.

Journal of Neuroscience Methods 70 1996 121129

Dye screening and signal-to-noise ratio for retrogradely transported
voltage-sensitive dyes
Yang Tsau
a,),1
, Peter Wenner
b
, Michael J. ODonovan
b
, Lawrence B. Cohen
a
,
Leslie M. Loew
c,d
, Joseph P. Wuskell
c,d
a
Department of Physiology, Yale Uniersity School of Medicine, New Haen, CT 06510, USA
b
The Laboratory of Neural Control, National Institute of Neurological Disease and Stroke, NIH, Bethesda, MD 20892, USA
c
The Department of Physiology, Uniersity of Connecticut Health Center, Farmington, CT 06032, USA
d
Marine Biological Laboratory, Woods Hole, MA 02543, USA
Received 27 October 1995; revised 13 June 1996; accepted 17 June 1996
Abstract
Using a novel method for retrogradely labeling specific neuronal populations, we tested different styryl dyes in attempt to find dyes
whose staining would be specific, rapid, and lead to large activity dependent signals. The dyes were injected into the ventral roots of the
isolated chick spinal cord from embryos at days E9E12. The voltage-sensitive dye signals were recorded from synaptically activated
motoneurons using a 464 element photodiode array. The best labeling and optical signals were obtained using the relatively hydrophobic
dyes di-8-ANEPPQ and di-12-ANEPEQ. Over the 24 h period we examined, these dyes bound specifically to the cells with axons in the
. ventral roots. The dyes responded with an increase in fluorescence of 13% DFrF in response to synaptic depolarization of the
motoneurons. The signal-to-noise ratio obtained in a single trial from a detector that received light from a 14=14 mm
2
area of the
motoneuron population was about 10:1. Nonetheless, signals on neighboring diodes were similar, suggesting that we were not detecting
the activity of individual neurons. Retrograde labeling and optical recording with voltage-sensitive dyes provides a means for monitoring
the activity of identified neurons in situations where microelectrode recordings are not feasible.
Keywords: Voltage-sensitive dye; Optical recording; Embryonic chick spinal cord; Multi-site recording; Retrograde staining; Styryl dye; Dye synthesis;
Hydrophobic vs hydrophilic dye
1. Introduction
In the preceding paper we showed that injecting volt-
age-sensitive dyes extracellularly into the ventral root of
the embryonic chicken spinal cord preparation would retro-
gradely label only those populations of neurons projecting
axons through the root. Here we report testing of 10 styryl

dyes Ross et al., 1977; Gupta et al., 1981; Loew and

.
Simpson, 1981; Grinvald et al., 1983; Loew et al., 1992
and two cyanine dyes as optically active retrograde tracers.
Several of the styryl dyes were synthesized for these
experiments to allow us to test dyes with a range of
hydrophobicities and a localized charge that was either
positive, negative, or double negative. We anticipated that
)
. . Corresponding author. Tel.: 202 687-1617; Fax: 202 687-0617.
1
Present address: Georgetown Institute of Cognitive and Computa-
tional Sciences, Georgetown University, Washington, DC, USA.
dyes which were relatively hydrophilic would mainly move
by diffusion in the extracellular space while hydrophobic
dyes might label the neurons and cells with which they
made their initial contact.
We evaluated each of the dyes for three characteristics.
First, we compared the dyes for specificity of staining; did
the dye remain bound to the cells with which it made
initial contact or was the staining diffuse? Second, we
determined the signal size and the signal-to-noise ratio for
each dye. Third, we determined the rate of dye movement
from the ventral root injection site to the motoneuron cell
bodies.
A major goal of this report is to describe the efforts we
made to synthesize and test a variety of voltage-sensitive
dyes in order to optimize the signal-to-noise ratio for
retrograde transported dye. In the past, a number of unsuc-
cessful attempts have been made to label neurons specifi-
cally with voltage-sensitive dyes. Because it now appears
that these efforts failed because hydrophobic dyes were not
0165-0270r96r$15.00 Copyright q 1996 Elsevier Science B.V. All rights reserved.
. PII S0165- 0270 96 00109- 4
( ) Y. Tsau et al.rJournal of Neuroscience Methods 70 1996 121129 122
tested, we present enough structural information to allow
further improvement in the dye structure and to be explicit
about the best dyes for this purpose.
Preliminary reports of these experiments have appeared
.
Wenner et al., 1994; Tsau et al., 1994 .
2. Methods
2.1. Voltage-sensitie dyes
.
Many of the dyes we used Tables 1 and 2 were
synthesized for these experiments. The dyes designated
.
with the prefixes RH and NK see Section 4 were synthe-
sized by Rina Hildesheim and Amiram Grinvald at the
Weizmann Institute, Rehovot, Israel and at Nippon
Kankoh-Shikiso Kenkyusho, Co. Ltd., Okayama, Japan,
respectively.
The styryl dyes were prepared using variations of the
.
general procedures described by Hassner et al. 1984 see
.
synthetic scheme in Fig. 1 . These dyes were obtained as
very hygroscopic salts and were characterized by TLC, and
FAB Mass spectral analysis. Spectra were supplied by the
Midwest Center for Mass Spectrometry with partial sup-
port by the National Science Foundation, Biology Division
.
Grant No. DIR 9017262 . Detailed absorption and fluo-
rescence spectra were routinely obtained in ethanol and are
summarized in Table 1; all emission spectra were cor-
rected for wavelength dependence of the monochromator
and detector.
Table 1
Spectral characteristics of styryl dyes
abs em
. . . Dye designation l nm Log e l nm
max max
. Di-8-ANEPEQ JPW-1229 506 4.51 712
a
. Di-8-ANEPPQ JPW-2045 514 4.55 714
. Di-8-ANEPPS JPW-1153 500 4.50 705
. Di-8-ANEPEP JPW-1231 474 4.51 704
. Di-12-ANEPEQ JPW-1215 500 4.55 710
. Di-16-ANEPPQ JPW-2051 514 4.54 712
. Di-18:2-ANEPPQ JPW-2061 498 4.61 708
. Di-4-ANEPPS JPW-211 502 4.50 706
. Di-4-ANEPPQ JPW-1294 514 4.53 715
a
. w w . 1- 3-Trimethylammoniopropyl -4- b- 2- di-n-octylamino -6-
x x naphthyl vinyl pyridinium dibromide.
A description of the preparation of di-8-ANEPPQ fol-
lows and is illustrative of the general procedures employed
for all dyes. The numbers in parentheses refer to the
numbered structures in Fig. 1.
( )
6-Bromo-2-aminonaphthalene 2 . 6-Bromo-2-amino-
.
naphthalene 2 was obtained from 6-bromo-2-naphthol
.
using the Bucherer reaction Drake, 1942 .
( ) ( )
6-Bromo-2- di-n-octylamino naphthalene 3 . A mix-
.
ture of 5.6 g. 25.2 mmol 6-bromo-2-amino-naphthalene
. . .
2 , 13.2 g 55 mmol 1-iodooctane, 3.83 g 27.7 mmol
anhydrous potassium carbonate in 23 ml of anhydrous
N, N-dimethylformamide was heated under gentle reflux
Fig. 1. General synthetic scheme that was used for the synthesis of the styryl dyes used in these experiments. Additional details are provided in Section 2.
( ) Y. Tsau et al.rJournal of Neuroscience Methods 70 1996 121129 123
Table 2
Dye screening
Dye Number of preparations X n R Signal size
Hydrophobic dyes
q
. Di-8-ANEPEQ 16 - N CH 2 octyl qqq
3 3
q
. Di-8-ANEPPQ 19 - N CH 3 octyl qqq
3 3
y
Di-8-ANEPPS 6 -SO 3 octyl -
3
2y
Di-8-ANEPEP 1 -PO 2 octyl NS
3
q
. Di-12-ANEPEQ 6 - N CH 2 dodecyl qq
3 3
q
. Di-16-ANEPPQ 2 - N CH 3 hexadecyl q
3 3
q
. Di-18:2-ANEPPQ 7 - N CH 3 linoleyl qq
3 3
a
DiI 5 q
b
Fast-diI 3 q
c
Hydrophilic dyes
y
Di-4-ANEPPS 4 -SO 3 butyl -
3
q
. Di-4-ANEPPQ 4 - N CH 3 butyl - - -
3 3
q
. Di-2-ANEPEQ 4 - N CH 2 ethyl - -
3 3
a X X X
. 1,1-Dioctadecyl-3,3,3 ,3 -tetramethylindocarbocyanine perchlorate Molecular Probes, Eugene, OR .
b X X X
. 1,1-Dilinoleyl-3,3,3 ,3 -tetramethylindocarbocyanine perchlorate Molecular Probes, Eugene, OR .
c
These relatively hydrophilic dyes did not remain restricted to motoneurons; they stained the cord uniformly.
.
1308C for a period of 188 h. The mixture was cooled to
ambient temperature and partitioned with 50 ml H O and
2
40 ml ethyl acetate. The aqueous layer was extracted twice
with 20 ml ethyl acetate, the combined extracts were
.
washed with H O 30 ml , saturated NaCl solution 30
2
.
ml , dried over MgSO and concentrated on a rotary
4
evaporator to leave a brown oil. The crude oil was heated

to 75808C with a hot water bath and vacuum applied 0.1

.
mmHg to distill off unreacted iodooctane. The remaining
pot residue was purified by flash chromatography on silica
.
gel. Elution with hexane afforded 8.56 g. 76% yield of a
.
light yellow oil which was pure 3 by TLC analysis.
[ ] ( )
b- 2-Di-n-octylamino-6-naphthyl -4-inylpyridine 4 .

A heavy walled pyrex tube was charged with 3.0 g 6.7

. . .
mmol 6-bromo-2- di-n-octylamino naphthalene 3 , 0.96
.
g 9.13 mmol 4-vinylpyridine, 23 mg of palladium ac-
etate, 56 mg tri-o-tolylphosphine, and 6.0 ml dry trieth-
ylamine. The tube was flushed with nitrogen, capped, then
heated with magnetic stirring at 1051158C by means of
an oil bath for a total of 96 h. The cooled reaction mixture
was then partitioned with water and chloroform. The water
layer was extracted twice with 10 ml of chloroform. The
combined chloroform extracts were washed with water,
dried over MgSO , and concentrated on a rotary evapora-
4
tor to leave 2.89 g of a dark yellow oil. The crude product
was purified by chromatography on silica gel. Unreacted
starting materials were eluted with hexane and pure prod-
. .
uct 4 2.34 g, 75% of theoretical yield was eluted with

chloroform and ethyl acetate. TLC analysis silica gel,

.
ethyl acetate , showed one yellow fluorescent spot, R
f
0.433.
. Fig. 2. Dye spectra. Excitation left spectrum . The curve shows the
excitation spectrum of the dye di-8-ANEPEQ when bound to multi-
lamellar lipid vesicles. An emission wavelength of 638 nm was used to
generate this curve. The thick line below the curve represents the
transmission band of the incident-light interference filter we used. Emis-
. sion right spectrum . The emission spectrum of the dye bound to lipid
vesicles. An excitation wavelength of 465 nm was used. The solid line
below the curve indicates the wavelengths passed by the RG-610 sec-
ondary filter.
( ) Y. Tsau et al.rJournal of Neuroscience Methods 70 1996 121129 124
( ) [ [ (
1- 3-Trimethylammoniopropyl -4- b- 2- di-n-oc-
) ] ] ( )
tylamino -6-naphthyl inyl pyridinium dibromide 5 .
. w
A mixture of 0.320 g 0.679 mmol of b- 2-di-n-oc-
x .
tylamino-6-naphthyl -4-vinylpyridine 4 and 0.274 g 1.0
.
mmol of 3-bromopropyltrimethylammonium bromide in
3.0 ml of anhydrous dimethylformamide was overlaid with
nitrogen, then heated at 951058C in an oil bath for a
period of 48 h. The solvent was then removed on a rotary
evaporator and the dark red residue recrystallized from
absolute ethanol-diethyl ether, as a dark red resin. TLC
.
analysis silica gel, MeOH-CHCl 1:4 showed one rose
3
fluorescent spot just off the origin. Due to its hygroscopic
properties, the dye was stored and handled in ethanol
solution.
Fig. 2 shows the excitation and emission spectra of
di-8-ANEPPQ. The solid curve on the left is the excitation
spectrum of the dye when bound to multilamellar lipid
.
vesicles. The absorption spectrum not shown is very
similar to the excitation spectrum. The solid line below
this spectrum represents the transmission band of the
incident-light interference filter we used. The curve on the
right is the emission spectrum of the dye bound to lipid
vesicles. The solid line below this spectrum represents the
transmission band of the barrier filter, RG-610. Table 1
presents a summary of the spectra of several styryl dyes
.
in ethanol whose structures are shown in Table 2. The
spectra of styryl dyes is relatively independent of hy-
drophobicity and the sign of the localized charge. The
spectra in Table 1 were measured with the dye in ethanol.
As can be seen from a comparison of Fig. 2 and Table 1,
the excitation spectra in lipids are shifted to the blue by
3040 nm in comparison with the spectra in ethanol and
the emission spectra in lipids are shifted to the blue by
6070 nm in comparison with the spectra in ethanol.
2.2. Preparation and optical recording
The preparation and dye injection methods are de-
.
scribed in the preceding paper Wenner et al., 1996 .
2.3. Optical recording
Optical recordings were made using the 464 element
photodiode array. Except where mentioned, the traces il-
lustrate results from single trials. Except for two of the
traces in Fig. 4, the traces represent the outputs of individ-
ual detectors from single trials. The data was recorded at a
rate of one frame per 1.6 ms. The microscope field of view
was reasonably evenly illuminated; using a test slide with
a large drop of dye we found that the resting intensity on
all pixels was within "20% of the mean.
Because the resting light intensity in these measure-
ments was relatively low, dark noise could be a problem.
To minimize the effect of dark noise and to reduce the
relative shot noise the following three steps were taken to
. . Fig. 3. A hydrophobic dye di-8-ANEPEQ stains specifically while a hydrophilic dye di-4-ANEPPS stains diffusely. Photographs and schematic
. . drawings of a preparation stained with two different styryl dyes, di-8-ANEPEQ left-hand side of the preparation and di-4-ANEPPS right-hand side . The
. . two dyes were injected into the same root on opposite sides of the caudal lumbosacral spinal cord. A photomicrograph A and schematic drawing C of a
. . transverse cross-section of the cord demonstrating specific labeling of parasympathetic preganglionic neurons PGN and motoneuronal dendrites with the
. . . . dye di-8-ANEPPQ left side but diffuse labeling with di-4-ANEPPS right side . A photomicrograph B and schematic drawing D illustrating labeled
. . ventral roots VR , and motoneurons LMC in the isolated spinal cord viewed through the ventral white matter. Again, the staining with di-8-ANEPPQ
. left side is specific and the labeling with di-4-ANEPPS is non-specific. DREZ, dorsal root entry zone; WM, white matter. E8.
( ) Y. Tsau et al.rJournal of Neuroscience Methods 70 1996 121129 125
maximize the fluorescence intensity reaching the photodi-
ode array.
.
1 The light source was a 150 W xenon short-arc lamp
.
Osram, XBO 150 WrCR ORF in a model 770U lamp-

house powered by a model 1600 power supply Opti-Quip,

.
Highland Mills, NY . The light intensity from this arc
lamp was about 3-times brighter than that of a tungsten-
halogen source. Noise from arc wander was not detected.
.
2 A wide bandwidth incident light filter 520"45
.
nm was used. As can be seen from the bar in Fig. 2, this
filter passes wavelengths that include most of the long-
wavelength excitation band of the dye and thus should be
nearly optimal for a signal that in part results from a shift

in the peak wavelength of the absorption Loew et al.,

.
1985 . Smaller signals were obtained with narrower band-
pass filters.
.
3 We attempted to use lenses with a large numerical
.
apertures N.A. because the light intensity in an epifluo-
rescence measurement is proportional to the fourth power
.
of the N.A. Grinvald et al., 1983 . The numerical aper-
tures of the objective lenses are indicated in Table 1 of
.
Wenner et al. 1996 .
The wavelengths passed by the RG-610 secondary filter
are indicated by the solid bar in the right panel of Fig. 2.
These wavelengths include the long-wavelength portion of
the emission spectrum; from the spectrum of the fluores-
.
cence change of another styryl dye Loew et al., 1985 ,
using this portion of the spectrum is expected to optimize
the signal-to-noise ratio.
3. Results
3.1. Dye screening
We screened several different voltage-sensitive dyes
.
Table 2 for their properties as retrograde tracers; three
aspects of the dyes behavior were monitored; labeling
specificity, signal size, and rate of movement. The dyes
were injected into ventral roots and the spinal cord was
examined starting 30 min later for evidence of neuronal
labeling and for activity dependent signals. Optical signals
were measured through the ventral white matter with the
orientation of the preparation in the microscope field of
view similar to that in Fig. 3B,D. We measured fluores-
cence changes from labeled motoneuron populations re-
sulting from orthodromically activated afferent inputs from
the rostral cord. The signal size is expressed as the per-
.
centage change of the total fluorescence DFrF . Table 2
provides the chemical structures of the dyes as the indi-
cated substitution on the common backbone shown at the
top. Table 2 is divided into two sections: the nine dyes at
the top are relatively hydrophobic, the three dyes at the
bottom are relatively hydrophilic. We could not distinguish
the signals we obtained with di-8-ANEPEQ and di-8-
ANEPPQ; these dyes differ by only one -CH group. We
2
use the name di-8-ANEPPQ to refer to both dyes.
Labeling specificity was very dependent on the hy-

drophobicity of the dye determined by the hydrocarbon

.
chain length, R in Table 2 . Fig. 3 illustrates this difference
by comparing the staining with two styryl dyes which
differ in the lengths of the hydrocarbon chains attached to
the anilino nitrogen. The right-hand side of the preparation
was stained with a styryl dye with two butyl groups
.
di-4-ANEPPS and the left-hand side of the preparation

was stained with a dye with two octyl groups di-8-

. .
ANEPPQ . In both the transverse A,C and the longitudi-
.
nal B,D view it is clear that the more hydrophilic dye
stained very diffusely in comparison to the hydrophobic
dye. All three dyes from Table 2 that are relatively hy-
.
drophilic Rsdiethyl or dibutyl labeled the cord dif-
fusely. With these dyes it is not possible to observe either
axon tracts or cell bodies in the cord except during the first
hour after the dye injection into the ventral root. We do not
know whether this early apparent specificity of labeling is
real; it might reflect restricted extracellular diffusion path-
ways. Thus, these hydrophobic dyes do not appear to be
attractive for monitoring membrane potential from specific
neuronal populations. By contrast, labeling with dyes like

di-8-ANEPPQ that are more hydrophobic dioctyl or longer

.
carbon chains was specific over the 24 h period we
examined. Only the axons, cell bodies, and dendrites of the

neurons projecting through the root were stained also see

evidence presented in the preceding paper Wenner et al.,

..
1996 .
The signal size and polarity depended dramatically on
.
both the hydrophobicity R and the sign of the localized
.
charge X in Table 2 . The size and time course of the
signals from five hydrophobic dyes are illustrated in Fig.
4. The largest signals were found with di-8-ANEPPQ and
di-12-ANEPPQ. Smaller signals were found with di-18:2-
ANEPPQ and di-16-ANEPPQ. Di-8-ANEPPS, with a neg-
ative localized charge, had a small and slow signal that
was a decrease in fluorescence. The results presented in
Fig. 4 show that di-8-ANEPPQ not only had the largest
.
fractional fluorescence change DFrF but also the great-
est signal-to-noise ratio. The fluorescence intensity mea-
sured with this dye was relatively large.
The last column of Table 2 shows the relative signal
size obtained with these five and seven additional dyes. No
signal was found using di-8-ANEPEP with a double nega-
tive phosphonate; only small signals were found with di-I

and fast-di-I. We used a scale of signal sizes from NS no

.
signal to qqq or - - - where the number of symbols
qualitatively indicates signal size and the sign indicates the
direction of the signal. This comparison of optical signal
sizes must be considered provisional because we do not
have a direct measure of the size of the motoneuron
electrical response. While a supramaximal stimulus to the
rostral cord should result in a similar depolarization in
motoneurons in different preparations, the size of the
response and the number of cells activated will depend on
the health and age of the preparation. In addition, the
( ) Y. Tsau et al.rJournal of Neuroscience Methods 70 1996 121129 126
Fig. 4. Signals from motoneuron populations labeled with five different
hydrophobic dyes. For each of these dyes the signal illustrated was one of
the largest we measured. Di-8-ANEPPQ had the largest fractional fluores-
cence changes as well as the largest signal-to-noise ratio. The analogue
with a negative localized charge, di-8-ANEPPS, had a small, slow, signal
that was opposite in direction. A dashed line drawn through the prestimu-
lus data for di-8-ANEPPS is used to make it easier to detect this
relatively small and slow signal. In this and subsequent figures fluores-
cence signals were measured through the ventral white matter using
transverse illumination with the orientation of the preparation in the
microscope field of view similar to that shown at the bottom of Fig. 5.
Each trace was filtered with two passes of the 1-2-1 digital filter. The
result for di-16-ANEPPQ is the mean of three detectors; the results for
the other dyes are from a single detector. The result for di-8-ANEPPS is
from the average of eight trials; the results for the other dyes are from a
single trial. The y-axis for each trace was adjusted so that the same scale
of fractional change in fluorescence could be used for all traces. The
resting fluorescence values were corrected by subtracting the background
fluorescence. In this and following figures the direction of the arrow
indicates the direction of an increase in fluorescence; the filter used to
remove low-frequency noise had a time constant of 500 ms; this filtering
will have only a small effect on the time course of the signals shown.
degree of compartmentalization of the dye between the
membrane and the cytosol may vary from preparation to
preparation and we presume that cytosolic dye does not
respond to transmembrane potential changes. However, in
some cases, the signals from different dyes were compared
in the same embryo to improve the validity of the compari-
son.
The rate of movement of dye from the injection site to
the cell bodies in the lateral motor column was inversely
proportional to hydrophobicity. Di-2-ANEPPQ and di-4-
ANEPPQ moved very quickly; small voltage-sensitive dye
signals could be detected in the cell body region 30 min
after the injection. Di-8-ANEPPQ moved at an intermedi-
ate speed. Di-8-ANEPPQ signals could be detected after 2
h but cellular labeling of ventral motoneurons required
about 6 h. Di-16- and di-18-ANEPPQ took the longest to
stain the motoneurons. To allow complete labeling of the
motoneuron cell bodies with di-8- and di-12-ANEPPQ,
ventral roots were injected and the preparation left
overnight at room temperature. After longer waiting peri-
ods the staining became slightly more diffuse in appear-
ance. It is possible that this resulted from more complete
staining of the motoneuron dendritic trees.
The most common signal from motoneurons retro-
gradely labeled with di-8-ANEPPQ, and all of the more
hydrophobic styryl dyes with a positive localized charge, is

an increase in fluorescence during depolarization last

.
column, Table 2 . This positive-going signal is opposite in
direction to the signals found on other preparations where
the staining periods were brief and the dye was bath-ap-

plied Ross et al., 1977; Orbach et al., 1985; Orbach and

.
Cohen, 1983; Kauer et al., 1987 , opposite to the signal
obtained from motoneurons retrogradely labeled with di-
8-ANEPPS, and also opposite in direction to the signals
.
found with the three hydrophilic dyes Table 2, bottom .
In all of the presented results an increase in fluores-
cence was found during motoneuron depolarization using
di-8-ANEPPQ. This was true of the vast majority of

recordings. However, in rare instances recordings from 3

.
out of more than 200 stained motoneuron pools , an in-
verted signal was found with this dye.
3.2. Diode array recording
Using one of the best dyes, We have examined the
signals detected with the 464 element photodiode array to
assess the possibility of resolving the action potentials of
individual motoneurons. The inset at the bottom of Fig. 5
shows schematically the relative position of the image of
the spinal cord and the pixels of the array. Fig. 5 illustrates
recordings from a labeled motoneuron population using the
464 element array and di-8-ANEPPQ. The array received
light from an area of spinal cord that was about 360 mm in
diameter; each pixel received light from a square area of
14 mm
2
. The area enclosed by the dashes in the figure
indicates the approximate extent of the labeled motoneuron
population. The fluorescence intensity reaching the pixels
inside the dashed outline was about 10-times larger than
that outside the outline. The rostral spinal cord was given a
single electrical stimulus at the time indicated by the
triangles below the three traces in the center of Fig. 5. This
stimulus is expected to produce a synaptic response in
motoneurons; the early part of this response was detected
optically. A rapid increase in fluorescence is seen through-
out the region that was stained. The results illustrated in
Fig. 5 show that the optical signals are easily recorded
without signal averaging. Moreover, the signals on individ-
( ) Y. Tsau et al.rJournal of Neuroscience Methods 70 1996 121129 127
ual diodes probably originate from a small number of cells
because of the area of the object plane contributing light to
an individual diode was only 14=14 mm
2
.
However, the signals on adjacent detectors in Fig. 5 are
similar; this point is examined in more detail in Fig. 6
where the signals from a group of 16 detectors from a
second preparation are displayed with expanded time and
Fig. 5. Optical recordings made using a 464 diode array from motoneu-
rons retrogradely labeled with di-8-ANEPPQ. A schematic diagram show-
ing the approximate position of the photodiode array elements on the
image of the chick spinal cord projected onto the array is shown at the
bottom. Relatively large optical signals are seen in the area of cord that
had neurons that were retrogradely labeled by the dye. Each trace
represents the output of one photodiode. The pixels are arranged so that
their relative position in the figure corresponds to the relative position of
the areas of the cord projected onto the octagonal array as shown in Fig.
. 1A of the preceding paper . The time of the stimulus to the rostral cord is
indicated by the small arrows under three of the detectors in the center of
the array. Top is medial, left is caudal. In this recording four of the
photodiode amplifiers were not functioning. For each of those pixels, the
data displayed is the mean of the abutting four pixels. The data in this
. figure are presented as the change in fluorescence DF ; they are not
scaled by dividing by the resting fluorescence. However, DFrF, the
change in fluorescence divided by the resting fluorescence indicated by
. the vertical calibration line is shown for the detectors in the center of the
field. The photocurrent across the 1 GV resistor in the first stage
amplifier for the detectors in the center of the field was 10y
9
amps. No
digital filtering was used. E10 embryo; the 40=, 0.75 N.A. objective was
used. The recording was made 8 h after dye injection.
Fig. 6. Expanded signals from a portion of an array recording similar to
that illustrated in Fig. 5 but from a different preparation. The signals on
adjacent detectors are quite similar suggesting that the signals on the
individual detectors do not result from the potential change in an individ-
ual motoneuron. Each trace was filtered with two passes of the 1-2-1
digital filter. E12 embryo; the 40=, 0.75 N.A. objective was used. The
recording was made 11 h after dye injection.
vertical axes. In this expanded presentation it is clear that
the fluorescence change on neighboring detectors is simi-
lar. Furthermore, there is not a large increase in the noise
following the stimulus that might result from random
spiking by individual neurons. Thus, we think that these
fluorescence signals are likely to be population signals
representing the average of the potential change of a group
.
of neurons see Section 4 .
4. Discussion
To determine which voltage-sensitive dyes applied to
the ventral root label neurons in a specific manner and
produce the largest optical signals we screened several
dyes. Our results demonstrate that retrograde movement of
relatively hydrophobic voltage-sensitive dyes results in
specific labeling of identified neuronal populations. This
result makes it possible to assign, unambiguously, the
optical signals obtained with these dyes to a well-defined
neuron type in a vertebrate central nervous system. The
largest signals and best signal-to-noise ratios were ob-
tained using di-8-ANEPPQ and di-12-ANEPEQ; we ob-
tained a signal-to-noise ratio of about 10:1 in a single trial
from a pixel that received light from a 14=14 mm
2
area
.
of the motoneuron population Figs. 46 .
The location of the styryl dyes in the neuron membrane
( ) Y. Tsau et al.rJournal of Neuroscience Methods 70 1996 121129 128
may be inferred from the sign of the fluorescence change.
When styryl dyes with either positive or negative localized

charge were bath applied with brief staining periods tens

. .
of minutes to squid axons Ross et al., 1977 , rat cortex
.
Orbach et al., 1985 , or salamander olfactory bulb Orbach
.
and Cohen, 1983, Kauer et al., 1987 , their fluorescence
always decreased during depolarization. Because these dyes
are not rapidly membrane permeant and the incubations
were brief, it was assumed that the dyes were mainly
bound to the external leaflet of the neuron membrane.
Furthermore, in experiments using squid axons, signals of
opposite direction were obtained when dyes were injected
.
internally into the axoplasm Ross et al., 1977 . The
finding that the signals obtained with retrogradely trans-

ported dyes of opposite localized charge di-8-ANEPPQ

.
and di-8-ANEPPS, Table 2 , are opposite in sign then
suggests the possibility that these two dyes are mainly
located on opposite leaflets of the plasma membrane. The
sign of the di-8-ANEPPS signal suggests that the dye is
mainly in the outer leaflet while the sign of the di-8-
ANEPPQ signal suggests that this dye is mainly in the
inner leaflet. Partitioning of di-8-ANEPPQ on the inner
leaflet may occur because its positive localized charge
favors an internal localization. Even a relatively low rate
of flipping across the membrane might be enough to
account for an internal localization after an incubation
period of many hours. Alternatively, this dye could ini-

tially insert into the membrane in the inner leaflet based

.
on positive charge at the injection site where the axons
were damaged by the electrode or by the solvent used to
dissolve the dye or the dye molecules could be actively
transported back to the cell body via intracellular or-
ganelles and then partition into the inner leaflet.
.
The hydrophilic dyes bottom three dyes of Table 2
had negative going signals suggesting an external localiza-
tion. In addition, these dyes labeled the cord in a diffuse
.
manner Fig. 3 . Presumably these dyes did not bind
tightly to motoneuron axons and traveled by diffusion in
the extracellular space, eventually binding to the outer
leaflet of most neuronal and glial membranes.
Two groups made earlier attempts to measure voltage-
sensitive dye signals from dye that had been transported
along axons. In each case, the dyes used were relatively

hydrophilic. Orbach and London personal communication,

.
1984 injected two hydrophilic styryl dyes, RH160 and
RH414, and a merocyanine dye, NK2367, into the sala-
mander olfactory nerve. Consistent with our findings that
.
hydrophilic dyes would stain indiscriminately Table 2 ,
they did not detect specific staining in the olfactory bulb.
.
Lev-Ram and Grinvald 1987 injected the dye RH461, an
analogue of di-2-ANEPPQ, intraocularly in the rat and,
after a 410 day waiting period, monitored signals from
dye which had moved in the anterograde direction into the
optic nerve. Their signals, like those we obtained with
di-8-ANEPPQ, were an increase in fluorescence during
depolarization. Thus, RH461 is another example of a
positively charged dye whose equilibrium position appears
to be on the inner leaflet of the neuron membrane.
Two dyes, di-4-ANEPPS and di-8-ANEPPS, that yield
relatively large signals after brief extracellular application

to a variety of preparations Loew et al., 1985, Bedlack et

.
al., 1992 , had small signals in our experiments. One
possible explanation is that these dyes become evenly
distributed on the outer and inner membrane leaflets after
the long incubation period we used and this leads to a
cancellation of their signal.

In a minority of the cases recordings from 3 out of

.
more than 200 stained motoneuron pools the fluorescence
changes found with di-8-ANEPPQ were negative-going
.
opposite to normal . One possibility is that a negative
signal was recorded because more dye was bound to the
external rather than the internal leaflet of the membrane.
This could happen if the dye failed to insert into the
damaged axonal membranes at the injection site and leaked
around the axons labeling them externally or, if the dye
had been accidentally injected directly into the spinal cord.
Alternatively, the relative partitioning of the dye between
inner and outer membrane leaflets may be variable.
Dil is an indolenine-cyanine; several less hydrophobic
members of this group are known to be voltage-sensitive
.
Cohen et al., 1974; Ross et al., 1977 . However, in
comparison with the styryl dyes, only small signals were
obtained with Dil and Fast-Dil.
Although the retrogradely transported dyes appear to be
reliable indicators of membrane potential, asynchronous

spiking activity was not found in the optical signals Fig.

.
6 . One explanation is that the optical signals come from
motoneuron populations and asynchronous spiking of an
individual cell is lost in the noise from the background
fluorescence of the population. In the top trace of Fig. 4
and in Figs. 5 and 6, the signal-to-noise ratio for an
individual detector in a single trial was about 10:1. These
detectors probably received light from about 10 labeled

motoneurons an estimate of 2 cell bodiesrdetector in the

X-Y plane=5 overlapping cell layers in the Z-direction. If
the cell bodies were the major source of light in the
recording, and a single cell made an action potential, then
the signal-to-noise ratio for that spike would be about 1:1

individual motoneuron signal: noise of all 10 motoneu-

.
rons . However, another source of fluorescence is from
dendritic branches from other motoneurons in the labeled
population. This contribution would reduce the expected
signal-to-noise ratio to less than 1:1. The signal-to-noise
ratio associated with individual spikes will be even lower
in the photomultiplier recordings because hundreds of
neurons contributed to the optical signal.
Additional experiments will be required to determine
how close we are to being able to detect the signal from an
individual neuron. One strategy to reduce the number of
neurons per diode would be to crush the root and do the
injection distal to the crush. This would label fewer mo-
toneurons and they would be more sparsely grouped. An-
( ) Y. Tsau et al.rJournal of Neuroscience Methods 70 1996 121129 129
other possibility would be to record from the cut face of
the spinal cord, focusing on isolated neurons. In any case,
.
it is important to note that ODonovan et al. 1993 have
already shown that changes in calcium from single action
potentials could be detected using calcium sensitive dyes.
While, it is not yet clear which kind of dye will be most
useful for monitoring spiking activity of individual neu-
rons in various circumstances, voltage-sensitive dyes could
allow detection of both subthreshold and inhibitory poten-
tials.
Acknowledgements
Meide Wei kindly provided the dye spectra shown in
Fig. 2 and Table 1. We are grateful to Avrum Cohen for
the analysis software. We thank David Senseman for the
loan of the 464 element photodiode array. Vic Pantani and
Henrik Abildgaard of the Physiology electronics shop de-
signed and constructed the amplifiers and analog-to-digital
converter used to record the output of the diode array.
Supported in part by Grant No. NS08437 from NINDS and
GM35063 from NIGMS.
References
. Bedlack, R.S., Wei, M.-D. and Loew, L.M. 1992 Localized membrane
depolarizations and localized calcium influx during electric field-
guided neurite growth, Neuron, 9: 393403.
Cohen, L.B., Salzberg, B.M., Davila, H.V., Ross, W.N., Landowne, D.,
. Waggoner, A.S. and Wang, C.H. 1974 Changes in axon fluores-
cence during activity: molecular probes of membrane potential, J.
Membr. Biol., 19: 136.
. . Drake, N.L. 1942 The Bucherer reaction. In R. Adams Ed. , Organic
Reactions, Vol. 1, Wiley, New York, pp. 105127.
. Grinvald, A., Fine, A., Farber, I.C. and Hildesheim, R. 1983 Fluores-
cence monitoring of electrical responses from small neurons and their
processes, Biophys. J., 42: 195198.
Gupta, R.K., Salzberg, B.M., Grinvald, A., Cohen, L.B., Kamino, K.,
. Lesher, S., Boyle, M.B., Waggoner, A.S. and Wang, C.H. 1981
Improvements in optical methods for measuring rapid changes in
membrane potential, J. Membr. Biol., 58: 123137.
. Hassner, A., Birnbaum, D. and Loew, L.M. 1984 Charge shift probes of
membrane potential: Synthesis, J. Org. Chem., 49: 25462550.
. Kauer, J.S., Senseman, D.M. and Cohen, L.B. 1987 Odor elicited
activity monitored simultaneously from 124 regions of the salamander
olfactory bulb using a voltage sensitive dye, Brain Res., 418: 255261.
. Lev-Ram, V. and Grinvald, A. 1987 Activity-dependent calcium tran-
sients in central nervous system myelinated axons revealed by the
calcium indicator fura-2, Biophys. J., 52: 571586.
. Loew, L.M. and Simpson, L.L. 1981 Charge-shift probes of membrane
potential: a probable electrochromic mechanism for p-aminostyryl-
pyridinium probes on a hemispherical lipid bilayer, Biophys. J., 34:
353365.
Loew, L.M., Cohen, L.B., Salzberg, B.M., Obaid, A.L. and Bezanilla, F.
. 1985 Charge-shift probes of membrane potential. Characterization
of aminostyrylpyridinium dyes on the squid giant axon, Biophys. J.,
47: 7177.
Loew, L.M., Cohen, L.B., Dix, J., Fluhler, E.N., Montana, V., Salama, G.
. and Wu, J.Y. 1992 A naphthyl analog of the aminostyryl pyridinium
class of potentiometric membrane dyes shows consistent sensitivity in
a variety of tissue, cell, and model membrane preparations, J. Membr.
Biol., 130: 110.
. ODonovan, M.J., Ho, S., Sholomenko, G. and Yee, W. 1993 Real-time
imaging of neurons retrogradely and anterogradely labelled with
calcium sensitive dyes, J. Neurosci. Methods, 46: 91106.
. Orbach, H.S. and Cohen, L.B. 1983 Optical monitoring of activity from
many areas of the in vitro and in vivo salamander olfactory bulb: a
new method for studying functional organization in the vertebrate
central nervous system, J. Neurosci., 3: 22512262.
. Orbach, H.S., Cohen, L.B. and Grinvald, A. 1985 Optical mapping of
electrical activity in rat somatosensory and visual cortex, J. Neurosci.,
5: 18861895.
Ross, W.N., Salzberg, B.M., Cohen, L.B., Grinvald, A., Davila, H.V.,
. Waggoner, A.S. and Wang, C.H. 1977 Changes in absorption,
fluorescence, dichroism, and birefringence in stained giant axons:
optical measurement of membrane potential, J. Membr. Biol., 33:
141183.
Tsau, Y., Wenner, P., Kleinfeld, D., Friedman, B., Stepnoski, R.A., Falk,
. C.X., Cohen, L.B., Loew, L.M. and ODonovan, M. 1994 Hy-
drophobic voltage-sensitive dyes for monitoring the activity of spe-
cific populations of central neurons, Abstr. Soc. Neurosci., 20: 700.
Wenner, P., Tsau, Y., Cohen, L.B., ODonovan, M.J. and Wuskell, J.P.
. 1994 Optical monitoring of rhythmic, synaptic, and orthodromic
activity from identified spinal neurons retrogradely labeled with volt-
age sensitive dyes, Abstr. Soc. Neurosci., 20: 700.
Wenner, P., Tsau, Y., Cohen, L.B., ODonovan, M.J. and Loew, L.M.
. 1996 Voltage sensitive dye recording using retrogradely transported
dye in the chicken spinal cord: staining and signal characteristics, J.
Neurosci. Methods, 70: 111120.