Hi everyone, @BrainAWG
Gregory Nelson’s talk yesterday got me thinking. After irradiation, microglia look resting again within about a week, but the population itself does not recover. Density drops over the following weeks, self-renewal does not bring it back, and the niche is gradually filled by peripherally derived macrophages. The cells there a month later are not the ones that were there before.
That made me wonder whether, at least in some cases, what looks like resolution is actually replacement of the cell population, and whether behaviour tracks which cells are present rather than how much signal they emit at a given moment. So the question I want to raise: do behavioural outcomes track the composition of the microglial population ( density, turnover, share of peripherally derived cells ) rather than levels of an inflammatory marker at a single time point? Has anyone here thought about this in relation to our datasets, or come across it in discussion? I would like to know before I try to take it further.
From an architectural perspective, this also highlights the need to reduce radiation exposure before biological damage occurs through appropriate habitat design, shielding materials and protected internal zones.
Really excellent question! This is very difficult to measure with the datasets available on OSDR but it is something people thinking about in the context of neurodegeneration. I am sure Greg would be open to discussing!
To echo the talk, an important facet to examine (per Dr. Nelson) is the cellular membrane (in response to irradiation, as opposed to only nuclear structure). There likely needs to be more radiation dose rate/time testing to parallel these experimental configurations (with greater longitudinal sampling)
In terms of controlling the radiation exposure solar particle events can be shielded from reasonably well. For galactic cosmic rays, its a much more difficult problem because they are so penetrating and create a lot of secondary radiation. Shielding optimization can gain marginal benefit but you need to launch too much mass to make a large impact. There is the possibility of using regolith on the lunar surface but this would need significant amount of infrastructure which won’t be available in the near future. ITs a option for later missions though. There is no indication that active shielding using electromagnetic fields will be technologically feasible from a practical perspective in the forseeable future
One possibility I find interesting is that programmed cell death may occur alongside cellular replacement mechanisms. This replacement could involve the proliferation of surviving cells or the differentiation of other cell populations. In this context, it would be interesting to investigate whether a single surviving cell could proliferate and contribute to restoring the cell population, even if the original cells are no longer present.
I began studying and investigating different shielding options capable of providing this protection as a substitute for lunar regolith. The mass ratio is highly relevant.
Yes, there are definitely better materials for shielding than regolith. The main advantage of regolith is that it is already there. When you look at launch costs per kilogram to get to the moon its the main limitation for shielding
Another point is that I don’t see the means available today to build up a mass of regolith on the Moon, I believe that, even if it were possible, the cost would be considerable.
yeah, I thought the same thing until the last year or so when I got pulled into a study talked to some of the people working this. let me see if I can find relevant papers to send
Also you may want to join the regolith AWG if you haven’t already. One of the people there is from Kennedy space center and is involved in this work (moving the regolith and ISRU more generally)
Data: 8 spaceflight brain datasets from 4 independent missions (RR-3, RR-23, RR-10, RRRM-2), plus 3 radiation-only datasets. Stats: Welch t and Hedges g per dataset; tissues from the same mission averaged, then Stouffer across missions; 2,000 label permutations.
The composition index moves away from the peripheral/border signature in all 4 missions: Stouffer Z = −2.76, p = 0.006; permutation p ≈ 0.001. Dropping any single marker leaves Z between −2.64 and −2.88.
The shift comes from the peripheral set (Z = −2.57, p = 0.01). The homeostatic set does not move (Z = 0.08, p = 0.94).
The inflammatory set does not reach significance (Z = −1.84, p = 0.07). Its within-animal correlation with composition flips sign across datasets (r from −0.84 to +0.80). In these data, composition and marker level dissociate, which is the distinction you raised.
Radiation only (OSD-374 proton, OSD-611 Si-28, OSD-202 0.04 Gy Co-57 in normally loaded mice): all three point toward a higher peripheral share (g = +0.01, +0.29, +0.52). None is significant, and n is small.
Limits. This is bulk tissue, so it is a composition proxy, not a cell count. Border-associated macrophages and infiltrating monocyte-derived cells share markers. There is no behavioural readout in these animals. The direct test is the OSD-612 snRNA-seq (RR-10 brain, same animals as OSD-563/564) and the OSD-561–564 Visium data, both released as raw FASTQ only. That processing is the next step, and I’m working on it.
@michellewiens — follow-up with the single-nucleus data (OSD-612, RR-10 brain, same animals as OSD-563/564).
How: I processed the snRNA-seq without downloading the 284 GB of raw FASTQ. For each of the 10 animals (5 flight, 5 ground) I streamed the first 20M read pairs straight from OSDR and pseudoaligned them with kallisto|bustools. The reference was Ensembl GRCm39 cDNA plus repeat-masked gene bodies for a 58-gene cell-type marker panel, so intronic nuclear reads count too. That gives about 3–11k nuclei per animal.
Per-animal result, flight vs ground (threshold of ≥3 immune markers per nucleus; the same pattern holds at ≥4 and ≥5):
Microglia per 1,000 nuclei: 9.5 vs 9.4 (g = +0.04, p = 0.95). No loss of microglia in flight.
Border-associated macrophages per 1,000 nuclei: 0.36 vs 0.47 (g = −0.25, p = 0.67).
BAM share of immune nuclei: 0.043 vs 0.046 (g = −0.06, p = 0.93).
Homeostatic program within microglia (P2ry12, Tmem119, Sall1, Siglech, Cx3cr1 UMIs per 1k): 2.85 vs 2.57 (g = +0.60, p = 0.33).
How this fits the bulk result from my previous post: the directions agree (peripheral/BAM share down, homeostatic program up in flight), but none of the single-nucleus contrasts is significant at n = 5 per arm. Neither layer shows the population-replacement pattern in RR-10 spaceflight. That is a statement about spaceflight at the RR-10 dose and duration. It is not a test of acute irradiation, where the bulk radiation datasets leaned the other way.
Limits: I used about 5% of each library. Median depth is 180–630 UMIs per nucleus. Microglia are under-detected (roughly 1% of nuclei called vs. several % expected), immune-call purity is 47–78%, and there are only 0–5 BAM nuclei per animal, so the BAM comparisons are underpowered. Running the full libraries with a genome-wide intron-aware reference is the obvious next pass.
