About five years ago, preeminent neuroscientist Eric Kandel of Columbia University was asked by a radio interviewer what mysteries remained about the brain. “Almost everything,” Kandel responded. Such a statement does not diminish the considerable progress neuroscience has made in the more than a century since Italian physician Camillo Golgi and Spanish anatomist Santiago Ramón y Cajal created the first drawings of neurons. But people like Kandel stay humble. They know that every time neuroscientists increase the resolution with which they peer into the brain, they discover new, unimaginable levels of complexity.

A new study by researchers at the Allen Institute for Brain Science and their colleagues, published on August 21 in Nature, did just that. It revealed what the Allen Institute calls “the most detailed ‘parts list’ of the human brain to date.” As a reminder of how much we still do not know, that list focused on just one part of the human brain. But by using gene expression—the RNA transcribed from DNA to make proteins—as a way of categorizing cells, the study’s authors defined 75 distinct cell types in that brain region, and they now think there may be on the order of a few thousand cell types across the whole brain, far more than previously believed. In addition, when the scientists compared cell types from humans and mice, they found that the majority of human brain cells have counterparts in mouse brains but that there are also remarkable differences. Those differences are likely to influence brain circuitry and may explain why the great majority of clinical trials for therapeutic drugs that were conducted in mice have not succeeded when moved into human patients, making it clear how essential it is to study human brains directly.

“What this is going to lead to is human cellular neuroscience,” says Christof Koch, chief scientist and president of the Allen Institute and a co-author of the paper. “We believe that many of the psychological and neurological diseases are diseases of specific cell types. It’s not good enough to know there’s something wrong in the amygdala. You need to know, of the 100 different cell types expressed in the amygdala, which one, specifically, is overexpressed or underexpressed, which synapse is not functional anymore.”

Koch and his colleagues’ study amounts to a look at the cutting edge of genomics and neuroscience, says Nenad Sestan, a neuroscientist and geneticist at the Yale School of Medicine, who was not involved in the work. “The database they have generated is not just an incredibly useful research tool but also a harbinger of the sort of data sets we will soon be able to develop and use to probe brain disorders.”

Because of the difficulty of studying living human brain tissue, most of what we know about human brains so far comes from either postmortem brain tissue or from brain– scanning techniques such as functional magnetic resonance imaging. In the new paper, the researchers investigated human brain cells from the middle temporal gyrus (MTG), an area associated with memory and the integration of information. It was chosen because some of the human brain tissue the Allen Institute uses comes, with permission, from surgeries performed on epilepsy patients. The rest comes from postmortem tissue from donors.

The Allen Institute’s team took a molecular approach to understanding cell diversity. It used a methodological advance, called single-cell transcriptomics, that measures the RNA content of a cell. “Instead of using the anatomy or the physiology as the way to think about cell types, one can think about a cell as defined by the genes that are active in that cell at some point in time,” says neuroscientist Ed Lein, an Allen Institute investigator and senior author of the study. Genes code for proteins that become the cell constituents that carry out those cells’ myriad tasks. By measuring the genes being actively used or the gene expression in individual cells, scientists can discern the function of each cell. In this study, the researchers profiled the parts list for only the nuclei of cells because they can be easily separated from one another, and their gene expression patterns can be individually profiled. In adult human brain tissue, whole cells are often large and interconnected, so they are frequently damaged when separated for profiling. The use of nuclei greatly expanded the number of cells they could analyze (16,000 in all). The combination of genetic technology and computational ability on display in this study is exciting, Sestan says. “We can see everything coming together.”

The researchers were thrilled by the results from their cross-species comparison. A human brain has roughly 1,000 times more cells than that of a mouse, but the basic components turn out to be the same. Lein describes seeing the classification of the two sets of cells line up so neatly as an “aha moment.” Still, critical differences were discovered in the relative proportion of cell types among the two species and in where the cells were found in the body, their gene-expression patterns and their structural properties. As an example, Lein points to pyramidal cells—the main “excitatory” neurons that release a signaling molecule to spur neurons to fire. “In the human cortex, they’re 20 times less abundant than in the mouse circuit,” he says. “The whole circuit has been remodeled by using the same Lego pieces but [then] assembling them in a different way.”

Ideally, the researchers would have compared the human MTG cells with a similar, or homologous, area in the mouse brain, but it is not clear that there is one. Instead they contrasted the human MTG cells with two very different areas of the mouse cortex: a motor area and a sensory (visual) area. “Comparing to both of them and showing strong homology to both of them, we’ve kind of covered the range of what one would see in the mouse cortex,” Lein says.

Such comparisons are important because we often use mice to model human diseases. “We justify that by noting that many anatomical and functional features are conserved between rodents and primates. And to the extent they can be observed, they are also similar in humans,” says neurobiologist Edward Callaway of the Salk Institute for Biological Studies, who did not work on the new paper. “Even if the same cell types exist, differences in gene expression between the homologous cell types could point to important functional differences.” One important implication is that nonhuman primates might be a much better model for certain human disorders than mice.

An example cited in the study is serotonin receptors. Both mice and humans have plenty of such receptors, but the researchers wrote that they found “highly divergent expression between species.” That means that drugs such as selective serotonin reuptake inhibitors, used to treat psychiatric disorders, are likely to function very differently in humans than in mice. Tests in the animals will not reveal much about how those drugs will work in us.

“We have the data to say whether or not the thing you’re trying to target is expressed in the same way or not,” Lein says. “In many cases, the gene will be used in a mouse and a human in the same brain region but not in the same cells. It completely changes function. Now I think you can predict why these things might not translate.”