In vivo three-photon imaging of activity of GCaMP6-labeled
neurons deep in intact mouse brain
Dimitre G Ouzounov
1,5
, Tianyu Wang
1,5
, Mengran Wang
1
, Danielle D Feng
2
, Nicholas G
Horton
1
, Jean C Cruz-Hernández
2
, Yu-Ting Cheng
2
, Jacob Reimer
3
, Andreas S Tolias
3,4
,
Nozomi Nishimura
2
, and Chris Xu
1
1
School of Applied and Engineering Physics, Cornell University, Ithaca, New York, USA
2
Nancy E. and Peter C. Meining School of Biomedical Engineering, Cornell University, Ithaca,
New York, USA
3
Center for Neuroscience and Artificial Intelligence, Department of Neuroscience, Baylor College
of Medicine, Houston, Texas, USA
4
Department of Electrical and Computer Engineering, Rice University, Houston, Texas, USA
5
These authors contributed equally to this work
Abstract
High-resolution optical imaging is critical to understanding brain function. We demonstrate that
three-photon microscopy at 1,300-nm excitation enables functional imaging of GCaMP6s-labeled
neurons beyond the depth limit of two-photon microscopy. We record spontaneous activity from
up to 50 neurons in the hippocampal stratum pyramidale at ~1-mm depth within an intact mouse
brain. our method creates opportunities for noninvasive recording of neuronal activity with high
spatial and temporal resolution deep within scattering brain tissues.
Optical imaging provides the high spatial resolution necessary to resolve individual neurons
and neuronal processes, but widefield images acquired deep within scattering biological
tissues are blurred. Multiphoton microscopy (MPM) has substantially extended the
penetration depth of high-resolution optical imaging
1,2
. When combined with activity-
sensitive fluorescent indicators, MPM enables functional imaging of populations of neurons
in their native environment. While two-photon microscopy (2PM) has been successful in
recording neuronal activities within the cortex
3–7
of mouse brain, 2PM imaging of
subcortical neurons currently requires invasive procedures that either remove overlying brain
tissue
8
or insert penetrating optical elements
9–11
.
Reprints and permissions information is available online at
http://www.nature.com/reprints/index.html.
Correspondence should be addressed to D.G.O. (ouzounov@cornell.edu) or C.X. (cx10@cornell.edu).
AUTHOR CONTRIBUTIONS
C.X. conceived the study. D.G.O., T.W., M.W., D.D.F., N.G.H. J.C.C.-H., and Y.-T.C. performed the experiments. T.W., D.G.O., and
J.R. analyzed the data. C.X. and N.N. supervised the project. J.R. and A.S.T. provided transgenic mice for this study. D.G.O., T.W.,
and C.X. prepared the manuscript.
COMPETING FINANCIAL INTERESTS
The authors declare no competing financial interests.
Note: Any Supplementary Information and Source Data files are available in the online version of the paper.
HHS Public Access
Author manuscript
Nat Methods
. Author manuscript; available in PMC 2019 March 30.
Published in final edited form as:
Nat Methods
. 2017 April ; 14(4): 388–390. doi:10.1038/nmeth.4183.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
The imaging depth for MPM in scattering biological tissue is limited by the signal-to-
background ratio (SBR)
12,13
. When the SBR approaches 1, the image contrast is
substantially degraded. For
in vivo
2PM imaging in densely labeled mouse cortex, the image
contrast is lost at depth of 1 mm (refs. 13–16). Because of its higher order of nonlinear
excitation and weaker scattering at longer excitation wavelengths, three-photon microscopy
(3PM) reduces out-of-focus excitation and has an SBR orders of magnitude higher than that
of 2PM at the same depth
17
.
The optimal wavelengths
18–20
for minimum excitation attenuation in mouse brain tissue
occur at two narrow regions centered at 1,300 nm and 1,700 nm (ref. 17). 3PM at the 1,300-
nm spectral window enables three-photon excitation (3PE) of a variety of blue and green
fluorophores, such as the current generations of protein-based genetically encoded calcium
indicators (GECIs; e.g., GCaMP6 (ref. 21)), for which researchers have found increasingly
broader applications in neuroscience research.
Here, we demonstrate that 3PM at 1,300-nm excitation in combination with GCaMP6s
21
is
capable of structural and functional imaging of populations of neurons as deep as the SP
layer of the CA1 hippocampal region within intact mouse brains. The demonstrated method
will introduce potential applications where noninvasive, high spatial- and temporal-
resolution imaging deep within tissue is required.
We used a custom-built multiphoton microscope as well as a noncolinear optical parametric
amplifier (NOPA) and a mode-locked Ti:sapphire laser for 3PM and 2PM, respectively
(Supplementary Fig. 1a–c).
To validate functional imaging with 3PM, we compared it with 2PM, which is capable of
detecting single action potentials
21
. We used a time-division multiplexing (TDM) scheme
(Supplementary Fig. 1d), which enabled us to alternatively image neuronal activity using
2PM (at 920 nm) and 3PM (at 1300 nm) within ~1 μs of each other. We recorded activity in
cortical layer 2/3 (L2/3) neurons in a transgenic mouse (CamKII-tTA/tetO-GCaMP6s) (Fig.
1a,b). Neuronal activity traces recorded by 2PM and 3PM were essentially the same, with a
Pearson’s correlation coefficient of 0.976 ± 0.004 (mean ± s.e.m.) calculated from four
neurons over a total of 19 traces (each 75 s long).
Although the SBR of 2PM is adequate for sparsely labeled samples or in shallow regions,
many experiments require densely labeled brain tissue—for example, while recording the
activities from a large number of neurons simultaneously at multiple depths of a cortical
column. We compared 2PM and 3PM SBR by imaging densely labeled vasculature and
neurons in deep mouse cortex at the same site and time. For fluorescein-labeled vasculature
in an 11-week-old C57BL/6J mouse, SBR approached ~1 for 2PM at 780 μm (Fig. 1c,d). In
comparison, the SBR for 3PM at the same imaging depth was above 40. In densely labeled
transgenic mice (CamKII-tTA/tetO-GCaMP6s, 20 weeks) it is difficult to identify neurons of
average brightness using 2PM at the depth of 780 μm (Fig. 1e,f). The 3PM image has better
contrast, and more neurons are visible in the same field of view (FOV). These results show
the advantages of the larger SBR of long-wavelength 3PM in densely labeled tissues; in fact,
SBR does not limit 3PM imaging capability until much deeper layers
17
.
Ouzounov et al.
Page 2
Nat Methods
. Author manuscript; available in PMC 2019 March 30.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
We performed 3PM of both the structure and activity of neurons in the SP layer of the mouse
hippocampus (Fig. 2 and Supplementary Figs. 2–6). The hippocampal neurons were infected
by AAV virus expressing GCaMP6s (synapsin promoter, AAV2/1 encapsulation) in wild-
type mice (C57BL/6J). The imaging took place in the third to fifth week after injection (11–
16-weeks-old mice). For structural imaging, we acquired a stack from 500 to 1,100 μm (3D
reconstruction in Fig. 2a and Supplementary Video 1). Labeled neurons in L6 are visible
immediately above the external capsule (EC), where myelinated axons display strong third
harmonic generation (THG) signals (Fig. 2a,b). The EC features fibrous structure (Fig. 2b)
extending approximately from 730 to 860 μm below the dura (Fig. 2a). GCaMP6s-labeled
neurons in the SP layer span from 940 to 1,040 μm below the surface and begin ~80 μm
below the EC. Labeled neurons display clear nuclear exclusion and the characteristic
honeycomb arrangement of the SP layer (Fig. 2c). The distinctly resolved neuron
morphology at ~1 mm below the dura shows that 3PM is capable of high-spatial-resolution
imaging deep within an intact mouse brain.
We imaged spontaneous activity (without averaging) from the hippocampal neurons at the
site shown in Figure 2c. We were able to image a population of up to 150 neurons, acquiring
fluorescence time traces of neurons of different brightness (Fig. 2d). A high signal-to-noise
ratio was achieved, which was indicated by the absolute photon counts per neuron per frame
(Supplementary Figs. 2b and 5b). We recorded continuously at a frame rate of 8.49 Hz (256
× 256 pixels/frame) for up to 48 min with a FOV of 200 × 200 μm (Supplementary Video 2,
16-min long). The imaging area per unit time reported here is similar to that of calcium
imaging using 2PM at shallower imaging depths of the mouse brain tissue
4,6,8
. The average
power used was ~50 mW, and no noticeable photobleaching was observed throughout the
entire recording session (Supplementary Figs. 3 and 4). Higher time resolution for a subset
of neurons can be achieved by reducing the FOV or using line scan.
To obtain the activity trace, we integrated the signal from the entire neuron body, which
usually consists of more than 100 pixels with the FOV reported. Our results show that the
repetition rate of the excitation source is adequate for imaging calcium transients produced
by typical GECIs. For structural imaging, longer pixel dwell time or averaging over multiple
frames eliminates the impact of the low repetition rate.
Imaging the mouse brain at 1,280 nm with average power up to 120 mW did not result in
any noticeable tissue damage
19
. However, the higher intensities required by the nonlinear
excitation might induce sample change or damage. Yet, no damage has been observed during
functional imaging at L5 at 925 nm with 4- to 5-nJ pulses at the focus
6
. Longer wavelength
excitation (~1,300 nm) is less phototoxic
22
. We performed deep cortex and hippocampus
imaging by scaling the pulse energy with imaging depth to ensure less than 1.5 nJ at the
focal plane. Furthermore, we revisited the same site multiple times many days after the first
imaging session, and we successfully recorded neuronal activity from the same population
of neurons each time (Supplementary Figs. 5 and 6).
We imaged the hippocampal SP layer in a number of mice using ~50-mW average power.
Our experimental observation was that hippocampal SP layer imaging will be successful as
long as there is insignificant tissue growth under the cranial window, which dramatically
Ouzounov et al.
Page 3
Nat Methods
. Author manuscript; available in PMC 2019 March 30.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
increases the amount of laser power necessary for imaging the neurons even for the
superficial cortical layers.
An additional feature of 3PM is the intrinsic THG signal
23
generated at various interfaces
(e.g., blood vessels and myelinated axons; Fig. 2), which provides additional structural
information without an exogenous stain.
Reliable turnkey laser systems at 1,300 nm for 3PM are commercially available, and it is
relatively straightforward to incorporate such lasers into an existing multiphoton
microscope. While the laser systems are more expensive than mode-locked Ti:S lasers for
2PM, it is likely that their cost will decrease and performance improve as the technology
matures, which will make long- wavelength 3PM widely available to complement 2PM for
deep imaging or imaging densely labeled samples.
3PM can record activity from a large number of neurons with high spatial and temporal
resolution deep within scattering brain tissues. Imaging the spontaneous activity in the
hippocampal SP layer within an intact mouse brain is just one such demonstration, and the
application of our method will enable studies that require noninvasive, high-resolution
imaging deep within scattering tissue.
METHODS
Methods, including statements of data availability and any associated accession codes and
references, are available in the online version of the paper.
ONLINE METHODS
Experimental set up.
Excitation source.—The excitation source for 3PM is a noncollinear optical parametric
amplifier (NOPA, Spectra Physics) pumped by a one-box regenerative amplifier (Spirit,
Spectra Physics). A two-prism (SF11 glass) compressor is used to compensate for the
normal dispersion of the optics of the light source and the microscope, including the
objective. The NOPA operates at wavelength centered at 1,300 nm and provides an average
power of ~500 mW (1,250 nJ per pulse at 400 kHz repetition rate). The pulse duration
(measured by second-order interferometric autocorrelation) under the objective is ~55 fs
after adjusting the prism compressor. An optical delay line is set up to effectively double the
laser repetition rate to 800 kHz. The total optical path length of the delay line is ~3 m, and
the two pulses are separated by ~10 ns, which is much longer than the fluorescence lifetime
(~1 ns). A half-wave plate (HWP) and a polarizing beamsplitter cube (PBS) are used for
power control.
The excitation source for 2PM is a mode-locked Ti:Sapphire laser (Tsunami, Spectra
Physics) centered at 920 nm. The output of the Ti:sapphire laser is modulated with an
electro-optical modulator (Conoptics, model 350–160), which is capable of sub-μs intensity
modulation. The 920-nm beam and 1,300-nm beam are combined by a dichroic mirror
(DM), spatially overlapped, and directed to the microscope.
Ouzounov et al.
Page 4
Nat Methods
. Author manuscript; available in PMC 2019 March 30.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
Imaging setup.—The images were taken with a custom-built multiphoton microscope
with a high-numerical aperture objective (Olympus XLPLN25XWMP2, 25X, NA 1.05). The
signal is epicollected through the objective and then reflected by a dichroic beam splitter
(FF705-Di01–25 × 36, Semrock) to the detectors. The detection system has two channels:
one for green fluorescence signal emitted by the calcium indicator (GCaMP6s) and the other
for violet third harmonic generation (THG) signal generated at the interfaces of various
structures such as blood vessel walls, neuronal processes, and myelinated axons in the white
matter. We use a photomultiplier tube (PMT) with GaAsP photocathode (H7422–40) for the
fluorescence signal and an ultra bialkali PMT (R7600–200) for the THG signal. The optical
filters for the fluorescence and THG channels are 520/60 (i.e., transmission wavelength
centered at 520 nm with a FWHM of 60 nm) band-pass filter (Semrock) and 420/40 band-
pass filter (Semrock), respectively. A 488-nm dichroic beam splitter (Di02-R488–25×36,
Semrock) is inserted in the signal path at 45° between the two PMTs to separate the THG
and fluorescence. The mouse is placed on a motorized stage (M-285, Sutter Instrument
Company). A computer running the ScanImage 3.8 (ref. 24) module on MATLAB
(MathWorks) software is used to control the stage translation and image acquisition. The
PMT current is converted to voltage, amplified and low-pass filtered using a transimpedance
amplifier (C9999, Hamamatsu) and an additional 1.9 MHz low-pass filter (Minicircuts,
BLP-1.9+). Analog-to-digital conversion was performed by a data acquisition card (NI
PCI-6115, National Instruments). For depth measurement, the slightly larger index of
refraction in brain tissue (1.35 to 1.43 for the cortex and as high as 1.467 for the white
matter), relative to water (~1.33), results in a slight underestimate (5–10%) of the actual
imaging depth within the tissue because the imaging depths reported here are the raw axial
movement of the objective
18
.
Activity recording.—For imaging activity, we imaged the neurons continuously using 256
× 256-pixel frame size at 8.49 Hz frame rate for up to 48 min in each imaging session. The
average power was ~50 mW under the objective lens. To locate the same imaging site in
multiple imaging sessions, we used the surface blood vessel map (imaged in the THG
channel) as a guide to navigate the
XY
location. After reaching the same depth as that of the
previous imaging sessions, further adjustments in the
XY
position were made according to
the tissue morphology (cell, blood vessel, and shadow structure) to ensure that the FOV
within the labeled area (total ~400 μm in diameter) was approximately the same as before.
Simultaneous two- and three-photon imaging.—Simultaneous recording of
spontaneous activity with 2PE and 3PE is implemented by time division multiplexing the
920-nm Ti:Sapphire laser beam and 1,300-nm NOPA beam. The 920-nm and 1,300-nm
beams are first spatially overlapped to have the same focal position after the objective. The
920-nm beam is switched on only for ~1 μs between two adjacent NOPA pulses by intensity
modulation with an electro-optic modulator (EOM). The EOM is driven by a radio
frequency signal derived from the NOPA pule train so that the sample is alternatively
illuminated either with 920-nm beam for 2PE or 1,300-nm beam for 3PE. By recording the
NOPA pulses and the PMT signal at 5 MHz sampling rate (200 ns pixel time), the two- and
three-photon excited fluorescence signals are separated in time according to the recorded
1,300 nm laser pulse train with a postprocessing MATLAB script.
Ouzounov et al.
Page 5
Nat Methods
. Author manuscript; available in PMC 2019 March 30.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript