlyWire data release from a Princeton-led team whose set of simulations and statistical assessment, grounded by the fly connectome data, led to their concise title “The fly connectome reveals a path to the effectome”.

Our early articles have been exploring neural effects revealed by the techniques of optogenetics and/or chemogenetics. These can artificially stimulate a specific molecularly-identified set of neurons, which serves to prove sufficiency: that a specific small set of neurons has a big effect, such as triggering fainting or controlling sepsis. Some studies went further, using genetic knockout technologies to establish necessity, i.e., abolish the function if a key connection to or from that small neural set is severed. The combination of necessity and sufficiency has strong implications in formal logic and mathematics. Here, more practically, it strongly suggests the neural circuits being identified are quite likely the specific circuits for the specific functions studied.

One of our early articles extrapolated from these examples: there might be a broader ‘dictionary’ of such one-to-one mappings between neural circuits and medical symptoms and unwanted states. We might call these “mal-effectoids”, i.e., specific individual, countable bad effects. As a patient stakeholder, this extrapolation is admittedly grounded in hope.

The Princeton-led analysis, on the other hand, was grounded in math. While the fly connectome itself was postmortem, and hence not a measurement of any physiological effects, it did reveal that connectivity was sparse – on average any neuron was only physically connected to ~.01% of other neurons – in addition to providing data regarding how strongly two neurons were connected (number of synapses) and the sign of that connection (predictable from the shape of those synapses). This information was formulated into a sparse matrix of over a billion entries whose properties could be studied with techniques of applied mathematics, seeking to identify which small sets of source neurons would have the largest global effects. The results were summarized in Figure 3 of the Princeton paper.

Snippet of Figure 3 from The fly connectome reveals a path to the effectome | Nature, licensed under CC-BY. Paper’s conclusion was largely based on a mathematical eigenvalue decomposition of the connectome matrix, whose eigenvalues and eigenvectors (summarized at left) correspond to putative ‘eigencircuits’ illustrated at right. The large number of roughly equal-valued eigenvalues and the low neuron counts of and correlation between these eigencircuits support the paper’s conclusions per the authors.

Their team conducted random stimulations of select neurons in the digital model and observed their effects over time, reaching the conclusion that “fly whole-brain dynamics are generated by a large collection of small circuits that operate largely independently of each other”. Their digital model was achieved by formulating the sparse connectome matrix as a linear dynamical operator, whose resulting eigenvalues and eigenfunctions reveal the dominant dynamics of the connectome ‘brought to life’ with random small stimuli.

Importantly, this large-scale virtual experiment could be grounded by real-world experimental data. Fly neuroscientists have been doing the same kinds of optogenetics experiments – one neural circuit at a time – as in the rodent experiments highlighted by our early posts. This Princeton computational study “rediscovered” some of these known neural circuits, shown in their Figure 4, including the “ring attractor” circuits discovered to act like a compass for fly navigation.

The heart of their conclusion – “...a large collection of small circuits that operate largely independently of each other” – concisely connected several dots in my head. A “large collection” jibed with my guess there may be a broad ‘dictionary’ of effects, and hence mal-effects (aka symptoms), identifiable within our neural circuitry. That they “operate largely independently” jibed with my own experience with symptoms and dysfunctions seemingly having ‘minds of their own’ with somewhat predictable ‘logic’.

These ideas of piecemeal physiological logic went into my Idea submission during the spring 2024 public comment period for the ME/CFS Research Roadmap, which was also highlighted in the year-in-review post by the NINDS director. Public commentary on the scientific landscape appears to be expanding, e.g., a new philanthropy-backed “gap map” from Convergent Research that launched last month. Ansyme thus aspires to make commentary on public scientific “maps” one of our activities, so those who plan and fund future scientific activities can see the connections between today’s advanced neuroscience and tomorrow’s systemic medicine.

In 2024, the conversation changed after this breakthrough, also published in Nature. The phenomenon of a “cytokine storm” entered public consciousness during the height of the pandemic, unfortunately all too often: these storms preceded many of the millions of deaths. This study discovered something hardly anyone conceived during the pandemic: some cytokine levels are significantly controlled by the lower brain. 

This work came from the team of Charles Zuker, a pioneer in the sense of taste. For this study, his team turned its focus to internal sensing, aka interoception, of cytokines. They pinpointed specific neurons within the vagus nerve that responded to the injection of cytokines into the bloodstream. In fact, they found two distinct sets of neurons: one set responding to pro-inflammatory cytokines (IL-6, IL-1β, TNF) and another set responding to an anti-inflammatory cytokine (IL-10).

Several neuroscience technologies were used: activity mapping, activity-based genetic labeling, chemogenetics, and live calcium imaging. Activity mapping, using a molecule called cFos that brightens with activity, is what first clued Zuker’s team into the role played by ascending body-to-brain vagal neurons. Whole-brain microscopy pinpointed a nucleus within a nucleus:  the caudal nucleus of the solitary tract nucleus (abbreviated cNTS due to Latin name), as the single most activated brain region following an artificial immune stimulus with lipopolysaccharides (LPS), a bacterial component known to trigger the innate immune system. NTS is the brainstem region where the vagus nerve enters the brain from the body. The authors further confirmed the cNTS brainstem activity arises from body to brain interoception, not LPS permeating the brain, by abolishing the activity through vagus nerve ablation.

Once this interoception pathway became clear, the authors wondered: does the cNTS itself exert influence back from brain to body? Today’s neuroscience allowed them to specifically “tag” the activated cNTS neurons with chemogenetic receptors for physiological testing. They found the answer to be Yes, with two key experiments:

• silencing the body-to-brain signaling led to runaway inflammation (Figure 2)

• activating cNTS suppressed/boosted pro-/anti-inflammatory cytokines (Figure 3)

From there, they proceed to what scientist-me considers the centerpiece: Figure 4, excerpted above, revealing specific cell types within the vagus nerve that sense the pro- and anti-inflammatory cytokines reaching the cNTS. This worked because the vagus nerve “balloons” at what’s called the nodose ganglion, making it accessible to today’s neuromicroscopy. They went further (Figure 5) to identify which molecules are differently expressed in those distinct vagal neurons, i.e., their respective molecular “barcodes”. 

Finally they reached what everyone-but-me would consider the centerpiece: a pathway to preventing LPS-induced sepsis! Chemogentically activating the now molecularly pinpointed anti-inflammatory pathway, either in the vagus nerve or the brainstem, could rescue mice from an otherwise lethal dose of LPS:

Overall, the lower nervous system appears to have evolved accelerator- and brake-like control over the “fuel” and “rubber” components of the immune system, with each pathway identifiable down to cells and molecules by using advanced neuroscience. This detailed understanding appears poised for clinical use, as neural interfaces could be used to “pump the brakes” to prevent fatalities from acute infections. While I’m no immunologist, a quick check shows the cytokine pathways in this study may be similar to those of the COVID-19 cytokine storms. Imagine - this bacteria model study could help limit casualties in a future viral pandemic!

Although focused on acute illness dynamics, this study notably opened up the chronic illness community to discussing the brain’s role more. But much of the chatter, and even Nature itself, drifted towards images and ideas of the brain being inflamed, sometimes under the broad banner of neuroinflammation. In fact, the brain was not directly affected by LPS in this study, even at lethal doses. That’s why Ansyme exists: to better communicate what today’s neuroscience is capable of and what the lower nervous system can do.

As a PhD student, I worked on the zebrafish version of the cortex. With a background in physics, I specialized in calcium imaging and single-cell electrophysiology–methods for mapping the activity of neurons in living animals. I hoped that by listening closely to these cortical cells, I could unlock their secrets. The widespread use of calcium imaging is based on the idea that monitoring as many neurons as possible will enable a better understanding of the brain. But even after three decades of technical progress, we can often only explain a fraction of what's going on. Despite these challenges, I stayed focused on the cortex, until a series of coincidences shifted my perspective and pulled me towards the brainstem.

The first twist of fate was that I stayed in Switzerland after my PhD. Like many academics I had intended a bigger move at this career stage, but my best opportunity turned up nearby in Fritjof Helmchen's lab in Zurich. The second coincidence occurred just before I started on my originally planned project. Fritjof approached me because a PhD student in his lab, Steward Berry, had died in a tragic accident, and Fritjof asked whether I would continue the project and keep Steward’s intellectual legacy alive.

Steward had been studying astrocytes - non-neuronal brain cells I knew little about. After some thought, I agreed to take it on, at least for a year or two. Being open to the unexpected had paid off for me before, so why not this time?

To study astrocytes, I started using a set of techniques that sound like science fiction but are now standard in “systems neuroscience”. We use viruses to deliver genetic information to astrocytes in a mouse’s brain, making them produce a fluorescent protein (GCaMP) that lights up when calcium levels rise, an indicator of activity. While the mouse walks or rests on a treadmill, we use laser-scanning microscopy to observe these signals through a glass window in the skull.

Excerpt of Figure 1 (modified) from our study, licensed under CC-BY. Shows the methods used to record from astrocytes: monitoring of awake mice (left), transgenic expression of a fluorescent protein using a virus (center), and two-photon microscopy of astrocytes in the living brain.

Over time, I became deeply familiar with astrocytic calcium signals in the mouse hippocampus. Interestingly, these signals grew stronger when the animal was aroused or alert. I began tracking arousal by recording pupil diameter, which, as in humans, reflects not only lighting conditions but also internal states - relaxation, attention, stress. I also noticed that calcium signals tended to flow from astrocytic processes to their cell bodies, but only during heightened arousal (a finding we published last year). Arousal seemed to control this integration of signals in astrocytes.

At this point, I began wondering: What exactly is arousal? In working so closely with these signals, I had unknowingly entered a new world, like a child who touches a magic portkey and finds themself in another realm, a world governed by the brainstem. But it would take more time to fully understand that.

Excerpt of Figure 2 (modified) from our study, licensed under CC-BY. Shows the pupil of the mouse during the experiment. The pupil appears bright since laser light entering the brain for imaging exits via the pupil. Examples of different pupil sizes are shown. Similar fluctuations also occur in humans.

In another fortuitous coincidence, my colleague Xiaomin Zhang happened to know Sian Duss, a PhD student at ETH Zurich, and suggested I talk to her. Sian worked in Johannes Bohacek’s lab, which focuses on stress and the locus coeruleus (or LC) - a tiny brain region (about 3,000 neurons in mice, 50,000 in humans) that plays an outsized role in regulating arousal and stress. Sian is an expert in optogenetics, a technique that allows precise activation of neurons with light. Together, we optogenetically activated the LC to induce “artificial arousal.”

The effects were stunning. We observed widespread activation of astrocytes across the hippocampus, and others have made similar observations in the cortex. The LC therefore can control millions of cells in mice (in humans, it's billions).

Calcium imaging recording of astrocytes in the hippocampus when artificially stimulating the locus coeruleus. Data recorded together with Sian Duss.

More than that, stimulating the LC changed the animal’s behavior. The pupils dilated. Sometimes it would freeze for several seconds, or its tail shot straight up. The LC has descending connections to the spinal cord, but I suspect that its broad influence on the cortex also contributes to these behavioral effects. Sian and I are now exploring what happens not just to astrocytes but also to neurons during artificial arousal.

These results reminded me of research I encountered during my PhD in the neighboring lab of Silvia Arber, where optogenetic stimulation of other small brainstem nuclei evoked very specific movement patterns. I came to understand that the brainstem has tremendous power–not only to control the body via descending pathways, but also to shape cortical activity via ascending projections.

Since then, I've been following the growing interest in the brainstem's role in controlling the body and brain. The case studies that Vijay has shared on this blog are only a subset of larger efforts to identify brainstem nuclei and to understand their connections to brain states and physiological events.

Gradually, I realized these small but powerful brainstem regions might be the ideal gateways for therapeutic intervention. Currently, brain-machine interface (BCI) research mostly focuses on the cortex, attempting to control a few thousand neurons through electrical stimulation. But wouldn’t it make more sense to target control hubs in the brainstem, which themselves influence billions of brain cells and interface directly with the body?

I’m convinced that this approach could, over the next decades, become more clinically relevant than cortical BCIs. Disorders associated with neurodegeneration (like Alzheimer’s or Parkinson’s) or imbalanced brain states (like depression) may not be effectively treated through cortical BCIs. But targeted stimulation of specific brainstem nuclei could help restore balance and mitigate symptoms.

The critical question is: how can we achieve such targeted stimulation in the human brain? Broad electrical stimulation is unlikely to work for such small nuclei. That’s why the first step must be thorough investigations in animal models: What is the diversity of brainstem nuclei and subnuclei? How can we define and distinguish them using molecular tools (like transcriptomics and genetic markers) and functional signatures (like activity patterns)? How do perturbations of these nuclei affect the entire organism–its body, cortex, astrocytes, and neurons? Only with these foundational insights can we begin to use the brainstem as an interface to precisely influence the brain and body.

In spite of the coincidences that led me to the brainstem, I still believe that what makes us most human–the seat of our intelligence–lies in the cerebral cortex. But if we want to interact with that cortex (and the body) in an effective way, it might be wise to use the tools the brain itself built for that purpose. And those tools sit deep in the brainstem.

It’s important to remember that not all neural functions reside in the brain and spinal cord. We have a second nervous system called the enteric nervous system that controls digestion. Back in 2019, I started to develop chronic constipation, a condition that sounds relatively benign, but can lead to painful bloating, cramps, and more serious conditions like diverticulitis or fecal impaction. At the time, I was able to manage it through dietary changes and increased exercise, but then in 2021, I developed ME/CFS as a result of COVID-19, and any real physical activity became impossible. Needless to say, my gut became quite unhappy.

Another common gut dysfunction is Irritable Bowel Syndrome, which causes chronic bloating, diarrhea, constipation, and abdominal pain. As with other complex and systemic medical issues, the etiology of IBS and similar disorders is poorly understood and attributed to a wide range of possibilities. Therapeutic options are limited to lifestyle changes or medications that address specific symptoms, but because of the varied nature of these disorders, treatments are often unsuccessful.

Research on the enteric nervous system is essential for understanding many of the functions that underlie gut disorders like IBS. The macro physiological structures are well known, but there remains much to learn about the cellular-level details, especially in terms of enteroendocrine cells, which form synaptic connections to nerves in the digestive tract. These nerves originate from the spinal cord dorsal root ganglia as well as the vagal nodose ganglia, both of which play roles in the lower nervous system. Recent advances in neuroscience techniques are providing more insight into the workings of this “second brain” and one of its key players, the enterochromaffin cell (EC), a subset of enteroendocrine cells. These cells produce serotonin and regulate a variety of gut functions, including motility, secretion, and pain.

Figure 1 from Enterochromaffin Cells–Gut Microbiota Crosstalk: Underpinning the Symptoms, Pathogenesis, and Pharmacotherapy in Disorders of Gut-Brain Interaction, licensed under CC-BY-NC, illustrating the structures and dynamics of enterochromaffin cells and gut mechanisms.

Research by Kaelberer and Bohórquez has revealed a neural circuit mediated by glutamate that directly links the gut to the nodose ganglia. Synaptic links between specific enteroendocrine cells and vagal neurons were confirmed in both mouse models and in vitro organoids. Bohórquez dubbed these special gut cells “neuropods”, and some scientists credit him as being the first to discover these connections, though more study is necessary to establish the term.

Excerpt of Figure 2 in An Enteroendocrine Cell – Enteric Glia Connection Revealed by 3D Electron Microscopy, licensed under CC-BY. This shows a 3D reconstruction of an enteroendocrine cell with an axon-like basal process, also called a neuropod, obtained via serial block face scanning electron microscopy (SBEM) of a mouse EC cell.

An April, 2025, paper by Yan Song and others from UC San Diego identified 14 different EC cell types in the gut, including a subset of EC cells in the colon that demonstrate “molecular and cellular characteristics reminiscent of neurons.” Specifically, the Piezo2 EC cells have transcriptome factors common to neurons as well as axon-like projections, similar to the neuropods identified by Bohóroquez. The UCSD group used mice models to confirm that these colonic EC cells affect gut motility, which is often dysregulated in people with IBS.

Figure 8 in Stratification of enterochromaffin cells by single-cell expression analysis, licensed under CC-BY. This shows different EC clusters identified in the gut, including the piezo2 cells that have neuronal characteristics and regulate motility in the colon.

In a November 2024 paper in Nature, Tongtong Wang and the Oka group at Caltech identified two distinct classes of gut neurons that separately control motility and secretion, the two primary functions of EC cells. Using a combination of transcriptomics, optogenetics, and chemogenetics–three techniques that have enabled a lot of advancements in recent neuroscience–they examined the celiac-superior mesenteric ganglia (CG-SMG), specifically looking at the sympathetic neuron population in this complex.

The CG-SMG complex is known to interface between the brain and the gut. The Caltech group examined 371 genes and identified two classes that are expressed by neurons, relaxin receptor type 1 (rxfp1), and homeobox gene shox2. Using genetically modified mice and optogenetic stimulation, they further determined that rxfp1 neurons down-regulate the secretion of bile and glucagon, which control digestion. The shox2 neurons down-regulate gastrointestinal transit, thereby slowing gut motility.

An illustration of the differentiation of enterochromaffin neurons in the CG-SMG complex. The rxfp1 cells innervate and control the secretory organs of the digestive system. The shox2 neurons innervate the intestinal tract and affect gut motility. Credit: Divya Srinivasan Breed

While these various findings don’t immediately point to solutions for chronic disorders like IBS, they illustrate that the digestive system contains its own complex neural circuitry, including cell types that have yet to be identified. Further work needs to be done to connect the dots between the enteric “brain”, the preganglionic neurons of the spinal cord, and the “ventral brain”, to understand the complete (and complex) circuit that regulates digestive function. My gut tells me that this will eventually lead us to more effective treatments for disorders such as mine.

This article revealed  two specific neural pathways associated with one form of fainting, vasovagal syncope, also known as the Bezold-Jarisch Reflex. First described in the 19th-century, this form of fainting has a characteristic set of three observable signs: reduced heart rate (bradycardia), reduced blood pressure (hypotension), and a backwards eye roll moments before physical collapse.

The first neural pathway runs along the vagus nerve from the heart to the lowest region (medulla) of the brain and was shown in mice, via optogenetics technology, to reliably trigger the full Bezold-Jarisch Reflex along with additional physiological signs, such as reduced cerebral blood flow. In this case, “pinpointing” a neural pathway meant three specific things: 1) beginning at a specific subregion of the heart, the ventricular walls, 2) molecularly-identified neurons within the vagus nerve, namely those expressing neuropeptide Y receptor Y2, aka NPY2R, and 3) ending at a specific subregion of the medulla known as area postrema. 

Snippet of Figure 1 from Vagal sensory neurons mediate the Bezold–Jarisch reflex and induce syncope | Nature, licensed under CC-BY. Conveys the level of spatial, molecular, and temporal specificity the authors achieved within the vagus nerve. 

This is the level of specificity we can now expect from today’s neuroscience. The authors went further to establish this neural pathway as both necessary and sufficient for vasovagal syncope, i.e., this appears to be the pathway for vasovagal syncope.  Several approaches were used to reach this strong level of evidence, such as targeted ablation. Finally, they went on to distinguish the discovered pathway from a neighboring one involved in sleep-wake regulation. Fainting and sleeping may both involve “laying low” but we know they are distinct physiologically. Now this can be pinpointed to specific neural wires!

The second new neural pathway identified determined how long the animal was laid low. It turns out this “post-syncope” has a neural timer on the upper end (hypothalamus) of the lower brain. Again it was pinpointed to a specific sub-region, the periventricular zone, and with mechanistic/causal precision: how long the animal was both physically laid low/slow and cognitively impaired could be modulated (made either better or worse) with particular (activating or inactivating) chemogenetic stimulation of this region, all without affecting the other physiological signs of syncope.

Snippet of Extended Data Figure 9 from Vagal sensory neurons mediate the Bezold–Jarisch reflex and induce syncope | Nature, licensed under CC-BY. Shows the ability to modulate the apparent syncope duration, via manipulation in the hypothalamus.

While I’ve never experienced vasovagal syncope per se, my own recumbent episodes come to an end (I call it “rebooting”) at a certain unpredictable point, and also (sometimes) with a slowness that persists after I’ve arisen. Perhaps this mechanism is in play, or another like it.

This research was remarkable not only for its findings but also for being a technical tour de force leveraging several advanced neuroscience technologies: optogenetics, chemogenetics, single-cell RNA sequencing, tissue-cleared microscopy, Neuropixels probes, small-animal EEG, and quantitative behavioral analysis. 

For me, this work was instantly a watershed moment. It was also encouraging that at least one physician, Dr. Robert Wilson at the Cleveland Clinic, started to connect the dots between these findings and the recent rise of dysautonomia cases secondary to Covid, as profiled in just three worthwhile minutes on National Public Radio (NPR).

These are exactly the kinds of bridges that we at Ansyme aspire to build. We believe that clever 21st-century detective work is what’s needed to beat chronic illness and other medical mysteries. To borrow the phrase of a famous fictional 19th-century detective, this research article was the moment I knew “the game is afoot”.