Molecular organization of CA1 interneuron classes

Kenneth D. Harris, Lorenza Magno, Linda Katona, Peter Lönnerberg, Ana B. Muñoz Manchado, Péter Somogyi, Nicoletta Kessaris, Sten Linnarsson, Jens Hjerling‐LefflerbioRxiv 2015 · DOI

TL;DR

This paper leverages single-cell RNA sequencing (scRNA-seq) to create a detailed, 5-level classification of CA1 hippocampal interneurons based on gene expression patterns. This approach not only validates existing interneuron types but also identifies a novel population, the "R2C2 cells," and predicts new molecular markers for various classes, advancing our understanding of neural circuit complexity.

What Problem Does This Paper Address?

The hippocampus, a crucial brain region for memory and learning, relies heavily on a diverse array of GABAergic interneurons to regulate its activity. However, classifying these interneurons has been a long-standing challenge, with existing methods often yielding controversial or incomplete results. While traditional approaches have identified around 21 distinct types based on morphology, electrophysiology, and neurochemistry, it's widely believed that this diversity is even greater, with many "dark matter" classes lacking clear molecular identifiers. The complexity arises because interneuron identity isn't determined by a single gene but by intricate combinatorial patterns of gene expression, making it difficult to capture with traditional histology. This paper aims to resolve this by using scRNA-seq to build a comprehensive molecular map of CA1 interneurons, enabling a more precise and data-driven classification.

Methods

The core of this study is the application of a novel algorithm called the "GeneTeam algorithm" for analyzing scRNA-seq data. This algorithm employs a divisive hierarchical clustering approach to split cells into branches based on mutually exclusive gene expression patterns. Essentially, it identifies "teams" of genes where cells either strongly express genes from one team or the other, but not both. This method is designed to be robust to variations in absolute expression levels and to identify discrete classes rather than a continuum of properties. The algorithm was applied recursively to build a 5-level classification hierarchy. To validate their findings, the researchers used in situ hybridization and immunohistochemistry to confirm the predicted expression patterns and locations of novel markers and cell classes.

Datasets

The primary dataset used in this study consists of scRNA-seq data from 126 dissociated mouse CA1 interneurons (ages P21-P31, both sexes). This data was previously described and made available by Zeisel et al. (2015). The researchers also utilized publicly available in situ hybridization data from the Allen Mouse Brain Atlas for validation.

Key Insights and Ideas

  • Combinatorial Gene Expression is Key: The study strongly reinforces the idea that interneuron identity is defined by complex combinations of gene expression, not single markers. This makes scRNA-seq a powerful tool for dissecting this complexity.
  • A New Interneuron Class: The identification of the "R2C2 cells" is a significant finding, highlighting that even well-studied brain regions like CA1 still harbor novel neuronal populations. Their unique gene expression signature (Rgs12, Reln, Cxcl14, Cpne5) and location at the stratum radiatum/lacunosum-moleculare border are key characteristics.
  • Hierarchical Classification: The 5-level hierarchy provides a structured way to understand interneuron diversity, moving from broad categories to finer distinctions, mirroring the complexity of neural circuits.
  • Predictive Power of scRNA-seq: The study demonstrates how scRNA-seq can predict novel molecular markers for known and new cell types, which can then be experimentally verified, accelerating the discovery process.
  • MGE vs. CGE Origins: The initial classification splits align with the known developmental origins of interneurons (medial and caudal ganglionic eminences), providing a biological grounding for the molecular distinctions.

Key Results

  • The GeneTeam algorithm successfully classified CA1 interneurons into a 5-level hierarchy comprising 14 distinct classes.
  • Most of these classes showed strong correspondence with previously identified interneuron types, validating the algorithm's effectiveness.
  • A novel interneuron population, termed "R2C2 cells," was identified, characterized by the combinatorial expression of Rgs12, Reln, Cxcl14, and Cpne5, and located at the border of strata radiatum and lacunosum-moleculare.
  • The study predicted and experimentally verified novel molecular markers for various interneuron classes, including Cxcl14 for R2C2 cells and Npy for certain Cck-positive interneurons.
  • In situ hybridization confirmed the predicted laminar distribution of several novel markers, such as Cxcl14 at the R-LM border.

Limitations and Future Work

  • Dissociation Bias: The authors acknowledge that scRNA-seq data can be biased by differential survival rates of cell types during tissue dissociation, potentially under-representing certain interneuron populations (e.g., Pvalb-positive cells).
  • Contamination: Minor contamination by RNAs from other cell types can occur during dissociation, which the researchers attempted to mitigate by removing known glial and pyramidal cell markers.
  • Resolution of Subclasses: While the algorithm identified a 5-level hierarchy, some branches at deeper levels contained too few cells to reach statistical significance, suggesting that further subdivisions might exist.
  • Continuum vs. Discrete Classes: While the study favors discrete classes defined by mutually exclusive gene sets, it acknowledges that expression levels within these classes can vary dynamically and may represent a continuum.
  • Future Work: The authors suggest that their classification scheme provides a rich set of predictions for novel molecular markers that can be tested in future systematic studies. They also propose that similar levels of complexity likely exist in other cortical regions and throughout the brain, warranting further investigation. The identified R2C2 cells, in particular, represent a new avenue for exploring the functional roles of interneurons in specific hippocampal layers.

Summary updated

Abstract

Abstract GABAergic interneurons are key regulators of hippocampal circuits, but our understanding of the diversity and classification of these cells remains controversial. Here we analyze the organization of interneurons in the CA1 area, using the combinatorial patterns of gene expression revealed by single-cell mRNA sequencing (scRNA-seq). This analysis reveals a 5-level hierarchy of cell classes. Most of the predicted classes correspond closely to known interneuron types, allowing us to predict a large number of novel molecular markers of these classes. In addition we identified a major new interneuron population localized at the border of strata radiatum and lacunosum-moleculare that we term “R2C2 cells” after their characteristic combinatorial expression of Rgs12, Reln, Cxcl14 , and Cpne5 . Several predictions of this classification scheme were verified using in situ hybridization and immunohistochemistry, providing further confidence in the gene expression patterns and novel classes predicted by the single cell data.
BibTeX
@misc{harris2015molecular,
  author = {Kenneth D. Harris and Lorenza Magno and Linda Katona and Peter Lönnerberg and Ana B. Muñoz Manchado and Péter Somogyi and Nicoletta Kessaris and Sten Linnarsson and Jens Hjerling‐Leffler},
  title = {{Molecular organization of CA1 interneuron classes}},
  year = {2015},
  doi = {10.1101/034595},
  howpublished = {bioRxiv},
}
Figure