Two-photon single-cell optogenetic control ofneuronal activity by sculpted lightBertalan K. Andrasfalvy, Boris V. Zemelman, Jianyong Tang, and Alipasha Vaziri1

Howard Hughes Medical Institute, Janelia Farm Research Campus, Ashburn, VA 20147

Communicated by Charles V. Shank, Howard Hughes Medical Institute, Janelia Farm Research Campus, Ashburn, VA 20147, May 14, 2010 (received for reviewApril 6, 2010)

Recent advances in optogenetic techniques have generated newtools for controlling neuronal activity, with a wide range of neu-roscience applications. The most commonly used approach hasbeen the optical activation of the light-gated ion channel channelr-hodopsin-2 (ChR2). However, targeted single-cell-level optogeneticactivation with temporal precessions comparable to the spiketiming remained challenging. Here we report fast (≤1 ms), selec-tive, and targeted control of neuronal activity with single-cellresolution in hippocampal slices. Using temporally focused laserpulses (TEFO) for which the axial beam profile can be controlledindependently of its lateral distribution, large numbers of channelson individual neurons can be excited simultaneously, leading tostrong (up to 15mV) and fast (≤1ms) depolarizations. Furthermore,we demonstrated selective activation of cellular compartments,such as dendrites and large presynaptic terminals, at depths up to150 μm. The demonstrated spatiotemporal resolution and the se-lectivity provided by TEFO allowmanipulation of neuronal activity,with a large number of applications in studies of neuronal micro-circuit function in vitro and in vivo.

high-resolution neuronal stimulation | channelrhodopsin |temporal focusing | circuit mapping | electrophysiology

Artificial stimulation and inhibition of neuronal activity hasapplications ranging from fundamental neurobiology ques-

tions (1–4) to potential clinical treatment of neuropsychiatricdisorders (5, 6). Historically, this task has been achieved mainlyby electrical stimulation. Yet the recent development of genetictechniques for sensitizing neurons to optical stimulation (7–12)and silencing (13–16) have, for the first time, provided cell type–specific control of neuronal activity, which has been successfullyused to address significant biological questions (1–3, 17–19). Inthe most widely used approach, genetically expressed light-gatedion channels, such as channelrhodopsin-2 (ChR2) (20, 21), or ionpumps, such as halorhodopsin or archaerhodopsin-3 (22, 23), andoptical methods for triggering their function are combined tocontrol neuronal activity. However, although optical activation ofChR2 has been suitable to induce population activity in a largenumber of neurons, directed, efficient, and fast stimulation ofsingle cells has not been feasible. The main reason for that is thelow channel conductance of ChR2 (21, 24). To achieve sufficientdepolarizations, a large number of channels have to be activatednearly simultaneously, which typically extend over an area of tensof square micrometers. Although conventional one-photon exci-tation satisfies this requirement, light scattering and the extendedaxial beam parameter of the excitation area lead typically to in-evitable activation of neurons in an untargeted fashion. Somesuccess in addressing these issues has recently been reported (25,26) using a two-photon scanning (27) approach. Although thesestudies (25, 26) could show high spatial resolution, the necessarytime to activate a large surface area sufficient to fire actionpotentials (APs) was ≈30 ms. Thus, neither the one-photon northe two-photon scanning excitation provides the necessary com-bination of high spatial selectivity and the ability to stimulatea membrane area that is large enough to produce simultaneousand significant rapid depolarizations on a single neuron.

ResultsChR2 Activation by Two-Photon Temporal Focusing. We took a dif-ferent approach for addressing the shortcomings of both the one-photon and the two-photon method in a fundamentally uniqueway and have demonstrated reliable single neuron–specific acti-vation with temporal resolution (≤1 ms) in hippocampal slices.This was done by effectively decoupling the lateral (i.e., the waistsize) and the axial (i.e., the Rayleigh length) beam parameters fortwo-photon absorption, which are usually coupled for a Gaussianbeam, and thereby sculpting the spatial light distribution at thefocal plane. For a pulsed laser source, the decoupling can effec-tively be achieved by using the spectrum of the pulse to control itstwo-photon absorption probability in the axial direction. This hasbeen demonstrated in the technique of temporal focusing (28,29), which has also found applications in widefield two-photonimaging (30) and in 3D multilayer super-resolution microscopy(31). Two-photon temporal focusing (TEFO) can be experi-mentally realized by first spatially broadening a femtosecondoptical pulse using a diffraction grating. The spot on the grating isthen imaged onto the sample plane using a telescope that consistsof the microscope objective and an additional lens. This config-uration leads to an axial geometry in which the pulse is broadenedeverywhere in the sample except at the image plane, where thespatially separated paths of the different frequency componentsin the pulse meet again and the pulse reaches its minimum width.(Fig. S1 A and B). The two-photon excitation probability inTEFO is inversely proportional to the pulse width squared, whichleads to a depth of field ≈2 orders of magnitude shorter (31) thanthe widefield one-photon epifluorescence technique. This allowsa sectioning capability close to a confocal or two-photon micro-scope (27), while providing simultaneous excitation of an areathat is ≈3 orders of magnitude larger than the diffraction-limitedspot (Fig. S1C). Therefore, TEFO can be used for simultaneousexcitation of a thin disk-like volume in a biological sample. Inthe present work we took advantage of this fact and demonstratedthat TEFO can be used for targeted single-cell excitation ofChR2 with high spatiotemporal resolution in mouse and rathippocampal slices.After confirming the above parameters of the excitation vol-

ume for our temporally focused beam (Figs. S1B and S2), weinvestigated the activation of ChR2 expressed in cultured humanembryonic kidney (HEK293) (Fig. 1A) cells and measuredChR2-mediated currents with whole-cell voltage clamp record-ings. Fast (≤2 ms rise time) and large (≤1 nA) inward currents

Author contributions: B.K.A. designed and performed all electrophysiological experiments;B.V.Z. prepared plasmids, designed AAV-ChR2-sfGFP and Cre recombinase-dependent rAAV-FLEX-rev ChR2-sfGFP viruses, and helped with the manuscript; J.T. developed software fordata collection; A.V. designed and led research, built experimental setup, and performedoptical characterization experiments; B.K.A. and A.V. analyzed data; and B.K.A. and A.V.wrote the paper.

The authors declare no conflict of interest.

Freely available online through the PNAS open access option.1To whom correspondence should be addressed. E-mail: vaziria@janelia.hhmi.org.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1006620107/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1006620107 PNAS | June 29, 2010 | vol. 107 | no. 26 | 11981–11986

NEU

ROSC

IENCE

APP

LIED

PHYS

ICAL

SCIENCE

S

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 1,

202

1

mediated by ChR2 were detected using pulses between 1 and 100ms (Fig. 1 B and C). We aimed to systematically identify theoptimal parameters for two-photon temporal focusing excitationand investigated the activation of ChR2 as a function of laserpower, wavelength, and beam size (Fig. 1 D–I). Keeping a con-stant beam diameter of 3.8 μm and pulse duration of 100 ms fora range of wavelengths between 800 nm and 960 nm, we firstmeasured the peak amplitude and rise time of the inward currentas a function of laser power (Fig. 1 D–F). At each wavelength forlower powers we observed a nonlinear dependence of ChR2 ac-tivation on power, which is the signature of the two-photon effect(27). This quadratic dependence on power is not evident at firstsight over the studied range of powers, because it is overshad-owed by other parameters of the channel kinetics, such as satu-ration and desensitization. However, under our experimentalconditions it could clearly be seen for the power range between0 and 50 mW (Fig. 1D, Inset). At high powers the relationshipshowed near-asymptotic behavior, likely indicating the saturationof channel activation (26). These experiments revealed themaximum two-photon activation efficiency to be at 880 nm, whichis slightly blue shifted compared with the double of the one-photon absorption peak (Fig. 1F). Once the optimum wavelengthwas identified, we examined the dependence of the response onthe beam size. Using 100-ms-long pulses at 880 nm we varied thepower at three different spot sizes (6 μm, 10 μm, and 14 μm). The

peak inward current increased with power for all beam sizes (Fig.1G); however, by further reducing the spot size to 4 μm we couldshow that the highest efficiency of activation (as shown by thelarger peak inward current and faster rise time; Fig. 1H) could beachieved with the ≈6-μm spot (Fig. 1I). This fact reflects the in-trinsic tradeoff between the quadratic dependence of the proba-bility for opening a channel on the intensity, which increases withdecreasing spot size, and the total number of channels on theilluminated surface area, which decreases with the spot size. Theinduced peak current (or the total charge influx) is proportionalto the intensity squared and the area of illumination. Therefore,for a constant photon number (or a constant power) the peakcurrent should be inversely proportional to the square of the potsize (Fig. 1G, Inset). All cells were tested for activation of ChR2with blue light (488 nm), which revealed comparable biophysicalparameters to TEFO excitation (Fig. S3 A–E).

Temporal Resolution of TEFO. These results motivated us to in-vestigate the potential of this method as a tool with high spa-tiotemporal resolution for neuronal circuit mapping and otherneuroscience studies. We turned to hippocampal slices and useddifferent types of ChR2-expressing neurons: CA1 pyramidalcells (CA1PCs) in Thy1-ChR2-YFP transgenic mice (32) andparvalbumin-positive interneurons (PV-INs) in PV-Cre mice(33) that were infected with a Cre-dependent ChR2-GFP con-

0

200

Pea

k cu

rren

t (pA

)

400

600

800

E

H I

F

A B C

100 pA20 ms

D

control

820 nm488 nm

6 8 10Spot size (µm)

4 14120

4

Rise tim

e (ms)

8

12

16

20

0

200

Pea

k cu

rren

t (pA

)

400

600

100 200Power (mW)

0 4003000

10Ris

e tim

e (m

s)

20

30

100 200Power (mW)

400300

4050

G

120

160

Pea

k cu

rren

t (pA

)

200

240

280

840 880

Wavelength (nm)

800 9609200

200

Pea

k cu

rren

t (pA

)

400

600

100 200Power (mW)

0 4003000

8

Ris

e tim

e (m

s)

12

16

100 200Power (mW)

400300

20

24

110100

ms

100 pA20 ms

6 µm10 µm14 µm

4

0

24

0

12

400200

800 nm880 nm800

840880920960

nm

100

200

300

6 8 10 12 14Spot size (µm)

Pea

k cu

rren

t (pA

)0 5025

30

60

90

Fig. 1. Characterization of the biophysical properties of TEFO activation of ChR2 in HEK293 cells. (A) Superimposed fluorescent and differential interferencecontrast (DIC) image of a HEK293 cell during recording. (B) Representative traces of induced inward current by one-photon excitation of ChR2 (blue), TEFOexcitation of ChR2-expressing (red), and of control uninfected cell (black), using 100-ms-long pulses. (C) Representative traces of TEFO-activated ChR2-mediated inward current with different pulse durations. (D and E) Peak current amplitude (D) and rise time (E) at different wavelengths measured asa function of average power at the sample for a ≈3.8-μm spot and 100-ms duration (n = 22 cells). Inset in D shows the nonlinear dependence of peak inwardcurrent on the power at lower power levels, which is the result of two-photon excitation of ChR2. (F) Wavelength-dependency of TEFO excitation of ChR2 (n =16 cells). Inset in F emphasizes the differences in activation efficiency of 800-nm and 880-nm light. (G and H) Peak current (G) and rise time (H) as a function ofspot size at 880-nm wavelength and 100-ms duration (n = 13 cells). Inset in G shows the induced peak current with the TEFO for a constant photon number(constant power at 200 mW) for different spot sizes. Maximum amplitude and fastest rise time could be achieved with a spot size of ≈5–6 μm (I). Dots anderror bars represent means ± SEM.

11982 | www.pnas.org/cgi/doi/10.1073/pnas.1006620107 Andrasfalvy et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 1,

202

1

taining adeno-associated virus (ChR2-GFP-AAV) (34) (Materi-als and Methods and SI Materials and Methods). First we aimed todetermine the efficacy and basic biophysical characteristics ofour technique to explore the optimal parameters for differenttypes of applications. Somatic current-clamp recordings wereperformed in ChR2-expressing neurons while the soma was il-luminated by a temporally focused spot of ≈5-μm diameter with1-, 2-, 5-, 10-, and 100-ms pulse duration at 880 nm with ≈460-mW power at the sample (Fig. 2A). Similar to HEK293 cells, inboth types of neurons fast (≈1–5 ms) and large (≈5–15 mV)ChR2-mediated depolarizations could be evoked using TEFOexcitation (Fig. 2A). At a holding potential of −80 mV (to pre-vent generation of APs), the ChR2-induced membrane de-polarization continuously increased with pulse duration andreached a saturation value of ≈15–20 mV for pulses longer than≈10 ms (Fig. 2B). In these experiments we used a static beampath and moved the specimen to position the temporally focusedspot for targeting a certain area. However, to produce fast spa-tiotemporal activation patterns, it would be advantageous to beable to move the excitation spot on a submillisecond time scaleto any desired number of target points. To achieve this we in-tegrated the temporal focusing beam into a microscope equippedwith a pair of laser scanning mirrors (Materials and Methods andSI Materials and Methods). We tested this technical advance byrecording from CA1PCs and subiculum pyramidal cells (SubPCs)

in Thy1-ChR2-YFP mice and scanning variable spots in the sliceon and around the somata and dendrites of the patched cells. Asexpected, ChR2-mediated responses were detected only whenthe excitation spot was placed precisely on the soma or thedendrite, but no response was evoked when the beam was placedaway from the recorded cell (Fig. 2C). Using very short pulsedilluminations (0.1 ms) with short interpulse intervals (0.1 ms) atmultiple points positioned on the soma, we could effectivelyincrease the activated surface area and decrease the necessarytime to achieve the desired depolarization level to evoke APswithin a submillisecond time scale (Fig. 2D). (Further advantagesof the multiple-spot activation along single dendrites are dis-cussed in Fig. S4).

Spatial Resolution of TEFO. Precise spatial resolution of excitationwas indicated by the elimination of ChR2-mediated de-polarization when the beam was moved a few microns away fromthe dendrite (Fig. 3 A and B). Given that dendrites providea spatially confined and a much smaller excitable region (≈1–2μm diameter dendritic compartments) than the soma, they wereideal for precise measurement of the spatial resolution of ourTEFO technique. We used thin apical dendrites (50–400 μmfrom the soma, in 50–160 μm depth) and quantitatively com-pared the spatial resolution of our techniques with that acquiredwith 488-nm light. We displaced the temporally focused spotrelative to the dendrite in 3D while recording voltage signals atthe soma (Fig. 3 A–D). When the TEFO excitation spot wasmoved laterally (x-direction) or axially (z-direction), the peakdepolarization dropped off sharply to 50% of the peak responseafter ≈10 μm and was almost eliminated at ≈20 μm distance [Fig.3 B, C, and D (red trace)]. However, in comparison, responsesinduced by 488-nm light only dropped off by ≈50% at ≈50 μmaway from the dendrite (x-direction) [Fig. 3B (blue trace)] andremained almost unaltered at 100 μm above or below the den-drite (z-direction) (Fig. 3 C and D). This result demonstrates a5-fold improvement in the lateral excitation precession and atleast a 10-fold improvement in the axial excitation precession.

0 2 4 6 8 10 120

5

10

15

20

Pulse duration (ms)

Pea

k de

pola

rizat

ion

(mV

)

2 mV20 ms

12510100

ms488nm

880 nm

10 mV

A B

4

2

1 3

0.1 s2 mV

D

1 3 2 4

C

50 ms20 mV

50 ms2 mV

0.1 ms3x

5x4x

20 µm

// //

20 µm

Fig. 2. Temporal resolution of TEFO activation of ChR2 in hippocampalneurons of acute brain slices. (A) Representative traces and summary graph(B) (n = 5) of depolarization evoked by 1- to 100-ms-long TEFO pulses at thesomata of hippocampal neurons. (A, Inset) Representative response to 488-nm light. (C) Single focal section of a ChR2-expressing CA1PC loaded withAlexa594 in hippocampal slice from Thy1-ChR2-YFP mouse, with numberedTEFO illumination spots indicated by red dots. Bottom traces show mem-brane potential responses to 2-ms-long stimulations at the spots indicated inC. Note that somatic stimulation fires the cell reliably, whereas dendriticstimulation results in smaller depolarization, and stimulation at a spot out-side of the cell area has no effect. (D) CA1PC soma with five two-photontemporal focusing spots indicated by dots. (Lower) Response to 1-ms-longTEFO excitation at a single spot (black) as well as to 0.1-ms-long two-photontemporal focusing illuminations at three to five spots with 0.1-ms intervals.Individual excitations are also shown (Lower, interval = 300 ms). Note thatscale bar for the uppermost trace differs from that related to the four lowertraces. Depolarization was more effective with multiple short illuminationsthan that obtained with a single 1-ms pulse, despite the shorter total exci-tation time.

0 20 40Lateral distance (µm)

0 50 100Axial distance (µm)

-50-100

0

0.2

0.4

0.6

0.8

1.0

0

0.2Nor

mal

ized

dep

ol.

0.4

0.6

0.8

1.0

1.2

1 mV50 ms

20 µm

B

C

A

Nor

mal

ized

dep

ol.

0

50

100

20

-20

-100 µm

-50

D

2 mV20 ms

488 nm880 nm

488 nm880 nm

100 % 50 %

488 nm

880 nm

Fig. 3. Spatial resolution of TEFO activation of ChR2. (A) Stack image ofa CA1PC loaded with Alexa594, showing location of laterally moved TEFOactivation spots. (B) Lateral localization of response induced by 488-nm light(blue, n = 4) and TEFO (red, n = 3). (C) Axial localization of response inducedby 488-nm light (blue) and TEFO with 50% (≈130 mW, red) and 100% (≈260mW, black) power (dashed lines indicate Gaussian fit). (Inset) Individualresponses at the focal plane of the dendrite (color-coded respectively). (D)Representative responses to 488-nm (blue) and TEFO-evoked (red) responsesat different distances above and below the dendrite.

Andrasfalvy et al. PNAS | June 29, 2010 | vol. 107 | no. 26 | 11983

NEU

ROSC

IENCE

APP

LIED

PHYS

ICAL

SCIENCE

S

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 1,

202

1

The axial resolution could be further improved by decreasing theapplied power by 50% (Fig. 3C). This is a strong indication, asalso discussed in a recent publication (26), that saturation ofChR2 in the center of the target plane leads to an additionalcontribution of the out-of-focus regions to the evoked response.

Depth Penetration. We next quantified how the efficacy of TEFOexcitation depended on the depth in the tissue. The maximumdepolarization evoked by single 1-ms pulses, placed on somatalocated at variable depths in the slice, showed a linear attenuationwith depth (Fig. 4A). However, as demonstrated above, the totalinduced depolarization could be increased by using the multispottemporal focusing excitation approach. As a result, when a mul-tispot temporal focusing excitation pattern with pulsed illumina-tion durations of 0.1–0.2 ms and interpulse intervals of 0.1 ms wasused, the efficiency of the excitation was increased. As shown inFig. 4A, this led for the studied cells to a more than 2-fold in-crease of depth of excitation for the same total illumination du-ration. This dependence of the spatiotemporal patterning ofexcitation was used to induce depolarizations that were strongenough to evoke APs even at the apical tuft at >150 μm depth,where depolarization induced by a single short-pulse illuminationwas barely detectable (Fig. 4B).

Induction of Suprathreshhold Activity. In current clamp, at theresting membrane potential (−58 to −67 mV), we set out todetermine the parameters necessary to induce the minimal de-polarization that was sufficient to reach AP threshold in a givencell of all three neuron types included in our study (Fig. 5A). Weused either one spot with variable pulse durations (PV-INs andtwo of six CA1PCs; total t = 1–5 ms), or five spots with sub-millisecond pulses (four of six CA1PC, and all SubPC, total t =0.1–0.5 ms × 5). In all cells tested (n = 13), total pulse durationof ≤5 ms evoked suprathreshold depolarization and AP firing(PV-INs: 3.0 ± 1.2 ms, n = 3; CA1PCs: 2.5 ± 0.8 ms, n = 6;SubPCs: 1.3 ± 0.1 ms, n = 4; Fig. 5A). We next tested the abilityof our method to induce high-frequency firing in PCs and PV-INs. We found that all cell types usually fired APs at the first fewstimulations in the train. Subsequently, PCs responded to thetrain of stimuli only intermittently (Fig. 5 B–D). Increasing pulseduration partially overcame AP failures and improved firing

fidelity (Fig. 5C). This reduced firing probability is the conse-quence of the spatial confinement of the TEFO excitationleading to an inactivation of ChR2 at higher frequencies, aneffect previously also observed for blue light excitation (10, 32,35, 36). Only recently a new genetic manipulation of ChR2 (12)could overcome the ChR2 inactivation problem at higher fre-quencies of stimulation.

Large Boutons Activation by TEFO. Our early data showing the lackof effect of glutamate receptor blockers on ChR2-evoked de-polarization in Thy1-ChR2-YFP mice (Materials and Methodsand Fig. 2) suggested that excitatory axons of ChR2-expressingprincipal cells, innervating the recorded cell, are not activated bytwo-photon temporal focusing within our standard parameterrange. Further investigations of excitatory axonal activation inthe CA1 region confirmed this fact (Fig. S5). Although we couldnot evoke synaptic responses from Schaffer collaterals, we hy-pothesized that certain types of terminals/axons with more fa-vorable properties could be excited by TEFO. Synaptic boutonsof PV-INs are possible candidates, because they are relativelylarge and they form multiple contacts on target cells (37), thusthese could provide more surface area to the temporally focusedtwo-photon illumination. To investigate that, we recorded from(ChR2-negative) CA1PCs that were innervated by ChR2-expressing PV-INs (mostly basket, axo-axonic, and bistratifiedcells) in AAV-ChR2-GFP virus-infected PV-Cre mice (Materialsand Methods and SI Materials and Methods) in the presence ofionotropic glutamate receptor antagonists. The temporally fo-cused beam spot was first placed around the soma and proximaldendrites of the recorded CA1PC, overlapping with the locationof ChR2-expressing boutons provided by PV-INs. Our experi-mental conditions were optimized to detect GABAergic post-

A 0.1 ms1 ms

0.1 ms x 5

0

5

10

15

Dep

olar

izat

ion

(mV

)

0.2 ms0.2 ms x 5

Depth of soma (µm)-80-100 -60 -40

50 ms20 mV

10 x 3

50 ms0.5 mV

10 µmB

Fig. 4. Dependence of induced depolarization on tissue depth. (A) De-pendence of induced depolarization at the somata of CA1PCs from Thy1-ChR2-YFP mice on tissue depth and spatiotemporal patterning of TEFO ex-citation for three different pulsed illumination durations and interpulsetimes. Red, single 1-ms pulses (n = 6, fitted linearly); open black dot andsquares (total n = 3), single submillisecond pulses; filled circle and black dots(total n = 3), 5× 0.1-ms spatiotemporally patterned excitation. Note thatrepeating the 0.1-ms pulses five times on the same cells (filled black dot andsquares, total excitation time ≤1 ms) yields a higher total depolarizationcompared with the single-spot 1-ms excitation. This can be used to inducebigger depolarizations at larger tissue depths. (B) Ten pulses (0.1 ms long)repeated three times with 0.1-ms intervals at the indicated locations in theapical tuft (distance from soma ≈400 μm, depth in slice ≈150 μm) evokedsufficient depolarization to evoke somatic AP (upper trace), whereasresponses to single activations were almost undetectable (bottom traces).

10 mV50 ms

A B

C

10 mV20 ms

20 Hz

50 Hz

100 Hz

D

20 mV20 ms

20 mV20 ms

20 Hz

50 Hz

100 Hz

20 Hz

20 Hz

50 Hz

0

5

10

15

Dep

olar

izat

ion

(mV

)

0

2

4

6

8

Pul

se d

urat

ion

(ms)

CA1 Subiculum

2 ms

5 ms

2 ms

0.2 ms x 5

PV-IN PCPC

5 ms CA1-PCSub-PC

PV-IN

Fig. 5. Fast and reliable AP responses by TEFO ChR2 excitation. (A) ChR2-mediated depolarization efficiently evoked APs in CA1 PV-INs (n = 3) andpyramidal cells (PC) in CA1 (n = 6) and subiculum (n = 4). ChR2-induced de-polarization was measured by hyperpolarizing the soma as necessary toprevent AP firing. Graphs indicate the difference between resting mem-brane potential and AP threshold (green circles), amplitude of ChR2-inducedsuprathreshold depolarization (red squares), and the corresponding pulseduration (black squares) for each cell type. (B) Action potentials evoked inChR2-expressing PV-IN by 1-ms-long pulses at 20–100 Hz. Insets: Sub-threshold responses at hyperpolarized holding potential. Similar resultswere obtained in n = 3 cells. (C and D) Representative responses of a CA1PC(C, n = 7) and a subiculum PC (D, n = 6) to 20-, 50-, or 100-Hz stimulation withdifferent pulse durations using single or multiple spots with different illu-mination time.

11984 | www.pnas.org/cgi/doi/10.1073/pnas.1006620107 Andrasfalvy et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 1,

202

1

synaptic hyperpolarization (holding potential, −60 to −65 mV;reversal potential of chloride, −80 mV). Strikingly, using pulsedurations of 1–2 ms, we could evoke inhibitory synaptic responsesof 0.5- to 4.0-mV amplitude that were completely eliminated bya specific GABAA receptor antagonist (SR-95531, 10 μM, n = 3)(Fig. 6A). Increasing the pulse duration did not lead to any furtherincrease of the inhibitory postsynaptic potential (IPSP) amplitude,suggesting that local synapses making contact with the post-synaptic soma were activated in an all-or-none fashion (Fig. 6B).One-photon excitation at 488 nm resulted in ≈6–9-mV hyperpo-larization (Fig. 6 A and B), presumably because the longer axialsize of the beam and scattering lead to activation of significantlymore synapses. To examine whether activation of boutons elicitedAPs in the axon, we tried to evoke postsynaptic responses atdifferent spatial locations of the activation beam. We marked 5spots around the recorded CA1PC soma (Fig. 6 C–E, Soma), 10spots in stratum pyramidale (Fig. 6 C–E, Pyr), 10 spots in proximalstratum radiatum (Fig. 6 C–E, Rad), and finally 5 spots on aChR2-expressing interneuron soma (Fig. 6 C–E, IN). In all of thelocations, synchronous activation of the spots evoked IPSPs in the

postsynaptic cell, and the responses were eliminated by applica-tion of tetrodotoxin (TTX, 0.5 μM; Fig. 6D). Notably, the latencyof the responses varied depending on the distance and activationtime of the excitation spot (Fig. 6E), further indicating that axonalAPs were elicited and propagated to evoke the synaptic responses.These results altogether demonstrate that the TEFO technique isefficient to directly activate the large (≈2–3 μm diameter) synapticboutons (or population of boutons) arising from PV-INs (37).

DiscussionAlthough the use of genetically expressed ChR2 in different typesof neurons has become widely popular in recent years, one-photon excitation does not allow for spatial precision of neuronalactivation using this method. Recognizing this caveat, recentstudies reported the possibility of two-photon excitation of ChR2using a scanning approach in cultured neurons (26). Although thespatial resolution indeed dramatically improved with two-photonexcitation, the long scanning time necessary for sufficient channelactivation and consequent depolarization significantly com-promises the temporal precision of neuronal activation (≈30 msfor evoking an AP). The TEFO technique, which does not requirescanning, provides reliable activation of cells within ≈1–3 ms andthereby provides ≈15 times faster activation than the scanningapproach. This improvement practically implies that only the two-photon temporal focusing approach allows quasi-synchronousactivation of neurons and their different cellular compartments infuture studies. The combination of TEFO with a conventionaldual galvanometer–based scanning system, as described in ourstudy, makes repositioning of the excitation spot fast (<0.2 ms toany point in a 100-μm field) and convenient. The precise spatialand temporal control of firing activity of individual or a desirednumber of individually selected cells, especially when combinedwith selective expression of ChR2 in particular cell populations,opens up the possibility (among others) for detailed, high-throughput studies of connectivity and dynamics of small- andlarge-scale neuronal networks and assessment of the functionalconsequences of their activation both in vitro and in vivo.Although the current TEFO technique can reliably stimulate

a few APs, further advances of the technique, such as 3D lightsculpting and development of ChR2 mutants with higher con-ductance or less desensitization (35), are expected to enable pre-cise induction of longer AP trains in a wide range of frequencies.Beside the reliable generation of APs at the soma, an importantaspect of our study is the demonstration of the possibility forspatially and temporally precise activation of certain cellularcompartments, such as dendrites and large synaptic boutons. Thisfeature can be exploited in many applications, for example forstudying dendritic properties and integration without or combinedwith uncaging methods (38), or for detailed investigation of thespatiotemporal characteristics and role of inhibition by geneticallylabeled interneuron types. On the other hand, the lack of activa-tion of excitatory synaptic terminals, such as Schaffer collaterals,shows that different types of synaptic boutons, presumably havingdiverse anatomical and biophysical properties, can exhibit differ-ent levels of excitability by the TEFO technique.The lack of axonal activation, at least under these experimental

conditions and within the explored wide parameter range, can becrucial in studies in which target neurons lie within a denselyintermingled network of thin ChR2-expressing axons with smallboutons, and therefore makes the TEFO technique a powerfultool for studies such as circuit mapping. On the other hand, theability to specifically activate axons with TEFO could be useful,for example, for mapping long-range connections in the brain.Further technical developments, such as combination of ourcurrent method with holographic techniques (39), are beingconsidered to allow targeted and specific stimulation of axonalcompartments with sculpted light. In summary, our method oftwo-photon excitation of ChR2 by temporal focusing, especially

A

C

E *

1 mV

20 ms

1 mV10 ms

PC somaPyrIN

B

IN

SomaPyr

Rad

20 µm

1 mV

40 ms

SR-95531

880 nm488 nm

-4

Hyp

erpo

lariz

atio

n (m

V)

-2

0

0 4 8Pulse duration (ms)

102 6 12

-6

1 mV300 ms

IN

2 mV300 ms

Soma

Rad

Pyr

2 mV300 ms

+TTX

+TTX**

*

*

**

+TTX

* ** **** **

* * ** **** **

* ****

**

x5x5

x10

x10x10

x5x5

* ****

D

Fig. 6. Large inhibitory synaptic terminals are activated by TEFO ChR2 ex-citation. (A) Representative traces of optically evoked hyperpolarizingpostsynaptic potentials in a CA1PC innervated by ChR2-expressing PV-INs(red, TEFO excitation; blue, 488-nm excitation) by targeting the soma. SR-95531 (10 μM) blocked responses with both types of stimulations (blacktraces, representative for n = 3 cells). (B) Peak hyperpolarization evoked byTEFO (red) did not depend on pulse duration (n = 6). Excitation at 488 nm(blue) evoked larger responses than TEFO excitation (n = 5). Dots and errorbars represent means ± SEM. (C) Stack image showing a recorded CA1PC andindicating locations of series of TEFO illumination spots at the soma (5 spots),stratum pyramidale (Pyr; 10 spots), stratum radiatum (Rad; 10 spots), and atthe soma of a PV-IN (IN; 5 spots). (D) Representative traces showingresponses to illumination at the locations indicated in C, using 100-msstimulus intervals (green), 0.1-ms stimulus intervals (red), and 0.1-ms stimulusintervals in the presence of 0.5 μm TTX (black). (E) Overlaid traces evoked bystimulation at the soma, in stratum pyramidale (Pyr), and at the soma of thePV-IN (IN). Note the differences in response latency and the stepwise rise ofthe response evoked in stratum pyramidale.

Andrasfalvy et al. PNAS | June 29, 2010 | vol. 107 | no. 26 | 11985

NEU

ROSC

IENCE

APP

LIED

PHYS

ICAL

SCIENCE

S

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 1,

202

1

when combined with genetic and electrophysiological techniques,has broad potential to contribute to our understanding of thecellular and network basis of neuronal functions.

Materials and MethodsOptical Setup. Please refer to SI Materials and Methods for a detailed de-scription. Briefly, a tunable 140-fs pulsed laser was used as the source fortemporal focusing. After passing through a variable zoom telescope thebeam was diffracted by a diffraction grating. The spot on the grating wasimaged via a telescope consisting of an achromatic lens and the microscopeobjective onto the specimen plane. For the multispot excitation experimentsthe temporally focused beam was integrated into the scan head of a com-mercial two-photon scanning microscope.

Virus Preparation and Injection. Cre recombinase–dependent virus was as-sembled, harvested, and purified as previously described (40) and injectedinto the hippocampal CA1 region of parvalbumin-Cre (33) mice and hippo-campal CA3 region of Sprague-Dawley rats.

Hippocampal Slice Preparation. Transverse slices were prepared as previouslydescribed (41), according to methods approved by the Janelia Farm Institu-tional Animal Care and Use Committee.

ACKNOWLEDGMENTS. We thank J. Magee for supporting this project;A. Losonczy, J. Makara, and K. Svoboda for numerous insightful discussions,experimental suggestions, and comments on the manuscript; A. Losonczy fortesting cell type-specific viruses, for sample preparation, and for help withthe analysis; and D. O’Connor and G. Paez for sample preparation.

1. Sohal VS, Zhang F, Yizhar O, Deisseroth K (2009) Parvalbumin neurons and gammarhythms enhance cortical circuit performance. Nature 459:698–702.

2. Cardin JA, et al. (2009) Driving fast-spiking cells induces gamma rhythm and controlssensory responses. Nature 459:663–667.

3. Tsai HC, et al. (2009) Phasic firing in dopaminergic neurons is sufficient for behavioralconditioning. Science 324:1080–1084.

4. Luo L, Callaway EM, Svoboda K (2008) Genetic dissection of neural circuits. Neuron 57:634–660.

5. Gradinaru V, Mogri M, Thompson KR, Henderson JM, Deisseroth K (2009) Opticaldeconstruction of parkinsonian neural circuitry. Science 324:354–359.

6. Adamantidis AR, Zhang F, Aravanis AM, Deisseroth K, de Lecea L (2007) Neuralsubstrates of awakening probed with optogenetic control of hypocretin neurons.Nature 450:420–424.

7. Zemelman BV, Lee GA, Ng M, Miesenböck G (2002) Selective photostimulation ofgenetically chARGed neurons. Neuron 33:15–22.

8. Zemelman BV, Nesnas N, Lee GA, Miesenbock G (2003) Photochemical gating ofheterologous ion channels: Remote control over genetically designated populationsof neurons. Proc Natl Acad Sci USA 100:1352–1357.

9. Banghart M, Borges K, Isacoff E, Trauner D, Kramer RH (2004) Light-activated ionchannels for remote control of neuronal firing. Nat Neurosci 7:1381–1386.

10. Boyden ES, Zhang F, Bamberg E, Nagel G, Deisseroth K (2005) Millisecond-timescale,genetically targeted optical control of neural activity. Nat Neurosci 8:1263–1268.

11. Li X, et al. (2005) Fast noninvasive activation and inhibition of neural and networkactivity by vertebrate rhodopsin and green algae channelrhodopsin. Proc Natl AcadSci USA 102:17816–17821.

12. Gunaydin LA, et al. (2010) Ultrafast optogenetic control. Nat Neurosci 13:387–392.13. Lechner HAE, Lein ES, Callaway EM (2002) A genetic method for selective and quickly

reversible silencing of Mammalian neurons. J Neurosci 22:5287–5290.14. Slimko EM, McKinney S, Anderson DJ, Davidson N, Lester HA (2002) Selective

electrical silencing of mammalian neurons in vitro by the use of invertebrate ligand-gated chloride channels. J Neurosci 22:7373–7379.

15. Karpova AY, Tervo DGR, Gray NW, Svoboda K (2005) Rapid and reversible chemicalinactivation of synaptic transmission in genetically targeted neurons. Neuron 48:727–735.

16. HanX,BoydenES (2007)Multiple-coloroptical activation, silencing, anddesynchronizationof neural activity, with single-spike temporal resolution. PLoS ONE 2:e299.

17. Huber D, et al. (2008) Sparse optical microstimulation in barrel cortex drives learnedbehaviour in freely moving mice. Nature 451:61–64.

18. Petreanu L, Huber D, Sobczyk A, Svoboda K (2007) Channelrhodopsin-2-assistedcircuit mapping of long-range callosal projections. Nat Neurosci 10:663–668.

19. Varga V, et al. (2009) Fast synaptic subcortical control of hippocampal circuits. Science326:449–453.

20. Nagel G, et al. (2003) Channelrhodopsin-2, a directly light-gated cation-selectivemembrane channel. Proc Natl Acad Sci USA 100:13940–13945.

21. Nagel G, et al. (2005) Channelrhodopsins: Directly light-gated cation channels.Biochem Soc Trans 33:863–866.

22. Lanyi JK (1986) Halorhodopsin: A light-driven chloride ion pump. Annu Rev BiophysBiophys Chem 15:11–28.

23. Chow BY, et al. (2010) High-performance genetically targetable optical neuralsilencing by light-driven proton pumps. Nature 463:98–102.

24. Feldbauer K, et al. (2009) Channelrhodopsin-2 is a leaky proton pump. Proc Natl AcadSci USA 106:12317–12322.

25. Schoenenberger P, Grunditz A, Rose T, Oertner TG (2008) Optimizing the spatialresolution of Channelrhodopsin-2 activation. Brain Cell Biol 36:119–127.

26. Rickgauer JP, Tank DW (2009) Two-photon excitation of channelrhodopsin-2 atsaturation. Proc Natl Acad Sci USA 106:15025–15030.

27. Denk W, Strickler JH, Webb WW (1990) Two-photon laser scanning fluorescencemicroscopy. Science 248:73–76.

28. Zhu GH, van Howe J, Durst M, Zipfel W, Xu C (2005) Simultaneous spatial andtemporal focusing of femtosecond pulses. Opt Express 13:2153–2159.

29. Oron D, Tal E, Silberberg Y (2005) Scanningless depth-resolved microscopy. OptExpress 13:1468–1476.

30. Tal E, Oron D, Silberberg Y (2005) Improved depth resolution in video-rate line-scanning multiphoton microscopy using temporal focusing. Opt Lett 30:1686–1688.

31. Vaziri A, Tang J, Shroff H, Shank CV (2008) Multilayer three-dimensional superresolution imaging of thick biological samples. ProcNatl Acad Sci USA 105:20221–20226.

32. Arenkiel BR, et al. (2007) In vivo light-induced activation of neural circuitry intransgenic mice expressing channelrhodopsin-2. Neuron 54:205–218.

33. Hippenmeyer S, et al. (2005) A developmental switch in the response of DRG neuronsto ETS transcription factor signaling. PLoS Biol 3:e159.

34. AtasoyD,AponteY, SuHH, SternsonSM(2008)AFLEX switch targets Channelrhodopsin-2 to multiple cell types for imaging and long-range circuit mapping. J Neurosci 28:7025–7030.

35. Lin JY, Lin MZ, Steinbach P, Tsien RY (2009) Characterization of engineeredchannelrhodopsinvariantswithimprovedpropertiesandkinetics.BiophysJ96:1803–1814.

36. Wang H, et al. (2007) High-speed mapping of synaptic connectivity usingphotostimulation in Channelrhodopsin-2 transgenic mice. Proc Natl Acad Sci USA 104:8143–8148.

37. Cobb SR, et al. (1997) Synaptic effects of identified interneurons innervating bothinterneurons and pyramidal cells in the rat hippocampus. Neuroscience 79:629–648.

38. Lutz C, et al. (2008) Holographic photolysis of caged neurotransmitters. Nat Methods5:821–827.

39. Papagiakoumou E, de Sars V, Oron D, Emiliani V (2008) Patterned two-photonillumination by spatiotemporal shaping of ultrashort pulses. Opt Express 16:22039–22047.

40. Grieger JC, Choi VW, Samulski RJ (2006) Production and characterization of adeno-associated viral vectors. Nat Protoc 1:1412–1428.

41. Losonczy A, Makara JK, Magee JC (2008) Compartmentalized dendritic plasticity andinput feature storage in neurons. Nature 452:436–441.

11986 | www.pnas.org/cgi/doi/10.1073/pnas.1006620107 Andrasfalvy et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 1,

202

1