Professional Documents
Culture Documents
Light-gated ion channels and pumps have made it possible fibers7–9 or by miniaturized LEDs10. The limitations imposed
to probe intact neural circuits by manipulating the activity by the low penetration depth of blue light and the lack of optical
of groups of genetically similar neurons. What is needed sectioning inherent for single-photon excitation are partly miti-
now is a method for precisely aiming the stimulating light gated by the low excitation levels necessary for photoactivation
(~1 mW mm−2)7 and by genetic targeting of ChR2 expression,
© 2010 Nature America, Inc. All rights reserved.
1Wavefront-Engineering Microscopy Group, Neurophysiology and New Microscopies Laboratory, Centre National de la Recherche Scientifique, Unité Mixte de Recherche
8154, Institut National de la Santé et de la Recherche Médicale U603, Paris Descartes University, Paris, France. 2Department of Photonics Engineering, Technical
University of Denmark (Fotonik), Lyngby, Denmark. 3Department of Molecular and Cell Biology and Helen Wills Neuroscience Institute, University of California,
Berkeley, Berkeley, California, USA. 4Material Science Division and Physical Bioscience Division, Lawrence Berkeley National Laboratory, Berkeley, California, USA.
5These authors contributed equally to this work. Correspondence should be addressed to V.E. (valentina.emiliani@parisdescartes.fr).
reliably fire action potentials with millisecond temporal resolu- microscope, that is, the limiting axial resolution achievable with
tion and low excitation power when the light was shaped over temporally focused two-dimensional patterns18 (Fig. 1f). This
the cell body, one or more dendritic subdomains or multiple axial resolution was pattern-independent (data not shown).
cells simultaneously.
Shaped two-photon excitation of ChR2-expressing HEK 293 cells
RESULTS We used whole-cell patch-clamp recordings from HEK 293 cells
Spatiotemporal light patterning by TF-GPC transfected with a plasmid encoding GFP-tagged mutant of ChR2
The optical path combining the GPC light mapping scheme with (ChR2(H134R)-GFP) to measure ionic currents through ChR2
temporal focusing is schematized in Figure 1a. We illustrate three channels in response to two-photon illumination. We used wide-
examples of GPC illumination visualized by exciting a thin fluo- field fluorescence imaging to visualize cells expressing ChR2. We
rescent layer (Fig. 1b): a circular spot, an excitation shape tailored used the TF-GPC system to modulate photocurrent amplitude by
to the geometry of a fluorescence image of a dendrite and an exci- generating excitation spots of variable size and shape (Fig. 2a).
tation shape tailored to multiple cell bodies. The intensity profiles The average current amplitude for the illumination of the whole
Theory
excitation of a thin (~1 μm) fluorescent layer
(λexc = 780 nm; objective 60×, 0.9 numerical FWHM = 3 µm
aperture (NA)). Shaped patterns were based
on a confocal image of a Purkinje cell (center,
z CCD
top) and widefield fluorescence image of CA1 z z
–20 –10 0 10 20
hippocampal neurons loaded with Oregon Green
y y y z-axis position (µm)
Bapta (center, bottom), in selected regions of
interest (yellow outlines). (c,d) Theoretical (c)
and experimental (d) y-z section axial propagation of the 20 μm spot without temporal focusing. (e) Experimental y-z section axial propagation of the
spot in d with temporal focusing (left). Experimental propagation measured on double microscope setup 16,19 (right). Scale bars, 10 μm. (f) Axial profile
of fluorescence intensity in e (TF-GPC) compared to 20 μm circular spot generated by TF-DH and theoretical curve for the axial integrated intensity in
line-scanning two-photon microscopy (theory), where I ≈ (1 + (z / zR)2)−0.5, with I being the integrated light intensity, z the actual axial position and zR
the Rayleigh range for a focused Gaussian beam in our experimental conditions, that is, zR = 0.8 μm.
a b c Measured
–2 Intensity (a.u.)
1 Simulation
0
0
10
20
–1
0
z-axis position (µm)
50 pA
100 pA
50 ms
50 ms Antishape 0
Shape –30 –20 –10 0 10 20 30
z-axis position (µm)
Figure 2 | Two-photon photoactivation of ChR2 by TF-GPC in HEK 293 cells. (a) Fluorescence images (λexc = 488 nm, bandwidth 10 nm) of HEK 293
cells transfected with plasmids encoding ChR2(H134R)-GFP with superimposed excitation patterns (red) of increasing size (top; left to right spots have
5, 8, 10, 12 and 14 μm diameter and whole cell). Whole-cell photocurrents (bottom) evoked by 10-ms laser pulses (red bars) of excitation spots in
respective images above (λexc = 850 nm, 0.45 mW μm−2). (b) Widefield epi-fluorescence image of an HEK 293 cell transfected with plasmids encoding
ChR2(H134R)-GFP and superimposed excitation patterns illuminating around the cell (antishape; top left) or covering the whole cell (shape; top right).
Whole-cell photocurrents (bottom; 0.38 mW μm−2). Scale bars, 20 μm. (c) Normalized integrated currents elicited by moving shaped excitation pattern
shown in b along the z axis by steps of 2 μm (dots) (0.17 mW μm−2) and simulation with cell modeled as a parallelepiped of size x = 15 μm, y = 26 μm,
z = 10 μm, corresponding to measured (x, y) and estimated (z, distance between the two experimental peaks) cellular dimensions. Inset, excitation
volume represented as infinite sheet of light with axial distribution given by the experimental curve, measured by scanning the 40×, 0.8 NA objective,
through a fluorescent layer and integrating the light collected by the CCD camera (red line is a guide for the eye). λexc = 850 nm.
© 2010 Nature America, Inc. All rights reserved.
cell was −365 ± 166 pA (n = 3 cells). This corresponded to ~40% (± s.e.m.; n = 7 cells)). In some cells (4 of 7), the photocurrents
of the peak current generated by illumination of the whole cell reached a maximum near the top cell membrane, decreased near
(that is, including the bottom and top cell membranes) with blue the cell center and increased again to a second maximum at the
light (midpoint wavelength of excitation, λexc = 470 nm, bandwidth bottom cell membrane. This was not evident in all cells, perhaps
17 nm, 2.5 mW μm−2). Illumination of an area just outside the edge because of differences in cell flatness. The experimental results
of the cell (antishape) reduced the evoked current by ~87% compared could be simulated by convolving the geometry of the cell mem-
to the current evoked by illuminating the whole soma (−27 ± 10 pA brane, approximated with a parallelepiped, with the excitation
for antishape; −205 ± 137 pA (± s.e.m.; n = 5 cells) for shape covering volume represented by an infinite sheet of light with an axial dis-
the whole cell) (Fig. 2b). Some of the off-target current could be due tribution given by the experimental curve (Fig. 2c).
to the excitation of thin processes that were not visible. These results demonstrate that TF-GPC permits a precise two-
Because a key advantage of two-photon excitation is the ability photon activation of ChR2 channels at nonsaturating excitation den-
to confine the excitation narrowly along the z axis, we tested the sity, and that this stimulation evokes photocurrents that approach those
axial resolution by moving the whole cell–shaped excitation in achieved with blue-light, single-photon widefield illumination but that
z-axis steps of 2 μm. The photocurrent dropped off sharply at focal afford the axial resolution advantage of two-photon excitation.
points above and below the cell (Fig. 2c; FWHM = 13.5 ± 4.0 μm
Shaped two-photon excitation in cultured cortical neurons
a We used TF-GPC illumination patterns to photostimulate dissociated
mouse cultured cortical neurons transfected with plasmids encod-
ing ChR2(H134R)-GFP. We used widefield illumination to find the
ChR2-expressing neurons. We used fluorescence images of the cells to
generate local excitation patterns of different sizes and shapes (Fig. 3a).
In current clamp, a large illumination shape covering the whole cell
evoked action potentials (16 of 19 cells, 50 ms pulse). Action poten-
20 mV
Current (–pA)
Latency (ms)
for ChR2-YFP (λexc = 488 nm, 5 nm bandwidth).
4 mV
300 30
(b) Plot of the peak current (inward currents
200 10 ms 20
indicated in negative picoamperes) (n = 6 cells)
as a function of excitation spot diameter (average 10
100
on three trials in all cases, 0.52 mW μm−2, 10 ms 8 9 10 1112 13 14 15 10 11 12 13 14 15
pulse). (c) Voltage responses to photoexcitation Spot diameter (µm) Spot diameter (µm)
40 mV
20 mV
15 μm in diameter; 0.52 mW μm−2; 10 ms pulse). d
100 ms 50 ms
Action potential latency as a function of the 15 µm
spot 10 Hz
excitation spot diameter (right; n = 7 cells). Error
bars, s.e.m. Average of three trials is shown in 10 µm
spot
all cases. (d) Widefield epi-fluorescence image 20 Hz
of cell body loaded through the patch pipet with 7 µm
the fluorescent indicator Alexa Fluor 594 (left; λexc spot
5 μm spot (0.4 mW μm−2). Example of action potential firing after light stimulation at 10 Hz (5/5 trials), 20 Hz (5/5 trials) and 30 Hz (4/11 trials) (0.40 mW
μm−2; 10 ms pulse; 15 μm excitation spot; right). Scale bars, 20 μm. λexc = 920 nm, 40×, 0.8 NA objective, excitation depth of 50–70 μm.
From these results we conclude that two-photon TF-GPC gen- diameters: the average threshold was 9.8 ± 0.8 μm (± s.e.m.;
erates sufficiently large ChR2 photocurrents to reliably evoke n = 9 cells). The amplitude of the photocurrents increased, and
action potentials in primary neuronal cultures at excitation den- the latency to the action potential decreased as the diameter of the
sities well below the damage threshold25. excitation spot increased (Fig. 4b,c). Similar trends occurred with
increasing excitation density (Supplementary Fig. 4).
Shaped two-photon excitation in cortical brain slices Increasing the illumination area during a sustained (1 s) light
To explore the efficiency of TF-GPC in brain slices, we used coro- pulse increased action-potential firing (Fig. 4d), reaching a spik-
nal slices of somatosensory cortex from Thy1-ChR2-YFP trans- ing rate of 15 Hz (11.8 ± 0.8 Hz; 6 trials). Stimulation with a train
genic mice, in which wild-type ChR2 is expressed in pyramidal of brief light pulses evoked action-potential trains at 10 Hz or
neurons of cortical layer V (ref. 5). Using widefield imaging we greater (of eight cells, five reached 10 Hz, three reached 20 Hz
searched for YFP-ChR2–positive cells and used the fluorescence and two reached 30 Hz; Fig. 4d).
or transmission images to design and locate the excitation spot Finally, we tested the ability of TF-GPC to maintain the
(Fig. 4a). Most tested neurons (13/14) responded with an action two-photon excitation confinement in the greater depth and
potential to a 10-μm excitation spot (0.30–0.52 mW μm−2; scattering medium of the cortical slice. To test the lateral
10 ms pulse). To define the exact threshold for action-potential precision, we compared (Fig. 5a) the photocurrent evoked by
generation, we stimulated neurons with excitation spots of various an excitation shape covering the whole cell to the one evoked
by an illumination area covering the space surrounding the cell
(antishape). The shape-evoked current was about sevenfold
a Shaped Antishaped c larger than that evoked by the antishaped excitation in two
different cells (ratio between antishape- and shape-evoked cur-
rents was 7.7- and 7.1-fold, respectively). This demonstrated
an x-y axis contrast almost as sharp as what we observed in
monolayer cultures (Fig. 2b).
z = –5 µm
Shaped Figure 5 | TF-GPC provides lateral and axial precision in ChR2 activation
Antishaped
10 pA in brain slices. (a) Widefield fluorescence images of a ChR2-YFP positive
20 ms z = –3 µm
neuron filled with Alexa Fluor 594 and superimposed excitation patterns
(red) with shaped and anti-shaped profiles (top). Photocurrents evoked
b z = –2 µm
by shaped and antishaped excitation (bottom; 10 ms laser pulses,
∆z = 5 µm
Current (–pA)
20 mV 200
four identical spots placed elsewhere (bottom). White line indicates
10 ms the excitation field of 60 μm in diameter. Scale bars, 10 μm. (b) Action
100 potential evoked by single-spot (top) and multispot (bottom) illumination
(red bars) (0.29–0.34 mW μm−2; λexc = 920 nm). (c) Averaged (three trials per
Single-spot Multispot cell) amplitudes of photocurrents measured under voltage clamp in response
to single or multispot illumination (three spots of 12 μm diameter for
d four cells or five spots of 11 μm diameter in one cell). (d) Transmission
B A image of three ChR2-YFP–positive neurons (A, B and C) with 15 μm
excitation spots (red) superimposed. Scale bar, 10 μm. (e) Neurons A and
B were simultaneously patch clamped. A spot over neuron A triggered
C an action potential only in A. A spot over neuron B triggered an action
potential only in neuron B. When the three cells were illuminated
e f simultaneously both A and B fired an action potential. (f) After recordings
50 mV
shown in e, the patch pipet was removed from B and a patch recording
Stimulation Record Stimulation Record
50 ms was established in C. A spot over neuron A triggered an action potential
A A
A A
only in A. A spot over C triggered an action potential in C and a small
B C depolarization in A. When the three cells were illuminated simultaneously
both A and C fired an action potential (0.25 mW μm−2; λexc = 850 nm, 40×,
© 2010 Nature America, Inc. All rights reserved.
Photonics X10468-02), illuminated at oblique incidence by the by the software to the LCOS-SLM phase pattern (Supplementary
expanded laser beam (5×). The device was controlled by a cus- Fig. 6a–c). The thickness of the ring was adjusted so that the
tom-designed software16 that, given a target intensity distribution total phase-shifting area Aπ(tot) = Aπ(spot) + Aπ(ring) was always
at the focal plane of the microscope objective converted (in ~30 around one-quarter of the illuminated surface of the SLM, ASLM.
ms) the intensity map, A(x,y), into a binary phase map, Aφ(x,y)SLM The external diameter of the ring was adjusted to the size of the
and addressed (in ~15 ms) the output profile to the SLM. The tar- circular aperture placed in front of the SLM. As a result, the laser
get intensity distribution is typically obtained by selecting on the power at the output plane was distributed between the ring and
transmission or fluorescence picture a region of interest (ROI) and the excitation shaped area in such a way, that excitation density
automatically generated by the software using threshold detection in the excitation spot had approximately the same value, indepen
in the selected ROI. For excitation patterns precisely shaped on dently of its size.
cell or dendritic morphology, we increased contrast and defini- For temporal focusing, a blazed reflectance grating (830 lines
tion in primary fluorescence images by filling individual neurons per mm) was placed at the focal plane of the L2 lens, to geometri-
with Alexa Fluor 594 (Invitrogen Molecular Probes) through the cally disperse the frequencies of the ultrashort laser pulses and
patch pipet and imaging in wide-field (λexc = 590 nm, bandwidth temporally focus the beam on the sample plane. An illumination
10 nm). The total time for acquiring the fluorescence image and angle of 40.5° was chosen, such that the central frequency of the
generating the ROI was approximately 1 s. This time would not +1 order beam diffracted by the blazed grating (~80% of the inci-
slow the experiment, because a series of ROIs can be made during dent beam) was directed along the optical axis of the microscope
setting up the experiment, in advance of recording. while the zero order beam was blocked (for details concerning the
The beam reflected from the SLM was separated in its Fourier alignment of the grating, see reference 19). The tilted illumina-
components by a 400 mm focal length achromatic lens (L1) and tion of the grating caused a horizontal stretching of the original
focused on the PCF, positioned at the Fourier plane of the lens. An intensity pattern by a factor of cos (40.5°). To compensate for the
iris was placed in front of the SLM, whose radius (Rc = 5 mm) was tilt, the phase pattern was equally preshrunk along the x direction,
adapted to optimize the contrast of the output intensity distribu- at the SLM’s plane (Supplementary Fig. 6a–c). To block the ring
tion. The on-axis, low-spatial-frequency components were shifted added to balance the filling factor of the SLM from the excitation
in phase by the PCF and then, the second 300 mm achromatic field, an iris was placed right after the grating (not shown on the
lens (L2) recombined the high (signal wave) and low (synthetic layout for simplicity).
reference wave) spatial frequency components. The introduced The duration of the laser pulses for photoactivation was control-
phase shift caused the different components to interfere and led by a mechanical shutter (Uniblitz, VCM-D1 Shutter Driver)
produced an intensity distribution according to the spatial phase controlled either by the electrophysiology amplifier (Axon 200B)
information carried by the higher spatial frequencies. The lens L or by a home-made electronic circuit, based on a microcontrol-
(f = 500 mm) (Fig. 1a) and the microscope objective (Olympus, ler (PIC 18F4550, Microchip), driven by a PC interface through
LUMPLFL60XW/IR 60×, NA = 0.9 or LUMPLFL40XW/IR; 40×, a USB connection. In the last case the circuit synchronized the
NA = 0.8) formed the second 4-f lens system that scaled the inten- shutter, based on a trigger pulse sent by the Axon 200B amplifier.
sity distribution (~1/110) on the sample plane, via a dichroic mir- The shutter’s rise time was 1.9 ms (Supplementary Fig. 8).
ror (Chroma Technology 640DCSPXR). The theoretical calculation of the axial propagation of the exci-
The PCF is selected from a patterned phase mask, fabricated by tation beams around the objective focal plane was carried out by
depositing a photoresist on a glass optical flat (DanChip) contain- using the formalism described previously16.
ing circular pits of variable diameter (4–100 μm) arranged in a The experimental propagation was measured with a ‘double
squared array of 10 × 10 filters, which provide a half-wave phase microscope’ described in previous work16,17,19 using an upper
shift in the wavelength range 850–950 nm. The radius, R1, of the objective 60× (Olympus, LUMPLFL60XW/IR; NA 0.9). The axial
with 5% fetal bovine serum and 0.5% penicillin-streptomycin. Cells using a Digidata board (Axon instruments) and Clampex software
were transfected with 2 μg of plasmids encoding ChR2(H134R)- (Axon Instruments).
GFP33 using ExGen 500 (Euromedex). All recordings were per- D-APV was purchased from Tocris Bioscience, and 2,3-dioxo-
formed 24–48 h after transfection, at room temperature (22 °C). 6-nitro-1,2,3,4-tetrahydrobenzo[f ]quinoxaline-7-sulfonamide
Dissociated embryonic mouse cortical neurons (E16-17) were plated disodium salt (NBQX) was purchased from Ascent Scientific.
on poly(l-lysine)-coated glass coverslips at a density of 1.5 × 105 cells. Stock was prepared in water and dissolved in extracellular solu-
Neurons were maintained in MEM supplemented with 5% FBS, serum tion just before experiments. All other chemicals were purchased
extender (BD Biosciences) and 0.5% penicillin-streptomycin. Ara-C from Sigma.
(4 μM) was added after DIV6. Cells were transfected by the calcium Data analysis was performed using either the Clampfit analysis
phosphate method using 1 μg of plasmids encoding ChR2(H134R)- software (Axon Instruments) and custom routines within the Igor
GFP per 12 mm coverslip. Recordings were performed 24 h after trans- Pro environment (Wavemetrics) or in Origin 8.0 (OriginLab).
fection, at room temperature. Statistical data are reported as mean ± s.e.m. The significance
Coronal cortical slices (250–280 μm) were obtained from wild was calculated using paired Student’s t test.
type and Thy1-ChR2-YFP mice34 older than postnatal day 30
purchased from the Jackson Laboratory (B6.Cg-Tg(Thy1-COP4/ Photodamage assessment protocols. In slices, photostimulation
EYFP)18Gfng/J strain). of pyramidal neurons with twenty 10-Hz trains of 50 ms pulses
Somatosensory slices were prepared in an ice-cold slicing every 10 s (0.44 mW μm−2) or with thirty 400-ms pulses at the
solution35 containing 130 mM K-gluconate, 15 mM KCl, 0.2 mM maximum excitation power (0.53 mW μm−2) did not induce vis-
EGTA, 20 mM HEPES, 25 mM glucose, 50 μM D-APV and 50 nM ible cell damage. Neurons were still alive and able to fire action
minocycline (pH 7.4) and then transferred for 30 s to a recovery potentials at the end of each protocol (data not shown).
solution containing 225 mM d-mannitol, 2.5 mM KCl, 1.25 mM In the same preparation, illumination protocols similar to
NaH2PO4, 26 mM NaHCO3, 25 mM glucose, 0.8 mM CaCl2, those used in Figures 3 and 4 elicited no response in ChR2-
8 mM MgCl2, 50 μM D-APV and 50 nM minocycline (95% O2 and positive neurons if a block was inserted in the excitation path
5% CO2). Before recording slices were incubated for at least 30 min preventing the laser light from reaching the preparation or in
at 33 °C in external solution containing 125 mM NaCl, 3.5 mM ChR2-negative neurons.
KCl, 1.25 mM NaH2PO4, 26 mM NaHCO3, 25 mM glucose, 2 mM
CaCl2, 1 mM MgCl2 (95% O2 and 5% CO2).
Recordings were performed at room temperature.
31. Glückstad, J. & Mogensen, P.C. Optimal phase contrast in common-path
interferometry. Appl. Opt. 40, 268–282 (2001).
Electrophysiology and data analysis. Patch-clamp pipets (4–8 MΩ 32. Palima, D. & Glückstad, J. Multi-wavelength spatial light shaping using
for neurons and 2–4 MΩ for HEK 293 cells) were back-filled with generalized phase contrast. Opt. Express 16, 1331–1342 (2008).
intracellular solutions containing 145 mM CsCl, 10 mM HEPES, 33. Nagel, G. et al. Light activation of channelrhodopsin-2 in excitable cells
of Caenorhabditis elegans triggers rapid behavioral responses. Curr. Biol.
0.5 mM CaCl2, 1 mM MgCl2, 5 mM EGTA (pH 7.2) for HEK 15, 2279–2284 (2005).
293 cells; 131 mM K-gluconate, 10 mM NaCl, 10 mM HEPES, 34. Arenkiel, B.R. et al. In vivo light-induced activation of neural circuitry in
2 mM MgCl2, 2 mM MgATP, 1 mM EGTA (pH 7.4) for neurons transgenic mice expressing channelrhodopsin-2. Neuron 54, 205–218
(2007).
in culture; and 137 mM K-MeSO4, 6 mM NaCl, 5 mM MgCl2,
35. Otsu, Y. et al. Optical monitoring of neuronal activity at high frame rate
0.05 mM EGTA, 10 mM HEPES, 4 mM K2ATP and 0.4 mM with a digital random-access multiphoton (RAMP) microscope. J. Neurosci.
GTP-Na2 (pH 7.3) for brain slices. In some recordings, 10 μM Methods 173, 259–270 (2008).