P254 Synaptic Population Shapes Form Fingerprints of Brain Areas, Organize Along a Rostro-Caudal Axis
Martin Rehn*1, Erik Fransén1,2,3
1School of Electrical Engineering and Computer Science, KTH Royal Institute of Technology, Stockholm, Sweden.
2Digital Futures, KTH Royal Institute of Technology, Stockholm, Sweden.
3Science for Life Laboratory, KTH Royal Institute of Technology, Stockholm, Sweden.
*Email: rehn@kth.se
Introduction
Synaptic creation, modification, and removal underpin learning in the brain, balanced against homeostasis. The resultant distributions of synaptic sizes may reflect these processes.
We suggest that theshapesof distributions have biological relevance. This is suggested both empirically, by the prevalence of skewed distributions [1,2] and functionally, as large synapses may have particular importance [3]. We find a low-dimensional descriptor for such shapes. We proceed to explore and contrast brain regions at various spatial scales, and across the lifespan, using our proposed descriptor.Methods
We studied a measure of PSD95, a key postsynaptic protein [4–6] in parasagittal sections of mouse brains [7]. PSD95 correlates with EPSP amplitudes [8–10], spine volumes and synaptic face areas [11]. In contrast to previous work [7,12] we chose a scalar measure per synapse and considered synaptic populations.
We analyzed multiple anatomical levels, in ages ranging from one week to 18 months. Per 100 μm tiles, we computed a profile of the synaptic size distribution comprised of the arithmetic mean, normalized width, robust skewness, robust kurtosis, and synaptic density. Then we applied clustering methods and built a bi-linear model to compactly model variability.Results
Fig. 1 shows three parts of the profile descriptor. All five components differ between brain areas, and also by age. The upper tails of the distributions vary from relatively heavy-tailed (HT) regions, also more skewed, to less heavy-tailed (LT) ones. This is amplified in older animals. Regions in the hindbrain and midbrain tend to the HT-type; forebrain regions, in particular the cortex, and the hippocampus, to the LT-type. Mean intensity and spatial density follow the opposite trend. We thus found that the profiles seem to principally trace the anterior-posterior neuroaxis; our bi-linear and clustering models concur. The structure also parallells gene expression data [13].Discussion
We propose to analyze local brain regions using a fingerprint, a “distronomical signature”, based solely on the collective properties of synaptic distributions. This correlates with known anatomy and gene expressions, but exhibits striking differences in local heterogeneity (Fig. 1), and a rather dramatic evolution over the lifespan. We argue that this reflects underlying processes central to brain function, and that it may serve as a novel tool to characterize regular and perhaps anomalous structure in the brain.
Figure 1. Fig. 1: Global distributional structure. False color representation of three statistical moments, in a three month old individual. Tile size 25 µm x 25 µm. The tiles are color coded by arithmetic mean (red), normalized width (green) and robust kurtosis (blue), clipped at the 5th and 95th percentiles. Anatomical regions can be readily identified.
Acknowledgements
The Swedish Research Council grant no. 2022-01079.
References
1. doi:10.1371/journal.pbio.0030068
2. doi:10.1038/nrn3687
3. doi:10.1016/j.celrep.2022.111383
4. doi:10.1038/24790
5. doi:10.1523/JNEUROSCI.4457-06.2007
6. doi:10.1113/jphysiol.2008.163469
7. doi:10.1126/science.aba3163
8. doi:10.1016/S0092-8674(02)00683-9
9. doi:10.1073/pnas.0608492103
10. doi:10.1016/j.celrep.2021.109972
11. doi:10.1038/s41598-020-70859-5
12. doi:10.1016/j.neuron.2018.07.007
13. doi:10.1126/sciadv.abb3446
