Binary and analog variation of synapses between cortical pyramidal neurons

Sven Dorkenwald, Nicholas L. Turner, Thomas Macrina, Kisuk Lee, Ran Lu, Jingpeng Wu, Ágnes L. Bodor, Adam Bleckert, Derrick Brittain, Nico Kemnitz, William Silversmith, Dodam Ih, Jonathan Zung, Aleksandar Zlateski, Ignacio Tartavull, Szi-chieh Yu, Sergiy Popovych, William Wong, Manuel Castro, Chris S. Jordan, Alyssa M. Wilson, Emmanouil Froudarakis, JoAnn Buchanan, Marc Takeno, Russel Torres, Gayathri Mahalingam, Forrest Collman, Casey M Schneider-Mizell, Daniel J. Bumbarger, Yang Li, Lynne A. Becker, Shelby Suckow, Jacob Reimer, Andreas S. Tolias, Nuno Maçarico da Costa, R. Clay Reid, H. Sebastian SeungeLife 2022 · DOI

TL;DR

This paper presents the largest-ever map of synaptic connections between a specific type of neuron (layer 2/3 pyramidal cells) in the mouse visual cortex, reconstructed from electron microscopy. They discovered that synapse sizes aren't a continuous spectrum as previously thought, but rather a combination of a "binary" (on/off) state and a continuous analog variation, with the binary component showing strong correlations between co-innervating synapses.

What Problem Does This Paper Address?

Understanding how the brain learns and adapts hinges on understanding how neural connections, or synapses, change. For decades, the prevailing model for synapse strength was a continuous, analog variable, much like a dimmer switch. This view aligns with many artificial neural network learning algorithms. However, some earlier work hinted at discrete "small" and "large" synapse categories. This paper aims to resolve this dichotomy by creating a high-resolution, large-scale map of synaptic connections and analyzing synapse size distributions to see if they are truly continuous or if there's a discrete component.

Methods

The core of this work is the reconstruction of a large volume of mouse visual cortex (250 x 140 x 90 µm³) using serial section electron microscopy (ssEM). This involved significant advancements in handling image defects inherent in ssEM data, using a combination of automated deep learning and human-in-the-loop correction. Once the neuronal structures were traced and segmented, the researchers meticulously identified and quantified synapses between layer 2/3 pyramidal cells. They then employed statistical modeling, specifically mixture models, to analyze the distribution of synapse sizes. A key part of their analysis involved examining "dual connections" – pairs of synapses between the same presynaptic and postsynaptic neurons – to investigate correlations in synapse size.

Datasets

The primary dataset is a newly reconstructed 3D volume of mouse primary visual cortex (250 × 140 × 90 μm³). This dataset is publicly available on the MICrONS explorer. The paper also references the use of the SNEMI3D challenge dataset for preliminary experiments on data augmentation techniques.

Key Insights and Ideas

  • Synapses aren't just analog: The most striking finding is that synapse size isn't a simple continuum. Instead, it's better described as a sum of a binary state (like an on/off switch) and a continuous analog variation. This challenges the long-held view and has implications for how synaptic plasticity might work.
  • Binary states are correlated: When two synapses connect the same pair of neurons (dual connections), their "binary" states are highly correlated. This suggests a shared underlying mechanism that influences whether a synapse is in its "small" or "large" state.
  • Analog variation is independent: In contrast, the analog component of synapse size shows little to no correlation between these dual connections. This suggests that analog variations might be driven by more localized or stochastic processes.
  • Implications for learning: The binary nature of synapses could be the result of Hebbian or other activity-dependent plasticity rules that operate on more global signals. The independent analog variation might reflect more local influences or spontaneous fluctuations. This has implications for how learning algorithms might be implemented in the brain.
  • Bimodality is specific: The observed bimodality (two distinct peaks in synapse size distribution) is specific to the carefully selected population of L2/3 PyC to L2/3 PyC connections. When broader populations of synapses are considered, this bimodality disappears, highlighting the importance of defining specific neuronal populations for analysis.

Key Results

  • Reconstruction of a 250 × 140 × 90 μm³ volume of mouse primary visual cortex, enabling the largest map of connectivity between L2/3 pyramidal cells to date.
  • Synapse size distribution for L2/3 PyC-to-L2/3 PyC synapses is best modeled by a mixture of two log-normal distributions, indicating a binary component.
  • In dual connections between L2/3 PyCs, the binary components of the two synapses are highly correlated (Pearson's phi = 0.637), while the analog components are not.
  • Bimodality and strong correlations are observed specifically in dual connections between L2/3 PyCs, but not in broader ensembles of synapses.

Limitations and Future Work

  • Underlying mechanism of binary states: The paper acknowledges that it's unclear whether the binary latent variable reflects a true underlying bistable mechanism or is simply a statistical description. Further experimental and theoretical work is needed to distinguish between these possibilities.
  • Scenario for binary switches: While the authors lean towards a "binary switch" scenario for synapse plasticity, they admit it's tentative and more research is required.
  • Specificity of findings: The observed bimodality and correlations are specific to the L2/3 PyC to L2/3 PyC connections. The authors suggest that this heterogeneity might exist across different classes of neocortical synapses, and further investigation into other neuronal types and connections is warranted.
  • Broader implications: The findings have implications for understanding cortical computation and the role of infrequent strong connections, but more experimental and theoretical studies are needed to draw firmer conclusions.

Summary updated

Abstract

Learning from experience depends at least in part on changes in neuronal connections. We present the largest map of connectivity to date between cortical neurons of a defined type (layer 2/3 [L2/3] pyramidal cells in mouse primary visual cortex), which was enabled by automated analysis of serial section electron microscopy images with improved handling of image defects (250 × 140 × 90 μm<sup>3</sup> volume). We used the map to identify constraints on the learning algorithms employed by the cortex. Previous cortical studies modeled a continuum of synapse sizes by a log-normal distribution. A continuum is consistent with most neural network models of learning, in which synaptic strength is a continuously graded analog variable. Here, we show that synapse size, when restricted to synapses between L2/3 pyramidal cells, is well modeled by the sum of a binary variable and an analog variable drawn from a log-normal distribution. Two synapses sharing the same presynaptic and postsynaptic cells are known to be correlated in size. We show that the binary variables of the two synapses are highly correlated, while the analog variables are not. Binary variation could be the outcome of a Hebbian or other synaptic plasticity rule depending on activity signals that are relatively uniform across neuronal arbors, while analog variation may be dominated by other influences such as spontaneous dynamical fluctuations. We discuss the implications for the longstanding hypothesis that activity-dependent plasticity switches synapses between bistable states.
BibTeX
@inproceedings{dorkenwald2022binary,
  author = {Sven Dorkenwald and Nicholas L. Turner and Thomas Macrina and Kisuk Lee and Ran Lu and Jingpeng Wu and Ágnes L. Bodor and Adam Bleckert and Derrick Brittain and Nico Kemnitz and William Silversmith and Dodam Ih and Jonathan Zung and Aleksandar Zlateski and Ignacio Tartavull and Szi-chieh Yu and Sergiy Popovych and William Wong and Manuel Castro and Chris S. Jordan and Alyssa M. Wilson and Emmanouil Froudarakis and JoAnn Buchanan and Marc Takeno and Russel Torres and Gayathri Mahalingam and Forrest Collman and Casey M Schneider-Mizell and Daniel J. Bumbarger and Yang Li and Lynne A. Becker and Shelby Suckow and Jacob Reimer and Andreas S. Tolias and Nuno Maçarico da Costa and R. Clay Reid and H. Sebastian Seung},
  title = {{Binary and analog variation of synapses between cortical pyramidal neurons}},
  booktitle = {eLife},
  year = {2022},
  doi = {10.7554/elife.76120},
}
Figure