We have invented a novel lathe ultramicrotome to automatically slice a large volume of fixed brain tissue and mount it on a continuous strip of very thin tape. This tape is then imaged in a scanning electron microscope.

The problem

In order to understand the brain’s function neuroscientists must be able to map out its basic neuronal circuits. A neuronal circuit (for example, a thalamocortical circuit) typically spans quite a large volume of brain tissue (tens of cubic millimeters). At the same time, the axonal and dendritic processes comprising such a circuit are so fine and so tortuously interconnected that only electron microscopy of ultrathin (50nm) serial sections can resolve their connectivity. No current techniques can image so large a volume of tissue at such fine resolution.

Automating serial sectioning for large volume nanoscale imaging

We are developing a new type of microtome, an Automatic Tape-Collecting Lathe Ultramicrotome (ATLUM), which could potentially allow efficient nanoscale imaging over such large tissue volumes.  The ATLUM automatically sections an embedded block of brain tissue into thousands of ultrathin sections and collects these on a long carbon-coated tape for later staining and imaging in a scanning electron microscope (SEM). Because the process is fully automated, volumes as large as tens of cubic millimeters, large enough to span entire multi-region neuronal circuits, can be quickly and reliably reduced to a tape of ultrathin sections. Already our prototype ATLUM (see figures) has collected, in hands free operation, over 100 sections each 50nm thick and 1mm x 5mm in area of embedded mouse cortex. SEM images of these ATLUM collected sections can attain lateral resolutions of 5nm or better, sufficient to image individual synaptic vesicles (see figures) and to identify and trace all circuit connectivity.

Ultrathin Section Libraries

Following collection, the ATLUM tape can be stained with heavy metals (or any other markers), and cut into shorter lengths. This allows row after row of sections to be attached to large (200mm in diameter) imaging plates which can be loaded into a standard SEM for automated random access imaging of any location within any of the hundreds of sections on the plate’s surface. A set of a few dozen of these plates could hold an entire 10mm3 volume representing an incredible 8x1015 voxels of raw image data.

Bulk imaging of such a large volume at the highest resolution would take hundreds of years, but having the ultrathin sections laid bare on a set of tissue plates solves this imaging time problem. A researcher can quickly produce a lower-resolution image set of the entire volume, setting up a unified coordinate system for the sample volume and plates, and then use robotic loading and positioning of plates to zoom in on any part of the volume to obtain the highest resolution SEM images. We call such a set of tissue plates queued for robotic random access SEM imaging, an Ultrathin Section Library (UTSL).

Tracing multi-scale, multi-region neuronal circuits

Using an UTSL a researcher could, for instance, efficiently trace the dendritic arbor of a target cell by robotically loading and positioning each successive section in turn to trace only that cell’s processes. The researcher could then select a subset of the axonal contacts synapsing on this dendritic arbor and follow these back to their source cells in a distant brain region, in essence mapping out the cell’s receptive field in that region. Such a reconstruction could easily span tens of cubic millimeters of tissue volume, but because high resolution imaging is performed only where needed to follow individual processes, the total volume needed to be imaged (and the time required) is reduced by more than a thousand fold. What’s more, other researchers could follow up such a study with additional circuit mapping (involving the very same neurons) using the same UTSL.

A confluence of techniques

Over the next few years we see an incredibly productive confluence of the following five technologies:

  • Automated high-throughput serial sectioning (e.g. ATLUM)
  • Automated random-access nanoscale imaging (e.g. UTSL)
  • Intelligent neuronal tracing algorithms
  • In vivo cellular resolution imaging of neuronal activity (e.g. two-photon microscopy of calcium-sensitive indicators)
  • Large scale simulations of neuronal circuits (e.g. Blue Brain Project)

By combining these techniques, the activity patterns within a brain region could be overlaid on a detailed map of the synaptic circuitry within that region as well as its inputs and output to other regions. Large scale simulations based on the actual connectivity of these traced circuits could then be used to test computational models of the region’s function and be directly compared to the recorded activity patterns.

Future directions

Today only tiny volumes of neuropil can be traced at EM resolution using painstaking manual techniques. Mapping the neuronal network of the nematode worm C. elegans was a decade long Herculean task even though it is less than 0.01 cubic mm in volume. We anticipate that continuing advances in automating the serial sectioning, imaging, and tracing process will permit an exponential rise in the volume of brain tissue able to be mapped. Single cortical columns, small multi-region pathways, and entire insect brains are within reach of instruments like the ATLUM. Further advances in automation and tissue preparation will continue to push to ever larger volumes until… Human Connectome Project anyone?


This research is funded by a grant to Jeff Lichtman and Kenneth Hayworth from the McKnight Endowment Fund for Neuroscience and continuing support from the Center for Brain Science, Harvard University.