Articles

nAture methods  |  VOL.8  NO.4  |  APRIL 2011  |  319

recording electrical activity from identified neurons in intact tissue is key to understanding their role in information processing. recent fluorescence labeling techniques have opened new possibilities to combine electrophysiological recording with optical detection of individual neurons deep in brain tissue. For this purpose we developed dual-core fiberoptics–based microprobes, with an optical core to locally excite and collect fluorescence, and an electrolyte-filled hollow core for extracellular single unit electrophysiology. this design provides microprobes with tips <10 mm, enabling analyses with single-cell optical resolution. We demonstrate combined electrical and optical detection of single fluorescent neurons in rats and mice. We combined electrical recordings and optical ca2+ measurements from single thalamic relay neurons in rats, and achieved detection and activation of single channelrhodopsin-expressing neurons in Thy1::ChR2-YFP transgenic mice. the microprobe expands possibilities for in vivo electrophysiological recording, providing parallel access to single-cell optical monitoring and control.

The central nervous system is characterized by heterogeneous populations of cells with highly specialized phenotypes defined by their morphological, biochemical and physiological proper-ties. Efficient characterization of each cell type in intact tissue has remained challenging because the neurons of interest are often dispersed in these heterogeneous populations. There are several approaches that allow electrophysiologists to identify the cell types they record from, but most of these approaches have drawbacks. Antidromic activation of projection neurons is only applicable to a small subset of cells and requires preservation of connections between the recording and projection sites1,2. Labeling cells with dyes at the time of recording allows a posteriori identification, but this approach is inefficient when dealing with subpopulations that represent a small proportion of the overall neuronal population in an area3,4. Labeling with fluorescent markers allows for targeted recording, independently of connectivity5–9 and has been very instrumental for studies in central nervous system tissue slices10–12. However, it cannot be fully exploited for electrophysiological

investigations in live animals because light scattering limits optical microscopy to surface measurements (500–1,000 µm)13,14. Micro-endoscopes help overcome limitations for deep tissue imaging but do not allow for combined electrical and optical monitoring, especially for single unit recording15,16. The challenge remains to conduct combined electrophysiological recording and optical identification of individual neurons deep into central nervous system tissue in live animals. Combination of both approaches is important because purely optical techniques remain limited for resolving single action potentials and for prolonged functional measurements (>1 h).

Several strategies have been sought to combine microelec-trode and guided optical recording. One had used distinct probes brought together using two independent microdrives17,18. Such an approach is not practical for in vivo measurements. Other groups have relied on attaching large fibers (250 µm in diameter) to an electrode19,20. Even with smaller optical fibers (25–35 µm) inserted into a micropipette, the final tip diameter of the probe is 35–45 µm, precluding combined optical and electrical record-ing from a single neuron. Furthermore, this approach relies on an external source of illumination, which prevents efficient light delivery into deep tissue and produces tissue excitation over wide areas21.

To overcome these limitations, we designed a new type of opti-cal fiber, incorporating both an optical and a hollow core. We show that this type of microprobe provides, to our knowledge for the first time, sufficient spatial resolution to correlate electro-physiological and fluorescence signals that emanate from single fluorescently labeled neurons at a depth of >6,000 µm in the intact central nervous system.

resultsthe microprobeOur objective was to detect, in parallel, the electrical field and the optical signal from single fluorescent neurons in vivo. We designed an optical fiber composed of a hollow core and an opti-cal core (Fig. 1a,b and Supplementary Fig. 1a). We filled the hollow core with an electrolyte solution (1–3 M NaCl) to record

1Centre de recherche Université Laval Robert-Giffard, Québec, Canada. 2Centre d’optique, photonique et laser, Université Laval, Québec, Canada. 3Department of Psychiatry and Neuroscience, Université Laval, Québec, Canada. 4These authors contributed equally to this work. Correspondence should be addressed to Y.D.K. (yves.dekoninck@crulrg.ulaval.ca).Received 15 July 2010; accepted 20 JanuaRy 2011; published online 13 febRuaRy 2011; doi:10.1038/nmeth.1572

A microprobe for parallel optical and electrical recordings from single neurons in vivoYoan LeChasseur1,2, Suzie Dufour1,2,4, Guillaume Lavertu1,4, Cyril Bories1, Martin Deschênes1,3, Réal Vallée1,2 & Yves De Koninck1–3

© 2

011

Nat

ure

Am

eric

a, In

c. A

ll ri

gh

ts r

eser

ved

.

320  |  VOL.8  NO.4  |  APRIL 2011  |  nAture methods

Articles

electrical activity, and the optical core was a graded refractive index waveguide allow-ing efficient light delivery and collection through a tapered tip. To achieve single-cell resolution with the optical probe, we pulled the fiber to tip sizes smaller than that of the neuron somata under study, using a conventional laser micropipette puller. We generated probes with end tip diameters of 6–20 µm (the small-est probe tips had an optical core of 4 µm and a hollow core of 0.7 µm; Fig. 1a and Supplementary Fig. 1a). The all-glass dual- core fiber could be pulled into a variety of shapes as standard glass microelectrodes (Supplementary Fig. 1b). To link an optical

setup to the biological preparation, we connected the optical core to a multimode fiber (Fig. 1c). We injected light into the probe for local delivery at the tip end and also used this tip to collect fluorescence emanating from labeled cells.

With this design, fluorescence collection is asymmetrical, with only light in front of the tip being collected. Therefore, as the probe moves past a labeled neuron, the detected fluo-rescence signal should follow a characteristic intensity profile: growing gradually when approaching a labeled cell and drop-ping abruptly when the probe moves beyond the cell (Fig. 1d). In contrast, the electrical signal should grow and decline sym-metrically as the probe moves past a neuron. The peaks of both signals emanating from a cell are thus not necessarily aligned along the trajectory of the probe: the optical signal should peak when the probe is at nearest approach to the cell body, whereas the electrical signal should peak near the maximal current sink (typically the axon hillock22). The interpeak distance should thus be equal to or smaller than the length of the cell body and

a

Laser

Fiberlaunch

PMT

Acquisitionsystem

c

Signal amplitude

Pro

be p

ositi

on

d

Neutraldensityfilter

Light insulation box

DichroicmirrorMicroprobeGraded index

optical core

Gradedindex opticalcore

Hollowcore

Multimodeoptical fiber

Shoulder

Shank

Tip

50-µmwire

UV-light curableadhesive

b

Hollowcore Fluorescence

Excitation light

Electric field potential

Figure 1 | The microprobe. (a) Schematic of the microprobe (not drawn to scale), which is based on a dual-core optical fiber: a graded index optical core and a hollow core. The optical core is connected at one end to a multimode optical fiber, and the other end is tapered. (b) Schematic of the optical fiber preform design: a series of different-size silica rods are assembled to position the optical and hollow cores side by side in an inclusion tube. Scale bar, 2 mm. (c) Diagram of the experimental setup. Light delivery and collection as well as electrical recording are all achieved through the same fiber probe. PMT, photomultiplier tube. (d) Illustration of the directionality of optical recording and the nondirectionality of electrical recording, with intensity profiles along the trajectory of the microprobe and interpeak distance shown on the right.

50%

axi

al d

ista

nce

(µm

)

Inte

nsity

(a.

u.)

–100

10

50

100

0

0.5

1

Axial neuron-probe distance (µm)–20 –10 0 10 20

0

0.2

0.4

0.6

0.8

1

Transverse position (µm)

Inte

nsity

(a.

u.)

0

0.2

0.4

0.6

0.8

1 Neuron diameter = 14 µmProbe diameter = 9 µm

a b

Inte

nsity

(a.

u.)

Axialdc

e fTransverse position (µm)

0 20

0 2 4 6 8 10

40 60 80

Transverse

0

0.5

1

Intensity (a.u.)

Axi

al p

ositi

on (

µm)

0

10

20

30

40

50

0 0.5 1

Laser power (mW)

0

5

10

15

20

R = –0.05P = 0.92

3020100 40 50

Transverse position (µm)

Axial position (µm)

Figure 2 | Microprobe optical resolution. (a,b) Fluorescence intensity profile as a function of transverse (a) and axial (b) probe displacement around single Lucifer yellow–labeled neurons in brain slices. Zero displacement in a corresponds to the center of the cell body, zero displacement in b is at nearest approach to the cell body. Shown are seven superimposed detection profiles (thin black lines), average detection profile (thick black line) and the expected detection profile based on numerical simulations (red line). (c,d) Photomicrographs of two adjacent Lucifer yellow–labeled fluorescent neurons in fixed brain slices used for testing probe detection in transverse (c) and axial (d) translation axes. Microprobes are outlined in white, and the arrow shows the probe trajectories. Scale bars, 10 µm. Detected fluorescence during probe displacement is plotted; zero displacement corresponds to the start of the white arrows in the micrographs. (e) Rising distance from 50% to peak fluorescence plotted versus laser power. Dashed line indicates the fitted regression. (f) Simulated fluorescence collection field of a 9-µm-diameter probe in front of a 12-µm-diameter neuron plotted against the axial and transverse positions of the probe. Color coding is arbitrary.

© 2

011

Nat

ure

Am

eric

a, In

c. A

ll ri

gh

ts r

eser

ved

.

nAture methods  |  VOL.8  NO.4  |  APRIL 2011  |  321

Articles

could serve as a criterion to determine whether the electrical and optical signals emanate from the same neuron (Fig. 1d).

optical resolution and single cell detectionWe estimated the resolution of the microprobe assuming a fluo-rescence point source, as is commonly used for calculation of point spread functions in microscopy (Supplementary Data). For probe tip diameters of 10 µm (6 µm optical core) or 6 µm (4 µm optical core), the calculated full width half maximum (FWHM) of the detected signal distribution was 13 µm or 12 µm in the axial direction and 6 µm or 4 µm in the transverse plane, respectively. Our calculations revealed that numerical aperture (NA) values of 0.2–0.3 yielded an optimal compromise between axial and transverse resolution (Supplementary Data and Supplementary Fig. 2a,b). We then estimated numerically (Supplementary Data) the fluorescence decay profiles as a function of both fiber and cell diameters. With optical cores smaller than the cells, the decay profile of the detected signal in the transverse axis was sensitive to cell size but insensitive to probe tip size. Conversely, in the axial direction, the decay profile was sensitive to fiber diameter but not cell size, indicating that controlling tip size is important for axial resolution of single cells (Supplementary Fig. 2c,f).

To test the applicability of these resolution estimates based on physical parameters of the fibers (Supplementary Fig. 2a–f), we performed measurements in fixed brain slices with cortical neu-rons labeled using intracellular injection of Lucifer yellow. First we examined slices under epifluorescence to locate labeled cells.

Then we turned off the external light source and moved the probe in the transverse and axial directions across the labeled cells. The detected signal FWHM using a 9-µm-diameter probe tip was 15 ± 5 µm transversely and 11 ± 4 µm axially for 14 ± 1 µm-diameter neurons (mean ± s.d.; n = 6 cells; Fig. 2a,b), consistent with our predictions (Supplementary Fig. 2c–f). Using this tip diameter, we achieved sufficient resolution to distinguish adjacent cells in both transverse and axial directions (Fig. 2c,d). Although the detected fluorescence peak intensity varied from cell to cell, as expected, the axial resolution remained constant and independent of excitation power for a given microprobe (Fig. 2e,f). We confirmed this by repeated measurements with different excitation powers (Fig. 2e).

recording from single retrograde-labeled neurons in vivoTo test the optical and electrical performance of the microprobe in vivo, we first used it to detect spinothalamic tract neurons that we retrograde-labeled via injections in the thalamus of the fluores-cent dye DiI (Fig. 3a,b). As estimated from histological sections, spinothalamic tract neurons in spinal laminae III–VI had soma diameters of 24 ± 6 µm (mean ± s.d.; n = 153 cells; Fig. 3c).

We used microprobes with 14 ± 3 µm tip diameter to record from spinothalamic tract neurons down to 2,200 µm in the tis-sue. The electrical channel of the probes had 6–26 MΩ resistance, yielding spike recordings with a signal-to-noise ratio of 7.6 ± 3.7 (n = 155 cells; Supplementary Fig. 3) and stable recordings >> 1 h (Fig. 3d–g). The probes’ recording capabilities and stability were comparable to those of conventional micropipettes.

100 µV

0.5 ms

c

Cum

ulat

ive

prob

abili

tyN

euro

n co

unt

f

5 10 15 200–2

0

2

4

6

Inte

nsity

(a.

u.)

d

b

13 µm

0

50

100

0

10

20

Num

ber

of s

pike

sTime (s)

0

5 10 15 20 25 30 35 40 45

5 10 15 20 25 30 35 40 45

50Axial relative position (µm)

Neuron diameter (µm)

100 1500

0.2

0.4

0.6

0.8

113 µm

Inte

nsity

(a.

u.)

e

h

g

Optical S/N

30

20

10

00 20 40 60 80 100

Interspike interval (ms)

a

100 µV

* *

100 ms

Probe travel direction

Probe diameter (µm)

Opt

ical

S/N

1

10

0 5 10 15 20

100

1,000

R = –0.13P = 0.23

Num

ber

of n

euro

ns

0 100 200 300 400 500 6000

5

10

15

RecordingmicroprobeDiI injection

Naturalstimulus

FluorescenceSpikes

FluorescenceSpike amplitude

more than 90% of cells had a diameter >16 µm. (d) Fluorescence signal and spikes evoked from the stimulation of the receptive field as the microprobe descended into the spinal cord at constant speed (6 µm s−1). (e) Plot of the data shown in d, with the time scale converted to spatial position. (f) Data from d shown on an expanded time scale. Asterisks indicate digitally subtracted movement artifacts. (g) Distribution of interspike intervals for cell analyzed in d–f. Inset, 212 superimposed individual spikes. (h) Distribution of optical signal-to-noise ratio (S/N) obtained with 87 detected neurons (average S/N was 21; n = 87 cells). Inset, relationship between the optical S/N and microprobe tip diameter.

Figure 3 | Single-unit recordings from neurons in vivo. (a) Photomicrograph of a spinal cord transverse section in which spinothalamic tract neurons were retrograde-labeled with DiI. The overlaid drawing of a microprobe (gray) illustrates a typical descent (arrow). Inset, magnification of a stained neuron. Scale bars, 1 mm and 20 µm in inset. (b) Schematic representation of the experiment. (c) Distribution of labeled spinothalamic tract neuron diameters plotted as a cumulative probability (top) or a histogram (bottom). The dotted arrow indicates the distance between the optical and electrical signal maxima in e;

© 2

011

Nat

ure

Am

eric

a, In

c. A

ll ri

gh

ts r

eser

ved

.

322  |  VOL.8  NO.4  |  APRIL 2011  |  nAture methods

Articles

We lowered the microprobe at constant speed into the spinal cord (4–8 µm s−1), maintaining the excitation light on throughout the descent. We observed one of the following: (i) a gradual rise of fluorescence followed by a decay over several tens of micrometers without any clear spiking activity; (ii) spikes increasing gradually in amplitude followed by a decrease without noticeable change in flu-orescence; or (iii) a parallel growth and decay of fluorescence and spike amplitude (Fig. 3d,e). In the vast majority of cases, changes in optical and electrical signals occurred without signs of neuron damage. Peak optical signal-to-noise ratio was not correlated with tip diameter (Fig. 3h) indicating that there was no disadvantage in using smaller probes to increase optical resolution.

We analyzed quantitatively how the amplitude of the electri-cal and fluorescence signals changed as a function of the axial position of the probe. We performed this analysis in well-isolated single units as assessed by waveform analysis and the absence of short interspike intervals (<2 ms) in spike trains (Fig. 3f,g). Under these conditions, the spike amplitude rose and decayed symmetrically with a 48 ± 19 µm FWHM (n = 15 cells; Fig. 4a). In contrast, the fluorescence signal intensity rose and decayed in

a nonsymmetrical fashion, regardless of the probe travel direction (up or down), indicating no hysteresis (Fig. 4b,c). In 87 optical recordings, the mean fluorescence signal-to-noise ratio was 21. As the probe approached the cell, the fluorescence rose from 50% to peak over a distance of 21 ± 8 µm; it decayed from peak to 50% over 7 ± 3 µm when the probe passed by the cell (n = 14 cells; Fig. 4b). These values were consistent with our simulated predictions and measurements in slices (Supplementary Fig. 2b,d). The asym-metry between fluorescence rise and decay was consistent with our working model (Fig. 1d). Based on results obtained for many descents, the criterion we used to identify a neuron optically was that the fluorescence signal had to rise to a maximum within the distance predicted from the probe characteristics (Supplementary Fig. 2a,b,d) and drop back to baseline within less than half of the rising distance with a signal-to-noise ratio >9.

combined optical and electrical detection criterionTo determine whether the optical and electrical signals originated from the same cell, we measured the distance between the axial positions of the two signal peaks (Fig. 3e). We first considered that both signals arose from the same neuron when the interpeak distance was less than the mean neuron diameter according to our working model (Fig. 1d). In 16 spinothalamic tract neurons, the dis-tance between the optical and electrical signal peaks was < 24 µm, which corresponded to the mean spinothalamic tract neuron soma diameter. As a more stringent criterion, we used an interpeak distance below 90% of these soma diameters (16 µm; Fig. 3c); we considered the two signals to originate from the same cells in 12/16 cases. This likely increased the number of false negatives but minimized false positives, which is more critical.

recording from single eGFP-expressing GABAergic neuronsWe next targeted reticular thalamic neurons expressing enhanced GFP (eGFP) under the control of the glutamic acid decarboxylase promoter (isoform 65; GAD65) in transgenic mice. These cells’ average soma diameter was 12.4 ± 2.3 µm (mean ± s.d.; n = 41 cells; Fig. 4d and Supplementary Fig. 4a).

a

Neuron

c

50 µm 150 µm–50 µm –50 µm

25 a.u.

25 µm

Neuron MicroprobeMicroprobe

b

100

300

500

Axial relative position (µm)

Axial relative position (µm)Axial relative position (µm)

Spi

ke a

mpl

itude

(a.

u.)

–100 –50 0 50 1000

0.2

0.4

0.6

0.8

150%

Flu

ores

cenc

e in

tens

ity (

a.u.

)

0

0.5

150%50%

0

0.5

1

d

200 µV

0.5 ms

Time (ms)Axial relative position (µm)0–10 0 10 20

10

5

02 4 6 8 10

Num

ber

of s

pike

s

e f

Sig

nal i

nten

sity

(a.

u.)

0

0.5

1

200µV

1 s

50 100 150 20000

0.5

Inte

nsity

(a.

u.)

1

O1

4 ms0.1

mV

4 ms

0.1

mV

O2E1 E2

Displacement forward (µm) Displacement backward (µm)

E1 spikes

–200 –150 –100 –50 00

1

0.5

O1

E1O2E2E2 spikes

7 µm

Flu

ores

cenc

e in

tens

ity (

a.u.

)–50 0

–50 0

50 100 150

50 1000

OpticalElectrical

Figure 4 | Optical and electrical signal profiles of single cells in vivo. (a,b) Schematics (top) show microprobe displacement (arrow) relative to neuron position. Spike amplitudes (a) as the microprobe passed by a neuron were detected (curve indicates the average) and normalized spike amplitude was plotted as a function of probe travel distance (bottom, average in blue, n = 15 cells). Typical fluorescence detection patterns (b; middle) when the microprobe passed by four fluorescently labeled neurons and normalized fluorescence profiles (bottom), with the average (thick black line) (n = 14) and the expected profile from simulations (red line) shown. Dotted lines indicate value at 50% to peak. (c) Successive units detected by fluorescence and/or electrical signals during a 225-µm, bidirectional probe excursion. E1 and E2 represent two electrically recorded neurons and O1 and O2 represent two optically recorded ones. Insets, examples of spikes recorded in E1 and E2. (d) Photomicrograph of a GAD-GFP mouse brain section. The overlaid drawing of a microprobe (gray) illustrates a typical descent (white arrow) into the reticular thalamic nucleus. Inset, high-magnification photomicrograph of GAD-GFP neurons. Scale bar, 1 mm (20 µm in inset). (e) Magnitude of optical signal and spikes as the probe passed by a reticular thalamic GFP-expressing neuron (bottom). Zero value on the x axis is aligned with the peak of the fluorescence curve. Continuous electrophysiological signal as the probe traveled through the tissue (top). (f) Distribution of interspike intervals. Inset, consecutive superimposed action potentials (n = 26).

© 2

011

Nat

ure

Am

eric

a, In

c. A

ll ri

gh

ts r

eser

ved

.

nAture methods  |  VOL.8  NO.4  |  APRIL 2011  |  323

Articles

We thus used microprobes with 6–10 µm tips. Applying the stringent criteria described above, combined detection from the same cell required that the optical and electrical signal peaks were ≤10 µm apart (Supplementary Fig. 4a). For this neuron population, the optical signal rose from 50% to peak over a distance of 12 µm and decayed from peak to 50% over 6 µm (Supplementary Fig. 4b). Of 11 cases with overlapping electrical and optical signals, seven fell within the combined detection criterion (Fig. 4e and Supplementary Fig. 4a).

In these experiments, we identified single GAD65-eGFP neurons in a dense nucleus of inhibitory interneurons (Fig. 4d–f), which were not spatially distributed in any identifiable way; our tech-nique was particularly useful under such conditions.

Parallel recording and ca2+ measurements in single neuronsTo test whether optical and electrophysiological recordings with our microprobe originated from the same cell, we combined Ca2+ measurements with electrophysiological recording. We first recorded spontaneous activity in thalamic barreloids after loading neurons with Oregon Green BAPTA1 acetoxymethyl (AM) ester via local injections in vivo in adult rats. Single unit

a

c

b

250 ms

200 µV

100 ms

100

µ V

0°

90°

180°

270°

0°90°

270° Piezo180°

f

Ca2+

res

pons

e (a

.u.)

Number of spikes0 10 20 30

0.12

0.16

0.20

0.24

0.28

R = 0.92P < 0.0001

e

Spi

king

ang

ular

pref

eren

ce

Ca2+ angular preference0° 150° 300°

90°

0°

R = 0.98P < 0.001

d 5 events

100 ms

Whisker

50 ms

100 µV

Ca2+ response

Spiking

5%∆F/F

1% ∆F/F

5%∆F/F

5%∆F/F

0°

300°

150°

Flu

ores

cenc

esi

gnal

Spi

kebu

rsts

Figure 5 | Monitoring Ca2+ fluctuations confirm optical and electrical recording from the same neuron in vivo. (a,b) Spontaneous spiking and Ca2+ fluctuations in a thalamic relay neuron (recorded at a depth of 6 mm in the brain). The fluorescence signal and spike bursts (a), and Ca2+ response (quantified as the area under the curve) plotted as a function of the number of spikes in each burst (b) for the same cell. The two parameters were linearly correlated (n = 24 bursts). ∆F/F = (F – F0)/F0, in which F is the fluorescence and F0 = F in the stained region – F outside the stained region. (c) Schematic representation of the whisker deflection using a piezoelectric mechanical stimulator. Inset, superposition of three spiking and calcium responses following a whisker deflection. (d) Peristimulus time histograms of spiking responses to 20 whisker deflections (four directions indicated in degrees) and the associated averaged Ca2+ response over the same 20 stimuli for the same neuron. The duration of whisker deflection is represented in gray. (e) The angular preferences of spike response latency plotted against angular preference for Ca2+ response (n = 8 cells). Inset, polar representation of the normalized inverted spike latency after stimulation (black) and the normalized inverted latency of the Ca2+ response (green). (f) Comparison between Ca2+ responses (top) and spiking responses (middle and bottom) of a neuron-to-whisker deflection. Black traces indicate two representative successful responses, and gray traces indicate two representative failures to respond in the same cell.

Light intensity (mW mm–2)

200 µV

Micro-whisker Light stimulation

1 ms0

40

80

Res

pons

e (s

pike

s s–1

)

d

Spi

ke a

mpl

itude

(m

V)

0

0.2

0.4e

Th MC0

0.4

0.8

1.2

Del

ay s

.d. (

ms)

f

g

b ca

Relative position (µm)

Inte

nsity

(a.

u.)

0 10

0

0.3

0.6

0.9

5 µm

–10

Light stimulation

200 µV

40 ms

0.1 1 10

100 µV50 ms

Both combined

0.1 1 10

1 23

1 2 3

Figure 6 | Detection and activation of single ChR2-expressing neurons in vivo. (a) Photomicrograph of a sagittal brain section of a Thy1::ChR2-EYFP mouse. White boxes highlight recording sites. Scale bar, 1 mm. (b) Magnitude of optical and electrical signals of a ChR2-EYFP–expressing neuron as a function of probe position. Zero position was aligned to peak fluorescence value. (c) Response of the neuron in b to light stimulation through the probe. Arrowheads point to the depolarization and repolarization of the cells at the pulse onset and offset. (d) Response of an individual neuron versus light intensity. Individual responses are shown in insets. (e) Spike amplitude as a function of light intensity at probe tip (same cell as in d). Each inset represents seven superimposed spikes. (f) Mean ± s.d. of the response delay to light stimulus for motor cortical neurons (MC) (n = 15) and thalamic neurons (Th) (n = 9). (g) Responses of a thalamic ChR2-expressing neuron to micro-whisker deflection, light stimulation or both.

© 2

011

Nat

ure

Am

eric

a, In

c. A

ll ri

gh

ts r

eser

ved

.

324  |  VOL.8  NO.4  |  APRIL 2011  |  nAture methods

Articles

recordings from thalamic relay neurons revealed low-frequency (2–3 Hz) bursts, which coincided with Ca2+ transients (Fig. 5a). We found that over several minutes of recording in five different neurons, virtually all spike bursts coincided with a rise in Ca2+ and, inversely, Ca2+ fluctuations were always associated with spik-ing (Fig. 5a). Furthermore, the magnitude of the Ca2+ responses was highly correlated to the number of spikes per burst (Fig. 5b). We found the same linear relation for the five neurons tested (P < 0.01). Because population activity varies much less than indi-vidual neuron activity, this correlation indicates that, in these cases, the two signals most likely arose from the same cell.

Once we achieved single-unit recording of spontaneous activity in the stained area, we studied evoked responses to deflection of single whiskers (Fig. 5c). We recorded concurrent robust evoked electrophysiological and Ca2+ responses in 47 neurons. Thalamic relay cells in a barreloid respond differently to the direction of whisker deflection, and this angular preference is cell-dependent4. Thus, to correlate both signals we determined the angular pre-ference of recorded cells using the Ca2+ response latency and spikes elicited by moving the whiskers in four distinct directions (Fig. 5d)23. We found a linear relationship with a slope of 1 between the angular preferences for both measurements, again indicating that both signals emanated from the same cells (Fig. 5e; n = 8 cells; P < 0.001).

Finally, we analyzed spike ‘failures’ in response to whisker deflection. Using stimuli of constant repetition rate and strength, we found that whenever we elicited spikes in thalamic neurons, this caused a Ca2+ response (Fig. 5f). In contrast, when spike failures occurred, we recorded no significant Ca2+ response (amplitude decreased by 94 ± 8%; P > 0.5; 10 spikes and 10 spike failures in three cells). As spike failures are characteristics of indi-vidual neurons, as opposed to field (population) responses, coin-cidence of spiking and Ca2+ response failures provided further evidence that both signals were unitary and originated from the same neuron.

optogenetic activation of single cellsWe performed experiments on Thy1::ChR2-eYFP transgenic mice24, which express the light-gated channel ChR2 in various subpopulations of neurons, including in the thalamus and the cortical pyramidal layer 5 (Fig. 6a)25. We tested the ability of the microprobe to activate and record from individual neurons and thus confirm that they expressed ChR2. We used a 488-nm laser to simultaneously stimulate ChR2 and eYFP with brief (5–500 ms) pulses of light. We optically detected 27 cells in the motor cortex (layer 5) and the thalamus in six mice. Of the 27 cells, 22 satis-fied the combined detection criteria, and of these 19 showed clear excitatory responses to light pulses (Fig. 6b). We recorded spikes of constant amplitude and shape in response to local light pulses at different intensity, confirming single unit recordings (Fig. 6c–e). At the light pulse onset and offset, we detected small deflections with opposite polarities reflecting depolarization and repolarization at the beginning and end of the pulse (Fig. 6c). These deflections were neither present at subthreshold light intensities (Fig. 6d) nor in front of nonresponsive neurons (data not shown). The light intensity activation threshold (1.3 ± 0.3 mW mm−2; n = 13 neurons) was consistent with that previously reported26 (Fig. 6d). Response delay to light stimulation showed minimal jitter (Fig. 6f). These observations indicated direct cell

depolarization resulting from ChR2 activation in the target cell and not indirect activation via a neighboring neuron. Finally, because light stimulation could be used to directly drive spiking, we could observe postsynaptic inhibition (Fig. 6g; for example, microwhisker deflection silenced the cell during light-driven activity). These results indicate that the microprobe can be used to deliver sufficient light for local stimulation in deep brain areas to control activity from identified single neurons.

discussionOur technique is particularly useful for analyzing sparsely dis-tributed cells that are not otherwise identifiable based on spatial distribution (for example, inhibitory interneurons dispersed in a nucleus). With densely packed neurons such as pyramidal neu-rons in the hippocampal CA1 layer, the ability to unequivocally identify single neurons would be more challenging. If a specific phenotypic property distinguished a subset of these pyramidal neurons, however, they would probably be more sparsely distrib-uted and hence discernable by the microprobe, for instance, upon engineering to express a fluorescent protein.

The optical search protocol described here involves local light delivery and does not require prolonged light exposure; the system could also be adapted to further minimize excitation cycles using pulsed illumination. Optical probing allows localization of cells with-out electrically induced search stimuli such as antidromic or afferent pathway stimulation, minimizing potential physiological alterations of neural circuits. An important asset of the probe is that it comprises only glass; it can be easily shaped and hence can be tailored to dif-ferent approaches (for example, shorter or longer shanks, tips and so forth) and can also be used for microinjection (iontophoresis or pressure ejection) into an optically identified targeted area.

The simple optical design thus yields a versatile system adapt-able for multiple applications. It will allow exploitation of the expanding array of tools that use fluorescent sensors and light-activated channels24–27 and could be used for spectroscopic detection of intrinsic signals. Exploiting the Brainbow mouse28, it would be possible to identify each cell by adapting the system for multicolor recording. The combined optical and electrical capabilities of the microprobe described here vastly expand pos-sibilities for conventional in vivo electrophysiology with access to single-cell molecular technologies.

methodsMethods and any associated references are available in the online version of the paper at http://www.nature.com/naturemethods/.

Note: Supplementary information is available on the Nature Methods website.

AcknoWledGmentsThis work was supported by the Canadian Institutes for Health Research. Y.D.K. was funded by a Chercheur National award from the Fonds de la recherche en santé du Québec. Y.L. and S.D. received a Studentship from the Canadian Institutes for Health Research Neurophysics Training program grant. We thank A. Proulx and A. Croteau for the collaboration on the dual-core fiber design and fabrication; D. Côté and A. Castonguay for helpful comments on the manuscript; and S.L. Côté, K. Bachand, K. Vandal and M. Demers for expert technical assistance.

Author contriButionsY.L., R.V. and Y.D.K. designed the microprobe. Y.L., S.D., G.L., C.B., M.D. and Y.D.K. designed the experiments. Y.L., S.D., G.L. and Y.D.K. analyzed data. Y.L., S.D., G.L. and C.B. performed experiments. S.D. performed Ca2+ measurement and photostimulation experiments. Y.L. and S.D. performed numerical simulations. Y.L., S.D., G.L., C.B., M.D., R.V. and Y.D.K. contributed to writing the manuscript.

© 2

011

Nat

ure

Am

eric

a, In

c. A

ll ri

gh

ts r

eser

ved

.

nAture methods  |  VOL.8  NO.4  |  APRIL 2011  |  325

Articles

comPetinG FinAnciAl interestsThe authors declare no competing financial interests.

Published online at http://www.nature.com/naturemethods/. reprints and permissions information is available online at http://npg.nature.com/reprintsandpermissions/.

1. MacLean, J.N., Fenstermaker, V., Watson, B.O. & Yuste, R. A visual thalamocortical slice. Nat. Methods 3, 129–134 (2006).

2. Swadlow, H.A. Neocortical efferent neurons with very slowly conducting axons: strategies for reliable antidromic identification. J. Neurosci. Methods 79, 131–141 (1998).

3. De Koninck, Y., Ribeiro-da-Silva, A., Henry, J.L. & Cuello, A.C. Spinal neurons exhibiting a specific nociceptive response receive abundant substance P-containing synaptic contacts. Proc. Natl. Acad. Sci. USA 89, 5073–5077 (1992).

4. Timofeeva, E., Merette, C., Emond, C., Lavallee, P. & Deschenes, M. A map of angular tuning preference in thalamic barreloids. J. Neurosci. 23, 10717–10723 (2003).

5. Honig, M.G. & Hume, R.I. Dil and diO: versatile fluorescent dyes for neuronal labelling and pathway tracing. Trends Neurosci. 12, 333–335 (1989).

6. Feng, G. et al. Imaging neuronal subsets in transgenic mice expressing multiple spectral variants of GFP. Neuron 28, 41–51 (2000).

7. Spergel, D.J., Kruth, U., Shimshek, D.R., Sprengel, R. & Seeburg, P.H. Using reporter genes to label selected neuronal populations in transgenic mice for gene promoter, anatomical, and physiological studies. Prog. Neurobiol. 63, 673–686 (2001).

8. Callaway, E.M. Transneuronal circuit tracing with neurotropic viruses. Curr. Opin. Neurobiol. 18, 617–623 (2009).

9. Luo, L., Callaway, E.M. & Svoboda, K. Genetic dissection of neural circuits. Neuron 57, 634–660 (2008).

10. Ikeda, H., Heinke, B., Ruscheweyh, R. & Sandkuhler, J. Synaptic plasticity in spinal lamina I projection neurons that mediate hyperalgesia. Science 299, 1237–1240 (2003).

11. Labrakakis, C. & Macdermott, A.B. Neurokinin receptor 1-expressing spinal cord neurons in lamina I and III/IV of postnatal rats receive inputs from capsaicin sensitive fibers. Neurosci. Lett. 352, 121–124 (2003).

12. Yuste, R. Fluorescence microscopy today. Nat. Methods 2, 902–904 (2005).13. Helmchen, F. & Denk, W. Deep tissue two-photon microscopy.

Nat. Methods 2, 932–940 (2005).14. Margrie, T.W. et al. Targeted whole-cell recordings in the mammalian brain

in vivo. Neuron 39, 911–918 (2003).

15. Jung, J.C., Mehta, A.D., Aksay, E., Stepnoski, R. & Schnitzer, M.J. In vivo mammalian brain imaging using one- and two-photon fluorescence microendoscopy. J. Neurophysiol. 92, 3121–3133 (2004).

16. Levene, M.J., Dombeck, D.A., Kasischke, K.A., Molloy, R.P. & Webb, W.W. In vivo multiphoton microscopy of deep brain tissue. J. Neurophysiol. 91, 1908–1912 (2004).

17. Bowmaster, T.A., Davis, C.C. & Krauthamer, V. Excitation and detection of action potential-induced fluorescence changes through a single monomode optical fiber. Biochim. Biophys. Acta 1091, 9–14 (1991).

18. Duff Davis, M. & Schmidt, J.J. In vivo spectrometric calcium flux recordings of intrinsic Caudate-Putamen cells and transplanted IMR-32 neuroblastoma cells using miniature fiber optrodes in anesthetized and awake rats and monkeys. J. Neurosci. Methods 99, 9–23 (2000).

19. Kudo, Y., Takeda, K., Hicks, T.P., Ogura, A. & Kawasaki, Y. A new device for monitoring concentrations of intracellular Ca2+ in CNS preparations and its application to the frog’s spinal cord. J. Neurosci. Methods 30, 161–168 (1989).

20. Nakamura, T. et al. Increased intracellular Ca2+ concentration in the hippocampal CA1 area during global ischemia and reperfusion in the rat: a possible cause of delayed neuronal death. Neuroscience 88, 57–67 (1999).

21. Bradley, P.M., Murphy, D., Kasparov, S., Croker, J. & Paton, J.F. A micro-optrode for simultaneous extracellular electrical and intracellular optical recording from neurons in an intact oscillatory neuronal network. J. Neurosci. Methods 168, 383–395 (2008).

22. Humphrey, D.R. & Schmidt, E.M. Extracellular single-unit recording methods. Neuromethods 15, 1–64 (1991).

23. Bellavance, M.A., Demers, M. & Deschenes, M. Feedforward inhibition determines the angular tuning of vibrissal responses in the principal trigeminal nucleus. J. Neurosci. 30, 1057–1063 (2010).

24. Arenkiel, B.R. et al. In vivo light-induced activation of neural circuitry in transgenic mice expressing channelrhodopsin-2. Neuron 54, 205–218 (2007).

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

26. Gradinaru, V., Mogri, M., Thompson, K.R., Henderson, J.M. & Deisseroth, K. Optical deconstruction of parkinsonian neural circuitry. Science 324, 354–359 (2009).

27. Lima, S.Q., Hromadka, T., Znamenskiy, P. & Zador, A.M. PINP: a new method of tagging neuronal populations for identification during in vivo electrophysiological recording. PLoS ONE 4, e6099 (2009).

28. Livet, J. et al. Transgenic strategies for combinatorial expression of fluorescent proteins in the nervous system. Nature 450, 56–62 (2007).

© 2

011

Nat

ure

Am

eric

a, In

c. A

ll ri

gh

ts r

eser

ved

.

online methodsMicroprobe fabrication. We designed dual-core optical fibers in collaboration with the Institut National d’Optique, where the fibers were fabricated. The fiber preform was assembled through a procedure inspired from the stack-and-draw method previously reported in the literature for the fabrication of microstructured optical fibers29,30. The goal was to position a graded refractive index optical core and a silica capillary hole side by side in a silica inclusion tube (Fig. 1b). The optical core was created by modified chemical vapor deposition using GeO2 dopant31. The preform was pulled to obtain an optical fiber with a 900-µm exter-nal diameter, a 550-µm optical core (NA = 0.22) and a 95–105-µm hollow core. The fiber was cleaved by a diamond pen into 5–7-cm-long rods. Then the rods were heated and pulled with a standard laser puller (P-2000 laser-based micropipette puller, required for this type of glass; Sutter). Probes were generated with 2–3-mm-long shoulders, 3–8-mm-long shanks (Fig. 1a) with a tip pulled to sub-micrometer diameter and grinded back at 90° to the desired diameter on a rotating diamond abrasive plate (Fig. 1b; BV-10 micropipette beveler; plate 104C; Sutter). A wet-ting agent (Kodak Photo-Flo 200 solution) was added on the abra-sive plate to prevent small glass residues from blocking the hollow core. The hollow core was filled with 1–3 M NaCl solution using negative pressure. Then, a 50 µm-diameter tungsten or stainless steel wire was introduced into the large extremity of the fiber to make electric contact with the tip as with conventional glass microelectrodes. A 2-m-long multimode step index fiber (BFL22-550, core = 550 µm, cladding = 600 µm, NA = 0.22; Thorlabs) was aligned with the graded index core and fixed with a UV curing adhesive (NOA81, Norland products and Curing system UV75 light, Thorlabs) inside a borosilicate glass capillary (1.5 cm long, 1.5 mm inner diameter).

Optical setup. A laser diode was used as excitation light source (Lambda pro 50 mW 532 nm DPSS laser, Alltech Solutions or FCD488 20 mW 488 nm, JDS Uniphase). The light passed through a neutral density filter set (N.D. 1-3, NEK01, Thorlabs) and was reflected on a dichroic beam splitter (z543rdc or T495LP, Chroma). The beam was injected into a 10× microscope objec-tive (UIS-2 Plan-N, NA = 0.25; Olympus) fixed on a fiber launch system (KT110; Thorlabs) mounted on an xyz translation stage (PT3; Thorlabs). The laser beam was injected into the probe and propagated to the tip. The emitted fluorescence was collected through the microprobe tip, propagated back and collimated by the 10× microscope objective, passed through the dichroic beam splitter and entered in a 5.08-cm-long light isolator tube (SM1L30, Thorlabs). The fluorescent light was then filtered by two succes-sive band-pass filters (ET605/70M or ET525/50M, Chroma) and focused by a plano-convex lens (focal length = 50 mm, LA1131-A, Thorlabs) on a photomultiplier tube (PMT) detector (photosensor module H6780-20, Hamamatsu) mounted on an xyz translation stage. The entire optical system was enclosed in a light isolator box. The PMT output was amplified (gain = 0.5 to 10), filtered (30 Hz low pass, model 440, Brown lee precision), digitized and stored on a disk (Fig. 1c).

Brain-slice preparations. Brain slices were prepared from adult male Sprague-Dawley rats and bacterial artificial chromosome (BAC) transgenic mice expressing eGFP under the control of the

GAD65 promoter32. All procedures were performed according to the guidelines of the Canadian Council on Animal Care and were approved by the Animal Care Committee at Laval University. Mice and rats were deeply anaesthetized with ketamine and xylazine, and were decapitated. The brains were removed quickly (<60 s) and placed in a ice-cold solution containing 210 mM sucrose, 3 mM KCl, 0.75 mM CaCl2, 3 mM MgSO4, 1 mM NaH2PO4, 26 mM NaHCO3 and 10 mM glucose, saturated with 95% O2–5% CO2. Coronal slices, including the prelimbic area, were cut at 400 µm on a vibrating tissue slicer (VT 1000s, Leica) and kept into artificial cerebral spinal fluid (ACSF) containing 124 mM NaCl, 3 mM KCl, 1.5 mM CaCl2, 1.3 mM MgSO4, 1 mM NaH2PO4, 26 mM NaHCO3 and 20 mM glucose, saturated with 95% O2–5% CO2 at room temperature (21–23 °C). A slice was transferred to a submerge-type chamber where it was continuously exposed to ACSF at 30–32 °C saturated with 95% O2–5% CO2 and perfused at a rate of 2.0–2.5 ml min−1. The slice was viewed first with a 10× objective and the prelimbic area of the prefrontal cortex was localized as the area between the forceps minor corpus callosum and the midline. Neurons of the prelimbic area were viewed under near infrared illumination with a 40× water-immersion objec-tive (IR-ACHROPLAN, 40×, 0.80 W, Zeiss) and a charge-cou-pled device (CCD) camera (CCD-KP-M1AN, Hitachi). Labeled cells were identified by epifluorescence. The microprobes were mounted and moved with a MPC-200 microelectrode manipula-tor (Sutter Instrument).

Single-cell labeling in fixed tissue. In a subset of experiments, Lucifer yellow was injected in single neurons in fixed brain slices to control the distance between labeled cells. Briefly, the frontal part of a rat brain was fixed overnight in 4% paraformaldehyde (PFA) in 0.1 M phosphate buffer (PB) and then stored in 0.1 M PB at 4 °C. Coronal slices (400-µm-thick) were cut on a vibratome (Leica VT-1200S) and then placed in a patch-clamp recording chamber in 0.1 M PB for Lucifer yellow injection. Several pairs of pyramidal cell were thus injected, under visual control, with sharp electrode filled with 0.5% Lucifer yellow33.

In vivo extracellular single unit recording. Retrograde labeling of spinothalamic tract neurons was achieved using four injections of 1 µl DiI (2.5%, Invitrogen) under stereotaxic guidance in the thalamus of male Sprague-Dawley rats (200–250 g) under isoflu-rane anesthesia. Rats were anesthetized with isofluorane 7–10 d later and spinal segments L5–S1 were exposed by a laminaectomy and stabilized with two lateral clamps on the closest vertebra. A recording chamber was created on the back of the rat using agar (3%), and this chamber was perfused continually with buffered saline heated at 37 °C. A tracheotomy was performed, and the jugular vein was canulated. Rats were paralyzed with pancuro-nium bromide (Sigma, 5 mg ml−1, 0.05 ml h−1, intravenously) and ventilated to reduce breathing movements. Expiratory CO2 was maintained at 4% and body temperature at 37.5 °C using a heating blanket. The microprobe was mounted on a microman-ipulator (Burleigh, 6000). The extracellular signal was amplified on an ER-1 extracellular amplifier (Cygnus technology), filtered (bandwidth, 0.1–3.0 kHz), digitized and stored on disk simultane-ously with the PMT output signal.

In vivo experiments were also conducted on the GAD-eGFP BAC transgenic mice under ketamine (100 mg kg−1) and xylazine

doi:10.1038/nmeth.1572

© 2

011

Nat

ure

Am

eric

a, In

c. A

ll ri

gh

ts r

eser

ved

.

(10 mg kg−1) anesthesia. The mice were placed in a stereotaxic apparatus. The mice breathed freely and body temperature was maintained at 37.5 °C with a heating pad controlled thermostati-cally. Throughout the experiment, a deep level of anesthesia was maintained (stage III-3) using additional doses of anesthetics given at 1-h intervals (20 mg kg−1 ketamine plus 0.3 mg kg−1 xyla-zine intra muscularly) The microprobe was mounted on a micro-manipulator (Burleigh 8200, Inchworm) and descended into the brain with a 30° angle to reach GFP-expressing thalamic reticular neurons (2,200–2,500 µm deep). The extracellular electrophysio-logical signal was amplified (Neurodata IR183, Cygnus technol-ogy), filtered (band pass, 300–3,000 Hz, model 440, Brownlee Precision), digitized and stored on disk simultaneously with the PMT output signal.

Single-cell Ca2+ measurements. Ca2+ measurements were carried out in adult rats anesthetized as described above. A craniotomy was performed over the VPM region, and the dura matter was resected. In some experiments, another craniotomy was per-formed over the barrel cortex and the exposed tissue was lesioned with silver nitrate to abolish rhythmic thalamic activity induced by corticothalamic inputs. Vibrissae were cut at 5 mm from the skin. Before stereotaxic injection of Ca2+ indicators, the coordinates of the whisker responsive neurons were established by electrophysio-logical recording. The Ca2+ indicator Oregon Green BAPTA-1 AM (OGB-1; Invitrogen) was dissolved in dimethyl sulfoxide (DMSO) containing 20% pluronic (Invitrogen). This solution was then diluted (1/20) in 0.9% saline to obtain the final staining solution, which was loaded into 30-µm-tip injection micropipettes (Fischer Scientific, Drummond wiretrol II calibrated micropipettes). The injection micro-pipette was lowered to coordinates established by prior electrophysiological recording, and 200 nl of the OGB-1 AM solution was injected. Three to four whisker responsive sites were localized in the VPM and were labeled in each rat. When recording from single units, the whisker corresponding to the receptive field of the cell was deflected for 70 ms in four different directions by means of a piezoelectric stimulator.

Photostimulation in Thy1::ChR2-EYFP transgenic mice. Channelrhodopsin-2 transgenic mice were purchased from the Jackson Labs (B6.Cg-Tg(Thy1øCOP4/EYFP)18Gfng/J). In these experiments a craniotomy was performed over the motor cortex or over the VPM region of the thalamus. The microprobe was descended into the brain (800–4,000 µm deep). A 488-nm laser (FCD488 20 mW 488 nm, JDS Uniphase Corporation) was used to detect and activate ChR2-expressing cells. A set of different

neutral density filters ((ND) 0.1–4, Thorlabs) was used to con-trol the light stimulus strength and a mechanical shutter (LS3, Uniblitz) was placed in front of the laser beam to control stimulus duration. The light intensity at the microprobe tip (Itip) is given by (Pi × TND × Tfiber × Tp)/(πr2), where Pi is the laser power, TND is the transmission coefficient of the neutral density filters at 488 nm, Tfiber is the transmission coefficient after the injection, propagation in the relay fiber and across the optical junction with the probe, Tp is the transmission of the probe and r is the optical core radius at the tip. Tfiber and Tp were determined empirically to be 72% and 0.1%, respectively.

Histology. Brain and spinal cords were extracted from mice and rats under deep anesthesia. The entire mouse brain and the rat spinal segments L5–S1 were immersed in 4% paraformaldehyde and rinsed with a phosphate buffer solution. Tissue was cut in 80–100-µm-thick sections and mounted on glass slides. DiI- or GFP-labeled cell body diameters were measured as the average of two perpendicular diameters for each cell. Only cell bodies with visible nuclei in the section were used for the quantification.

Simulations and data presentation. Theoretical calculations and numerical simulations used to estimate the microprobe trans-verse and axial resolutions as well as its excitation and collection dynamics in tissue are described in Supplementary Data.

Statistical analysis. Correlation analysis was performed to ana-lyze relationship between FWHM and laser power, optical signal-to-noise ratio and probe diameter as well as Ca2+ responses and number of spikes discharged by the neurons. Regression analysis (ANOVA, F-test) was conducted to analyze the angular prefer-ence to test whether the relationship was linear and whether the slope was equal to 1.

29. Bennett, P.J., Monro, T.M. & Richardson, D.J. Toward practical holey fiber technology: fabrication, splicing, modeling, and characterization. Opt. Lett. 24, 1203–1205 (1999).

30. Knight, J.C., Birks, T.A., Russell, P.S. & Atkin, D.M. All-silica single-mode optical fiber with photonic crystal cladding. Opt. Lett. 21, 1547–1549 (1996).

31. MacChesney, J.B., O’Connor, P.B. & Presby, H.M. New technique for preparation of low-loss and graded-index optical fibers. Proc. IEEE 62, 1280–1281 (1974).

32. Lopez-Bendito, G. et al. Preferential origin and layer destination of GAD65-GFP cortical interneurons. Cereb. Cortex 14, 1122–1133 (2004).

33. Buhl, E.H. & Lubke, J. Intracellular Lucifer Yellow injection in fixed brain slices combined with retrograde tracing, light and electron microscopy. Neuroscience 28, 3–16 (1989).

doi:10.1038/nmeth.1572

© 2

011

Nat

ure

Am

eric

a, In

c. A

ll ri

gh

ts r

eser

ved

.