This yearly tradition is a way to express my admiration for humanity’s creativity and its desire to understand the world and make it a better place. Similar to the 2022 and 2023 lists, there’s no particular curation criteria - just breakthroughs (in neuro tech bio) that I find fascinating and consider incredibly impactful.
The immune system precisely regulates responses to threats and helps maintain the health of whole organisms. It is mainly divided into adaptive and innate. The function of the adaptive immune system is to learn about previous threats and develop antibodies for specific protection. The innate immune system acts immediately after a threat to build a non-specific response (like inflammation). Scientists have believed that the nervous system is involved in the control of immune response, but it hasn’t been clear how, until a study from this year, where researchers identified and characterized a body-brain circuit that senses and controls innate immune response.
First, the researchers identified a brain region that gets activated when mice get injected with bacteria that induce an innate immune response. This brain region is the caudal nucleus of the solitary tract (cNST) and is located in the brainstem. When neurons in this brain region were inhibited using chemogenetics, the immune response to the bacteria got out of control - increased pro-inflammatory and decreased anti-inflammatory response. When neurons were activated, instead of inhibited, also using chemogenetics, the effect was opposite - decreased pro-inflammatory and increased anti-inflammatory response. This means that these neurons in the brainstem perform neural control of immune response. Interestingly, when this brain region was activated in the absence of an immune challenge (no injection of bacteria), the immune response was not altered. This implies that the neural task is to monitor and regulate the immune response, not to start it. This suggests the sensing of the immune response starts elsewhere, likely transmitted via the vagus nerve.
The researchers imaged (calcium optical imaging) afferent neurons in the nodose vagal ganglia (where cell bodies of the vagus nerve reside, see image below). This imaging revealed that a population of neurons are activated in response to pro-inflammatory cytokines, and a different population are activated in response to anti-inflammatory cytokines (see image below). This indicates that neurons in the nodose ganglia directly sense cytokines and then transmit inflammatory information to the cNST. Researchers then activated neurons in the nodose ganglia, via chemogenetics, and showed immune control. Activation of a population of vagal neurons enhanced anti-inflammatory response while reducing pro-inflammatory response. The researchers then demonstrated that these vagal neurons have monosynaptic connections to the neurons in the cNST. Finally, the researchers showed that chemogenetic activation of this body-brain circuit increased the survival rate of mice to 90% after a lethal injection of bacteria.
Neurological conditions like spinal cord injury (SCI) involve complex neural interaction across multiple neural systems and time. Incomplete SCI is characterized by immediate paralysis after the injury followed by a partial recovery of walking due to reorganization of neural projections from the brain to the spinal cord. Researchers from Courtine’s lab developed an atlas looking at transcriptional activity in the brain and projections from the brain to the spinal cord. Due to the remodeling after the injury, researchers performed this at different time points (at 1 week and 8 weeks after injury). To identify which brain regions are involved in this remodeling, 4 requirements were established: 1) a decrease in transcriptional activity after SCI (1 week), 2) an increase in transcriptional activity after remodeling, when walking recovery occurs (8 weeks), 3) a decline in neural projections after SCI (1 week), and 4) an increase in neural projections after walking is partially recovered (8 weeks).
A single brain region met all four requirements: the lateral hypothalamus (LH). Optogenetic activation of the LH facilitated walking of mice with incomplete SCI. Projections from the LH to the spinal cord below the injury were not found, so the researchers looked for relay neurons. Projections were found from the LH to the ventral gigantocellular nucleus (vGi) and then to the spinal cord below the injury (see the beautiful projections in the video below). Precise optogenetic activation of neurons in the LH projecting to the vGi demonstrated immediate recovery of walking in mice with incomplete SCI. The researchers then performed deep brain stimulation (DBS) of the LH in mice and rats, and demonstrated improvements in walking. Finally, the researchers implanted DBS electrodes into the LH of two humans with SCI, and showed improvements in walking, including climbing stairs independently.
The understanding of light-tissue interactions has advanced many medical technologies. Laser surgery, ultraviolet therapy, near-infrared spectroscopy, optical coherence tomography, fluorescence imaging, optogenetics, to name only a few, all rely on light-tissue interactions. However, most of these technologies are limited by scattering, the redirection of photons as they encounter particles, instead of following a straight path. The refractive index (RI) of a material indicates the degree to which light slows down when passing through a material compared to its speed in vacuum. Light scatters when there’s an RI mismatch. Molecules in tissue, such as organelles, proteins, and lipids, all of which have different RIs, increasing scattering.
Scientists used food dyes (derived from a Lorentz oscillator model) to increase the RI of the tissue medium and reduce scattering. The researchers chose tartrazine, a common food dye approved by the FDA. When tartrazine was applied to tissue from living mice, tissue became transparent. Clearing the mouse abdomen with tartrazine enabled direct visualization of gut movements, while clearing of the scalp enabled visualization of cerebral blood vessels. Clearing of muscle tissue enabled imaging of sarcomeres using th a microscope. Impressively, when tartrazine was washed off, tissue transparency was reversed.
The best example of a foundational model is ChatGPT, a transformer-based architecture that acquires the capacity of performing well across multiple domains. These architectures learn general-purpose knowledge (like grammar, syntax, semantics) due to a pre-training process on absurdly large amounts of data (like the whole text on the internet). These models are then fine-tuned during training to adapt the pre-trained model for specific tasks (like question answering). These architectures are now inspiring foundational models in biology, where instead of text, DNA sequences are used for pre-training.
This year, a foundational model called Evo was released. Interestingly, Evo is not based on a transformer architecture, instead, is based on StripedHyena, which can handle larger context lengths (up to 32,000 tokens). Evo was trained on 2.7 million prokaryotic and phage genomes. Evo shows zero-shot prediction capabilities, meaning it can make accurate predictions without being explicitly trained on those examples. Evo showed zero-shot protein function prediction, being on par with protein language models like ESM. Evo also performed zero-shot prediction on non-coding RNA and regulatory DNA. In an impressive demonstration, the team showed that Evo can generate novel CRISPR-Cas and transposable systems. This is challenging because CRISPR systems require proteins and non-coding RNA, and transposable systems require proteins and DNA. Finally, the team showed that Evo can generate sequences at the scale of whole genomes, due to the long context enabled by the StripedHyena architecture.
The mammalian brain is an incredibly complex organ, thousands of cell types and billions to trillions of connections (synapses) between them. Our interpretation of reality and all of our experiences arise from the information flow (in the form of chemical and electrical signals) between neural cells. A connectome is a wiring diagram of the brain, and so far, there’s only a full connectome of the worm C. elegans, which has only 302 neurons. On the other extreme, it is estimated that the human brain has 80 billion neurons and 100 trillion synapses. At an intermediate size, the fruit fly (Drosophila) has one million times fewer neurons than a human brain. This year, an ultra-collaborative and open project called the FlyWire consortium, published the full connectome of the fruit fly. The researchers used electron microscopy to image the whole brain at nano-scale resolution. Then, the images were aligned to reconstruct the shape and position of every neuron and synapse. 54.5 million synapses and 140,000 neurons were identified. Based on this reconstruction, the researchers determined if synapses were likely excitatory or inhibitory, as well as the information flow from sensory to motor neurons across multiple brain regions.
Runner-ups:
-Mechanism for topical vaccines
-Lenacapavir for HIV prevention
-Ultrasound to open the blood-brain barrier
Shameless plug. This year we released a project that I’ve been working on for the past years. We showed that we can control living muscles artificially with great accuracy and incredible fatigue-resistance using optogenetics, opening the door for optogenetic motor prosthesis.
No posts
Comments
Nothing yet. Say the first thing.
Sign in to join the conversation.