Exploiting DNA Sequencing Technology for High-Throughput Neuroanatomy

Justus M Kebschull

The brain is the most complex organ of the body, formed by billions of neurons and trillions of synapses, all precisely connected by 100,000 miles of wiring. Understanding how the brain processes information relies, at least in part, on understanding these connections. However, in mammals, we still lack a fine-resolution map of neural connectivity. Inferences about structure-function relationships are therefore limited.

Current fluorescence or electron-microscope based methods for mapping neural connectivity have provided neuroscientists with unprecedented information on meso- and microscale connectivity1–3; however, microscopy-based brain mapping is labor intensive, involves trade-offs between resolution and scale, and can be prohibitively expensive. Over the last years, my colleagues and I have worked to overcome these bottlenecks by leveraging the high parallelization and cost effectiveness of modern DNA sequencing for brain mapping.

Our approach relies on the unequivocal identification of large numbers of neurons via labeling with unique nucleic acid sequences (“barcodes”) followed by barcode localization based connectivity mapping. The strength of this approach arises from the almost limitless diversity of unique labels possible using barcodes. A barcode formed by a string of only 30 nucleotides, for example, can take any of 1018 different sequences – billions of times greater than neurons in the mouse brain – each of which can be reliably, quickly, and inexpensively read out by high-throughput sequencing.

We first developed MAPseq, a technique that uses cellular barcoding and DNA sequencing to map mesoscale connectivity in the mouse brain at single cell resolution and thousands of cells at a time in a single one-week experiment – orders of magnitude faster than previously possible. In MAPseq, we virally deliver barcodes to neurons, where they are expressed and trafficked into axonal processes by a modified pre-synaptic protein. Potential target areas can then be dissected and sequenced in bulk to determine which barcodes are present in which brain areas. The task of projection mapping thus translates to that of matching a barcode in the soma to where the barcode is found in processes throughout the brain, avoiding all tracing. We first used MAPseq to map the projectome of locus coeruleus (LC)4, the major source of noradrenaline in the brain that is thought to dictate overall brain arousal. Here, we mapped hundreds of neurons in single mice to demonstrate that, contrary to dogma, LC does not broadcast axons throughout the brain. Instead, single neurons have distinct and preferred wiring patterns, indicating that noradrenaline may act as a pathway specific modulator of arousal rather than a global alarm signal.

Figure 1: Schematic of MAPseq.

Figure 1: Schematic of MAPseq. A high-diversity viral barcode library is injected into the brain region of interest. Individual viral particles infect, and thus uniquely label, neurons at the injection site. In every infected cell, the barcodes are expressed as barcode mRNAs. These mRNAs are then specifically trafficked by the engineered presynaptic protein MAPP-nl into axonal processes, where they can subsequently be detected by sequencing. Image adapted from ref 4.

We have now extended MAPseq for use in other systems including projectome mapping of hundreds of neurons in primary visual cortex (V1). In V1, we showed that we can use MAPseq to reveal the existence of specific anatomical motifs for intra-cortical information transfer5. We found that the majority of V1 neurons project to more than one target area, and that these neurons preferentially innervate specific combinations of areas while explicitly avoiding others. Previous bulk tracing efforts lacked the resolution to uncover these characteristic motifs, and our results rule out the often-assumed “one-neuron—one-target-area” model of information flow in visual cortex.

MAPseq permits the mass interrogation of single neuron projectomes, but is limited in its spatial resolution to the size of dissected areas. While without consequences in many applications, some studies require the very high spatial resolution provided by microscopy. To combine the throughput of MAPseq with the spatial resolution of imaging, we recently developed an in situ sequencing method to read out barcodes directly from brain slices, thus avoiding all dissection. We applied the resulting method, BARseq, to mapping the projections of thousands of neurons in the auditory cortex, while recording the location of all barcoded somata in 3D6. BARseq lays the foundation for a multimodal understanding of the properties of single neurons, by allowing to register barcode-based connectivity information with spatially resolved measurements of neural activity and gene expression.

In addition to MAPseq, we now are working on techniques that exploit barcodes for mapping the synaptic connectivity of brain circuits. In a feasibility study, we developed SYNseq, a method that relies on the joining of barcodes across the synapses of connected neurons and high-throughput sequencing of these joined barcodes to read out microscale connectivity7, albeit at low efficiency. Improvements to SYNseq will hopefully allow biologically relevant applications of this technique in the future.

Barcode based brain mapping approaches open the door to high-throughput neuroanatomy, and are already providing novel insights into the fine circuit structure of mammalian brains.


  1. Oh, S. W. et al. A mesoscale connectome of the mouse brain. Nature 508, 207–214 (2014). doi: 10.1038/nature13186.
  2. Zingg, B. et al. Neural networks of the mouse neocortex. Cell 156, 1096–1111 (2014). doi: 10.1016/j.cell.2014.02.023.
  3. Helmstaedter, M. et al. Connectomic reconstruction of the inner plexiform layer in the mouse retina. Nature 500, 168–174 (2013). doi: 10.1038/nature12346.
  4. Kebschull, J. M. et al. High-Throughput Mapping of Single-Neuron Projections by Sequencing of Barcoded RNA. Neuron 91, 975–987 (2016). doi: 10.1016/j.neuron.2016.07.036.
  5. Han, Y. et al. The logic of single-cell projections from visual cortex. Nature 556, 51–56 (2018). doi: 10.1038/nature26159.
  6. Chen, X. et al. Spatial organization of projection neurons in the mouse auditory cortex identified by in situ barcode sequencing. bioRxiv 294637 (2018). doi:10.1101/294637
  7. Peikon, I. D. et al. Using high-throughput barcode sequencing to efficiently map connectomes. Nucleic Acids Res. 45, aaf7907 (2017). doi: 10.1093/nar/gkx292.

About the Author

Justus M KebschullJustus M Kebschull: My main interest is in understanding the evolution of neural circuits. I completed my undergraduate degree in Systems Biology from the University of Cambridge in 2011, before joining the laboratory of Anthony Zador at Cold Spring Harbor Laboratory for my PhD research. All work described in this article was performed there. After graduating in 2017, I moved to Stanford University, where I am working as a postdoctoral fellow in Liqun Luo’s lab.