R ES E A RC H

◥ volume imaging and also lacks the speed for

RESEARCH ARTICLE SUMMARY effective brain-wide or cortex-wide imaging
of multiple specimens.
IMAGING TECHNIQUES RATIONALE: We combined two imaging tech-
nologies to address these issues. Expansion mi-

Cortical column and whole-brain croscopy (ExM) creates an expanded, optically

clear phantom of a fluorescent specimen that
retains its original relative distribution of fluore-
imaging with molecular contrast scent tags. Lattice light-sheet microscopy (LLSM)
then images this phantom
and nanoscale resolution
◥

ON OUR WEBSITE in three dimensions with

Read the full article
minimal photobleaching at
at http://dx.doi. speeds sufficient to image
Ruixuan Gao*, Shoh M. Asano*, Srigokul Upadhyayula*, Igor Pisarev, Daniel E. Milkie,
org/10.1126/ the entire Drosophila brain
Tsung-Li Liu, Ved Singh, Austin Graves, Grace H. Huynh, Yongxin Zhao, John Bogovic, science.aau8302 or across the width of the
Jennifer Colonell, Carolyn M. Ott, Christopher Zugates, Susan Tappan, ..................................................
mouse cortex in ∼2 to 3 days,
Alfredo Rodriguez, Kishore R. Mosaliganti, Shu-Hsien Sheu, H. Amalia Pasolli,
with multiple markers at an effective resolution of
Song Pang, C. Shan Xu, Sean G. Megason, Harald Hess, Jennifer Lippincott-Schwartz,
∼60 by 60 by 90 nm for 4× expansion.
Adam Hantman, Gerald M. Rubin, Tom Kirchhausen, Stephan Saalfeld, Yoshinori Aso,
Edward S. Boyden†, Eric Betzig† RESULTS: We applied expansion/LLSM (ExLLSM)

Downloaded from http://science.sciencemag.org/ on February 3, 2019

to study a variety of subcellular structures in
INTRODUCTION: Neural circuits across the teins and the speed to readily image multiple the brain. In the mouse cortex, we quantified
brain are composed of structures spanning specimens. Conversely, confocal fluorescence the volume of organelles, measured morpholog-
seven orders of magnitude in size that are as- microscopy offers molecular contrast but has in- ical parameters of ~1500 dendritic spines, de-
sembled from thousands of distinct protein sufficient resolution for dense neural tracing termined the variation of distances between
types. Electron microscopy has imaged densely or the precise localization of specific molecular pre- and postsynaptic proteins, observed large
labeled brain tissue at nanometer-level resolu- players within submicrometer-sized structures. differences in postsynaptic expression at ad-
tion over near-millimeter-level dimensions but Last, superresolution fluorescence microscopy jacent pyramidal neurons, and studied both
lacks the contrast to distinguish specific pro- bleaches fluorophores too quickly for large- the azimuthal asymmetry and layer-specific
longitudinal variation of axonal myelination. In
Drosophila, we traced the axonal branches of
olfactory projection neurons across one hem-
isphere and studied the stereotypy of their
boutons at the calyx and lateral horn across five
animals. We also imaged all dopaminergic neu-
rons (DANs) across the brain of another specimen,
visualized DAN morphologies in all major brain
regions, and traced a cluster of eight DANs to
their termini to determine their respective cell
types. In the same specimen, we also determined
the number of presynaptic active zones (AZs)
across the brain and the local density of all AZs
and DAN-associated AZs in each brain region.

CONCLUSION: With its high speed, nanomet-

ric resolution, and ability to leverage genetically
targeted, cell type–specific, and protein-specific
fluorescence labeling, ExLLSM fills a valuable
niche between the high throughput of conven-
tional optical pipelines of neural anatomy and
the ultrahigh resolution of corresponding EM
pipelines. Assuming the development of fully
validated, brain-wide isotropic expansion at
10× or beyond and sufficiently dense labeling,
ExLLSM may enable brainwide comparisons
of even densely innervated neural circuits
across multiple specimens with protein-specific

Nanoscale brain-wide optical imaging. ExLLSM images neural structures with molecular
contrast at 25-nm resolution or better.
▪
contrast over millimeter-scale volumes, including (clockwise from top right) mouse pyramidal The list of author affiliations can be found in the full article online.
neurons and their processes; organelle morphologies in somata; dendritic spines and synaptic *These authors contributed equally to this work.
proteins across the cortex; stereotypy of projection neuron boutons in Drosophila; projection †Corresponding author. Email: [email protected] (E.S.B.);
[email protected] (E.B.)
neurons traced to the central complex; and (center) dopaminergic neurons across the brain, Cite this article as R. Gao et al., Science 363, eaau8302
including the ellipsoid body (circular inset). (2019). DOI: 10.1126/science.aau8302

Gao et al., Science 363, 245 (2019) 18 January 2019 1 of 1

R ES E A RC H

◥ connections determined by EM connectomics

RESEARCH ARTICLE (15), and given that the anatomical circuits for
specific tasks can vary substantially between
individuals of the same species (16, 17), high-
IMAGING TECHNIQUES resolution three-dimensional (3D) imaging with
molecular specificity of many thousands of

Cortical column and whole-brain brains may be necessary to yield a comprehen-

sive understanding of the genesis of complex
behaviors in any organism. Here, we describe a
imaging with molecular contrast combination of expansion microscopy (ExM)
(18, 19), lattice light-sheet microscopy (LLSM)

and nanoscale resolution (20), and terabyte-scale image processing and

analysis tools (21) that achieves single-molecule
sensitivity and ~60- by 60- by 90-nm resolution
Ruixuan Gao1,2,3*, Shoh M. Asano1,2*†, Srigokul Upadhyayula3,4,5,6*, Igor Pisarev3, at volumetric acquisition rates ~700× and 1200×
Daniel E. Milkie3, Tsung-Li Liu3‡, Ved Singh3§, Austin Graves3¶, Grace H. Huynh1#, faster than existing high-speed SR (22) and
Yongxin Zhao1**, John Bogovic3, Jennifer Colonell3, Carolyn M. Ott3, EM (5) methods, respectively, at comparable or
Christopher Zugates7, Susan Tappan8, Alfredo Rodriguez8, Kishore R. Mosaliganti9, higher resolution (table S1). We demonstrate its
Shu-Hsien Sheu3, H. Amalia Pasolli3, Song Pang3, C. Shan Xu3, Sean G. Megason9, utility through multicolor imaging of neural
Harald Hess3, Jennifer Lippincott-Schwartz3, Adam Hantman3, Gerald M. Rubin3, subsets and associated proteins across the thick-
Tom Kirchhausen3,4,5,6, Stephan Saalfeld3, Yoshinori Aso3, ness of the mouse cortex and the entirety of the

Downloaded from http://science.sciencemag.org/ on February 3, 2019

Edward S. Boyden1,2,10,11,12,13††, Eric Betzig3,14,15,16,17,18†† Drosophila brain while quantifying nanoscale
parameters, including dendritic spine morphol-
Optical and electron microscopy have made tremendous inroads toward understanding the ogy, myelination patterns, stereotypic variations
complexity of the brain. However, optical microscopy offers insufficient resolution to reveal in boutons of fly projection neurons, and the
subcellular details, and electron microscopy lacks the throughput and molecular contrast to number of synapses in each fly brain region.
visualize specific molecular constituents over millimeter-scale or larger dimensions. We combined Combining expansion and lattice light-
expansion microscopy and lattice light-sheet microscopy to image the nanoscale spatial sheet microscopy (ExLLSM)
relationships between proteins across the thickness of the mouse cortex or the entire Drosophila
brain. These included synaptic proteins at dendritic spines, myelination along axons, and In protein-retention ExM (proExM) (19), fluorophore-
presynaptic densities at dopaminergic neurons in every fly brain region. The technology should conjugated antibodies (Abs) and/or fluorescent
enable statistically rich, large-scale studies of neural development, sexual dimorphism, degree proteins (FPs) that mark the features of interest
of stereotypy, and structural correlations to behavior or neural activity, all with molecular contrast. within a fixed tissue are chemically linked to an
infused polyacrylamide/polyacrylate gel. After pro-

T
tease digestion of the tissue, the gel can be ex-
he human brain is a 1.5-kg organ that, immunofluorescence, fluorescent proteins, or panded in water isotropically, creating an enlarged
despite its small size, contains more than fluorescence in situ hybridization (FISH) enables phantom of the tissue that faithfully retains the
80 billion neurons (1) that connect through high-sensitivity imaging of specific protein ex- tissue’s original relative distribution of fluorescent
approximately 7000 synapses each in a pression patterns in brain tissue (6, 7), brain- tags (fig. S1 and supplementary note 1). This yields
network of immense complexity. Neural wide tracing of sparse neural subsets in flies an effective resolution given by the original re-
structures span a size continuum greater than (8, 9) and mice (10), and in situ identification solution of the imaging microscope divided by
seven orders of magnitude in extent and are com- of specific cell types (11, 12) but has insufficient the expansion factor. Another advantage of di-
posed of more than 10,000 distinct protein types resolution for dense neural tracing or the precise gestion is that lipids, protein fragments, and
(2) that collectively are essential to build and localization of specific molecular players within other optically inhomogeneous organic com-
maintain neural networks. Electron microscopy critical subcellular structures such as dendritic ponents that are not anchored to the gel are suf-
(EM) can image down to the level of individual spines. Diffraction-unlimited superresolution (SR) ficiently removed so that the expanded gel has
ion channels and synaptic vesicles (3) across the fluorescence microscopy (13, 14) combines nano- a refractive index nearly indistinguishable from
~0.03 mm3 volume of the brain of the fruitfly scale resolution with protein-specific contrast but water and therefore can be imaged aberration-
Drosophila melanogaster (4, 5). However, EM bleaches fluorophores too quickly for large-volume free to a postexpansion depth of at least 500 mm
creates a grayscale image in which the segmen- imaging and, like EM, would require months (fig. S2) by using conventional water immersion
tation of specific subcellular components or to years to image even a single D. melanogaster objectives. ProExM has been applied to a range
the tracing of the complete arborization of spe- brain (table S1). of model animals, including mouse (19), zebra-
cific neurons remains challenging and in which Given the vast array of molecular species that fish (23), and Drosophila (24–28). Although up
specific proteins can rarely be unambiguously contribute to neural communication through to 20× expansion has been reported (29), at
identified. Optical microscopy combined with many mechanisms in addition to the synaptic 8× expansion by using an iterated form of

1
MIT Media Lab, Massachusetts Institute of Technology (MIT), Cambridge, MA 02139, USA. 2McGovern Institute for Brain Research, MIT, Cambridge, MA 02139, USA. 3Janelia Research Campus,
Howard Hughes Medical Institute, Ashburn, VA 20147, USA. 4Department of Cell Biology, Harvard Medical School, 200 Longwood Avenue, Boston, MA 02115, USA. 5Program in Cellular and
Molecular Medicine, Boston Children’s Hospital, 200 Longwood Avenue, Boston, MA 02115, USA. 6Department of Pediatrics, Harvard Medical School, 200 Longwood Avenue, Boston, MA 02115,
USA. 7arivis AG, 1875 Connecticut Avenue NW, 10th floor, Washington, DC 20009, USA. 8MBF Bioscience, 185 Allen Brook Lane, Suite 101, Williston, VT 05495, USA. 9Department of Systems
Biology, Harvard Medical School, 200 Longwood Avenue, Boston, MA 02115, USA. 10Department of Biological Engineering, MIT, Cambridge, MA 02139, USA. 11MIT Center for Neurobiological
Engineering, MIT, Cambridge, MA 02139, USA. 12Department of Brain and Cognitive Sciences, MIT, Cambridge, MA 02139, USA. 13Koch Institute, MIT, Cambridge, MA 02139, USA. 14Department
of Molecular and Cell Biology, University of California, Berkeley, CA 94720, USA. 15Department of Physics, University of California, Berkeley, CA 94720, USA. 16Howard Hughes Medical Institute,
Berkeley, CA 94720, USA. 17Helen Wills Neuroscience Institute, Berkeley, CA 94720, USA. 18Molecular Biophysics and Integrated Bioimaging Division, Lawrence Berkeley National Laboratory,
Berkeley, CA 94720, USA.
*These authors contributed equally to this work. †Present address: Internal Medicine Research Unit, Pfizer, Cambridge, MA 02139, USA. ‡Present address: Vertex Pharmaceuticals, 3215 Merryfield Row, San Diego, CA
92121, USA. §Present address: Intel, 2501 Northwest 229th Avenue, Hillsboro, OR 97124, USA. ¶Present address: Department of Neuroscience, Johns Hopkins University School of Medicine, Baltimore, MD 21205, USA.
#Present address: Microsoft Research Lab, 14820 NE 36th Street, Redmond, WA 98052, USA. **Present address: Department of Biological Sciences, Carnegie Mellon University, Pittsburgh, PA 15143, USA.
††
Corresponding author. Email: [email protected] (E.S.B.); [email protected] (E.B.)

Gao et al., Science 363, eaau8302 (2019) 18 January 2019 1 of 16

R ES E A RC H | R E S EA R C H A R T I C LE

the N,N-dimethylacrylamide-gel expansion pro-

tocol, we observed regions where the expansion
superficially appears accurate (fig. S3A) and
other regions of clear distortion, such as irreg-
ularly shaped somata and nuclei (fig. S3B). High
expansion ratios also require exceptionally high
fluorescence labeling densities to take advantage
of the theoretically achievable resolution and take
longer to image. Thus, for this work we chose to
focus only on applications (table S2) enabled by
4× expansion.
Several challenges emerge when attempting
to extend ExM to specimens at the millimeter
scale of the fly brain or a mouse cortical col-
umn. First, even 4× expansion requires effec-
tive voxel dimensions of ~30 to 50 nm on each
side to match the full resolution potential of
ExM, or ~20 trillion voxels/mm3/color. This in
turn necessitates imaging at speeds on the order
of 100 million voxels/s to complete the acquisition
in days rather than weeks or more, as well as an

Downloaded from http://science.sciencemag.org/ on February 3, 2019

image-processing and -storage pipeline that can
handle such high sustained data rates. Second,
photobleaching often extinguishes the fluores-
cence signal from deeper regions of 3D speci-
mens before they can be imaged—a problem
that becomes more severe with thicker spec-
imens, longer imaging durations, and/or the
higher illumination intensities needed for faster
imaging. Last, because ExM resolution is pro-
portional to imaging resolution, the latter should
be as high as possible within these other con-
straints while also striving for near-isotropic
resolution, so that neural tracing and quanti-
fication of nanoscale structures is not limited by
the axis of poorest resolution.
To address these challenges, we turned to
LLSM (20), which sweeps an ultrathin sheet
of laser light through a specimen and collects
the resulting fluorescence from above with a
high numerical aperture (NA) objective to image
it on a high-speed camera (supplementary note 2).
Confinement and propagation of excitation
light within the detection focal plane permits
parallel acquisition of data at rates of 10 million
to 100 million voxels/s at low intensities that
minimize photobleaching within the plane and
eliminates bleaching in the unilluminated regions
above and below. Consequently, we could image
large volumes of expanded tissue expressing Fig. 1. Comparing modalities to image-expanded mouse brain tissue. (A) 3D rendered volumes
yellow fluorescent protein (YFP) in a subset of at equal magnification of tissue sections from the primary somatosensory cortex of a Thy1-YFP
mouse cortical neurons with uniform signal from transgenic mouse, expanded ~4× by using the protein-retention expansion microscopy (proExM)
top to bottom (Fig. 1A, left). By contrast, at a protocol and imaged by means of (left to right) LLSM in sample scan mode [LLSM (SS), blue]; spinning
comparable signal in the acquired images, the disk confocal microscopy (Spinning Disk, green); and Airyscan in fast mode (Airyscan, orange). Scale
out-of-focus excitation and high peak power bars, 50 mm, here and elsewhere given in preexpanded (biological) dimensions. (B) (Top) xy and
at the multiple foci of a spinning disk confocal (bottom) xz maximum intensity projections (MIPs) of 25-mm-thick slabs cut from the image volumes in
microscope (SDCM) photobleached the expanded (A) at the locations denoted by the red and purple lines in the slabs perpendicular to them, respectively.
tissue ~10× faster than LLSM (Fig. 1C), rendering (Insets) Regions in the white rectangles at higher magnification. Scale bars, 50 mm, full MIPs; 5 mm,
deeper regions completely dark (Fig. 1, A and insets. (C) Comparative imaging and photobleaching rates for the three modalities (table S1).
B, center), while the sparse illumination of (D) (Top) xy and (bottom) xz spatial frequency content in the same three image volumes as measured
the SDCM focal array slowed volumetric ac- from mitochondria-targeted antibody puncta, with different resolution bands as shown (fig. S4).
quisition by ~7× (table S1). Another commer-
cial alternative, Airyscan, efficiently images the expanded tissue ~40× slower (table S1) and with move in discrete steps across the image volume,
fluorescence generated at the excitation focus ~20× faster bleaching (Fig. 1C) than LLSM. and sample scan (Fig. 1), in which the sample
and uses this information to extend the imaging LLSM can operate in two modes: objective is swept continuously through the light sheet.
resolution approximately 1.4× beyond the dif- scan (fig. S4), in which the sample is stationary Sample scan is faster (tables S1) but yields
fraction limit (30, 31). However, Airyscan imaged while the light-sheet and detection objective slightly lower yz resolution (fig. S4) than that of

Gao et al., Science 363, eaau8302 (2019) 18 January 2019 2 of 16

R ES E A RC H | R E S EA R C H A R T I C LE

objective scan because information in the sam- represented mitochondria or lysosomes (Fig. 2C)— taining large organelles within the narrow
ple scanning direction is slightly blurred by in the latter case, the specific subset with LAMP1 confines of the axon, or they may reflect func-
simultaneous image acquisition and sample that likely represented multivesicular bodies or tional differences in the regulation of calcium
movement. Of the methods above, Airyscan autolysosomes (supplementary note 6a) (32). As in axons versus dendrites.
should in principle achieve the highest lateral expected, we found that mitochondria were We next turned our attention to the myeli-
(xy) resolution, followed by SDCM (owing to generally both longer and larger in volume than nation of axons, which is essential for the rapid
pinhole filtering), and last, the two modes of lysosomes (Fig. 2D and table S3). Mitochondria (38, 39) and energy-efficient (40) propagation of
LLSM. In practice, however, dendritic spines ranged in length from 0.2 to 8.0 mm, which is action potentials (APs) and which, when dis-
and axons appeared more clearly and faithfully consistent with EM measurements in the cor- rupted, can lead to neurodegenerative diseases
resolved in lateral views with LLSM than with tex (33) or other regions (34) of the mouse such as multiple sclerosis (41). The propagation
SDCM or even Airyscan (Fig. 1B, top row), a brain, whereas the subset of LAMP1 compart- velocity is affected by the g-ratio, the diameter
conclusion corroborated by its higher lateral ments ranged from 0.1 to ~1.0 mm, which is also of the axon normalized to the diameter of its
spatial frequency content (Fig. 1D and fig. S4A, consistent with EM (35). surrounding myelin sheath (42). Most EM mea-
top rows) as measured from mitochondria- Given this agreement—and the important surements of the g-ratio come from 2D images
targeted Ab puncta. Likewise, the thinness of roles mitochondria play in dendrite develop- of single sections cut transversely to axonal tracts
the lattice light sheet contributes to the axial ment, synapse formation, calcium regulation, and (43–45) and therefore lack information on how
(z) resolution of LLSM (Fig. 1D and fig. S4A, neurodegenerative disease (34, 36, 37)—we the g-ratio might vary along the length of a
bottom rows) and therefore yielded xz views extended our analysis across ~100 by 150 by given axon. To address this, we used ExLLSM to
of spines and axons only slightly poorer than 150 mm of the mouse somatosensory cortex. We image a 320- by 280- by 60-mm volume in the
in the lateral plane and substantially sharper classified length, aspect ratio, and volume (Fig. primary somatosensory cortex of a Thy1-YFP
than those obtained with SDCM or Airyscan 2E and fig. S8) of 2893 mitochondria and 222 transgenic mouse immunostained against mye-

Downloaded from http://science.sciencemag.org/ on February 3, 2019

(Fig. 1B, bottom row). lysosomes across the somata and initial por- lin basic protein (MBP) (Fig. 2F and Movie 2).
One additional challenge in millimeter-scale tions (78 mm mean length) of the apical dendrite At every longitudinal position z along a given
ExLLSM involves the processing of multitera- of five-layer V pyramidal neurons, as well as the myelinated axon, we measured the local g-ratio
byte data sets. In LLSM, the lateral extent of the initial portions (95 mm mean length) of three at every azimuthal position q by dividing the
light sheet (table S2) is far smaller than an ex- descending axon segments. As noted previously radius raxon(q, z) of the axon along the radial
panded fly brain or cortical column, so the final in the hippocampus (36), we found that long vector from the axon center by the radius
image volumes had to be computationally stitched and high-aspect-ratio mitochondria were far rmyelin(q, z) of the outer edge of the myelin
together from as many as 25,000 (table S2) tiled more prevalent in apical dendrites than in sheath along the same vector (Fig. 2G, fig. S9,
subvolumes per color. However, because of sys- axons, with mitochondria longer than 3 mm and supplementary note 4e). Across one 56-mm-
tematic sample stage errors and slight swelling comprising 6.5% all dendritic mitochondria long segment, the mean g-ratio of 0.57 calcu-
or shrinking of expanded samples over many (~12 per 100 mm of dendrite length) versus 0.7% lated from mean axon and sheath diameters
hours, many tiles did not perfectly overlap with of all axonal ones. These differences may re- of 0.52 and 0.90 mm, respectively, fell at the
their neighbors on all six sides. To address this, present the difficulty in assembling and main- lower end of a distribution previously reported
we developed an Apache Spark–based high- in the central nervous system yet was con-
performance computing pipeline (supplementary sistent with a theoretical estimate of 0.60 for
note 3 and figs. S5 to S7) that first performed a the ratio that optimizes propagation velocity
flat-field correction for each tile to account for (42). However, these values do not reflect the
intensity variations across the light sheet and substantial variability we observed, with the
then stitched the intensity-corrected tiles to- outer axon–to–outer myelin distance ranging from
gether by using an automated and iteratively 0.12 to 0.35 mm (fig. S10) and the local g-ratio
refined prediction model of tile coordinates. In a ranging from ~0.4 to 0.8 (Fig. 2H and Movie 2).
separate track, each intensity-corrected tile was Furthermore, the axon and the sheath were rarely
deconvolved by using a measured point spread concentric (Fig. 2G), leading to rapid longitudinal
function (PSF) so that when the final set of changes in capacitance and impedance that may
coordinates for all tiles was available, the de-
convolved image volume of the entire specimen
could be assembled and visualized (supplementary
note 4 and 5) with minimal stitching artifacts.

Quantification of subcellular structures

in mouse cortical neurons
The protein-specific fluorescence contrast of
ExLLSM enabled rapid, computationally effi-
cient, and purely automated segmentation and
nanoscale quantification of subcellular neural
structures over large volumes. For example,
dense cytosolic expression of YFP under the
thy1 promotor in mouse pyramidal neurons
revealed sharply delineated voids (Movie 1) Movie 1. Organelle analysis of layer V pyramidal Movie 2. Axon myelination and local g-ratio
representing subcellular compartments (Fig. 2A) neurons in the mouse somatosensory cortex. of layer V pyramidal neurons of the mouse
of various shapes and sizes whose volumes we Segmentation of cytosolic voids in Thy1-YFP– primary somatosensory cortex. Thy1-YFP–
could quantify accurately (Fig. 2B and supple- expressing neurons, quantification of their vol- expressing neurons and immunostained myelin
mentary note 4d). Simultaneous immunofluo- umes, and immunostaining-based classification sheaths across 320 by 280 by 60 mm, with
rescence labeling against Tom20 and LAMP1, of those voids that represent mitochondria or quantification of the local g-ratio on the surface
although comparatively sparse (movie S1), was multivesicular bodies or autolysosomes (Fig. 2, of a specific myelin sheath (Fig. 2, F and G, and
sufficient to identify the subset of these that A to E; fig. S8; and movie S1). figs. S9 and S10).

Gao et al., Science 363, eaau8302 (2019) 18 January 2019 3 of 16

R ES E A RC H | R E S EA R C H A R T I C LE

Fig. 2. Nanoscale, protein-specific 3D imaging of subcellular neural 5 mm and (inset) 500 nm. (G) Same region as (F), with the myelin sheath Downloaded from http://science.sciencemag.org/ on February 3, 2019
structures. (A) Segmented compartments void of cytosolic YFP (gray), color coded according to the local g-ratio (fig. S10). (Inset) Azimuthal
color-coded by volume, in portions of the somata and apical dendrites of variation in g-ratio in the region within the rectangle. Scale bar, 5 mm.
two layer V pyramidal neurons from the somatosensory cortex of a Thy1-YFP (H) (Left) Distribution of axon radius and myelin outer radius and (right)
mouse (Movie 1). Scale bars, 5 mm and (inset) 1 mm. (B) Distribution of distribution of g-ratio at all points on the axon in (G). (I) xy MIP of a
the compartment volumes. (C) Same region as (A), with voids identified 9.3-mm-thick slab within a 75- by 100- by 125-mm volume from the primary
with immunostaining (movie S1) as either mitochondria (magenta) or somatosensory cortex of a Thy1-YFP mouse, immunostained against
multivesicular bodies or autolysosomes (yellow). (D) Scatter plots of the synaptic proteins Bassoon and Homer1 (Movie 3 and fig. S10). Only
major axis (long axis) length versus volume for the two organelle types. YFP-associated Bassoon/Homer1 pairs are shown for clarity. (Insets)
Point colors in (D) and (E) indicate relative data point density (blue, low; (Top) magnified xy MIP of a 2.2-mm-thick slab from boxed region at right.
red, high). (E) Similar scatter plots for mitochondria only, separated by (Bottom) All Bassoon/Homer1 pairs in the same region. Three pairs are
cellular region (fig. S8). (F) Axon of a layer V pyramidal neuron and its indicated with arrows. Scale bars, 10 mm and (insets) 1 mm. (J) Distribution
surrounding myelin sheath, from the primary somatosensory cortex of distances between paired Bassoon and Homer1 centroids across the
of another Thy1-YFP mouse, immunostained against myelin (Movie 2). entire volume. (K) Distribution when restricted to only those pairs
(Inset) A cross-sectional view through the white parallelogram. Scale bars, associated with YFP-expressing neurons.

Gao et al., Science 363, eaau8302 (2019) 18 January 2019 4 of 16

R ES E A RC H | R E S EA R C H A R T I C LE

Downloaded from http://science.sciencemag.org/ on February 3, 2019

Fig. 3. Characterizing dendritic spine morphologies and postsynaptic indicating relationships between (top) spine backbone length and head
Homer1 across the mouse primary somatosensory cortex. (A) Coronal diameter and (bottom) spine neck length and neck diameter in the four
MIP of a 1900- by 280- by 70-mm tissue section spanning the pia to the regions from (B) (figs. S13 to S15 and movie S2). (D) Two adjacent layer V
white matter of the primary somatosensory cortex of a Thy1-YFP mouse pyramidal neurons selected within the volume (magenta), one exhibiting
(Movie 4), additionally immunostained against Bassoon and Homer1. strong Homer 1 expression (neuron 1) and the other exhibiting weak
Boxes denote seven regions for quantitative morphological analysis of expression (neuron 2). (Insets) Homer1 localization or lack thereof at
dendritic spines. Scale bar, 100 mm. (B) (Top) Magnified MIPs of YFP- apical dendritic spines (fig. S17). Scale bars, 50 mm and (insets) 10 mm.
expressing neurons in four of the regions from (A), with (bottom) further (E) (Top) MIP of the local density of Homer1 puncta across a ~25-mm-thick
magnified subregions showing differing spine morphologies. Scale bars, coronal slab, and (bottom) the cumulative number of puncta in 50- by
(top) 50 mm and (bottom) 10 mm. (C) Scatter plots and histograms 50- by 25-mm subvolumes across the cortex.

Gao et al., Science 363, eaau8302 (2019) 18 January 2019 5 of 16

R ES E A RC H | R E S EA R C H A R T I C LE

influence the speed and efficiency of signal in part because it is related to synaptic strength 4f) (60) to segment (fig. S14 and movie S2)
propagation. We subsequently confirmed these (49), whose time- and activity-dependent change and measure spine ultrastructure. Across the
observations with EM (fig. S11 and supplemen- (plasticity) (50) is implicated in learning and ~1500 spines so measured, the range of spine
tary note 2h). memory consolidation (51). However, although head diameters, neck diameters, overall back-
ExLLSM is also well suited to study the na- optical methods such as Golgi impregnations bone lengths (spine root to tip), and neck back-
noscale organization of synaptic proteins over (52), array tomography (6), and confocal (53) bone lengths (Fig. 3C and figs. S13B and S15)
large tissue volumes. Imaging a 75- by 100- by and two-photon microscopy (54, 55) can image were consistent with those seen in an EM study
125-mm tissue section cut from layer IV/V of the complete arborization of neurons spanning of layer II/III pyramidal neurons in the mouse
the primary somatosensory cortex of a trans- the cortex, they lack the 3D nanometric resolu- visual cortex (56). Furthermore, the absence of
genic Thy1-YFP mouse, we identified 25,286 tion needed to measure the detailed morphol- spines in the initial segment of the distal apical
synapses that have closely juxtaposed concen- ogy of spines. Conversely, EM (56, 57) and SR dendrite, and prevalence of much larger spines
trations of immunolabeled pre- and postsynaptic fluorescence microscopy (58, 59) have the re- on smaller dendritic branches than on the re-
proteins Bassoon and Homer1 (fig. S12A), 2325 quisite resolution but not the speed to scale mainder of the distal apical dendrite (Fig. 3D),
of which had Homer1 localized at YFP-labeled readily to cortical dimensions. ExLLSM, however, were in line with an EM study of pyramidal
dendritic spines (Fig. 2I and Movie 3). These has both. neurons in the primary somatosensory cortex
tended to form nested caps, with major axis To demonstrate this, we imaged a 1900- by of the cat (61). Mean spine head diameter and
lengths of 856 ± 181 nm and 531 ± 97 nm for 280- by 70-mm tissue slice spanning the pia to mean neck backbone length each approximately
Bassoon and Homer1, respectively [median ± the white matter in the primary somatosensory doubled from layer II/III (position 1) to the re-
median absolute deviation (MAD)] (fig. S12, B cortex of a transgenic Thy1-YFP mouse expres- gions of layers IV and V (positions 3 and 4)
and C). The Homer1 distribution was consist- sing cytosolic fluorescence within a sparse subset nearest the somata before falling again in layer
ent with SR measurements in dissociated hip- of layer V pyramidal neurons. The slice was VI (positions 6 and 7) to levels similar to layer

Downloaded from http://science.sciencemag.org/ on February 3, 2019

pocampal neurons (DHN) (46), but our Bassoon additionally immunostained against Bassoon II/III (table S4). This is consistent with a lon-
values were slightly larger. The centroid-to- and Homer1 (Fig. 3A and Movie 4). In each of gitudinal in vivo study of spine morphology that
centroid distance we measured between Bassoon/ seven different regions across the cortex (Fig. found that spines closer to the soma, including
Homer1 pairs was 243 ± 69 nm for all pairs 3B and fig. S13A), we selected four 27- by 27- those on proximal apical dendrites, were more
within the volume (Fig. 2J) and 185 ± 70 nm by 14-mm subvolumes and used a modified com- mature and formed stronger synaptic connec-
for those associated with YFP-filled spines (Fig. mercial analysis pipeline (supplementary note tions than those on basal dendrites or the distal
2K). The difference between these values sug- apical dendrite (62). We also found that head
gests that mature glutamatergic synapses of diameter and backbone length or neck back-
layer V pyramidal neurons, which are the ones bone length were correlated across all layers
expressing YFP, are narrower than other types of the cortex (Fig. 3C, top row; figs. S13B, top
across the primary somatosensory cortex. The row, and S15; and table S4), but neck diameter
difference between these values and previous and neck backbone length were not correlated
SR measurements of 150 ± 20 nm in the ventral across all regions (Fig. 3C, bottom row; fig. S13B,
orbital cortex (n = 252 Bassoon/Homer1 pairs) bottom row; and table S4).
(47), 165 ± 9 nm in DHN (n = 43 pairs) (46), and Colabeling with Homer1-specific antibodies
179 ± 42 nm in the middle of the primary somato- allowed us also to map excitatory synapses and
sensory cortex (n = 159 pairs) (29) may reflect their density (Fig. 3E) across the primary somato-
natural variations in different brain regions (29) sensory cortex. In particular, when 4.5 million
or a systematic bias in these earlier studies arising Homer1 puncta were binned in 50- by 50- by
by measuring the distance between 1D Gaussian 25-mm subvolumes to average across local fluc-
fits to the Bassoon/Homer1 distributions in a Movie 3. Synaptic proteins and their tuations, their density was revealed to be ~1.5
manually selected slice through the heart of associations to neuronal processes in to 2.0× greater in layers II/III and V (~40 to
each synapse, versus our approach of calcu- layers IV and V of the mouse primary 50 puncta/mm3) than in adjacent layers I, IV,
lating the distance between the 3D centroids somatosensory cortex. Thy1-YFP–expressing and VI. Similar dual maxima in synaptic density
calculated across the complete distributions. neurons and immunostained pre- and are seen in sparsely sampled EM images of the
postsynaptic proteins Bassoon and Homer1 rat somatosensory (63) and mouse barrel cortex
Somatosensory cortex–spanning across 75 by 100 by 125 mm, sequentially (64), although in different cortical layers (rat, II
measurement of dendritic spines and showing all Bassoon and Homer1 puncta, and and IV; mouse, I and IV) than seen in this work.
excitatory synapses only YFP-associated Bassoon and Homer1 Focusing on the subset of Homer1 puncta
The combination of fast imaging (table S1) and pairs (Fig. 2, I to K, and fig. S12). colocalized with YFP-expressing dendritic spines,
targeted sparse labeling enables ExLLSM-based we found that thin spines were approximately
quantification of nanoscale neural structures to twice as likely to coexpress Homer1 as spines
be extended to millimeter-scale dimensions over classified as stubby, mushroom, or filopodial
multiterabyte data sets. This yields statistically (fig. S16). As a synaptic scaffold protein, Homer1
large sample populations that can reveal subtle plays an important role in the recruitment and
changes in the distributions of specific morpho- cross-linking of other proteins that lead to the
logical parameters across different regions of Movie 4. Relationship of postsynaptic maturation and enlargement of spines (65–67),
the brain. Homer1 to neuronal processes across the so Homer1’s relative abundance at thin spines
One such application involves the morphol- mouse primary somatosensory cortex. Thy1- may presage their transformation to more mature
ogy of dendritic spines in different layers of the YFP–expressing neurons and immunostained forms. Surprisingly, we also observed dramatic
mouse cerebral cortex. A spine is a small (~0.01 postsynaptic protein Homer1 across 1900 by variations in the expression of Homer1 within
to 1.0 mm3) membranous protrusion from a 280 by 70 mm in the primary somatosensory neighboring layer V pyramidal neurons: Homer1
neuronal dendrite that receives synaptic input cortex, with specific focus on two adjacent layer was present at nearly all spines and throughout
from the closely juxtaposed axon of another V pyramidal neurons that exhibit substantially the cytosol of one neuron (Fig. 3D, neuron 1),
neuron. Spine morphology has been extensively different patterns of Homer1 expression (Fig. 3, whereas a parallel neuron ~57 mm away of sim-
studied with a variety of imaging methods (48), figs. S13 to S17, and movie S2). ilar morphology exhibited very little Homer1,

Gao et al., Science 363, eaau8302 (2019) 18 January 2019 6 of 16

R ES E A RC H | R E S EA R C H A R T I C LE

even at its dendritic spines (Fig. 3D, neuron 2). Homer1 levels are known to change rapidly Visual cortex–spanning neuronal tracing
This difference did not result from differential under different neuronal states [for example, and myelination patterns
labeling efficiency because the density of Homer1 asleep versus awake (68)], it may reflect the Although the radial anisotropy of axonal mye-
puncta in the immediate surrounds of each different excitatory states of these two neurons lination (Fig. 2E) can affect the speed and
neuron was similar (fig. S17). Instead, because at the time the animal was sacrificed. efficiency of AP propagation, so too can its

Downloaded from http://science.sciencemag.org/ on February 3, 2019

Fig. 4. Neural tracing and longitudinal myelination analysis across distal apical dendrite and two of its branches and (right) a basal dendrite
the mouse primary visual cortex. (A) Coronal MIP of a 25-mm-thick slab and its spines, from boxed regions i and ii in (B), respectively. Scale bars,
within a 1100- by 280- by 83-mm tissue section spanning the pia to the white 1 mm. (D) MIP view of boxed region iii in (B), showing (left) the distal end of
matter of the primary visual cortex of a Thy1-YFP mouse (Movie 5), the PMAS; (middle) Caspr at the start of myelination; and (right) cross-
additionally immunostained against MBP and Caspr to highlight myelin sectional views of the axon (1) before and (2) after the start of myelination.
sheaths and nodes of Ranvier, respectively. Scale bar, 100 mm. (B) Traced Scale bars, 1 mm. (E) MIP view of boxed region iv in (B), showing (left) break in
arborization (Movie 6) of a specific layer V pyramidal neuron denoted by the myelination and two branching collateral axons at a node of Ranvier
arrowhead in (A), showing the soma (red), apical (magenta), and basal and (right) Caspr highlighting the two ends of the node. Scale bars, 1 mm.
(orange) dendrites; myelinated (yellow) and unmyelinated (cyan) axon (F) (Top) Segmented view of a collateral axon with myelinated and unmyelinated
segments; and collateral axon branches (green). Arrows indicate nodes of sections from boxed region v in (B). (Bottom) Three MIP views of breaks in
Ranvier. Scale bar, 100 mm. (C) Magnified segmented views of (left) the myelination with flanking Caspr. Scale bars, (top) 10 mm; (bottom) 1 mm.

Gao et al., Science 363, eaau8302 (2019) 18 January 2019 7 of 16

R ES E A RC H | R E S EA R C H A R T I C LE

longitudinal variation. The repeated gaps in from the primary visual cortex extending from 4F). All these features matched the known mor-
myelination at the nodes of Ranvier house the pia to the white matter of a Thy1-YFP mouse. phologies of layer V pyramidal neurons (73) and
ion channels that are essential to regenerate The tissue was additionally immunostained were recapitulated in a second neuron traced
the AP during saltatory conduction (69), the against MBP and contactin-associated protein throughout the volume (Fig. 5A and Movie 6).
hallmark of high-speed signal propagation in (Caspr) (71) to visualize myelin sheaths and Given this assurance, we traced the axons
vertebrates. Recently, however, high-throughput their terminations, respectively (Fig. 4A and and their longitudinal myelination patterns for
EM imaging and axonal tracing at 30 by 30 by Movie 5). Although the dense global staining 10 neurons in layer V and 11 more in layer VI
240 nm/voxel (70) has revealed additional gaps of EM makes long-range 3D tracing of small (Fig. 5B). Within the imaged volume, all of the
in the axonal myelination of layer II/III neurons neurites challenging, expression of YFP in a layer V axons in the primary visual cortex ex-
in the mouse primary visual cortex much larger sparse subset of layer V and layer VI pyramidal hibited continuous myelination beyond the end
(for example, 55 mm) than either the ~2 mm neurons (72) enabled rapid semiautomatic trac- of the PMAS, except for the expected small gaps
typical of the nodes of Ranvier or the shorter ing (supplementary note 4h) of axons, their mye- at the nodes of Ranvier. This is consistent with
and rarer gaps observed in layers III to VI of the lination, and the entire arborization of selected the myelination pattern seen previously for
primary somatosensory cortex. neurons across the tissue section (Fig. 4B and layer III to VI axons in the primary somato-
To determine whether these differences are Movie 6). This included the distal apical den- sensory cortex (70). The range of PMAS lengths
more reflective of the layer of origination of drite and its branches (Fig. 4C, i), basal dendrites we measured for these neurons (28 to 41 mm,
the axon or the functional role of the cortical and their spines (Fig. 4C, ii), the premyelin ax- mean = 34.9 ± 1.1 mm) was also consistent with
region studied (the somatosensory versus the onal segment (PMAS) (Fig. 4D), the nodes of the range found in layers V and VI of the pri-
visual cortex), we imaged at 27 by 27 by 50 nm/ Ranvier (Fig. 4E), and collateral branches of mary somatosensory cortex (25 to 40 mm, mean =
voxel a ~280- by 1100- by 83-mm tissue section the main axon originating at the nodes (Fig. 33.7 ± 2.4 mm). The internodal spacing of the

Downloaded from http://science.sciencemag.org/ on February 3, 2019

Fig. 5. Longitudinal myelination profiles of layer V and VI pyramidal at left. Scale bars, (left) 10 mm and (right) 50 mm. (C) Node spacing for
neurons in the mouse primary visual cortex. (A) Traced arborization of four layer V neurons from (B) (fig. S18). (D) Volumes of eight layer V and
a second layer V pyramidal neuron within the volume in Fig. 4A. Scale bar, nine layer VI somata fully within the image volume [no asterisks in (B)]
100 mm. (B) (Left) Segmented soma and axon of a pyramidal neuron (mean ± SEM). (E) Volumes of the three somata with intermittently
shown in the context of its surroundings in layer VI. (Right) Segmented myelinated axons and five somata with continuously myelinated axons in
somata (color coded by volume) and axons, showing myelinated (yellow) layer VI (mean ± SEM). The P values are calculated from a permutation
and unmyelinated (cyan) segments, for 10 pyramidal neurons from layer V test for medians. n.s., not significant. (F) Scatter plot of soma volume
(top row) and 11 more from layer VI (bottom row). Boxed neuron is shown versus PMAS length for the neurons in (B) (fig. S19).

Gao et al., Science 363, eaau8302 (2019) 18 January 2019 8 of 16

R ES E A RC H | R E S EA R C H A R T I C LE

Downloaded from http://science.sciencemag.org/ on February 3, 2019

Fig. 6. Long-range tracing and stereotypy of neuron bundles in brains (D1 to D5) near CA. Scale bar, 10 mm. (D) Number of DC3 PN
Drosophila. (A) MIP view of DC3 olfactory projection neurons (PNs) boutons in CA for D1 to D5 shown in (C). (E) Volume of DC3 PN boutons in
projecting from the antenna lobe of an adult Drosophila brain and partially CA for D1 to D5 shown in (C). (F) MIP view of individually traced PPM3
traced here (Movie 7) to the calyx (CA) and lateral horn (LH). Scale bar, DANs in the right hemisphere of an adult Drosophila brain (Movie 8),
10 mm. (Inset) (White box) Comparison of cross-sectional views of the innervating the fan-shaped body (FB) (green), ellipsoid body (EB)
axon bundle by means of (left) confocal microscopy and (right) ExLLSM. (magenta), and noduli (NO) (green). The fine neurites arborizing FB, EB,
Scale bar, 1 mm. (Inset) (Yellow box) A magnified view of DC3 PN boutons and NO are from both hemispheres of the brain. Scale bar, 10 mm.
in CA. Scale bar, 1 mm. (B) Volume of each individual DC3 PN bouton in (G) MIP view of the identified cell types of PPM3 DANs (fig. S20).
CA and LH. (C) Overlaid MIP view of DC3 PNs from five adult Drosophila Scale bar, 10 mm.

Gao et al., Science 363, eaau8302 (2019) 18 January 2019 9 of 16

R ES E A RC H | R E S EA R C H A R T I C LE

four layer V neurons that could be traced to the selected subsets of its ~100,000 neurons, such constructed completely in the L1 instar larva
white matter increased with increasing distance as the dorsal paired medial (DPM) neurons that and partially in the adult brain by using EM
from the soma (Fig. 5C and fig. S18). By con- innervate the mushroom bodies (MBs) (movie (5, 77). However, the variation among individ-
trast, in layer VI only six axons were continu- S3). Fluorescence imaging of thousands of such ual animals has not been well studied at the
ously myelinated, whereas two were completely subsets across thousands of transgenic flies and level of detailed subcellular circuitry. The speed
unmyelinated, and three exhibited intermittent collation of the results then yields brain-wide 3D of ExLLSM now makes this possible. We studied
myelination with long unmyelinated segments reconstructions of complete neural networks at the stereotypy of DC3 PNs by comparing their
more reminiscent of the layer II and III axons single-cell resolution (8, 9). However, to trace fine morphologies in the CA across five different
in the primary somatosensory cortex than the neuronal processes and identify synaptic con- animals (Fig. 6C). As expected, we consistently
layer VI axons there (70). Thus, myelination nections, nanoscale resolution is needed. For observed the restriction of boutons to the ends
patterns of axons in the primary visual cortex all these reasons, the Drosophila brain is well of the neurites in CA. However, we found that
and the primary somatosensory cortex can differ, matched to the capabilities of ExLLSM. both the number and size of boutons differed
even for neurons in the same cortical layer. We thus chose to start with a relatively sim- among the three cells from the same hemi-
Although the volumes of the somata and ple case: three olfactory projection neurons (PNs) sphere as well as between animals. For example,
the diameters of the PMAS in layer V of the originating at the DC3 glomerulus of the an- the total number of boutons in CA varied from 7
primary visual cortex were twice as large as tennal lobes that feed most prominent sensory to 12, and none of the bouton assignments to
those in layer VI (Fig. 5D and fig. S19, respec- inputs to the calyx (CA) of the MB and lateral each cell was the same among all five brains
tively), there was not a strong relationship horn (LH) (75, 76). Imaging a ~250- by 175- by studied (Fig. 6D). The bouton size also showed
between soma volume and myelination pattern 125-mm volume, we were able to trace the axonal substantial variability among the brains (Fig.
(for example, intermittent or continuous) within branches of all three DC3 PNs across one hemi- 6E). These variations might arise from the dis-
layer VI (Fig. 5E). However, the PMAS lengths of sphere (Fig. 6A and Movie 7), although tracing tinct developmental histories of the individual

Downloaded from http://science.sciencemag.org/ on February 3, 2019

the six continuously myelinated and the three of fine dendritic processes was still difficult animals. It is not yet clear whether they also
intermittently myelinated axons in layer VI of at 4× expansion. We were also able to precisely indicate differences in synaptic strength and
the primary visual cortex split into distinct assign boutons to each cell within the CA (cell 1, connection with Kenyon cells or how they
populations (Fig. 5F), with the intermittent 3 boutons; cell 2, 3 boutons; cell 3, 4 boutons) might affect processing of olfactory information
ones of mean length (30.3 ± 1.7 mm) similar and the LH (cell 1, 19 boutons; cell 2, 32 boutons; for associative learning in the MB. ExLLSM will
to the axons of layer V, and the continuous cell 3, 23 boutons) and determine the shapes and enable such questions to be answered, thanks
ones more than twice as long (70.6 ± 3.6 mm). sizes of the boutons in these regions (Fig. 6B). to its high throughput and its precise descrip-
Thus, continuously myelinated axons in differ- The neuronal circuits of the olfactory path- tions of neuronal morphology.
ent layers of the primary visual cortex need not ways to the MB have been extensively described Given our success with this relatively simple
have similar PMAS lengths. Given that the dis- by using light microscopy and have been re- example, we next applied ExLLSM to a much
tal end of the PMAS is the site of AP initiation more challenging sample by imaging a ~340- by
(74), perhaps PMAS length might be one mech- 660- by 90-mm volume covering nearly the entire
anism by which neurons control the AP to ac- brain of a TH-GAL4 transgenic Drosophila spec-
count for differences in myelination or overall imen. The sample was immunostained in one
axon length in different layers and cortical regions. color against the membranes of all dopamin-
ergic neurons (DANs) and in a second color with
Long-range tracing of clustered neurons nc82 antibodies against Bruchpilot (Brp), a
in Drosophila and their stereotypy major structural and functional component of
Although millimeter-scale tissue sections pre- presynaptic active zones (AZs) (78, 79). Among
sent no problem for LLSM, the entire mouse brain the ~110 DANs within the image volume, we
is far too large, given the short working dis- focused our efforts on tracing the protocerebral
tances of commercially available high-resolution posterior medial 3 (PPM3) cluster of DANs that
objectives. The brain of the fruitfly D. melanogaster, project to the central complex, a key brain re-
on the other hand, fits comfortably within the gion essential for navigation, visual memory,
microscope, even in its 4× expanded form. sleep, and aggression (80–82). With manual
Furthermore, a vast array of genetic tools have annotation, we identified and traced all eight
been developed for Drosophila, such as split-
GAL4 drivers and MultiColor FlipOut (MCFO)
(17), which enable precise labeling of user-

Movie 5. Neuronal processes and myelina-

tion patterns across the mouse primary
visual cortex. Thy1-YFP–expressing neurons
across 1100 by 280 by 83 mm, immunostained Movie 6. Segmentation of pyramidal Movie 7. Tracing of DC3 olfactory projection
against myelin and Caspr, a marker of the nodes neurons in layer V of the mouse primary neurons (PNs) in an adult Drosophila brain.
of Ranvier, with specific emphasis on the visual cortex. Segmentation of two neurons, Volumetric view of three individually traced
neuronal processes and longitudinal myelination with specific emphasis on their branches neurons projecting from the antenna lobe in a
profile of a selected layer V pyramidal neuron and axonal myelination patterns (Fig. 4 and 5 bundle, with magnified views of their boutons at
(Figs. 4 and 5 and figs. S18 and S19). and figs. S18 and S19). the calyx and lateral horn (Fig. 6, A to E).

Gao et al., Science 363, eaau8302 (2019) 18 January 2019 10 of 16

R ES E A RC H | R E S EA R C H A R T I C LE

Downloaded from http://science.sciencemag.org/ on February 3, 2019

Fig. 7. Whole-brain analysis of presynaptic sites and DANs in (green bars) and nonDAN-associated nc82 puncta (red bars), and the
Drosophila. (A) MIP view of the subset of nc82 puncta marking presynaptic percentage of nc82 puncta that are DAN-associated (green curve), within
sites that are associated with DANs (DAN-assoc nc82), color coded by the each of the 33 brain regions of the adult Drosophila brain (fig. S30).
local puncta density, in an adult Drosophila brain (Movie 9). Scale bar, (E) MIP view of DANs and DAN-associated nc82 puncta, color coded by
100 mm. (Inset) (Right) MIP view of all nc82 puncta, using identical color 13 representative brain region (Movie 10). Scale bar, 100 mm. (Insets)
coding of local density. Scale bar, 100 mm. (B) Distribution of local Magnified views of the (top, angled view) PB and (bottom) EB. Brain
densities of (green) DAN-associated nc82 puncta and (orange) nonDAN- regions are ME, medulla; LOP, lobula plate; LO, lobula; OTU, optical
associated nc82 puncta in (A) (fig. S28). (C) Distribution of distances tubercle; VLPR, ventrolateral protocerebrum; LH, lateral horn; CA, calyx;
from DAN-associated nc82 puncta (green) and nonDAN-associated nc82 MB, mushroom body; ATL, antler; PB, protocerebral bridge; EB, ellipsoid
puncta (orange) to the nearest nc82 punctum of any kind, and nearest- body; FB, fan-shaped body; NO, noduli; LAL, lateral accessory lobe;
neighbor distances from one DAN-associated nc82 to another (magenta) and SP, superior protocerebrum. “L” and “R” indicate the left and right
(fig. S29). (D) Volumetric density of DAN-associated nc82 puncta hemispheres of the brain, respectively.

Gao et al., Science 363, eaau8302 (2019) 18 January 2019 11 of 16

R ES E A RC H | R E S EA R C H A R T I C LE

individual cells within the cluster (Fig. 6F, costained neighbor was consistent with their ments of the MB (Fig. 7D), perhaps reflecting
figs. S20 and S21, Movie 8, table S5, and movie mutual incorporation in a single AZ (fig. S23). the distinct computational needs of different
S4). Although tracing of fine processes inside In addition, we imaged another brain sample brain regions. The high density in the MB, for
the central complex was difficult, we were able of the output neuron from the a1 compartment example, is likely beneficial for increasing ca-
to trace the main axonal branches and precisely of the MB (MBON-a1) to validate the specificity pacity and sensory specificity of memory in as-
determine the number of cell types and the of nc82 antibody. We measured a 70-fold–higher sociative learning.
number of cells belonging to each cell type. surface density of nc82 puncta at the axons and When focusing on only those nc82 puncta
Within the PPM3 cluster, we found that two boutons of MBON-a1 than at its dendrites (fig. associated with DANs, we found additional dif-
cells (PPM3-EB) mainly projected to the ellip- S24 and supplementary note 6d), which is con- ferences. For example, the distance between non-
soid body (EB) (82); two cells (PPM3-FB3) pro- sistent with the near-absence of dendritic pre- DAN nc82 puncta and DAN-associated nc82
jected to layer 3 of the fan-shaped body (FB); synaptic densities observed for the same neuron puncta differed substantially between brain
two cells (PPM3-FB2-NO) projected to layer 2 with EM (83). Furthermore, we counted ~44,000 regions (fig. S29), indicating that the propor-
of the FB and noduli (NO); and two cells, which nc82 puncta in the a3 compartment (fig. S25), tion of synapses that can be modulated by do-
could be further categorized into two cell types compared with ~34,000 presynaptic densities in pamine may differ between brain regions. We
(PPM3-FB3-NO-a and PPM3-FB3-NO-b), projected the EM study (fig. S22 and supplementary note also found that the percentage of puncta asso-
to layer 3 of the FB and NO (Fig. 6G, figs. S20 6e). The distribution of distances between the ciated with DANs was approximately 10-fold
and S21, table S5, and supplementary note 6f). presynaptic densities was also similar in the two higher in the MB than in the optic lobes (Fig. 7D),
Using stochastic labeling of individual neurons cases (figs. S22B and S25B). which is consistent with dopamine-dependent
and split-GAL4 intersection, we were able to To see whether these differences were with- heterosynaptic plasticity being the basis of as-
identify and confirm the individual cell types in typical specimen variability, we imaged three sociative learning in the MB (83, 86, 87). On
we assigned (figs. S20 and S21, table S5, and additional wild-type females and counted between the other hand, the FB and the EB, which are

Downloaded from http://science.sciencemag.org/ on February 3, 2019

supplementary note 6f). ~34,000 and ~49,000 n82 puncta in the a3 known for visual and place memory formation
compartments of four MBs (fig. S26). Con- (88), exhibited surprisingly low DAN association,
Whole-brain analysis of presynaptic versely, for the two animals in which we studied whereas the protocerebral bridge (PB) and the
sites and DANs both a3 compartments (the original TH-GAL4 antler (ATL), which are not particularly known
We next turned our attention to the nc82 chan- specimen and the wild type), the number of for heterosynaptic plasticity, showed high DAN
nel of this specimen because recent EM mea- nc82 puncta in the left and right compart- association second only to the MB. Despite these
surements of the nearest-neighbor distances ments were within ~10% of one another. This differences, the variation in surface density of
between synapses in the a lobe of the MB (fig. suggests that the variability we observed be- nc82 puncta on DANs in different brain regions
S22) (83) suggest that quantitative counting of tween animals, including the EM result, is in- was considerably less pronounced (fig. S30B) be-
synapses across the Drosophila brain should deed natural and not due to errors from our cause the percentage volume occupied by DAN
be possible with ExLLSM at 4× expansion. How- counting methodology. in each domain (fig. S30D) followed similar trends
ever, to have confidence in the results, we needed Given confidence from these results, we then to the percentage of DAN-associated puncta (Fig.
to show that nc82 puncta larger than 100 nm extended our analysis across nearly the entire 7D). This could also be seen directly in volume
represented true AZs and not nonfunctional brain (the medial lobes of the MB were not renderings of the DANs and DAN-associated
Brp monomers or nonspecific background. To imaged because TH-GAL4 does not express puncta in each brain region (Fig. 7E and Movie
do so, we imaged two additional nc82-stained in the DANs in that region). In total, we counted 10), although local intradomain variations in
brains: one coimmunostained against V5-tagged ~40 million nc82 puncta, ~530,000 of them the spatial distribution of nc82 were also seen.
Brp and the other coimmunostained against localized at DANs (Fig. 7A and Movie 9), and
the AZ protein Syd1 (supplementary note 6c) calculated the brain-wide distribution of puncta Discussion
(84, 85). In both cases, the distribution of dis- density (Fig. 7B) and nearest-neighbor distances Thanks to its combination of high imaging
tances from each nc82 punctum to its nearest between any puncta or only DAN-associated speed, low photobleaching rate, and 3D nano-
ones (Fig. 7C). scale resolution, ExLLSM extends, by at least
We observed substantial differences when we 1000-fold in volume, the ability of SR fluores-
further subdivided our analysis into 33 major cence microscopy to generate detailed images
brain regions (fig. S28 to S30 and table S6). of subcellular ultrastructure. This fills a valua-
The volume density of all puncta, for example, ble niche between the high throughput of con-
varied from ~2 to 3 per cubic micrometer in ventional optical pipelines of neural anatomy
the lateral accessory lobe (LAL) and superior
protocerebrmm (SP) to ~6 to 8 in the compart-

Movie 8. Tracing and classification of PPM3

dopaminergic neurons (DANs) in an adult
Drosophila brain. Section of brain near the Movie 10. DANs and DAN-associated pre-
central complex with eight neurons from the synaptic sites in different brain regions of
protocerebral posterior medial 3 (PPM3) cluster Movie 9. Local density map of DAN-associated an adult Drosophila brain. Volume rendered
in the right hemisphere (colored) shown in presynaptic sites across an adult Drosophila DANs, DAN-associated nc82 puncta, and all
relation to surrounding DANs (white), and brain. Color-coded brain regions and 3D nc82 puncta across the entire brain, color coded
tracing of the individual neurons to their paired color-coded map of the local density of by brain region, followed by magnified 3D and
innervations in different regions of the central DAN-associated nc82 puncta in each domain orthoslice views of DANs and DAN-associated
complex (Fig. 6, F and G, and figs. S20 and S21). (Fig. 7, A to D, and figs. S28 to S30). nc82 in each of nine different domains (Fig. 7E).

Gao et al., Science 363, eaau8302 (2019) 18 January 2019 12 of 16

R ES E A RC H | R E S EA R C H A R T I C LE

(8, 9) and the ultrahigh resolution of correspond- wide isotropic expansion can be validated at ticolor labels for error checking (104). With such
ing EM pipelines (5, 70, 83). With genetically higher ratios with newer protocols of iterated a pipeline in place, 10 or more specimens might
targeted cell type–specific labeling (17, 89–91) expansion (29), ExM is still heir to the prob- be imaged in a single day at 4× to 10× expansion,
and protein-specific immunostaining, ExLLSM lems that bedevil other forms of high-resolution enabling statistically rich, brain-wide studies
enables sparse neural subsets and dense synap- fluorescence microscopy. Chief among these with protein-specific contrast and nanoscale
tic connections to be recorded, visualized, and is that because of the stochastic nature of label- resolution of neural development, sexual dimor-
quantified at ~60- by 60- by 90-nm resolution ing, the mean separation between fluorophores phism, degree of stereotypy, and structure/function
with ~100 person-hours of effort over cortex- must be ~5× to 10× smaller than the desired or structure/behavior correlations, particularly
spanning volumes in the mouse or brain-wide resolution in each dimension in order to dis- under genetic or pharmacological perturbation.
volumes in Drosophila. This compares with tinguish with high confidence two or more struc-
5 weeks to image and ~16,000 person-hours tures for which no a priori knowledge exists Materials and methods
to trace all neurons and count all synapses in a (102). We met this requirement at the level of Preparation of ExM samples
volume only 1/80th of a fly brain encompassing ~60- by 60- by 90-nm resolution in most cases Mouse, D. melanogaster, and human samples
the a lobe of the MB in a recent EM study at owing to the dense expression of cytosolic label were dissected, fixed, and immunostained fol-
8-nm isotropic resolution (83). The fluorescence in Thy1-YFP transgenic mice and DAN mem- lowing the protocols in supplementary note 1.
contrast of ExLLSM also raises the possibility of brane label in a TH-GAL4 transgenic fly, as well Sample genotypes and antibodies are summa-
correlating (92) fluorescence-based genetic in- as the exceptional specificity of Abs targeting rized in table S2. Unless otherwise noted, all
dicators of neural activity (93, 94) with neural MBP and nc82. Other Abs in our study did not samples were processed by using a protein-
ultrastructure over much larger volumes and meet this standard but were sufficient to iden- retention ExM (proExM) protocol with minor
without the labeling compromises common to tify organelles responsible for voids of cytosolic modifications (19, 105) or an expansion pathol-
correlative EM/fluorescence studies (95). label, mark Homer1 at synapses and Caspr at ogy (ExPath) protocol (98). Prepared ExM sam-

Downloaded from http://science.sciencemag.org/ on February 3, 2019

Although we have focused on the mouse cor- nodes of Ranvier, and measure statistical dis- ples were stored in 1× phosphate-buffered saline
tex and the Drosophila brain in this work, we tributions of synapse breadth and pre- and post- at 4°C and expanded in doubly deionized water
have also applied ExLLSM to image the mossy synaptic separation. However, immunostaining immediately before imaging with LLSM.
fiber innervation of granule cells in glomeruli in any form is probably not dense enough to
in the cerebellum of the mouse (fig. S31 and achieve true 3D resolution much beyond that Lattice light-sheet imaging
movie S5) as well as a complete human kidney already obtainable at 4× expansion, and the With the exception of Fig. 1, all ExM samples
glomerulus section (fig. S32). However, the ap- long distance between epitope and fluorophore, were imaged in objective scan mode (20) by
plication of ExM to any biological system must particularly with secondary Abs, further limits using a LLSM described previously (106), except
be examined on a case-by-case basis through resolution. Likewise, loss of FP fluorescence with adaptive optics capability disabled. The
careful controls and comparisons with known upon linking and digestion, as well as the slow ExM sample in the left column of Fig. 1 was
aspects (such as with EM) of the specific ultra- continued loss of fluorescence, which we alle- imaged by using a LLSM optimized for ExM,
structural elements under investigation. In par- viated here with a highly basic imaging buffer featuring a broader 160-mm FOV, a 1.5-mm scan
ticular, extrapolating the faithful nanoscale (supplementary note 2, c and d), probably pre- range, and software optimized for rapid sample
expansion of delicate membranous structures clude study at high resolution of many FP-linked scan acquisition (supplementary note 2a). All
and vesicles in a specimen from images of more proteins at the endogenous levels produced expanded samples were large compared with
robust components such as cytoskeletal ele- through genome editing. Indeed, even at 4× the LLS FOV and were therefore imaged in a
ments, clathrin-coated pits, or nuclear histones expansion, we rarely found sufficient residual series of overlapping 3D tiles that covered the
(18, 29, 96, 97) should be avoided. Elastic in- fluorescence to image targets labeled with red desired sample volume (supplementary note 2b).
homogeneity of the specimen after digestion, FPs of the Anthozoa family, despite reports to For imaging sessions of several hours or more,
such as from collagen-rich connective tissue or the contrary (19). focus was maintained through the periodic
adhesion to a rigid substrate, can also interfere Despite these challenges and limitations, the imaging of reference beads (supplementary
with expansion, although newer protocols with high speed and nanometric 3D resolution of note 2c). Raw data from each tile were deskewed
more aggressive digestion may help (98). In this ExLLSM make it an attractive tool for compar- (for sample scan mode), flat-fielded, deconvolved,
regard, brain tissue may represent a best case ative anatomical studies, particularly in the and stored for subsequent processing.
for ExM studies, owing to its comparatively Drosophila brain. For example, although we
homogenous mechanical properties and ready imaged the entire TH-GAL4/nc82 brain in Computing pipeline for flat-field
digestion. It should always be remembered that 62.5 hours (3.2 × 105 mm3/hour), with subsequent correction, stitching, and export of 3D
any image of a once-living specimen is an im- improvements in scanning geometry and field image tiles
perfect representation of that specimen, and the of view (FOV) we imaged mouse brain tissue in Because automatic tools for 3D stitching (107–111)
more steps that intrude in the process from one two colors at 4.0 × 106 mm3/hour. If transfer- do not scale to datasets with thousands of 3D im-
to the other the more imperfect it becomes. rable to the fly, this would allow whole-brain age tiles, we developed a scalable high-performance
Overexpression, chemical fixation, permeabili- imaging in ~5.0 hours. This limit is not fun- computing (HPC) pipeline to robustly flat-field
zation, and immunostaining already introduce damental; with simultaneous multicolor imaging correct, deconvolve, and assemble 3D image tiles
numerous structural artifacts (99–101) in all and multiple cameras to cover even broader into the final volume (supplementary note 3).
forms of high-resolution fluorescence micros- FOVs, rates up to ~108 mm3/hour may be achie- First, we extended and parallelized CIDRE (107)
copy, including ExM, but ExM also requires ad- vable, or ~12 min/fly brain at 4× expansion. for 3D volumes to calculate 3D flat fields (figs.
ditional steps of polymer infusion, gelation, label Assuming the future development of (i) robust, S5 and S6). We then corrected the raw image
attachment, digestion, expansion, and handling isotropic expansion at 10× or greater; (ii) longer tiles using these flat-fields and deconvolved each.
that can perturb ultrastructure even more. Care- working distance high NA water immersion Next, we parallelized the globally optimizing 3D
ful controls are essential. objectives or lossless sectioning (103) of ex- stitching method (108) to automatically stitch the
At 4× expansion, the resolution of ExLLSM panded samples; and (iii) a ubiquitous, dense, thousands of raw image tiles, without manual
is close, but not quite sufficient, to trace fine, and cell-permeable fluorescent membrane stain intervention, in an iteratively refined prediction
highly innervated neuronal processes—such as analogous to heavy-metal stains in EM, even model that corrects for systematic stage coordi-
the PPM3 cluster, which terminates in the central densely innervated circuits might be traced, nate errors (fig. S7). Last, we exported the stitched
complex—and would therefore benefit from high- particularly when imaged in conjunction with datasets using the flat-field–corrected and decon-
er expansion ratios. However, even if specimen- cell type–specific or stochastically expressed mul- volved image tiles as multiresolution hierarchies

Gao et al., Science 363, eaau8302 (2019) 18 January 2019 13 of 16

R ES E A RC H | R E S EA R C H A R T I C LE

into a custom file format (N5) (21) that enabled 20. B. C. Chen et al., Lattice light-sheet microscopy: Imaging 44. L.-J. Oluich et al., Targeted ablation of oligodendrocytes
parallel blockwise export and compression on a molecules to embryos at high spatiotemporal resolution. induces axonal pathology independent of overt
Science 346, 1257998 (2014). doi: 10.1126/science.1257998; demyelination. J. Neurosci. 32, 8317–8330 (2012).
HPC cluster. Bindings for N5 format for the pmid: 25342811 doi: 10.1523/JNEUROSCI.1053-12.2012; pmid: 22699912
ImgLib2 library (112) are provided for the ImageJ 21. I. Pisarev, S. Saalfeld, Stitcher and N5 viewer; https://github. 45. M. Zonouzi et al., GABAergic regulation of cerebellar NG2 cell
distribution Fiji (113). For interactive visualiza- com/saalfeldlab/stitching-spark, https://github.com/ development is altered in perinatal white matter injury.
tion, we developed a BigDataViewer-based viewer saalfeldlab/n5-viewer. Nat. Neurosci. 18, 674–682 (2015). doi: 10.1038/nn.3990;
22. J. Tønnesen, V. V. G. K. Inavalli, U. V. Nägerl, Super- pmid: 25821912
plugin (114) including a crop and export tool resolution imaging of the extracellular space in living brain 46. O. O. Glebov, S. Cox, L. Humphreys, J. Burrone, Neuronal
to make arbitrary subvolumes available in legacy tissue. Cell 172, 1108–1121.e15 (2018). doi: 10.1016/ activity controls transsynaptic geometry. Sci. Rep. 6, 22703
formats such as TIFF image series. j.cell.2018.02.007; pmid: 29474910 (2016). doi: 10.1038/srep22703; pmid: 26951792
23. L. Freifeld et al., Expansion microscopy of zebrafish for 47. A. Dani, B. Huang, J. Bergan, C. Dulac, X. Zhuang,
neuroscience and developmental biology studies. Proc. Natl. Superresolution imaging of chemical synapses in the brain.
RE FE RENCES AND N OT ES Acad. Sci. U.S.A. 114, E10799–E10808 (2017). doi: 10.1073/ Neuron 68, 843–856 (2010). doi: 10.1016/
pnas.1706281114; pmid: 29162696 j.neuron.2010.11.021; pmid: 21144999
1. S. Herculano-Houzel, The human brain in numbers: A linearly
24. T. J. Mosca, D. J. Luginbuhl, I. E. Wang, L. Luo, Presynaptic 48. N. L. Rochefort, A. Konnerth, Dendritic spines: From
scaled-up primate brain. Front. Hum. Neurosci. 3, 31 (2009).
LRP4 promotes synapse number and function of excitatory structure to in vivo function. EMBO Rep. 13, 699–708 (2012).
doi: 10.3389/neuro.09.031.2009; pmid: 19915731
CNS neurons. eLife 6, e27347 (2017). doi: 10.7554/ doi: 10.1038/embor.2012.102; pmid: 22791026
2. K. Sharma et al., Cell type- and brain region-resolved mouse
eLife.27347; pmid: 28606304 49. H. Hering, M. Sheng, Dendritic spines: Structure, dynamics
brain proteome. Nat. Neurosci. 18, 1819–1831 (2015).
25. C. K. Cahoon et al., Superresolution expansion microscopy and regulation. Nat. Rev. Neurosci. 2, 880–888 (2001).
doi: 10.1038/nn.4160; pmid: 26523646
reveals the three-dimensional organization of the Drosophila doi: 10.1038/35104061; pmid: 11733795
3. J. E. Heuser, T. S. Reese, Evidence for recycling of synaptic
synaptonemal complex. Proc. Natl. Acad. Sci. U.S.A. 114, 50. E. A. Nimchinsky, B. L. Sabatini, K. Svoboda, Structure and
vesicle membrane during transmitter release at the frog
E6857–E6866 (2017). doi: 10.1073/pnas.1705623114; function of dendritic spines. Annu. Rev. Physiol. 64, 313–353
neuromuscular junction. J. Cell Biol. 57, 315–344 (1973).
pmid: 28760978 (2002). doi: 10.1146/annurev.physiol.64.081501.160008;
doi: 10.1083/jcb.57.2.315; pmid: 4348786
26. A. Tsai et al., Nuclear microenvironments modulate pmid: 11826272
4. C. S. Xu et al., Enhanced FIB-SEM systems for large-volume
transcription from low-affinity enhancers. eLife 6, e28975 51. M. Segal, Dendritic spines and long-term plasticity. Nat. Rev.
3D imaging. eLife 6, e25916 (2017). doi: 10.7554/eLife.25916;

Downloaded from http://science.sciencemag.org/ on February 3, 2019

(2017). doi: 10.7554/eLife.28975; pmid: 29095143 Neurosci. 6, 277–284 (2005). doi: 10.1038/nrn1649;
pmid: 28500755
27. N. Jiang et al., Superresolution imaging of Drosophila pmid: 15803159
5. Z. Zheng et al., A complete electron microscopy volume of
tissues using expansion microscopy. Mol. Biol. Cell 29, 52. S. Konur, D. Rabinowitz, V. L. Fenstermaker, R. Yuste,
the brain of adult Drosophila melanogaster. Cell 174,
1413–1421 (2018). doi: 10.1091/mbc.E17-10-0583; Systematic regulation of spine sizes and densities in
730–743.e22 (2018). doi: 10.1016/j.cell.2018.06.019;
pmid: 29688792 pyramidal neurons. J. Neurobiol. 56, 95–112 (2003).
pmid: 30033368
28. F. Guo, M. Holla, M. M. Diaz, M. Rosbash, A circadian output doi: 10.1002/neu.10229; pmid: 12838576
6. K. D. Micheva, S. J. Smith, Array tomography: A new tool for
circuit controls sleep-wake arousal threshold in Drosophila. 53. D. Dumitriu, A. Rodriguez, J. H. Morrison, High-throughput,
imaging the molecular architecture and ultrastructure of
bioRxiv (2018). doi: 10.1101/298067 detailed, cell-specific neuroanatomy of dendritic spines using
neural circuits. Neuron 55, 25–36 (2007). doi: 10.1016/
29. J. B. Chang et al., Iterative expansion microscopy. microinjection and confocal microscopy. Nat. Protoc. 6,
j.neuron.2007.06.014; pmid: 17610815
Nat. Methods 14, 593–599 (2017). doi: 10.1038/nmeth.4261; 1391–1411 (2011). doi: 10.1038/nprot.2011.389;
7. J.-C. Rah et al., Thalamocortical input onto layer 5 pyramidal
pmid: 28417997 pmid: 21886104
neurons measured using quantitative large-scale array
30. C. J. L. Sheppard, Super resolution in confocal imaging. 54. M. Jiang et al., Dendritic arborization and spine dynamics are
tomography. Front. Neural Circuits 7, 177 (2013).
Optik (Stuttg.) 80, 53–54 (1988). abnormal in the mouse model of MECP2 duplication
doi: 10.3389/fncir.2013.00177; pmid: 24273494
31. C. B. Müller, J. Enderlein, Image scanning microscopy. syndrome. J. Neurosci. 33, 19518–19533 (2013).
8. A. S. Chiang et al., Three-dimensional reconstruction of brain-
Phys. Rev. Lett. 104, 198101 (2010). doi: 10.1103/ doi: 10.1523/JNEUROSCI.1745-13.2013; pmid: 24336718
wide wiring networks in Drosophila at single-cell resolution.
PhysRevLett.104.198101; pmid: 20867000 55. X. Yu, Y. Zuo, Two-photon in vivo imaging of dendritic spines
Curr. Biol. 21, 1–11 (2011). doi: 10.1016/j.cub.2010.11.056;
32. X.-T. Cheng et al., Characterization of LAMP1-labeled in the mouse cortex using a thinned-skull preparation.
pmid: 21129968
nondegradative lysosomal and endocytic compartments in J. Vis. Exp. 87, e51520 (2014). pmid: 24894563
9. A. Jenett et al., A GAL4-driver line resource for Drosophila
neurons. J. Cell Biol. 217, 3127–3139 (2018). doi: 10.1083/ 56. J. I. Arellano, R. Benavides-Piccione, J. Defelipe, R. Yuste,
neurobiology. Cell Reports 2, 991–1001 (2012). doi: 10.1016/
jcb.201711083; pmid: 29695488 Ultrastructure of dendritic spines: Correlation between
j.celrep.2012.09.011; pmid: 23063364
33. N. Kasthuri et al., Saturated reconstruction of a volume of synaptic and spine morphologies. Front. Neurosci. 1, 131–143
10. M. N. Economo et al., A platform for brain-wide imaging and
neocortex. Cell 162, 648–661 (2015). doi: 10.1016/ (2007). doi: 10.3389/neuro.01.1.1.010.2007; pmid: 18982124
reconstruction of individual neurons. eLife 5, e10566 (2016).
j.cell.2015.06.054; pmid: 26232230 57. C. Bosch et al., FIB/SEM technology and high-throughput 3D
doi: 10.7554/eLife.10566; pmid: 26796534
34. Q. A. Liu, H. Shio, Mitochondrial morphogenesis, dendrite reconstruction of dendritic spines and synapses in GFP-
11. S. Shah, E. Lubeck, W. Zhou, L. Cai, In situ transcription
development, and synapse formation in cerebellum require labeled adult-generated neurons. Front. Neuroanat. 9, 60
profiling of single cells reveals spatial organization of cells in
both Bcl-w and the glutamate receptor d2. PLOS Genet. 4, (2015). doi: 10.3389/fnana.2015.00060; pmid: 26052271
the mouse hippocampus. Neuron 92, 342–357 (2016).
e1000097 (2008). doi: 10.1371/journal.pgen.1000097; 58. K. Takasaki, B. L. Sabatini, Super-resolution 2-photon
doi: 10.1016/j.neuron.2016.10.001; pmid: 27764670
pmid: 18551174 microscopy reveals that the morphology of each dendritic
12. J. R. Moffitt et al., High-throughput single-cell gene-
35. W. Kuehnel, Color Atlas of Cytology, Histology, and spine correlates with diffusive but not synaptic properties.
expression profiling with multiplexed error-robust
Microscopic Anatomy (Thieme Flexibook, ed. 4, 2003). Front. Neuroanat. 8, 29 (2014). doi: 10.3389/
fluorescence in situ hybridization. Proc. Natl. Acad. Sci. U.S.A.
36. V. Popov, N. I. Medvedev, H. A. Davies, M. G. Stewart, Mitochondria fnana.2014.00029; pmid: 24847215
113, 11046–11051 (2016). doi: 10.1073/pnas.1612826113;
form a filamentous reticular network in hippocampal dendrites 59. J. Tønnesen, G. Katona, B. Rózsa, U. V. Nägerl, Spine neck
pmid: 27625426
but are present as discrete bodies in axons: A three-dimensional plasticity regulates compartmentalization of synapses.
13. J. Tønnesen, U. V. Nägerl, Superresolution imaging for
ultrastructural study. J. Comp. Neurol. 492, 50–65 (2005). Nat. Neurosci. 17, 678–685 (2014). doi: 10.1038/nn.3682;
neuroscience. Exp. Neurol. 242, 33–40 (2013). doi: 10.1016/
doi: 10.1002/cne.20682; pmid: 16175555 pmid: 24657968
j.expneurol.2012.10.004; pmid: 23063602
14. H. Zhong, Applying superresolution localization-based 37. M. R. Duchen, Mitochondria in health and disease: Perspectives 60. D. L. Dickstein et al., Automatic dendritic spine quantification
microscopy to neurons. Synapse 69, 283–294 (2015). on a new mitochondrial biology. Mol. Aspects Med. 25, 365–451 from confocal data with neurolucida 360. Curr. Protoc.
doi: 10.1002/syn.21806; pmid: 25648102 (2004). doi: 10.1016/j.mam.2004.03.001; pmid: 15302203 Neurosci. 77, 1, 21 (2016). doi: 10.1002/cpns.16;
15. C. I. Bargmann, Beyond the connectome: How 38. S. G. Waxman, M. V. I. Bennett, Relative conduction velocities pmid: 27696360
neuromodulators shape neural circuits. BioEssays 34, of small myelinated and non-myelinated fibres in the central 61. E. G. Jones, T. P. S. Powell, Morphological variations in the
458–465 (2012). doi: 10.1002/bies.201100185; nervous system. Nat. New Biol. 238, 217–219 (1972). dendritic spines of the neocortex. J. Cell Sci. 5, 509–529
pmid: 22396302 doi: 10.1038/newbio238217a0; pmid: 4506206 (1969). pmid: 5362339
16. J. Lu, J. C. Tapia, O. L. White, J. W. Lichtman, The 39. K. A. Nave, Myelination and support of axonal integrity by 62. J. Grutzendler, N. Kasthuri, W. B. Gan, Long-term dendritic
interscutularis muscle connectome. PLOS Biol. 7, e32 glia. Nature 468, 244–252 (2010). doi: 10.1038/nature09614; spine stability in the adult cortex. Nature 420, 812–816
(2009). pmid: 19209956 pmid: 21068833 (2002). doi: 10.1038/nature01276; pmid: 12490949
17. A. Nern, B. D. Pfeiffer, G. M. Rubin, Optimized tools for 40. J. J. Harris, D. Attwell, The energetics of CNS white matter. 63. L. Anton-Sanchez et al., Three-dimensional distribution of
multicolor stochastic labeling reveal diverse stereotyped cell J. Neurosci. 32, 356–371 (2012). doi: 10.1523/ cortical synapses: A replicated point pattern-based analysis.
arrangements in the fly visual system. Proc. Natl. Acad. JNEUROSCI.3430-11.2012; pmid: 22219296 Front. Neuroanat. 8, 85 (2014). doi: 10.3389/
Sci. U.S.A. 112, E2967–E2976 (2015). doi: 10.1073/ 41. A. Compston, A. Coles, Multiple sclerosis. Lancet 372, fnana.2014.00085; pmid: 25206325
pnas.1506763112; pmid: 25964354 1502–1517 (2008). doi: 10.1016/S0140-6736(08)61620-7; 64. J. DeFelipe, L. Alonso-Nanclares, J. I. Arellano, Microstructure
18. F. Chen, P. W. Tillberg, E. S. Boyden, Expansion microscopy. pmid: 18970977 of the neocortex: Comparative aspects. J. Neurocytol. 31,
Science 347, 543–548 (2015). doi: 10.1126/science.1260088; 42. W. A. H. Rushton, A theory of the effects of fibre size in 299–316 (2002). doi: 10.1023/A:1024130211265;
pmid: 25592419 medullated nerve. J. Physiol. 115, 101–122 (1951). pmid: 12815249
19. P. W. Tillberg et al., Protein-retention expansion microscopy doi: 10.1113/jphysiol.1951.sp004655; pmid: 14889433 65. C. Sala et al., Regulation of dendritic spine morphology and
of cells and tissues labeled using standard fluorescent 43. R. La Marca et al., TACE (ADAM17) inhibits Schwann cell synaptic function by Shank and Homer. Neuron 31, 115–130
proteins and antibodies. Nat. Biotechnol. 34, 987–992 myelination. Nat. Neurosci. 14, 857–865 (2011). doi: 10.1038/ (2001). doi: 10.1016/S0896-6273(01)00339-7;
(2016). doi: 10.1038/nbt.3625; pmid: 27376584 nn.2849; pmid: 21666671 pmid: 11498055

Gao et al., Science 363, eaau8302 (2019) 18 January 2019 14 of 16

R ES E A RC H | R E S EA R C H A R T I C LE

66. U. Thomas, Modulation of synaptic signalling complexes by 88. L. Kahsai, T. Zars, Learning and memory in Drosophila: 111. A. Bria, G. Iannello, TeraStitcher—A tool for fast automatic
Homer proteins. J. Neurochem. 81, 407–413 (2002). Behavior, genetics, and neural systems. Int. Rev. Neurobiol. 3D-stitching of teravoxel-sized microscopy images. BMC
doi: 10.1046/j.1471-4159.2002.00869.x; pmid: 12065649 99, 139–167 (2011). doi: 10.1016/B978-0-12-387003- Bioinformatics 13, 316 (2012). doi: 10.1186/1471-2105-13-316;
67. A. Dosemeci, R. J. Weinberg, T. S. Reese, J.-H. Tao-Cheng, 2.00006-9; pmid: 21906539 pmid: 23181553
The postsynaptic density: There is more than meets the eye. 89. H. Luan, N. C. Peabody, C. R. Vinson, B. H. White, Refined 112. T. Pietzsch, S. Preibisch, P. Tomancák, S. Saalfeld, ImgLib2—
Front. Synaptic Neurosci. 8, 23 (2016). doi: 10.3389/ spatial manipulation of neuronal function by combinatorial Generic image processing in Java. Bioinformatics 28,
fnsyn.2016.00023; pmid: 27594834 restriction of transgene expression. Neuron 52, 425–436 3009–3011 (2012). doi: 10.1093/bioinformatics/bts543;
68. G. H. Diering et al., Homer1a drives homeostatic scaling-down (2006). doi: 10.1016/j.neuron.2006.08.028; pmid: 17088209 pmid: 22962343
of excitatory synapses during sleep. Science 355, 511–515 90. B. D. Pfeiffer et al., Refinement of tools for targeted gene 113. J. Schindelin et al., Fiji: An open-source platform for
(2017). doi: 10.1126/science.aai8355; pmid: 28154077 expression in Drosophila. Genetics 186, 735–755 (2010). biological-image analysis. Nat. Methods 9, 676–682 (2012).
69. D. Debanne, E. Campanac, A. Bialowas, E. Carlier, G. Alcaraz, doi: 10.1534/genetics.110.119917; pmid: 20697123 doi: 10.1038/nmeth.2019; pmid: 22743772
Axon physiology. Physiol. Rev. 91, 555–602 (2011). 91. M. J. Dolan et al., Facilitating neuron-specific genetic 114. T. Pietzsch, S. Saalfeld, S. Preibisch, P. Tomancak,
doi: 10.1152/physrev.00048.2009; pmid: 21527732 manipulations in Drosophila melanogaster using a split GAL4 BigDataViewer: Visualization and processing for large image
70. G. S. Tomassy et al., Distinct profiles of myelin distribution repressor. Genetics 206, 775–784 (2017). doi: 10.1534/ data sets. Nat. Methods 12, 481–483 (2015). doi: 10.1038/
along single axons of pyramidal neurons in the neocortex. genetics.116.199687; pmid: 28363977 nmeth.3392; pmid: 26020499
Science 344, 319–324 (2014). doi: 10.1126/science.1249766; 92. D. D. Bock et al., Network anatomy and in vivo physiology of
pmid: 24744380 visual cortical neurons. Nature 471, 177–182 (2011). AC KNOWLED GME NTS
71. S. Einheber et al., The axonal membrane protein Caspr, a doi: 10.1038/nature09802; pmid: 21390124
We thank D. Bock, K. Svoboda, N. Ji, N. Spruston, L. Scheffer,
homologue of neurexin IV, is a component of the septate-like 93. T.-W. Chen et al., Ultrasensitive fluorescent proteins for
E. Snapp, P. Tillberg, L. Lavis, E. Bloss, W. Legant, D. Hoffman, and
paranodal junctions that assemble during myelination. J. Cell imaging neuronal activity. Nature 499, 295–300 (2013).
K. Hayworth at Howard Hughes Medical Institute (HHMI) Janelia
Biol. 139, 1495–1506 (1997). doi: 10.1083/jcb.139.6.1495; doi: 10.1038/nature12354; pmid: 23868258
Research Campus (JRC) and B. Sabatini and D. Van Vactor at
pmid: 9396755 94. B. F. Fosque et al., Neural circuits. Labeling of active neural
Harvard Medical School (HMS) for invaluable discussions and
72. C. Porrero, P. Rubio-Garrido, C. Avendaño, F. Clascá, circuits in vivo with designed calcium integrators. Science
comments. We also thank K. Schaefer, T. Wolff, C.-L. Chang, and
Mapping of fluorescent protein-expressing neurons and axon 347, 755–760 (2015). doi: 10.1126/science.1260922;
H. Choi at JRC for help with sample preparation and imaging. We
pathways in adult and developing Thy1-eYFP-H transgenic pmid: 25678659
gratefully acknowledge the shared resources and project teams at
mice. Brain Res. 1345, 59–72 (2010). doi: 10.1016/ 95. P. de Boer, J. P. Hoogenboom, B. N. G. Giepmans, Correlated
JRC, including D. Alcor, J. Heddleston, and A. Taylor of the

Downloaded from http://science.sciencemag.org/ on February 3, 2019

j.brainres.2010.05.061; pmid: 20510892 light and electron microscopy: Ultrastructure lights up! Nat.
Advanced Imaging Center and Light Microscopy Facility for help
73. S. Ramaswamy, H. Markram, Anatomy and physiology of the Methods 12, 503–513 (2015). doi: 10.1038/nmeth.3400;
with imaging; I. Negrashov and jET for manufacturing expertise;
thick-tufted layer 5 pyramidal neuron. Front. Cell. Neurosci. 9, pmid: 26020503
and O. Malkesman, K. Salvesen, C. Christoforou, G. Meissner, and
233 (2015). doi: 10.3389/fncel.2015.00233; pmid: 26167146 96. T. J. Chozinski et al., Expansion microscopy with conventional
the FlyLight project team for sample handling and preparation.
74. L. M. Palmer, G. J. Stuart, Site of action potential initiation in antibodies and fluorescent proteins. Nat. Methods 13,
Last, we are grateful to C. Pama and R. Karadottir at the University
layer 5 pyramidal neurons. J. Neurosci. 26, 1854–1863 485–488 (2016). doi: 10.1038/nmeth.3833; pmid: 27064647
of Cambridge; J. Melander and H. Zhong at OHSU; T. Herman at
(2006). doi: 10.1523/JNEUROSCI.4812-05.2006; 97. T. Ku et al., Multiplexed and scalable super-resolution
the University of Oregon; and E. Karagiannis, J.-S. Kang, and
pmid: 16467534 imaging of three-dimensional protein localization in size-
F. Chen at MIT for help with sample preparation and H. Otsuna,
75. S. J. C. Caron, V. Ruta, L. F. Abbott, R. Axel, Random adjustable tissues. Nat. Biotechnol. 34, 973–981 (2016).
T. Kawase, and E. Bas at JRC; C. Wietholt at FEI Amira;
convergence of olfactory inputs in the Drosophila mushroom doi: 10.1038/nbt.3641; pmid: 27454740
M. Gastinger at Bitplane; and J. McMullen and T. Tetreault at MBF
body. Nature 497, 113–117 (2013). doi: 10.1038/nature12063; 98. Y. Zhao et al., Nanoscale imaging of clinical specimens using
Bioscience for data analysis and visualization. Funding: I.P., D.E.M.,
pmid: 23615618 pathology-optimized expansion microscopy. Nat. Biotechnol.
T.-L.L., V.S., A.G., J.B., J.C., C.M.O., J.L.-S., A.H., G.M.R., S.S.,
76. N. J. Butcher, A. B. Friedrich, Z. Lu, H. Tanimoto, 35, 757–764 (2017). doi: 10.1038/nbt.3892; pmid: 28714966
Y.A., and E.B. are funded by HHMI. E.S.B. acknowledges, for
I. A. Meinertzhagen, Different classes of input and output 99. U. Schnell, F. Dijk, K. A. Sjollema, B. N. G. Giepmans,
funding, John Doerr, the Open Philanthropy Project, NIH
neurons reveal new features in microglomeruli of the adult Immunolabeling artifacts and the need for live-cell imaging.
1R01NS087950, NIH 1RM1HG008525, NIH 1R01DA045549,
Drosophila mushroom body calyx. J. Comp. Neurol. 520, Nat. Methods 9, 152–158 (2012). doi: 10.1038/nmeth.1855;
NIH 2R01DA029639, NIH 1R01NS102727, NIH 1R41MH112318,
2185–2201 (2012). doi: 10.1002/cne.23037; pmid: 22237598 pmid: 22290187
NIH 1R01EB024261, NIH 1R01MH110932, the HHMI–Simons
77. K. Eichler et al., The complete connectome of a learning and 100. D. R. Whelan, T. D. M. Bell, Image artifacts in single molecule
Faculty Scholars Program, IARPA D16PC00008, U.S. Army
memory centre in an insect brain. Nature 548, 175–182 localization microscopy: Why optimization of sample
Research Laboratory and the U. S. Army Research Office under
(2017). doi: 10.1038/nature23455; pmid: 28796202 preparation protocols matters. Sci. Rep. 5, 7924 (2015).
contract/grant W911NF1510548, U.S.–Israel Binational Science
78. W. Fouquet et al., Maturation of active zone assembly by doi: 10.1038/srep07924; pmid: 25603780
Foundation Grant 2014509, and NIH Director’s Pioneer Award
Drosophila Bruchpilot. J. Cell Biol. 186, 129–145 (2009). 101. D. Li et al., ADVANCED IMAGING. Extended-resolution
1DP1NS087724. S.U. and T.K. are funded by grants from Biogen,
doi: 10.1083/jcb.200812150; pmid: 19596851 structured illumination imaging of endocytic and cytoskeletal
Ionis Pharmaceuticals, and NIH grant R01GM075252 (to T.K.).
79. N. Ehmann et al., Quantitative super-resolution imaging of dynamics. Science 349, aab3500 (2015). doi: 10.1126/
S.U. gratefully acknowledges the Fellows program of the Image and
Bruchpilot distinguishes active zone states. Nat. Commun. 5, science.aab3500; pmid: 26315442
Data Analysis Core at Harvard Medical School and the MATLAB
4650 (2014). doi: 10.1038/ncomms5650; pmid: 25130366 102. W. R. Legant et al., High-density three-dimensional localization
code repository received from the Computational Image Analysis
80. Z. Mao, R. L. Davis, Eight different types of dopaminergic microscopy across large volumes. Nat. Methods 13, 359–365
Workshop, supported by NIH grant GM103792. K.R.M. and S.G.M.
neurons innervate the Drosophila mushroom body neuropil: (2016). doi: 10.1038/nmeth.3797; pmid: 26950745
are funded by NIH grant R01DC015478. S.T. and A.R. are funded
Anatomical and physiological heterogeneity. Front. Neural 103. K. J. Hayworth et al., Ultrastructurally smooth thick
by NIH grant R44MH093011. Author contributions: E.B., E.S.B.,
Circuits 3, 5 (2009). doi: 10.3389/neuro.04.005.2009; partitioning and volume stitching for large-scale
and R.G. supervised the project and wrote the manuscript with
pmid: 19597562 connectomics. Nat. Methods 12, 319–322 (2015).
input from all coauthors. T.-L.L. and J.C. built the microscopes with
81. E. C. Kong et al., A pair of dopamine neurons target the doi: 10.1038/nmeth.3292; pmid: 25686390
input from E.B., D.E.M., and jET (JRC) and performed all
D1-like dopamine receptor DopR in the central complex to 104. Y.-G. Yoon et al., Feasibility of 3D reconstruction of neural microscope characterization experiments. D.E.M. created the
promote ethanol-stimulated locomotion in Drosophila. morphology using expansion microscopy and barcode-guided instrument control software. R.G., S.M.A., T.-L.L., V.S., J.C., and
PLOS ONE 5, e9954 (2010). doi: 10.1371/ agglomeration. Front. Comput. Neurosci. 11, 97 (2017). C.M.O. acquired all biological data with coauthors. G.H.H. provided
journal.pone.0009954; pmid: 20376353 doi: 10.3389/fncom.2017.00097; pmid: 29114215 the Thy1-YFP mice, and R.G. and S.M.A. prepared the ExM
82. O. V. Alekseyenko et al., Single serotonergic neurons that 105. S. M. Asano et al., Expansion microscopy: Protocols for samples. A.G. and A.H. provided the Slc17a7-cre X TCGO mice and
modulate aggression in Drosophila. Curr. Biol. 24, 2700–2707 imaging proteins and RNA in cells and tissues. Curr. Protoc. prepared the ExM samples. Y.Z. provided the human kidney
(2014). doi: 10.1016/j.cub.2014.09.051; pmid: 25447998 Cell Biol. 80, e56 (2018). doi: 10.1002/cpcb.56; sections and prepared the ExPath samples. Y.A. and G.M.R.
83. S. Y. Takemura et al., A connectome of a learning and pmid: 30070431 created the split-GAL4 fly strains, Y.A. optimized the IHC
memory center in the adult Drosophila brain. eLife 6, e26975 106. T. L. Liu et al., Observing the cell in its native state: Imaging conditions, and R.G. and S.M.A prepared the ExM samples. K.R.M.,
(2017). doi: 10.7554/eLife.26975; pmid: 28718765 subcellular dynamics in multicellular organisms. Science 360, S.G.M., S.M.A., and T.-L.L. provided the initial stitching software
84. D. Owald et al., A Syd-1 homologue regulates pre- and eaaq1392 (2018). doi: 10.1126/science.aaq1392; packages. I.P. and S.S. created the automated flat-field, stitching,
postsynaptic maturation in Drosophila. J. Cell Biol. 188, pmid: 29674564 and N5 visualization pipeline, and I.P., R.G., J.B., and S.S.
565–579 (2010). doi: 10.1083/jcb.200908055; 107. K. Smith et al., CIDRE: An illumination-correction method for deconvolved, flat-fielded, and stitched all image data. Y.A.
pmid: 20176924 optical microscopy. Nat. Methods 12, 404–406 (2015). performed segmentation and tracing and supervised analyses of all
85. S. Holbrook, J. K. Finley, E. L. Lyons, T. G. Herman, Loss of doi: 10.1038/nmeth.3323; pmid: 25775044 fly image data. C.Z., S.T., and A.R. provided customized
syd-1 from R7 neurons disrupts two distinct phases of 108. S. Preibisch, S. Saalfeld, P. Tomancak, Globally optimal commercial software packages and helped with segmentation,
presynaptic development. J. Neurosci. 32, 18101–18111 stitching of tiled 3D microscopic image acquisitions. tracing and reconstruction of neurites and dendritic spines using
(2012). doi: 10.1523/JNEUROSCI.1350-12.2012; Bioinformatics 25, 1463–1465 (2009). doi: 10.1093/ these packages. S.-H.S. and H.A.P. designed experimental
pmid: 23238725 bioinformatics/btp184; pmid: 19346324 protocols and performed sample preparation for FIB-SEM. S.P.,
86. Y. Aso et al., The neuronal architecture of the mushroom 109. D. Hörl et al., BigStitcher: Reconstructing high-resolution C.S.X., and H.H. performed FIB-SEM sample preparation,
body provides a logic for associative learning. eLife 3, e04577 image datasets of cleared and expanded samples. bioRxiv image acquisition, and data processing. J.L.-S. and T.K. provided
(2014). doi: 10.7554/eLife.04577; pmid: 25535793 (2018). doi: 10.1101/343954 essential discussion on the subcellular ultrastructure analysis
87. Y. Aso, G. M. Rubin, Dopaminergic neurons write and update 110. M. Emmenlauer et al., XuvTools: Free, fast and reliable stitching and access to instrumentation and computational resources. S.U.
memories with cell-type-specific rules. eLife 5, e16135 of large 3D datasets. J. Microsc. 233, 42–60 (2009). and R.G. processed and performed quantitative analysis of all
(2016). doi: 10.7554/eLife.16135; pmid: 27441388 doi: 10.1111/j.1365-2818.2008.03094.x; pmid: 19196411 image data. S.U., R.G., E.B., and Y.A. produced all figures and

Gao et al., Science 363, eaau8302 (2019) 18 January 2019 15 of 16

R ES E A RC H | R E S EA R C H A R T I C LE

movies. Competing interests: Portions of the technology patents related to ExM. Data and materials availability: All SUPPLEMENTARY MATERIALS
described here are covered by U.S. patent 7,894,136 issued to E.B., data needed to evaluate the conclusions in the paper are present in www.sciencemag.org/content/363/6424/eaau8302/suppl/DC1
assigned to Lattice Light of Ashburn, VA, and licensed to Carl Zeiss the paper or the supplementary materials. The compressed size Supplementary Text
Microscopy; U.S. patents 8,711,211 and 9,477,074 issued to E.B., of the datasets used in generating the figures and movies Figs. S1 to S33
assigned to HHMI, and licensed to Carl Zeiss Microscopy; U.S. exceeds 100 terabytes, and it is therefore not practical to upload Table S1 to S6
patent application 13/844,405 filed by E.B. and assigned to HHMI; to a public data repository. All data used in this paper will be References (115–133)
and U.S. patent 9,500,846 issued to E.B. and assigned to HHMI. made freely available to those who request and provide a Movies S1 to S7
E.S.B. is a co-inventor on multiple patents related to ExM and is mechanism for feasible data transfers (such as physical hard disk
also a cofounder of a company that aims to provide kits and drives or cloud storage). Documentation for construction of a
services relating to ExM to the public. R.G. is a co-inventor on LLSM can be obtained by execution of a research license 19 July 2018; accepted 30 November 2018
multiple patents related to ExM. Y.Z. is a co-inventor on multiple agreement with HHMI. 10.1126/science.aau8302

Downloaded from http://science.sciencemag.org/ on February 3, 2019

Gao et al., Science 363, eaau8302 (2019) 18 January 2019 16 of 16

Cortical column and whole-brain imaging with molecular contrast and nanoscale resolution
Ruixuan Gao, Shoh M. Asano, Srigokul Upadhyayula, Igor Pisarev, Daniel E. Milkie, Tsung-Li Liu, Ved Singh, Austin Graves,
Grace H. Huynh, Yongxin Zhao, John Bogovic, Jennifer Colonell, Carolyn M. Ott, Christopher Zugates, Susan Tappan, Alfredo
Rodriguez, Kishore R. Mosaliganti, Shu-Hsien Sheu, H. Amalia Pasolli, Song Pang, C. Shan Xu, Sean G. Megason, Harald
Hess, Jennifer Lippincott-Schwartz, Adam Hantman, Gerald M. Rubin, Tom Kirchhausen, Stephan Saalfeld, Yoshinori Aso,
Edward S. Boyden and Eric Betzig

Science 363 (6424), eaau8302.

DOI: 10.1126/science.aau8302

Combining expansion and the lattice light sheet

Downloaded from http://science.sciencemag.org/ on February 3, 2019

Optical and electron microscopy have made tremendous inroads into understanding the complexity of the brain.
Gao et al. introduce an approach for high-resolution tracing of neurons, their subassemblies, and their molecular
constituents over large volumes. They applied their method, which combines expansion microscopy and lattice
light-sheet microscopy, to the mouse cortical column and the entire Drosophila brain. The approach can be performed at
speeds that should enable high-throughput comparative studies of neural development, circuit stereotypy, and structural
correlations to neural activity or behavior.
Science, this issue p. eaau8302

ARTICLE TOOLS http://science.sciencemag.org/content/363/6424/eaau8302

SUPPLEMENTARY http://science.sciencemag.org/content/suppl/2019/01/16/363.6424.eaau8302.DC1
MATERIALS

REFERENCES This article cites 129 articles, 33 of which you can access for free
http://science.sciencemag.org/content/363/6424/eaau8302#BIBL

PERMISSIONS http://www.sciencemag.org/help/reprints-and-permissions

Use of this article is subject to the Terms of Service

Science (print ISSN 0036-8075; online ISSN 1095-9203) is published by the American Association for the Advancement of
Science, 1200 New York Avenue NW, Washington, DC 20005. 2017 © The Authors, some rights reserved; exclusive
licensee American Association for the Advancement of Science. No claim to original U.S. Government Works. The title
Science is a registered trademark of AAAS.