Action Potentials for September
1: Engram cells are specialized neurons that become active during experiences, undergo chemical and structural changes to store a memory, and can trigger a specific behavior when reactivated.
A new study uses context-specific freezing behavior in mice as a readout for changes in engram cells, focusing on the monosynaptic connection between the hippocampus and the amygdala.
First, they found they could decrease reactivation of the engram by tagging engram cells in the amygdala with an inhibitory opsin and then shining a light on them, consistent with previous results:
On the cellular level, engram cells were found to have a higher density of dendritic spines and a decreased expression of the protein PSD-95 after recall (in the figure below, “linked” refers to the condition in which the memory was formed):
The authors also note that the MAGUK family of proteins, to which PSD-95 belongs, is “functionally redundant” with several interacting proteins, such as PSD-93, SAP97, or SH3. The difficulty of studying the effect of synaptic molecules in isolation is bad for neuroscientists who are trying to isolate the functions of individual molecules, but good in brain preservation, wherein the redundancy of biomolecules makes the task of preserving engram information easier.
2: Inhibiting the axonal projections from the orbitofrontal cortex to the dorsomedial striatum causes impairment in a decision-making task, so that rats are not as sensitive to differences in reward volume between choices. Interesting example of how frontal lobe brain damage can affect decision-making:
3: Neural decoding of attempted speech for people with facial paralysis, via recording from the ventral premotor cortex. Amazing progress, but they note that they have a 24% word error rate on a 125,000-word vocabulary, which may not yet be sufficiently low for everyday use (compared to 4-5% with speech to text systems).
4: New imaging method for mapping the brain across spatial scales using light microscopy. Their fixed tissue staining method was developed by testing the binding of polysaccharide-binding proteins in the extracellular space, which has a diversity of biomolecules. They settled on using wheat germ agglutinin, which can be used as a stain to outline cell membranes. They found that this led to good staining across the entire mouse brain:
This paper is a tour de force of neural imaging, making it hard to summarize, but it has lots of interesting findings.
5: New method for correlative light and electron microscopy in brain tissue using fluorescent aptamers:
6: Using single-cell transcriptomics and immunohistochemistry to map paracrine peptidergic signals that function alongside the synaptic connectome to make the Drosophila circadian clock tick:
Just another example of how some people have claimed that “but the molecules” is somehow a critique of the idea of brain preservation, when in reality the molecules can simply be mapped from the preserved brain as well (for example, the brains shown in D-F were preserved with formaldehyde).
7: Norepinephrine stimulates the differentiation of oligodendrocyte precursor cells. More evidence for “adaptive myelination”, wherein nearby neuronal signals modulate the extent of local oligodendrogenesis and myelination.
8: Progress in our understanding of astrocyte gliotransmission, as astrocyte-targeted deletion of VGLUT1, a protein important in glutamate transmission, is found to lead to alterations in electrophysiology:
9: Konrad Kording on a research program towards simulating C elegans. One of his key points is that if neuronal input-output functions are determined in a smaller organism, the basic rules might scale so that they don't all needed to be determined in as complicated of a way in larger organisms. At that point, you could estimate the neuronal input-output functions from the static molecular annotated connectomes.
10: Study on how Gene Ontology-based semantic modeling can be used to functionally annotate the C. elegans connectome in an automated matter. Related new-to-me term, not sure how I feel about it: “lo-fi uploading”.
11: The human brain project, launched in 2013, failed to reach its goal of simulating the human brain by 2023. Not even close!
12: Gwern on the benefits of artificial compared to biological neural networks. Also Gwern on how LLMs are naturally agentic and this is not surprising because they were trained to reproduce text written by agents - i.e. via behavior cloning. (Behavior cloning is a supervised learning approach in which a model is trained to mimic the actions of an expert agent based on observed state-action pairs, commonly used in fields like robotics and autonomous driving.)
13: Nora Belrose on why an AI pause will likely backfire and end up being net harmful.
14: Jacob Cannell argues that whole brain emulation (WBE) has economic advantages over artificial general intelligence (AGI) due to the negative externality of AGI being more challenging to control. He also suggests that narrow AI could serve as a stepping stone to achieving WBE.
15: It's Steve Horvath's world, we're just living in it. Another banger article from Horvath et al looks at using pig plasma to markedly reverse biomarkers of aging in rats.
16: Evidence that APOE e4 with a loss of function mutation is protective against Alzheimer's disease pathology. Suggests that APOE e4 increases the risk of Alzheimer’s disease due to some kind of active function.
17: Quest announced it will offer a direct to consumer Alzheimer’s disease blood test measuring the Aβ42/Aβ40 ratio for $399. This is possible because diagnostic tests do not require FDA approval. They can just become CLIA-approved if they have good enough test-retest reproducibility. “Alzheimer’s researchers contacted by Alzforum universally decried Quest's decision. In addition to concerns about the test’s accuracy, they stressed the need for blood tests to be interpreted by a knowledgeable dementia specialist.” 🤔
18: Plasma levels of dopa decarboxylase may be a biomarker for Parkinson's disease and related disorders (LBD, MSA, PSP, CBS). Very important because there currently is no good blood test for these conditions.
19: Genomics study predicts that there was a severe population bottleneck between 800,000 and 900,000 years ago, leading to a reduction in population size from 100,000 to 1000 people, that lasted around 100,000 years. Back when extended family reunions basically included all humans on earth.
20: Review on overdose detection technologies. For example, passive motion detection systems in high-risk environments like bathrooms can autonomously monitor for reduced micromovements linked to unconsciousness, triggering emergency alerts for timely intervention, without compromising user privacy through audio or video recordings. Some potential overlap with cryonics monitoring systems.
21: Comprehensive review of randomized control trials (RCTs) using psychodynamic psychotherapy. There have been more than 300 RCTs, most of relatively brief duration (i.e. <40 weeks), and 40% have been published within the last decade.
22: Analysis of an RCT initially started in the 1990s and followed for many years finds that stimulant prescription is not associated with adulthood substance use or substance use disorder.
23: Nitrous oxide reduces suicidal ideations at 24 hours after treatment in an analysis of small trials of people with treatment resistant depression.
24: Should PCPs routinely ask their patients if they are suicidal? A new USPSTF review evaluated the evidence and found that there was only one relevant trial, limited to patients who had screened positive for depression, and it had inconclusive results. Sadly, there are also real potential harms that could result, like denial of insurance payments. So their conclusion was that more research is needed into the benefits and harms before it can be recommended.
25: A Phase I study by BlueRock Therapeutics found that their stem cell-derived therapy, bemdaneprocel, was safe and well-tolerated in all participants over one year, and also improved Parkinson's disease severity scores, especially in the high-dose cohort.
26: ASOs to prevent abnormal slicing events in Fragile X syndrome, restoring normal protein production. Very promising.
27: Method to improve the cryopreservation of C. elegans using vitrification, improving the recovery rate to 98.4% for larvae and 84.3% for adults. Impressive advance, but I’ll note — not yet 100%.
28: David Wallace Croft has a proposal for creating a cryonics industry consortium.
29: Cryopets has made some updates to their website. They are not currently offering services, but are planning to start with pet cryopreservation.
30: A Tom Scott video profiling Tomorrow Biostasis has 2M+ views:
Thousands of negative comments. I think this is partially just the nature of the internet though. I could make a video about saving baby whales and people in the comments would tell me I’m a terrible person for it.
Action Potentials for September
The link to Konrad Kording's talk is broken. Here's a working link: https://youtu.be/r3UMIyhzew0
So our species came close to extinction 😮💨 what a way to start my morning