Professor Yoshiyuki Kubota:
Mapping the Microcircuitry of the Cortex.
For more than forty years, one question has shaped Professor Yoshiyuki Kubota’s research: how is the cortex actually wired? This is the story of that pursuit – and why array tomography became the tool that could finally begin to answer it.
Ask Professor Yoshiyuki Kubota why he does what he does, and the answer is disarmingly simple. “I would like to just find out the microcircuit of the cortex,” he says. Everything else – four decades of technique, instrumentation and painstaking reconstruction – follows from that single ambition.
Choosing the brain over the black box
Kubota came to neuroscience at a fork in the road. As a student he weighed two of biology’s great frontiers, the genome and the brain, and chose the brain. At the time it was very much the harder bet. Four decades ago the cortex was, in his words, “still like a black box.” Researchers knew the broad categories – pyramidal and non-pyramidal cells, glia, and the six cortical layers catalogued in the era of Santiago Ramón y Cajal a century earlier – but the wiring between them remained a mystery. “That’s all we knew,” he recalls. “And then it just seemed like still chaos.”
Bringing order to that chaos would become his life’s work.
From single synapses to whole circuits
Kubota began where the tools of the day allowed: with immunohistochemistry, labelling the cortex’s many cell types one molecular marker at a time. Working alongside his mentor Edward “Ted” Jones at the RIKEN Institute, he set about classifying the cortex’s non-pyramidal cells, eventually resolving them into three major subgroups. It was foundational work – but it was also, he admits, only the beginning of what he really wanted to know. “I really want to know the connection,” he says.
Connections, though, cannot be read from a single image. A lone electron microscopy section reveals a synapse, but not where its partners lead. To follow an axon to its target, Kubota needed to see through the tissue in three dimensions – which meant imaging many serial sections and reconstructing them into a single volume. By the late 1990s he had made the leap into 3D electron microscopy, and into the grind that came with it.
That early work was relentlessly manual. Sections were collected twenty at a time onto slotted grids, imaged by transmission electron microscope, and traced by hand with a contour-drawing tool. To follow a single synapse, Kubota searched for it section after section after section, matching the same structure across each one by eye. “Everything I had to do manually,” he says. It was, by any measure, a labour of conviction.
Figure provided by Prof Yoshiyuki Kubota
The right tool for the question
As larger-scale volume EM became possible, Kubota did what any rigorous scientist does: he evaluated everything. FIB-SEM offered superb Z-resolution, ideal for resolving individual synapses and vesicles, but it was not built for the sheer volume his connectomics questions demanded. Serial block-face SEM promised automation, yet at the time forced a trade-off between resolution and section thickness, and struggled with charging artefacts at the settings he needed.
His question – how is a whole microcircuit wired? – called for large volumes above all else. So when a research grant let him pursue his ambition of imaging 1,000 serial sections at 200 x 200 micrometres and 5 nm per pixel, he chose array tomography, pairing an ATUMtome with a scanning electron microscope. It was the technique that matched the scale of the problem. The dataset that had once seemed out of reach was captured in around three months.
Improving the tools, not just using them
Kubota has never been a passive user of his instruments. Along the way he helped develop the carbon nanotube-coated tape now used for serial-section collection with the ATUMtome – work he published in Nature Communications – and contributed to the software that automates array tomography image capture. It is a telling pattern: when the tools did not quite fit the science, he helped reshape them.
Where the work is heading
Array tomography has since become a cornerstone of Kubota’s research. Today his group applies related serial-section approaches to two questions at the frontier of the field. One is the marmoset connectome – mapping the microcircuitry of a primate cortex. The other reaches into how circuits change as we learn: using a mouse trained, over just a few days, to reach through a narrow slot and grasp a seed, his team watched the primary motor cortex physically remodel itself, with new synaptic connections forming, competing and being pruned as the animal mastered the task.
Running through all of it is the same thread that started him down this path – and a growing sense that the destination is finally within reach. As automated, dense segmentation tools mature, the reconstruction work that once consumed years is becoming faster and more accessible. “I like it now,” Kubota says, “because my dream almost comes true.”
For a field built on seeing the unseeable, that is what progress sounds like: a decades-long question, and the tools at last catching up to it.
Relevant publications:
- Kubuta et al (2025 A wide variety of techniques for a volume electron microscopy. Microscopy, 74 (3): 215–222, https://doi.org/10.1093/jmicro/dfaf011
- Sohn et al (2022) Presynaptic supervision of cortical spine dynamics in motor learning. Sci Adv. 8(30):eabm0531. doi: 10.1126/sciadv.abm0531
- Urakubo et al (2019) UNI-EM: An Environment for Deep Neural Network-Based Automated Segmentation of Neuronal Electron Microscopic Images. Sci Rep. 9:19413. doi: 10.1038/s41598-019-55431-0
