pISSN 2287-5123·eISSN 2287-4445

https://doi.org/10.9729/AM.2016.46.4.176

Review Article

Light-Microscopy-Based Sparse Neural Circuit Reconstruction:

Array Tomography and Other Methods
Jong-Cheol Rah*
Korea Brain Research Institute, Laboratory of Neurophysiology, Daegu 41062, Korea

*Correspondence to: Efficient neural circuit reconstruction requires sufficient lateral and axial resolution to
Rah JC, resolve individual synapses and map a large enough volume of brain tissue to reveal the
Tel: +82-53-980-8350 molecular identity and origin of these synapses. Sparse circuit reconstruction using array
Fax: +82-53-980-8339 tomography meets many of these requirements but also has some limitations. In this
E-mail: [email protected]
minireview, the advantages and disadvantages of applicable imaging techniques will be
Received December 21, 2016 discussed.
Revised December 23, 2016
Accepted December 23, 2016 Key Words: Neural circuit reconstruction, Array tomography, Light microscopy

Mapping the neural circuits of the brain is a challenging task. microscopy techniques such as serial block-face scanning
Neurons communicate with each other through specialized microscopy (Briggman et al., 2011) and TEM camera array
structures called synapses, significant fraction of which (Bock et al., 2011) as well as multi-beam scanning electron
are close to or smaller than the light diffraction limit. The microscopy combined with the development of large volume
physiological properties of synapses can vary depending on tissue processing (Mikula et al., 2012) and automated
many factors, such as their size and molecular composition. section collection (Hayworth et al., 2006), have enabled
To make the problem even more difficult, synapses are the acquisition of large volumes (Lee et al., 2016; Morgan
packed in an extremely high density (approximately 1 per et al., 2016). The impact of these efforts is tremendous and
m3) in most of brain regions (DeFelipe et al., 1999). To reveals more information about neural networks than was
understand how circuits are wired, identifying the origins previously possible. However, the limited compatibility of
of the synapses, which can be over a millimeter away, is these techniques with labeling for molecular identity or
also necessary. Therefore, an ideal technique for efficient input sources, as well as experimental cost leaves room for
circuit reconstruction requires (1) sufficient lateral and axial improvement.
resolution to resolve individual synapses, (2) the ability to On the other hand, most conventional microscopy techniques,
reveal the molecular identity and source of these synapses, such as confocal microscopy, does not provide sufficient
and (3) coverage of a large enough volume of brain tissue resolving power for reliable synapse detection (Mishchenko,
to map an entire unit of the neural circuit of interest (e.g., 2010). Per theoretical predictions, much of this unreliability
entire cortical layers). In addition, the readiness of image is due to limited axial resolution. Improved axial resolution
segmentation technique is also critical for the acquisition dramatically improves synapse detection accuracy (Rah,
of statistical information. In this minireview, I discuss new 2013).
toolsets for neural circuit mapping. Array tomography (AT) is a high-resolution imaging techni­
For many years, transmission electron microscopy (TEM) que based on wide-field fluorescent imaging of arrays
has been the gold standard for circuit reconstruction because of ultrathin serial sections followed by computational
of its resolution although it lacks throughput capacity. reconstruction (Micheva & Smith, 2007; Fig. 1). Because
Recently developed automated three-dimensional electron the imaging is performed on ultrathin sections, the Z-axis

CCThis is an open-access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/4.0) which permits
unrestricted noncommercial use, distribution, and reproduction in any medium, provided the original work is properly cited.
Copyrights © 2016 by Korean Society of Microscopy

176 Appl Microsc 2016;46(4):176-178

LM Based Circuit Reconstruction

resolution of this technique is determined by the thickness of interferometric photoactivated localization microscopy
the section instead of by Abbe’s rule. Synapses can be resolved (iPALM), three-dimensional stochastic optical reconstruction
with approximately 80% accuracy by AT compared to TEM microscopy (3D-STORM), one can acquire better axial and
(Rah, 2013). Multiple axonal origins can be distinctively lateral resolution than one can possibly slice without the hassle
labeled using different fluorescent proteins, which articulate of laborious serial sectioning (Huang et al., 2008; Shtengel
the origins of the synapses. Since every voxel of the tissue has et al., 2009). A potential drawback of these techniques is
the same chance of antibody labeling, the molecular identity the severe limitations in depth and field-of-view. The use of
of the synapses can be examined in a quantitative manner. selective fluorophores and slow imaging speeds also require
Moreover, antibody staining on thin sections can be easily improvement before these techniques can be used for large-
removed and the sections can then be restained. Repeated volume circuit reconstruction. It is important to remember
cycles of antibody staining and stripping allows for detailed that imaging thick specimens with high axial resolution
investigation of the proteomic diversity of synapses of interest is accompanied by antibody penetration issues. Clever
(Micheva et al., 2010). Gathering the necessary information sample preparation followed by conventional microscopy
about a neural circuit from AT images is relatively simple provides a good alternative for this issue. GRASP or green
because of high contrast and isotropic resolution. In terms of fluorescent protein (GFP) Reconstitution Across Synaptic
imaging area, AT is applicable to a sufficiently large volume Partners utilizes overexpressed GFP-fusions of interacting
of brain tissue to cover entire cortical layers (Rah, 2013) or presynapse- and postsynapse-specific proteins (Feinberg
hippocampal CA1 pyramidal neurons (Bloss et al., 2016). et al., 2008; Kim et al., 2011). Although there are concerns
However, circuit reconstruction with AT is a rather fragile and about the synaptogenic feature of overexpressed synaptic
laborious process that involves handling 100-nm thick serial proteins (Graf et al., 2004; Scheiffele et al., 2000), the number
sections. On top of that, a synapse detection accuracy of 80% of synapses as well as the subcellular localization of synapses
may be enough for some biological questions, but certainly were in agreement with previous studies at least in gross level
not for all. In the following sections, I review the pros and (Kim et al., 2011). Expansion microscopy uses the physical
cons of other imaging techniques for studying neural circuits. expansion of a polymer network that is covalently anchored
Selective-plane illumination microscopy (SPIM) is now to a specific location within the specimen (Chen et al., 2015,
widely used for efficient optical sectioning because of its 2016). Considering the expansion factor of the specimen, an
high-contrast with reduced photo damage. The volume effective resolution of approximately 70 nm lateral and 200
of reconstruction can be limited by the working distance nm axial could be achieved even with conventional confocal
of the lens. For that reason, conventional SPIM is more microscopy. There is no obvious evidence that this technique
commonly used with brain clearing methods for area-to- affects the overall structure of neurons, which would suggest
area connectivity reconstruction where the observation of cell uneven expansion or distortion of the neuronal structure.
bodies and bundles is sufficient (Tomer et al., 2015, but see The methods described above and other techniques may be
Cella Zanacchi et al., 2011). perfectly adequate for testing some hypotheses but none of
Many super-resolution microscopy techniques now also them is a magic bullet. The combination of these technologies
provide improved axial resolution. For instance, with would lead to a powerful tool for large-volume neural circuit

1. Label inputs and neurons 2. Dissect and embed in resin

3. Serial section (on ultramicrotome)

Input label
Post-synaptic
neuron
Immunostaining
5. Image reconstruction
(collate and align images)
4. Image sections
(multi-tiles and sections)

Fig. 1. Schematic diagram of the array

tomography.

177
 Rah JC

reconstruction (Nanguneri et al., 2012; Punge et al., 2008; ACKNOWLEDGMENTS

Sigal et al., 2015).
This study was supported by the KBRI Program (2231-
CONFLICT OF INTEREST 415) and the Basic Science Research Program (NRF-
2014R1A1A2057804).
No potential conflict of interest relevant to this article was
reported.

REFERENCES

Bloss E B, Cembrowski M S, Karsh B, Colonell J, Fetter R D, and Spruston Micheva K D, Busse B, Weiler N C, O'Rourke N, and Smith S J (2010)
N (2016) Structured dendritic inhibition supports branch- selective Single-synapse analysis of a diverse synapse population: proteomic
integration in CA1 pyramidal cells. Neuron 89, 1016-1030. imaging methods and markers. Neuron 68, 639-653.
Bock D D, Lee W C A, Kerlin A M, Andermann M L, Hood G, Wetzel A W, Micheva K D and Smith S J (2007) Array tomography: a new tool for
Yurgenson S, Soucy E R, Kim H S, and Reid R C (2011) Network imaging the molecular architecture and ultrastructure of neural
anatomy and in vivo physiology of visual cortical neurons. Nature circuits. Neuron 55, 25-36.
471, 177-182.
Mikula S, Binding J, and Denk W (2012) Staining and embedding the
Briggman K L, Helmstaedter M, and Denk W (2011) Wiring specificity in whole mouse brain for electron microscopy. Nat. Meth. 9, 1198-
the direction-selectivity circuit of the retina. Nature 471, 183-188. 1201.
Cella Zanacchi F, Lavagnino Z, Perrone Donnorso M, Del Bue A, Furia Mishchenko Y (2010) On optical detection of densely labeled synapses in
L, Faretta M, and Diaspro A (2011) Live-cell 3D super-resolution neuropil and mapping connectivity with combinatorially multiplexed
imaging in thick biological samples. Nat. Meth. 8, 1047-1049. fluorescent synaptic markers. PLoS One 5, e8853.
Chen F, Tillberg P W, and Boyden E S (2015) Optical imaging. Expansion Morgan J L, Berger D R, Wetzel A W, and Lichtman J W (2016) The fuzzy
microscopy. Science 347, 543-548. logic of network connectivity in mouse visual thalamus. Cell 165,
192-206.
Chen F, Wassie A T, Cote A J, Sinha A, Alon S, Asano S, Daugharthy E R,
Chang J B, Marblestone A, Church G M, Raj A, and Boyden E S (2016). Nanguneri S, Flottmann B, Horstmann H, Heilemann M, and Kuner
Nanoscale imaging of RNA with expansion microscopy. Nature T (2012). Three-dimensional, tomographic super-resolution
Methods 13, 679-684. fluorescence imaging of serially sectioned thick samples. PLoS One 7,
e38098.
DeFelipe J, Marco P, Busturia I, and Merchán-Pérez A (1999) Estimation
of the number of synapses in the cerebral cortex: methodological Punge A, Rizzoli S O, Jahn R, Wildanger J D, Meyer L, Schönle A, Kastrup
considerations. Cerebral Cortex 9, 722-732. L, and Hell S W (2008) 3D reconstruction of high-resolution STED
microscope images. Microsc. Res. Tech. 71, 644-650.
Feinberg E H, VanHoven M K, Bendesky A, Wang G, Fetter R D, Shen
K, and Bargmann C I (2008) GFP reconstitution across synaptic Rah J C (2013) Thalamocortical input onto layer 5 pyramidal neurons
partners (GRASP) defines cell contacts and synapses in living measured using quantitative large-scale array tomography. Front.
nervous systems. Neuron 57, 353-363. Neural Circuits 7, 177.
Graf E R, Zhang X, Jin S X, Linhoff M W, and Craig A M (2004) Neurexins Scheiffele P, Fan J, Choih J, Fetter R, and Serafini T (2000) Neuroligin
induce differentiation of GABA and glutamate postsynaptic expressed in nonneuronal cells triggers presynaptic development in
specializations via neuroligins. Cell 119, 1013-1026. contacting axons. Cell 101, 657-669.
Hayworth K J, Kasthuri N, Schalek R, and Lichtman J W (2006) Shtengel G, Galbraith J A, Galbraith C G, Lippincott-Schwartz J, Gillette J M,
Automating the collection of ultrathin serial sections for large volume Manley S, Sougrat R, Waterman C M, Kanchanawong P, Davidson M
TEM reconstructions. Microsc. Microanal. 12, 86-87. W, Fetter R D, and Hess H F (2009) Interferometric fluorescent super-
resolution microscopy resolves 3D cellular ultrastructure. Proc. Natl.
Huang B, Wang W, Bates M, and Zhuang X (2008). Three-dimensional
Acad. Sci. U.S.A. 106, 3125-3130.
super-resolution imaging by stochastic optical reconstruction
microscopy. Science 319, 810-813. Sigal Y M, Speer C M, Babcock H P, and Zhuang X (2015) Mapping
synaptic input fields of neurons with super-resolution imaging. Cell
Kim J, Zhao T, Petralia R S, Yu Y, Peng H, Myers E, and Magee J C (2011)
163, 493-505.
mGRASP enables mapping mammalian synaptic connectivity with
light microscopy. Nat. Meth. 9, 96-102. Tomer R, Lovett-Barron M, Kauvar I, Andalman A, Burns V M, Sankaran S,
Grosenick L, Broxton M, Yang S, and Deisseroth K (2015) SPED light
Lee W C A, Bonin V, Reed M, Graham B J, Hood G, Glattfelder K, and Reid
sheet microscopy: fast mapping of biological system structure and
R C (2016) Anatomy and function of an excitatory network in the
function. Cell 163, 1796-1806.
visual cortex. Nature 532, 370-374.

178