How Brain Mapping Helps Scientists Study the Brain
Christopher Dunne
There is widespread variance in peoples’ natural abilities and inclinations toward different subject areas from mathematics to the arts. Neuroscientists have often been interested in what differences in the brain lead to such differences in skills and have sometimes looked to outliers in a particular domain to try to elucidate what made those individuals stand out. However, this approach has so far failed to produce concrete links between differences in brain structure and intellectual ability. Take Albert Einstein’s brain for instance, which was photographed and preserved after his death in 1955. While several teams of scientists have noted certain peculiarities in Einstein’s brain that give some hint to his abilities, by and large, Einstein’s brain was similar to “typical” brains. What’s more, his brain was smaller than average by mass and there is an open question as to whether the noted differences were responsible for his genius, simply correlated with it, and/or a function of chance. While disheartening in some regards, the lack of determinative conclusions from these studies makes sense given that scientists acknowledge that any comprehensive understanding of brain function will require analyses deeper than those based simply on geometry and gross morphology.
This is not to say that morphology and geometry tells us nothing about brain function. In fact, many neuroscience discoveries were derived from patients who suffered discrete brain injuries and lost only specific functionality. Over the years, the process of observing behavioural changes following brain injuries has helped to map many brain functions (i.e., aspects of vision, sensation, and language) to specific regions. However, even when we know the overall function of a brain region, we usually do not know the specific roles and interplay of subgroups of neurons within those regions. To move forward on this question, it will be essential to look into the microscale connections between neurons in a given brain area.
To understand neuronal computations, we must start with how neurons generally receive and pass on information. Firstly, they receive neurotransmitters (i.e., dopamine or serotonin) from other neurons at sites of connectivity known as chemical synapses. The receiving neurons then react to these neurotransmitters, which alters their electrical potential. If the neuron’s electrical potential reaches a certain threshold after pooling across many synapses, an electrical spike (an action potential) is generated that propagates down the length of the neuron until it reaches a point of connectivity with another neuron (a new synapse). At that point, the electrical signal causes a new release of neurotransmitter, which interacts with a new neuron and the process repeats itself. Importantly, the generation of this electrical signal is probabilistic at the synapse level and generally relies on combining inputs from hundreds of neurons. In effect, this means that the same stimulus can produce a different output in a given neuron based on the other inputs the neuron receives, the current behavioural state, or other mediating factors. Despite these complexities, given that synapses are where neurons exchange and pass on information, scientists have been studying them for decades. However, it has only been in the past decade or so, with massive improvements in imaging techniques and computational bandwidth, that we have been able to start to image and map a wide range of connections in one organism.
The field of neuroscience that maps and investigates this neural connectivity is aptly called connectomics, and it can be pursued at a range of scales, from microscopic (synapse level) to macroscopic (region level). While macroscopic connectomics contributes to cutting edge research and findings on overall brain function, to understand the details of neuronal processing the microscopic scale is also required. However, microscale connectomics is complicated by the sheer numbers and scale involved. Consider for instance that synapse level connectomics requires imaging with an electron microscope rather than a regular light microscope, as the wavelength of light is too big to image neuronal components that are in the range of a few billionths of a metre. To further put the challenge in perspective, note that preparing and imaging a 1 mm3 section of a mouse brain (consisting of 100,000 neurons) took scientists at the Allen Institute for Brain Science almost a year and generated 50% more data than 30 years’ worth of satellite images from NASA’s Landsat missions (DeWeerdt 2019). Given that a human brain consists of around 86 billion neurons, this means that a similar human dataset would be orders of magnitude larger and require thousands of years to complete if the same technology was used. Nonetheless, advances in microscopy, artificial intelligence, machine learning, and software tools are making processing larger datasets faster and more automated, which will slowly allow more complicated datasets to be analysed.
For example, our research group at Cambridge, in collaboration with scientists at HHMI Janelia Research Campus, Oxford University, Princeton University, and the MRC Laboratory of Molecular Biology, is working on completing a full brain connectome of the fruit fly (Drosophila melanogaster), involving around around 200,000 neurons and 200 million synapses. Although surprising to some, fruit flies have been one of the most important organisms for neuroscience research (Kumar & Honasoge, n.d.). They display many complex behaviours such as learning to remember smells and sounds in order to attain a reward or fighting for territory in order to increase their chances of mating. Furthermore, due to decades of previous research, scientists have a huge array of genetic tools to manipulate individual genes in flies, which means that nuances of neuron function can be specifically tested in ways not possible in other organisms. Importantly, fruit flies also share 60% of human genes, and as such, are an excellent research organism for translational studies.
Coming all the way back to the example of Albert Einstein’s brain from a connectomics approach, in the future we may be able to formulate hypotheses about what makes someone a genius by focusing on variable connections between and within brain regions rather than mainly looking at the size and shape of the regions. While many specifics are yet to be determined, connectomic variability is already observable in neurological conditions such as schizophrenia, autism, and depression (van den Heuvel & Sporns, 2019). Not all differences are pathological however, and a person’s connectome can change over time via learning and development and is also impacted by genetic and environmental factors. Relatedly, differences in macroscale connectomes have been correlated with differences in scores on various cognitive tests (Li et al., 2009) and recent studies have shown that a single fMRI can be used to uniquely identify a person as accurately as a fingerprint (Valizadeh et al., 2018). While fascinating, these findings were based on macroscale connectomes and generally describe patterns in data more than propose why such patterns exist; to truly understand the processes going on under the hood, we still need to further explore connectomes at the synapse level.
In this light, much of the work in our lab is investigating the microscale of neural circuitry, function, and design so that scientists can make informed hypotheses about individual circuits and their impacts on behaviour. We are also trying to catalogue the expected variation across connectomes within a species to understand how generally applicable a single connectome will be for all members of a species. This will be essential to understand as scientists begin to tackle larger scale connectomes such as the mouse brain, where each dataset will be orders of magnitude larger than the fly brain and cost many billions of dollars. Luckily, so far our research has shown a similar degree of variability between two brains compared to the variability across the two hemispheres in one brain (left hand side versus right hand side), which means a small number of datasets may suffice for many research questions.
As an example of observed variability, below is a comparison of two “singleton” neurons from different brains. (Singleton neurons are uniquely identifiable and are found in one pair per brain, one neuron on each side). The magenta neurons are a left-right pair from a whole brain dataset (FAFB) while the cyan neuron is the same neuron from a half brain dataset (Hemibrain) that is overlaid on the first. All three neurons look quite similar, and without colour coding, it would be difficult to tell which neuron on the left side of the image corresponded to the FAFB neuron on the right side of the image. In fact, in the top image it (subjectively) appears that the Hemibrain neuron matches up better with the FAFB neuron on the right than the two FAFB neurons match up to each other.
While the above example is purely qualitative, we also quantified differences across a set of olfactory neurons. To perform this analysis, we grouped neurons into “cell types”’ based on their morphology and function and then compared these types between brains. (Cell types are recognisable across brains and hemispheres and are homologous to genes in genomes). Across this set of neurons, we generally found similar numbers of cells in each type between all three available hemispheres. For instance, we found almost the same number of neurons in the broad cell type grouping between all three hemispheres (Figure 4A). There was more variability when comparing granular cell types (Figure 4B); however, the number of times FAFB left and right hemispheres differed from each other (10) was similar to the number of times each hemisphere differed from the hemibrain dataset (13 and 11 respectively). This trend also held when we compared morphology and connectivity between cell types, which together suggests that the differences between a single brain are similar to the differences between two separate brains. We are in the process of continuing this comparison across all neurons in the Hemibrain dataset to determine how much this finding can be generalised.
It is important to note that there is of course real and meaningful variation between any two brains. However, our findings build on decades of previous research suggesting that this variation occurs on top of a highly conserved set of circuits and motifs that would normally be similar between any two organisms of the same species. This would be analogous to the story of DNA, where individual differences make each human unique, and yet, the vast majority of our genome is identical from person to person. In fact, it is this very similarity that has allowed genetics to contribute so much to advancing science and healthcare.
As with genetics, ideally the inherent similarity in brains will allow connectomics to eventually create a scaffolding of neural connectivity that can then be used to better investigate a whole range of behaviours and disease states, which is already the case to some extent (van den Heuvel & Sporns, 2019). For instance, scientists can start with a known neural population of interest (i.e., olfactory neurons) and look at connectomic data to find connected neurons that have not previously been explored but are plausible candidates to impact that function. This ability to focus an analysis offers a huge advantage when dealing with the complexity present in the brain, which may one day allow for more tailored treatment plans for multifaceted diseases such as depression or Alzheimer’s. It is important to note however, that difficult and time-consuming experimental work will also need to be done to confirm findings from connectomics.
While there is much to be done and many hurdles to overcome, the last decade has seen massive progress in mapping the brain and increasing our understanding of neural circuitry underlying brain functions. The hope is that this progress will eventually allow scientists to develop a causal understanding of neural disorders, just like genomics has helped us discover the genes that underlie many diseases.
References
DeWeerdt, S. (2019, July). How to map the brain. https://www.nature.com/articles/d41586-019- 02208-0
Falk, D. (2009). New Information about Albert Einstein’s Brain. Frontiers in Evolutionary Neuroscience, 1, 3. https://doi.org/10.3389/neuro.18.003.2009
Kumar, N. H., & Honasoge, K. (n.d.). How the humble fruit fly changed science. Google Arts & Culture. https://artsandculture.google.com/story/how-the-humble-fruit-fly-changed- science- national-centre-for-biological-sciences/5wWhgVL6dcxYKA?hl=en
Li, Y., Liu, Y., Li, J., Qin, W., Li, K., Yu, C., & Jiang, T. (2009). Brain anatomical network and intelligence. PLoS Computational Biology, 5(5), e1000395.https://doi.org/10.1371/journal.- pcbi.1000395
Neurons: How the brain communicates. (n.d.). Mental Health America. https://mhanational.org/neurons-how-brain-communicates
Schlegel, P., Bates, A. S., Stürner, T., Jagannathan, S. R., Drummond, N., Hsu, J., Serratosa Capdevila, L., Javier, A., Marin, E. C., Barth-Maron, A., Tamimi, I. F., Li, F., Rubin, G. M., Plaza, S. M., Costa, M., & Jefferis, G. S. X. E. (2021). Information flow, cell types and stereotypy in a full olfactory connectome. eLife, 10. https://doi.org/10.7554/eLife.66018
Valizadeh, S. A., Liem, F., Mérillat, S., Hänggi, J., & Jäncke, L. (2018). Identification of individual subjects on the basis of their brain anatomical features. Scientific Reports, 8(1), 5611. https://doi.org/10.1038/s41598-018-23696-6