Aberration-free three-dimensional multiphoton
imaging of neuronal activity at kHz rates
Edward J. Botcherbya, Christopher W. Smitha, Michael M. Kohlb,c, Delphine Débarred,a, Martin J. Bootha,
Rimas Juškaitisa, Ole Paulsenb,c, and Tony Wilsona,1
aDepartment of Engineering Science, University of Oxford, Parks Road, Oxford OX1 3PJ, United Kingdom;
Genetics, University of Oxford, Parks Road, Oxford OX1 3PT, United Kingdom;
Cambridge, Downing Street, Cambridge CB2 3EG, United Kingdom; and
bDepartment of Physiology, Anatomy and
cDepartment of Physiology, Development and Neuroscience, University of
dLaboratory for Optics and Biosciences, Ecole Polytechnique, 91128 Palaiseau,
Edited by Jennifer Lippincott-Schwartz, National Institutes of Health, Bethesda, MD, and approved December 19, 2011 (received for review July 26, 2011)
Multiphoton microscopy is a powerful tool in neuroscience, pro-
mising to deliver important data on the spatiotemporal activity
within individual neurons as well as in networks of neurons. A
major limitation of current technologies is the relatively slow scan
rates along the z direction compared to the kHz rates obtainable in
the x and y directions. Here, we describe a custom-built microscope
system based on an architecture that allows kHz scan rates over
hundreds of microns in all three dimensions without introducing
aberration. We further demonstrate how this high-speed 3D multi-
photon imaging system can be used to study neuronal activity at
millisecond resolution at the subcellular as well as the population
fluorescence microscopy ∣ multiphoton imaging ∣ multiphoton microscopy ∣
and translating them into action potential output. In order to un-
derstand the neural code it is important to study both the activity
in dendrites and the action potential outputs of large populations
of neurons. However, direct electrical recordings from thin den-
drites and simultaneous recordings from multiple identified neu-
rons are difficult to achieve and often not feasible, especially in
the in vivo setting. Instead, optical sectioning techniques using
two-photon excitation of fluorescent indicators of neuronal activ-
ity have established themselves as the method of choice to moni-
tor the activity in thin dendrites and large neuronal populations
(1–6). Current technology, however, does not allow the spot of
light to be scanned sufficiently fast in all three dimensions to
fully address the questions of dendritic integration and neuronal
population activity. While itis relatively straightforward to scanthe
focal spot in the x-y plane at kHz rates, scan rates along the
z-axis are limited to approximately 20 Hz with conventional ima-
ging approaches (7). This is due to the mechanical inertia of the
objective lens and specimen during refocusing. Higher-speed
refocusing was recently achieved using a carefully designed electri-
cally tunable lens (8), but at the expense of the numerical aperture.
Recently we overcame these fundamental problems and
showed that neither speed nor numerical aperture needs to be
compromised if we use a different microscope architecture (9).
Using this method, the spot can now be scanned along the z
axis at high speed while still maintaining diffraction-limited
performance. In our design, scanning is carried out by moving
a lightweight mirror instead of the objective and high-speed axial
scanning is achieved without introducing significant spherical
aberration. To demonstrate the practical feasibility of this princi-
ple, we have implemented our unique scan approach in a custom-
built two-photon microscope. With it, we monitored calcium
transients both along the dendritic arbor of individual pyramidal
neurons and in a population of bulk-loaded neurons from a
large volume of brain tissue (10). We are able to show functional
responses measured without aberration from different neurons
eurons process information by integrating many thousands
of synaptic inputs arriving at thin processes called dendrites
separated by more than 60 μm in depth with a time resolution
of 1 ms.
Remote Scanning Two-Photon Microscope. The two-photon micro-
scope designed for these experiments is shown in Fig. 1A. The
microscope has a lateral scan unit (LSU) to scan the focal spot
in the x-y plane and an axial scan unit (ASU) to independently
scan the focal spot in the z direction. Unlike traditional systems,
which are only capable of high-speed scanning in the x-y plane,
this microscope can scan the focal spot along any three-dimen-
sional trajectory at high speed. Both the objective lens L2and
the specimen remain stationary during imaging because the
scanning process is carried out remotely using elements upstream
of the objective.
The LSU comprises a pair of orthogonally mounted galvan-
ometer mirrors spaced closely together so they can be considered
to lie in a single plane. This plane is optically conjugated to the
entrance aperture of the ASU using standard optics. Plane waves
entering the LSU are thereby imaged onto the entrance aperture
of the ASU with an angle that depends on the orientation of the
galvanometer mirrors. The ASU, which comprises a lens and a
mirror, can be set to retro-reflect light entering it by placing
the mirror in the focal plane of the lens. In this configuration,
plane waves reemerge from the ASU through the entrance aper-
ture and are further imaged into the pupil plane of L2, once again
using standard optics. In terms of optical rays passing through the
system, this arrangement sets up a bundle of parallel rays that
enter the pupil of L2. As this objective lens is designed according
to the sine condition (11), all these parallel rays converge to a
single point in the specimen. By altering the orientation of the
galvanometer mirrors in the LSUit is possible to changethe angle
of the rays passing into L2and consequently the focal spot is
displaced laterally. Due to the sine condition, the rays still con-
verge to a single point and diffraction-limited performance is
The drawback with using sine-condition lenses, however, is
that it can be very hard to scan the spot in the axial direction
without distorting its shape. For instance, if we try to use a simple
quadratic converging or diverging wavefront to refocus the spot,
rays entering the pupil at its periphery would focus to a different
point along the axis to the rays entering the pupil at the center.
The focal spot therefore becomes less confined, has lower peak
Author contributions: E.J.B., C.W.S., M.M.K., D.D., M.J.B., R.J., O.P., and T.W. designed
research; E.J.B., C.W.S., and M.M.K. performed research; E.J.B., C.W.S., and M.M.K.
analyzed data; and E.J.B., C.W.S., M.M.K., R.J., O.P., and T.W. wrote the paper.
The authors declare no conflict of interest.
This article is a PNAS Direct Submission.
Freely available online through the PNAS open access option.
1To whom correspondence should be addressed. E-mail: email@example.com.
This article contains supporting information online at www.pnas.org/lookup/suppl/
www.pnas.org/cgi/doi/10.1073/pnas.1111662109PNAS ∣ February 21, 2012 ∣ vol. 109 ∣ no. 8 ∣ 2919–2924
intensity, and is no longer diffraction-limited. This effect, known
as spherical aberration (12), significantly reduces resolution and
This is not to say that objective lenses cannot produce a dif-
fraction-limited focal spot at a different position along the axis
—they can. They just need a beam with a very specific wavefront
profile. In this system we generate this specific wavefront profile
in the ASU using a second objective lens L1to mimic the optical
properties of L2. Crucially, this second objective also obeys the
sine condition. If the mirror M is displaced axially from the focal
plane, a wavefront emerges from the ASU with the exact profile
required for L2to produce a single diffraction-limited spot dis-
placed along the axis in the specimen. All that is needed then is to
reimage this wavefront profile into the pupil of L2, which is done
here with standard optics. In other words, L1modifies the wave-
front with equal and opposite aberrations in anticipation of those
that L2will introduce when focusing away from the focal plane.
Fig. 2 shows how the rays are focused by the imaging objective L2
for two different settings of the mirror M. In these cases the rays
striking the pupil are no longer parallel but still converge to a
We now explore this effect further by considering the wave-
front of the beam rather than its ray constituents. To produce a
diffraction-limited spot of light a distance z from the nominal
focal plane, the mathematical expression for the wavefront
required in the pupil of L2is given by (13):
ΔΨobjðρ2;zÞ ¼ kNA2z
where k ¼ 2π∕λ is the wavenumber and NA2, α2and ρ2are re-
spectively the numerical aperture, angular aperture, and normal-
ized pupil radius of L2. If a low-NA lens were used in the ASU
then moving the mirror would produce either a converging or
diverging wavefront with a simple quadratic profile that does
not fit this requirement exactly. The resulting wavefront error
would contain higher-order polynomial terms that can be inter-
preted as corresponding to spherical aberration (12). By “mi-
micking” the optics of the imaging objective, however, the ASU
produces a wavefront that is very similar in form:
ΔΨASUðρ1;zÞ ¼ kNA1· 2zM
where zMis the displacement of the mirror M from the focal
plane of L1and NA1, α1, and ρ1are respectively the numerical
aperture, the angular aperture, and the normalized pupil radius
of L1. The extra factor of 2 in this equation comes from the
reflection geometry: for any displacement of the mirror M the
optical path between lens and mirror is increased or decreased
by twice that displacement. With a simple magnification this
can be transformed into the exact wavefront required by L2for
diffraction-limited refocusing. In practice, this magnification is
provided by the 4f system linking the pupils of L1and L2. Using
a magnification factor that asserts
generated in the pupil of L2becomes:
sinα2, the wavefront
ΔΨðρ2;zÞ ¼ kNA2· 2zM
where n1and n2are the refractive indices of the immersion media
for L1and L2respectively.
Comparing this expression with Eq. 1, we see that this specific
situation produces a diffraction-limited focal spot a distance
z ¼ 2zMn1∕n2from the nominal focal plane in the specimen.
By varying zm, the position of the mirror, the focal spot is propor-
tionally displaced to a different axial position in the specimen.
The simplest system that could be built on this principle would
have identical lenses for L1and L2. Strictly speaking, perfect
compensation of aberrations only happens when focusing into
a medium of uniform refractive index so it is desirable to use
a water-immersion lens for L2as this leads to the best match
of refractive index when imaging biological tissues. On the other
hand, it is desirable to use a dry lens for L1as viscous forces from
an immersion medium would otherwise impede the motion of the
mirror. This is not a problem as nonidentical lenses may be used
so long as the magnification between them is chosen correctly.
For this, it is necessary to calculate the normalized pupil radius
for each lens using ρ ¼ rfNA, where r is the physical radius in the
objective pupil and f is the focal distance, which in turn is related
to the nominal tube length, F, and magnification of the objective,
M, by f ¼F
0.9 NA dry objective for L1and an Olympus LUMPlanFL N 40X
M. In our system we used an Olympus UApo/340, 40X,
scanning was carried out by two galvanometers in the lateral scan unit (LSU)
and z-scanning by the mirror M in the axial scan unit (ASU). A polarizing
beam splitter (PBS) and quarter-wave plate (QWP) directed all light through
the ASU and into the imaging objective (L2). (B) Theoretical wavefronts and
intensity distributions produced for various settings of the LSU and ASU. In
the central position, where mirror M lies in the focal plane of L1, the resulting
wavefront is flat. For positive and negative displacements of the mirror,
curved wavefronts are produced and the spot refocuses along the axis.
The effect of the LSU is to tilt the wavefronts, which displaces the spot lat-
erally in the specimen and the combination of both ASU and LSU together
can shift the spot to any 3-dimensional location in the specimen.
(A) Schematic of the fast focusing two-photon imaging system. x-y
for two different focal settings. In each case all rays focus through a single
point in the specimen irrespective of the focal setting.
Ray diagrams showing the ray paths in the focal regions of L1and L2
www.pnas.org/cgi/doi/10.1073/pnas.1111662109Botcherby et al.
0.8 NA water dipping lens for L2so the magnification between L1
and L2required for aberration-free imaging wasr2
Fig. 1 B shows the theoretical wavefronts and corresponding
focal spot intensity distributions in the x-z plane generated for
a series of different LSU and ASU settings. As can be seen in
the final pair of plots, we can use the LSU and ASU together
to move the focal spot to an arbitrary position in three-dimen-
sional space, away from both the focal plane and the optical axis.
As a result of employing this remote scanning technique, the
peak intensity of the focal spot, image resolution, and sectioning
strength are all maintained even when imaging at substantial dis-
tances from the nominal focal plane of the microscope objective
lens. To scan the lightweight mirror M at high speed a custom-
built actuator was constructed using a pair of galvanometer
motors (Fig. 3 A). This arrangement permitted a flat frequency
response up to 2.7 kHz (Fig. 3 B). Further details concerning the
construction of the actuator are given in the Methods section.
Using a previously developed method (13) to measure the
focal spot intensity distribution we found the spot could be
scanned more than 200 μm in the z direction without significant
degradation (Fig. 4). The peak intensity in each case was greater
than 85% of that measured in the nominal focal plane across the
whole 200 μm range. As a final point, these measurements show
that the system resolution is still set by the imaging properties of
the imaging objective L2and that the ASU does not adversely
affect it to any great extent.
Arbitrary Line Scanning. Neuronal processes extend over many
hundreds of microns in three dimensions so to monitor activity
in these processes, or over a large number of neuronal cell bodies,
it is desirable to scan the focal spot over large distances to record
activity from different locations with a high temporal resolution.
To test the performance of our system we defined a series of
complex trajectories through which we scanned the focal spot
and observed how faithfully these trajectories were reproduced at
different scan frequencies.
For each trajectory a set of points was defined in three-dimen-
sional space. Periodic cubic splines were then fitted through these
to define a scan trajectory and galvanometer drive signals were
calculated for each scan axis. Due to the nonperfect frequency
response characteristics of the galvanometers these signals had
to be modified using a precompensation algorithm; otherwise
the focal spot would follow a distorted trajectory and potentially
miss the points of interest completely. The algorithm made use of
the galvanometers’ optical feedback signals and computed phase
and amplitude corrections that were applied to the driving wave-
forms (see Supporting Information). The focal spot could thus be
scanned at high frequency with far greater accuracy through the
desired points. Three example trajectories are shown in Fig. 5
with and without precompensation. Each spans more than
100 μm in the z direction, and corrected trajectories are accurate
to within 0.5 μm of each control point. As the complexity of the
trajectories increase, the maximum achievable scan rate de-
creases because of the galvanometer torque limit.
Three-Dimensional Imaging of Neuronal Activity. To validate this
technology we considered two classes of problems in neural
imaging. We looked first at the activity of a single cell by scanning
different dendritic processes. The second case looked at multiple
neurons distributed widely in three dimensions in a tissue speci-
men, scanning at high speed along a cyclic trajectory that passed
through the soma of each cell. We demonstrate that the system
can be used to scan along the dendritic arbor of individually Fluo-
4 dye-loaded neurons to follow dendritic calcium signals in real
time and sample the activity of a large number of neurons in an
OGB1-AM bulk-loaded sample (see Methods).
To test the ability of this remote scanning microscope to
provide fast, multisite measurements of dendritic activity, we
monitored localized calcium transients in response to backpropa-
gating action potentials that were triggered by current injection
through the somatic patch pipette. In order to simulate experi-
ments carried out in vivo, we obtained whole-cell patches from
neurons in a cortical slab (Fig. 6A). Neurons, about 200–300 μm
below the cortical surface, were filled with the calcium-sensitive
fluorescent dye Fluo-4 and the fluorescent structural dye Alexa
594 to visualize the fine structure of neuronal processes. Dendri-
tic activity could then be monitored by measuring relative inten-
sity changes (ΔF/F) over time. The remote scanning method was
used to obtain a stack of images of the neuron allowing a three-
dimensional structural reconstruction (Fig. 6 B and C). This re-
construction was then used to select points for functional imaging
on two dendrites separated by at least 30 μm in depth (Fig. 6D).
In addition to single-cell investigations, our system can also
measure the behavior of multiple neurons in a cluster. Neurons
in the somatosensory cortex were bolus-loaded with a calcium dye
OGB1-AM. As before, a three-dimensional reconstruction of the
loaded neurons was used to mark neurons in the stack for func-
tional imaging (Fig. 6 E and F). Extracellular electrical burst sti-
mulation resulted in calcium transients in all neurons that were
1001k Frequency (Hz)
u t i l p
) . u . a ( e
mirror M is mounted on a flexible beryllium copper bridge and two further
galvanometers, G1 and G2, rotate synchronously to produce axial motion.
(B) The experimentally measured frequency response curve for this actuator
when scanning the focal spot 25 μm along z direction in the specimen.
(A) Custom-built actuator for scanning the mirror M in the ASU. The
00 1 -m
1.0 a.u. 0.96 a.u.0.85 a.u.
sity distribution in the x-z plane for different focal settings of the microscope
with corresponding peak intensities labeled in arbitrary units. Scale bar 2 μm.
Non-normalized experimental measurements of the focal spot inten-
1 kHz 500 Hz
mark points of interest through which it is desired to scan the focal spot.
A continuous trajectory is defined by fitting periodic cubic splines through
these and a correction algorithm is applied to compensate for the response
characteristics of the galvanometers. Dashed blue and solid red lines indicate
respectively the trajectory followed before and after correction. (A) Scanning
two points at a sampling rate of 1 kHz. (B) Scanning points on a three-dimen-
sional Lissajous figure with a sampling rate of 500 Hz. (C) Scanning points on
a complex arbitrary trajectory at 300 Hz. After correction, the focal spot
passes within 0.5 μm of each desired point in all three trajectories.
Experimental verification of arbitrary line scanning. Black squares
Botcherby et al. PNAS
February 21, 2012
monitored over a depth of more than 60 μm at 1 kHz acquisition
rate (Fig. 6G).
We have built a two-photon microscope that can be used to image
neuronal activity in three dimensions with a high sampling
rate. This is achieved by the use of a fundamentally new optical
approach, the ASU, which allows an aberration-free focal spot to
be repositioned axially at high speed. Compared with previous
inertial-based scanning systems, which permit absolute maximum
scan speeds of 20 Hz in the z direction (7), we have demonstrated
that rates of up to 2.7 kHz are achievable with our system. This
is comparable with galvanometer scan rates in the x-y plane and
we believe further development of this technology will eventually
allow all axes to be scanned at the same rate.
Reddy et al. (14) have developed an interesting alternative
to this approach which uses noninertial scanning, whereby a
combination of four acousto-optic deflectors (AODs) are used
to generate wavefront profiles so as to position the focal spot
arbitrarily in three dimensions. The attractive feature of this
approach is that, aside from a short transition time on the order
of microseconds, the focal spot can be translated an arbitrary
distance in the specimen without passing through any intermedi-
ate points. This is not possible with galvanometer-based scanners
as the response characteristics essentially prohibit the spot from
being repositioned instantaneously so instead a continuous tra-
jectory must be scanned through the points of interest. The draw-
back of AOD scanning, however, is that a certain amount of
spatial and temporal dispersion is introduced by the AODs them-
selves, and while these effects can be compensated for, this adds
to the complexity of system design (15, 16). Kirkby et al. have
refined this approach considerably and have achieved a scan
range of 137 μm in the z direction (17). Nonetheless, aberrations
remain and we therefore suggest a useful compromise would be
to combine these techniques in a hybrid arrangement that per-
forms x-y positioning with AODs and z scanning using the ASU
approach described in this paper. In this way, we would be able to
address random points while scanning quickly over a large range
in the z direction, without introducing aberrations.
The approach taken here is not limited to applications in neu-
roscience and will also prove useful for imaging tasks where it is
absolutely necessary that the specimen remains stationary and
undisturbed and/or where high scan speeds along the z direction
are of importance. This includes the possibility for scanning other
trajectories such as planes with arbitrary orientation (18, 19) or
even along curved surfaces at high speed. As a final point we also
note that this architecture lends itself to parallelization for situa-
tions where multiple points must be imaged simultaneously (20).
In summary, the system we have developed will provide neu-
roscientists with the tools necessary to study neuronal activity at
the population and subcellular levels over large enough distances
and at a speed that allows the activity of single action potentials
to be followed. This will enable us to study synaptic integration
and the neuronal population activity at an unprecedented level
of detail. Our remote scanning technology also prevents motion
artifacts from being introduced by the mechanical agitation of the
tissue as a result of moving the objective lens directly. Further-
more, we have demonstrated in cortical slabs that high-speed
z scanning with near diffraction-limited resolution is possible
allowing for detailed reconstructions and functional imaging in
three dimensions, making this technique ideal for in vivo applica-
tions (21). We therefore believe this technique presents an impor-
tant improvement of existing two-photon scanning methods and
will significantly facilitate the study of neuronal integration and
Two-Photon Imaging. We custom-built a two-photon imaging system around
an Olympus BX60M upright microscope stand using mechanical components
from the Linos microbench range. A two-dimensional x-y stage (Luigs &
Neumann) was fitted to the optical bench on which the specimen and patch-
clamp head stage could be mounted. The laser source used was a Ti:Sapphire
laser (Tsunami, Spectra Physics), producing ultrafast pulses with central
wavelength 850 nm (Δλ ∼ 50 nm, pulse length approximately 100 fs). Initially,
this was expanded to form a plane wave with Gaussian width 5 mm and di-
rected into the lateral scan unit (LSU), comprising two orthogonally mounted
Projection of three-dimensional image stack in x-z plane showing a single cell loaded with Fluo-4 and Alexa 594. (C) Three-dimensional rendering of a single
cell with two points marked that are separated by more than 30 μm in the z direction. (D) Fluorescence measurements from points of interest taken with a
temporal resolution of 500 Hz. (E) Three-dimensional rendering of neuronal population bolus-loaded with OGB1-AM. (F) Three-dimensional representation of
the neurons in the scanned volume (100 × 100 × 100 μm). (G) Fluorescence measurements from different points in the stack in response to extracellular elec-
trical burst stimulation taken with a temporal resolution of 1 kHz and over more than 60 μm z distance. Traces represent the average of 2–4 trials.
Functional imaging of cortical neurons. (A) Imaging was performed in cortical slabs to emulate the use of this method in the in vivo setting. (B)
www.pnas.org/cgi/doi/10.1073/pnas.1111662109 Botcherby et al.
galvanometer mirrors (VM1000+, Cambridge Technology) that controlled the
angular orientation of the wavefronts. From here the wavefronts were im-
aged into the pupil of L1using a 4f imaging system comprising two achro-
matic doublet lenses, with focal lengths 120 mm and 160 mm, to produce a
magnification of 4∕3. L1was an Olympus UApo/340, 40X, 0.9 NA dry objec-
tive, chosen for its favorable transmission characteristics at 850 nm. As this
objective lens is coverslip-corrected a standard 170 μm glass coverslip was
fixed to the front for optimal performance. Light passing through L1
reflected off the mirror M (PF03-03-P01, Thor labs) and passed back through
the lens. The emerging wavefront was then reimaged into the pupil plane of
L2using a further 4f system of achromatic doublets, this time with focal
lengths 150 mm and 200 mm, to produce a magnification of 4∕3. All experi-
ments were performed using an Olympus LUMPlanFL N 40X 0.8NA water dip-
ping objective for L2. A polarizing beam splitter (PBS) and quarter-wave plate
(QWP) ensured all light entering the ASU was transmitted into the final stage
of the system. A dichroic beamsplitter (DBS) (720DCSPXR, Chroma) and emis-
sion filter (EF) (ET720SP-2P8, Chroma) were used to separate fluorescence
photons produced in the specimen in the range 400–720 nm for measure-
ment on the photomultiplier tube (PMT) (P30PC-54, Sens-Tech). System scan-
ning and photon counting were carried out with a reconfigurable I/O card (NI
PCI-7830R, National instruments), which also produced digital pulses for sti-
mulating action potentials in patch-clamp experiments. Software to control
the microscope and interface with a desktop PC was written in Labview (Na-
Construction of Fast Moving Mirror. To provide the stroke and bandwidth
required for high-speed refocusing we developed a unique actuator
(Fig. 3 A) for scanning the mirror M. This comprised two galvanometer
motors, G1 and G2 (VM1000+, GSI), mounted in opposite directions with their
axes parallel and offset by 10 mm. Two custom-made aluminum clamps were
attached to the ends of the galvanometer shafts that gripped the ends
of a thin strip of beryllium copper, of thickness 100 μm, folded to produce
a flexible raised platform between the galvanometers on which the mirror
was glued. In operation the galvanometers rotated synchronously, giving
the mirror a smooth and repeatable motion. We measured the frequency
response (Fig. 3 B) of this actuator up to 5 kHz, while scanning the focal spot
25 μm axially in the specimen, using optical feedback signals from both
galvanometers. This indicates a bandwidth of 2.7 kHz.
Determining the Mirror Damage Threshold. As the mirror M (Figs. 1 A and 2)
lies in the focal region of L1, it is routinely exposed to high peak intensities
during imaging, particularly when in focus. We therefore carried out a study
to find a mirror with a high damage threshold suitable for use in our system.
A number of mirrors were tested by placing them in the focal region of L1
and exposing them to the full power of the laser beam. The mirrors were also
scanned in a slow sinusoidal fashion along the axis (Δz ∼ 20 μm, rate approxi-
mately 10 Hz) to ensure worst-case conditions in terms of intensity were
invoked at some point in the cycle. One mirror from Thorlabs (PF03-03-
P01) was found to be particularly suitable for our system, showing no signs
of damage after a prolonged exposure of 1 h under these conditions. During
this time the average Ti:Sapphire laser power (pulses approximately 100 fs)
entering the back aperture of imaging lens L2was found to be 120 mW.
Calibrating the Line Scans. At the beginning of each experiment a through
focus series of images was acquired from the specimen volume and x-y,
y-z, and x-z image projections displayed on a computer using custom-soft-
ware. On these projections, points of interest could be marked at arbitrary
locations in three dimensions through which it was desired to scan the focal
spot. A continuous scan trajectory was then generated using periodic cubic
spline functions using these points of interest as nodes. A calibration algo-
rithm was then run to correct for amplitude and phase distortions introduced
by the galvanometer scanners so as to ensure that the focal spot would
be driven through these points of interest. This involved pixelating the con-
tinuous trajectory into a series of discrete points that were equally spaced in
three dimensions and calculating a series of command voltages for each of
the x-, y-, and z- axes. The number of pixels as well as a pixel dwell time could
be selected on the software. Calibration then proceeded for each galvan-
ometer in turn. The command signal was sent to the galvanometer and a
feedback signal recorded. This was cross-correlated with the command signal
and a best-fit phase lag, ΔΦ, was determined for the axis (Fig. S1). The reverse
of this lag was then applied to the whole command signal to generate a
phase-corrected command signal. The galvanometer was then driven afresh
using this and a phase-corrected feedback recorded. Deviations, ΔVi, of
the phase-corrected feedback from the initial command signal were calcu-
lated at the scan pixels corresponding to each point of interest (Fig. S2),
and by correcting each node with ΔViand refitting the splines, an ampli-
tude-and-phase-corrected command signal was generated. By iterating this
procedure a number of times it was possible to generate a trajectory that
passed far closer to the desired points of interest (Fig. S3). This process
was repeated for all three galvanometer axes.
Tissue Preparation. Cortical slabs and thalamocortical slices (350 μm thick)
were prepared from postnatal day 13–28 C57BL/6 mice of both sexes after
decapitation under deep isoflurane-induced anesthesia, in accordance with
British Home Office regulations. Slabs were ca. 5 × 5 mm and 3 mm deep
and cut from the cortical surface in the region of the somatosensory cortex.
To anchor the slabs and slices were anchored by mounting them on cover slips
(coated with 0.1% poly-L-lysine in ultrapure H2O). The mounted slices were
kept at room temperature (20–25°C) in small round petri dishes between hu-
midified carbogen gas (95% O2, 5% CO2) and artificial CSF (aCSF, containing
(in mM): 126 NaCl, 3 KCl, 1.25 NaH2PO4, 2 MgSO4, 2 CaCl2, 26 NaHCO3, and
10 glucose, pH 7.2–7.4) for at least 1 h for recovery and then until used for
recording. For recordings, the coverslip with slice was transferred to a sub-
merged style recording chamber, and superfused with aCSF, bubbled
with carbogen gas.
Electrical and Optical Measurements. Whole-cell patch-clamp recordings
were performed with glass pipettes, pulled from standard borosilicate glass
(5–8 MOhm). Signals were acquired at 5 kHz and low-pass filtered at 2 kHz
using an Axon Multiclamp 700B amplifier (Molecular Devices). Pipette solu-
tion contained (in mM): 110 potassium-gluconate, 40 Hepes, 2 ATP-Mg,
0.3 GTP, and 4 NaCl (pH 7.2–7.3; osmolarity 270–285 mosmol l−1) and a com-
bination of a structural label, 40 μM Alexa 594, and a calcium-sensitive dye,
200 μM Fluo-4 (both Molecular Probes). Neurons in the somatosensory cortex
were patched using infrared differential-interference contrast imaging (22)
at depths approximately 200–300 μm below the surface of the cortical slab
and 100–150 μm below the surface of the slice, respectively. Calcium imaging
was begun about 30 min after establishing whole-cell configuration, to allow
for dye wash-in and diffusional equilibration. Cell populations were labeled
with the calcium indicator Oregon Green BAPTA-1 (Molecular Probes) using
the multicell bolus loading technique (3, 23). Briefly, 50 μg of the membrane-
permeant acetoxymethyl (AM) ester form of OGB-1 was dissolved in DMSO
plus 20% Pluronic F-127 (Molecular Probes) and diluted in aCSF to a final
concentration of about 1 mM. This solution was pressure-ejected into neo-
cortical layer 2∕3 in somatosensory cortex using a micropipette. A tungsten
electrode was used to deliver extracellular stimulation bursts.
Data Analysis. Somatic fluorescence signals were analyzed using custom-
written scripts in MATLAB (MathWorks) and are expressed as relative fluor-
escence changes (ΔF∕F). Fluorescent signals are presented as averages from
2–4 consecutive trials and smoothed digitally using a 50 Hz low-pass Butter-
worth filter. Structural renderings were made using Volocity (Perkin Elmer).
ACKNOWLEDGMENTS. This work was supported by the Engineering and
Physical Sciences Research Council (EPSCR) (EP/H018565/1), the Biotechnol-
ogy and Biological Sciences Research Council, and the Wellcome Trust.
E.J.B. held an EPSRC postdoctoral research fellowship (EP/F042647/1) and
M.M.K. was supported by a Wellcome Trust Prize PhD Studentship.
1. Denk W, Strickler JH, Webb WW (1990) Two-photon laser scanning fluorescence
microscopy. Science 248:73–76.
2. Svoboda K, Denk W, Kleinfeld D, Tank DW (1997) In vivo dendritic calcium dynamics in
neocortical pyramidal neurons. Nature 385:161–165.
3. Stosiek C, Garaschuk O, Holthoff K, Konnerth A (2003) In vivo two-photon calcium
imaging of neuronal networks. Proc Natl Acad Sci USA 100:7319–7324.
4. Helmchen F, Denk W (2005) Deep tissue two-photon microscopy. Nat Methods
5. Svoboda K, Yasuda R (2006) Principles of two-photon excitation microscopy and its
applications to neuroscience. Neuron 50:823–839.
6. Göbel W, Kampa BM, Helmchen F (2007) Imaging cellular network dynamics in three
dimensions using fast 3D laser scanning. Nat Methods 4:73–79.
7. Göbel W, Helmchen F (2007) New angles on neuronal dendrites in vivo. J Neurophysiol
8. Grewe BF, Voigt FF, van’t Hoff M, Helmchen F (2011) Fast two-layer two-photon
imaging of neuronal cell populations using an electrically tunable lens. Biomed
Opt Express 2:2035–2046.
9. Botcherby EJ, Juškaitis R, Booth MJ, Wilson T (2007) Aberration-free optical refocusing
in high numerical aperture microscopy. Opt Lett 32:2007–2009.
10. Yuste R, Konnerth A (2005) Imaging in neuroscience and development. A laboratory
manual (Cold Spring Harbor Laboratory Press, Cold Spring Harbor, New York).
Botcherby et al. PNAS
February 21, 2012
11. Born M, Wolf E (1999) Principles of optics: Electromagnetic theory of propagation,
interference and diffraction of light (Cambridge University Press, Cambridge, UK),
12. Gross H, Zügge H, Peschka M, Blechinger F (2007) Handbook of Optical Systems,
Volume 3, Aberration Theory and Correction of Optical Systems (Fig 33–15) (Wiley
VCH, Weinhein, Germany).
13. Botcherby EJ, Juškaitis R, Booth MJ, Wilson T (2008) An optical technique for remote
focusing in microscopy. Opt Commun 281:880–887.
14. Reddy GD, Kelleher K, Fink R, Saggau P (2008) Three-dimensional random access
multiphoton microscopy for functional imaging of neuronal activity. Nat Neurosci
15. Grewe BF, Langer D, Kasper H, Kampa BM, Helmchen F (2010) High-speed in vivo
calcium imaging reveals neuronal network activity with near-millisecond precision.
Nat Methods 7:399–405.
16. Saggau P (2006) New methods and uses for fast optical scanning. Curr Opin Neurobiol
17. Kirkby PA, Naga Srinivas Nadella KM, Silver RA (2010) A compact acousto-opticlens for
2D and 3D femtosecond based 2-photon microscopy. Opt Express 18:13720–13744.
18. Smith CW, Botcherby EJ, Booth MJ, Juškaitis R, Wilson T (2011) Agitation-free multi-
photon microscopy of oblique planes. Opt Lett 36:663–665.
19. Smith CW, Botcherby EJ, Wilson T (2011) Resolution of oblique-plane images in
sectioning microscopy. Opt Express 19:2662–2669.
20. Hoover EE, et al. (2010) Remote focusing for programmable multi-layer differential
multiphoton microscopy. Biomed Opt Express 2:113–122.
21. Grewe BF, Helmchen F (2009) Optical probing of neuronal ensemble activity. Curr Opin
22. Dodt HU, Zieglgänsberger W (1990) Visualizing unstained neurons in living brain
slices by infrared DIC-videomicroscopy. Brain Res 537:333–336.
23. Garaschuk O, Milos RI, Konnerth A (2006) Targeted bulk-loading of fluorescent
indicators for two-photon brain imaging in vivo. Nat Protoc 1:380–386.
www.pnas.org/cgi/doi/10.1073/pnas.1111662109Botcherby et al.