The number of cells in the human brain is a staggering figure. There are over a trillion cells if you count both neurons (the better-known cells that compute with electrical signals) and glia (cells that provide support and protection for neurons). But the complexity of the brain only becomes more mind-boggling once you recognize that these cells have remarkably different appearances, genetic signatures, and electrical functions. This fact was first appreciated over a century ago by Santiago Ramón y Cajal who painstakingly sketched the appearance of neurons in the brain and spinal cord using a staining technique developed by Camillo Golgi. Even just a small sample of his drawings is enough to illustrate breathtaking complexity and diversity of neurons. (We don't have enough space to even get started on glia.)
Even though a hundred years have elapsed, and scientists have developed a plethora of tools and techniques for studying the brain, a complete and detailed catalogue of brain cell types remains elusive. In 2013, President Obama's BRAIN research initiative designated enumerating the cell types of the brain as one of seven high priority areas for funding. Since then, research papers have hit the presses at a breakneck pace on this topic. This renewed interest in cell types rapidly became an international and multidisciplinary effort, involving research groups from Seattle to Switzerland, and drawing talent from the genomics, neuroscience, and machine learning communities.
Many of these efforts culminated in a recent paper spearheaded by Xiaolong Jiang and Andreas Tolias. The authors aimed to not only obtain a census of the neuron types in a brain region, but also characterize how they are woven together to form a complete circuit. In particular, they recorded the electrical activity of neurons in groups of eight using a difficult but highly precise technique called patch clamp. Even more impressively, the team was able to gather data from over 2,000 neurons using this procedure and test for over 11,000 possible connections between cells. Importantly, they able to identify the cell type after each recording by staining the neuron with a fluorescent dye and reconstructing its shape -- not unlike Cajal's essential approach. (The shapes of different cell types can be visualized in figure 1 from the paper, shown here below)
Through this impressive experimental tour de force, the team reached two major findings. First, they identified 17 fundamental cell types within the cortical region they targeted. Many of these cell types have been previously identified, but this study comprehensively characterized each of them in remarkable detail. Second, and much more importantly, they were able to estimate how the cell types wired together. The eight cell patch clamp protocol was essential to reach this second result -- it allowed them to deliver a precise electrical impulse to each neuron and simultaneously record which of the seven remaining neurons showed a response. In total, 15 of the 17 cell types were inhibitory neurons (which decrease the activity of other neurons) and 2 were excitatory (which increase activity). This is in line with previous work showing that inhibitory cells exhibit greater variety.
The researchers found that certain cell types formed promiscuous connections with all other cell types, while others only formed very specific connections with other cell types. For example, two cell types -- Martinotti cells and neurogliaform cells -- diffusely targeted all other cell types, whereas others like single bouquet cells selectively targeted other inhibitory cells and did not form connections to excitatory cells. The figure below shows a schematic of the full circuit that the authors discovered:
Overall, these results highlight how crucial it is to characterize and examine brain circuits in terms of cell types. Neurons do not connect to each other at random; they connect in a highly structured fashion, which suggests that brain circuits can only be understood by decomposing the circuit into its fundamental building blocks -- that is, cell types.
The experiments described in this post were performed in mouse visual cortex -- a region in the back of the brain that supports the sense of sight, and which has been extensively characterized by decades of research. The authors hope that similar neural circuits will be uncovered in other regions of the cerebral cortex, which would bolster the long-held, but contentious, view that each sub-region in the cerebral cortex contains the same generic circuit components and short-range connectivity patterns. Thus, understanding the circuit computations in one area in a mouse brain may provide crucial insight into how much of the human brain operates.
But the brain may turn out to possess even more complicated and varied architecture. To test this, the heroic circuit mapping done in this study needs to be applied in other species and in other brain areas.