- Email: [email protected]
Information processing in the system from light to horizontal cell response of the carp retina
INFORMATION PROCESSING IN THE SYSTEM FROM LIGHT TO HORIZONTAL CELL RESPONSE OF THE CARP RETINA 0. Lsuso Department
of Physiology.
Tokyo
Women’s
and K.
WAT.ASABE
Medical College. Japan
Kawada-cho.
Shinjuku-ku.
Tokyo
162.
Abstract-in
this paper. two types of experiments in the carp retina are presented. One is the analysis of responses to white-noise-modulated light at drfferent average mean intensity levels recorded from L-type horizontal cells. For weak light, the responses fluctuated symmetrically around mean values. For brighter hght. ho\\e\rr. the distribution of amplitude of the response departed from the normal distribution. The deviation from the normaI distribution was characterized in terms of symmetry rather than peakedness. The deviation also implies the existence of a nonlinearity in the system. The other type IS with barium Ion. Ba2- produced dramatic effects on the spectral response patterns of horizontal cells; the depolarizmg response IO long ua~elrngth lieht of the R G cell and the hypzrpolarizing response to deep red light of the Y RB cell were eliminated-in BaL”-containing solutions. The response of the L cell to whitr-noisrmodulated hght HIS also chanced by application of Ba?-. In Ba”-containing solution, the response became iar~er ;S well as slower than that in normal solution. The results of Ba” are discussed in relation IO the interactions
Carp retins
betwern
cone and horizontal
Horizontal ceil
B;$- iun
cells. White
INTRODUCTION
In a previous paper (Umino and Watanabe, 1981 b), considering a horizontal cell response to light as the response of an entire system composed of photoreceptors and a horizontal ceil, the system was divided into four subsystems, A, 8, C and D. The subsqstsm A relates principally to the first event of the system involving the absorption of photons, the subsystem B to the properties of the photoreceptor membrane, the subsystem C to those of synaptic transmission and the subsystem D to the horizontal cell itself. Among them. characterisiics of a system compused of the subsystems C and D were investigated by urihzing an electrically evoked response (E-response) of the horizontal cell as an impulse response of the sbstem to input from photoreceptor terminals (Umino and Watanabe, 1981b). since the E-response hits been thought to be the summated EPSP caused b) an increase in transmitter release evoked hg’ ;I tranricnt depolarization of the photorL’ccp:i): trrmlii;ii\ b> ;i rran~ctiiial electrical current pulse (Trifonov and Byzov. 1965; Watanabe ef nl., 197s). In the present paper. some properties of the response IO light of the entire system up to horizontal ceils \\erc e\;umincd by applying uhite-noisemodulated hght stimuli at dift‘erent average mean intensity levels. Probability density functions. also kno\\n as distribution of amplitudes of the response, were analyzed. and the etl-ect of barium ion on them was also examined. Although effects of barium ion on retinal cells have already been reported by several Lvorkers (Piccolino and Grrschenfeld, lY7S, 1980; I229
noise
Statistical
properties
Gerschenfeld and Piccolino, 1980; Tachibana. 1981). newly found dramatic effects of this ion on the spectral response patterns of horizontal cells will be presented and its influence to the probability density functions will be discussed. METHODS Prepmotions,
recording
cd
iighl
stinidi
Experiments were performed on the carp Cyprinrcs Recordings were obtained from the isolated retina, superfused with an artificial sahne solution. contained NaCl(iO0 mM), Normal solution NaHCO,(20 mM), KCI(2 mM), &Cl, (2 mM), MgCIz( 1 mM) and glucose (10 mM) and pH was adjusted to 7.8 by bubbling a gas mixture of 977,; O? and 3”,, CO?. Ions to be tested were added to the perfustng fluid without compensating for osmolarity. The horizontal cells were impaled with superfine microelectrodes (50-150 MQ) filled with 3 M KCI. Photic stimulation \vas provided by a t\vo-channel p~~t~)~titnul~t~r, both beam> soi to deli\e:r a spot of 4 mm diameter centered at the microelectrode tip. The unattenuated intensity of light was 7 x lO-‘~_~W/~rn’ for white light. In the experiments with white-noise stimuli, the light was modulated in m-sequence fashion (maximum-length linearly recurring sequence). According to Naka and Marmareiis (1975). the whitenoise approach nearly maximizes the rate of information gathering about the system behavior and results in a complete quantitative description. The whitenoise method employed was described in detail in a previous paper (Umino tr ul., lYS3). To obtain the carpio.
tar
Fig. I. Schematic diagram of the two-channel optical stimulator. Light source is a SOOW xenon arc lamp. Two light beams are symmetrically arranged except for a device placed in the test beam for modulating light in white-noise fashion. In each beam an electromagnetic shutter for conventional use is furnished. but it is not shown in the figure. To modulate the test light, a rotatory disk with twelve holes through which the light passes is directly driven by a stepping motor (PA-62 Nippon Pulse Motor Co.) which is controlled by the m-sequence (maximum-length linearly recurring sequence) signal. The sequence generator is a 5msec clock driven through I2 shift registers. Therefore, cycle of m-sequence employed is 20.475 set one [S msec x (2” - I)]. There were three neutral density filters (ND). ND filter I attenuates the test beam and ND filter II the background, so that the ratio of ND filter I to ND filter II determines the depth of modulation. ND tilter III serves to change the total light intensity and to determine the average mean intensity level keeping the modulation depth unchanged. The brightest modulated light (test beam + background) was composed of intensities 7 x lO-‘j~W/~rn~ for high level and I x IO-‘/rW/~m’ for low level (background light alone), and the ratio of the high level to the low one of modulated light is 7 in all experiments. The modulated light has a cutoff frequency around 90 Hz, and can be regarded as the analog signal which has the normal distribution for the retinal neuron system (see Methods). Inset shows a detailed sketch of the device for modulating the light stimuli. To improve the rise and fall times of light. a plate with an aperture stop smaller than one-tenth of the holes on the rotatory disk is placed behind the disk. Accordingly, rise and fall times are determined by the time the edge of holes on the disk passes across the aperture stop. Test light is monitored by a photomicrosensor composed of a photodiode and a detector (EE-SC Omron Tateisi Electronics Co.). (Produced by permission from Umino er 01.. 1983.)
white-noise-modulated light, a rotatory disk with I2 holes through which light passed was placed in the beam and driven by a stepping motor (PA-62, Nippon Pulse Motor Co.) vvhich was controlled by the m-sequence signal (Fig. I). The sequence generator was a 5 msec clock driven 12 shift registers. Therefore, one cycle of m-sequence employed was 20.475 set [5 msec x (2” - I)]. A neutral density filter designated ND filter I determined the intensity of test light and ND filter II placed in the second beam the intensity of background light. The ratio of ND filter I and ND filter II gave the depth of modulation, and ND filter III determined the total intensity of light under the condition of a constant modulation depth. The brightest light in the white-noise experiments was composed of intensities 7 x lO-‘~~Wi~~rn’ for high
level and I x !Om‘pU’ jlrn’
Some statistical functions were calculated to obtain a quantitative description of the response to whitenoise light. All calculations were made for the responses to the one cycle white-noise-modulated light. since the statistical properties of a part of the entire (one cycle) m-sequence is worse than those of the entire m-sequence (Umino er ul., 19S3). The probability density function describes the amplitude distribution of the response. It is useful to calculate this function to detect nonlinearity in the system, if it exists, since the deviation of the distribution of the response amplitude from the normal distribution implies its existence. The number of classes (NC) in the probability density function was calculated from the equation IV<= I + 3.3 x log ,t’ (IV: sample size) (Sturges. 1926). To describe the deviation from the normal distribution, skewness and kurtosis were calculated. Skeuness means lack of symmetry, and a measure of skewness shows how far the distribution departs from symmetry. Skewness ( fi) corresponds to the Third Moment and is defined \‘&
=/I,‘&
where N is the sample size, 11 is the mean value, x is the observed value, /lr is the Third Moment and CTis the standard deviation. For positively skewed distributions. .:/ii; :. I), for :I s~nlmi‘t~~z distribution ,_,x = 0 and for negative skewness, V/jy < 0. Kurtosis may be defined as “peakedness” and a measure of kurtosis serves to differentiate between a flat distribution curve and a sharply peaked curve. Kurtosis (IL) corresponds to the Fourth Moment and is defined
where pJ is the Fourth Moment. For a normal distribution kurtosis is equal to 3. If the distribution is more sharply peaked than the normal, kurtosis is
Informatton
processing
than 3, and if Ratter than the normal. kurtosis is less than 3. It is necessary to test these values. since the degree of freedom is different in each datum. The test adopted was reported by Kikkawa et al. (1981). The number of degrees of freedom (X-) is given by
m the horizontal
cell
vvhere T is the observation time and equivalent spectral width defined by
,
. \
I+‘, is the
where P(j) is the power spectrum. If we get the number of degrees of freedom from the data. we can calculate the level of significance in line with the traditional method (see Kikkawa et al., 1981). For skewness, a single-sided test was adopted, because Weber’s law exists in the biological system and it would be expected that the function might be distorted in one direction. For kurtosis. a two-sided test was adopted, since kurtosis might be changed in either direction. Cross-correlation was performed between the input (light) signal monitored (see Fig. I) and the resulting response. Gain characteristics were obtained by calculating Fourier transforms of the cross-correlation function
When the average mean intensity level of vvhitenoise-modulated light was increased. keeping the depth of modulation constant. both static (DC) and dynamic (fluctuating) components of the horizontal cell response became larger. Figure 2(A) shovvs an example of the Ructuating components of the responses to modulated white lights at different average mean intensity levels increased stepwise from (a) to (f), and the relation of the magnitudes of the DC component to the stimulus intensity IS illustrated in Fig. 2(B). For dim light, the response fluctuated around the mean level. For brighter light. however. upward fluctuations became larger than downward fluctuations. To describe this feature quantitatively. distributions of the amplitude (probability density function) of the responses were calculated and were plotted in Fig. 3. The average mean intensity level is increased from top to bottom. It is apparent that the probability density function of the response shows the normal distribution for the lowest average mean intensity level of light (a), but the function departed from the normal distribution with increases in the average mean intensity level. To evaluate the normality of the probability density function of the response, skewness and kurtosis were calculated (see Data analysis). Skewness means lack of symmetry and is 0 for the normal distribution. Skewness values of the responses at different intensity
(mV)
or :
L10
Light
intensity
(lag)
mV
5 set
Fig. 2. Responses of L-type horizontal cell to light modulated in white-now fashion (one cycle). The intensities7 x IO-‘pW!pm’ for high level and I x IO ktW)}lrn’ brightest light (0 log) was composed for low level. and the ratio of the high lebel to the low one was kept constant throughout the experiment. 1‘4) Dynamic (tluctuatmg) components of responses. Average mean intensity level of light is increased by 0.5 log untt steps from top (a) to bottom (f). Letters (a-f) correspond to those in Fig. Z(A) in the following tigurss. (B) R&non of the mean values of the response amplitudes to the average mean intensittos of the hght stimuli. Soltd circles indicate the mean value of responses and vertical poles standard devtatton of modulated response. Stimult were white light covet-in, 0 a circle of 4 mm tn dtametcr.
of
I
RESULTS
greater
,-
I23
-80
-70
-60
Msmbmnc
-so
potsnnol Imv)
Fig. 3. Probability density function of the responses (in other words, distribution of amplitude) of Fig. 2. Light intensity was increased from top (a) to bottom (f). Letters (a-f) correspond to those in Fig. 2. Solid triangle indicates the dark membrane potential of the cell.
normal distrtbution indxates that the nonhnexrr~ GL the system from light to honzontai cr!l r?:sp~-in% becomes stronger with increases in the average IIIZLLI: intensity level of the stimulus light. It is evident that variance ot the response becomes larger with increases in the average mean mtensity level of light except in the region of the saturation of the response. On the other hand, variance of the light signal input also increases proportionally to the increase in the light intensity with constant modulation of the depth. In comparison with the light signal inputs, the increase of the variance of the response is small, and the difference in variance between the input and output signals is magnified as the average mean intensity level of modulation hght is increased. The power spectra of horizontal cell responses to light showed Iowpass filter characteristics though not shown. Regarding this particular cell, the corner frequency of the power spectrum for the dimmest light was lower (5.4Hz) than those of ail other spectra, which showed no significant change in the corner frequency value (b, 6.5 Hz; c, 7.2 Hz:
A
1.0 levels of light are shown in Fig. 4(A). For the lowest average mean intensity level of light, the value of skewness is 0.047, indicating that the modulated response fluctuated symmetrically around the mean value. When the average mean intensity level of light is increased, however, the value of skewness became larger (b, 0.145; c, 0.017; d, 0.415; e, 0.510; f, 0.772). It is necessary to test these values, because the degree of freedom of each datum is different at different average mean intensity level of light. A dotted line in Fig 4(A) indicates a 50,; level of significance (see Data analysis). The test indicates that the measures of skewness at the three higher intensity levels depart from symmetry at the 57, level of significance. Kurtosis may be defined as “‘peakedness” and is 3 for the normal distribution. Kurtosis values are shown in Fig. 4(B) at different intensity levels of light (a, 3.08; b, 2.70; c, 2.97; d, 2.78; e, 2.84; f, 3.25). The test made for kurtosis indicates that none of the values deviate from the normai distribution (in terms of peakedness) at the 5% level of significance. We have cafculated the probability density functions of the responses at different intensity levels of light from I2 cells. All the data showed that the probability density function of the response departed from the normal distribution with increases in average mean intensity level of light, and the deviation was characterized in terms of symmetry rather than of peakedness. As already mentioned, the modulated light input can be regarded as the signal which has the normal distribution. Therefore, the deviation of the probability density function of the response from the
t
n” t
f3
.
0.5
0)
/!
l A
/
,,
0
*,‘A, __ -2.0
0
-1.0
Light mtensify (iogl
8
4.0
t _“-.-..
. . . . . . . .._._.............,.,,,,..,,.......
(,
.$ ,o s Y
_ _-
30
i 20
t
1
____ ,wK.z;d.:_ __ _
\ _ . . .. . ,.....-. - ___.___.
,..,,,_,,,.
/
-20
-to
__Lc
Light intensity (log) Fig. 4. Skewness (A) and kurtosis (B) of horizontal ceil responses. Cell is the same as in Figs 2 and 3. Skewness corresponds to the Third Moment in statistics and presents how far the distribution departs from symmetry. Kurtosis corresponds to the Fourth Moment and serves to differentiate a flat distribution curve from a sharply peaked one. Dotted lines indicate the 5% level of significance. Test indicates that the measures of skewness at the three higher levels departed from symmetry at the 5”,6 level of significance, and the kurtosis values did not deviate from the normal distribution (in terms of peakedness) at the 51,; level of significance (see Methods). Dashed line (B) indicates the kurtosis value for the normal distribution.
informarton processing in the honzontal d. 7.5 Hz: e. 6.3 Hz; f. 6.4 Hz). In some other celfs. however, the corner frequency shifted toward higher frequencies with increases in light intensity (see Umino er al.. 19833. E$LTIS of‘harium
ion on thr horixvmi
cell responses
It is cvell known
that. in the carp there are three types of photopic horizontal ceils classified by their spectral response patterns. i.e. L, R/G and Y, RB cells (Tomita, 1963). L cells produce hyperpolarizin~ responses to all monochromatic illuminations when the wavelength is scanned, R,‘G cells are characterized by a biphasic spectral response pattern. hyperpolarization in the short-wavelength range with depolarization in the long-wavelength range, and Y,‘RB cell has a triphasic spectral response pattern, hyperpolarization in blue, depolarization in orange and again hyperpolarization in deep red lights. In our recent study a dramatic effect of Ba’+ on the horizontal cell responses. especially on the spectral response patterns was found and was briefly reported (Watanabe and Umino. 1952). To confirm the previous results. experiments illustrated in Fig. 5 were carried out. Figure 5(A) shows an example of the spectral responses of a L cell scanned from 420 to 700 nm in JO nm sreps. After application of2 mM Ba’” the dark
cell
1133
membrane potential was gradually depolarized by about 15 mV. and the photoresponses became larger with little change in the spectral response pattern. Generally. each enhanced photoresponse in Ba”-containing medium is characterized by a slow rising phase and a large off response, the latter can be observed in Fig. 6. Application of Ba” depolarizes the dark membrane potential not only of the L cells but also of the chromatic type cells as seen in Fig. 5(B) and (C). Figure S(B) illustrates an example of the spectral responses of a R#G cell scanned from 420 to 740nm in 40 nm steps. It is noted that the depolarizing photoresponses to red light disappeared upon application of ?. mM Ba’+. In Fig. S(C) is shown an example of the effects of :! mM Ba’- on the spectral responses of a Y,‘RB cell. In this case. wavelength was scanned at the beginning from 400 to 760 nm in 20 nm steps. then alternate illuminations of 440 and 630 nm were given, and after the depolarization of the dark potential was brought about by Ba” application. the spectral response pattern was examined again. It is evident that, in Ba”-containing medium, the h~perpoiar~z~ng responses to deep red disappeared completely resulting tn a diphasic spectral response pattern. In some cases. the depolarizing responses to the middle-wavelength stimuli were also
A
B Normol
C Normol
Bo*+
contained
.-
Normol
Bo*+contained
)
Normal
Ba*+contoined
Normal
-+4-
Normal
Fig. 5. Etfrcts of Bit’- (2 mM) on photopic horizontal cells. Continuous application of Ba’- is indicated between arrows. (A) Spectral response of L cell. Scanning was made in steps of 40 nm from 420 (the left response III each scanning) to 700 nm with monochromatic lights adjusted to equal quanta (2 x IO’ photons:secipm’). Duration of light was 0.25 set at each wavelen_gth and that of followin_p intermission 3.73 sec. (B) Spectral response of R/G cell. Scanning was made III steps of 40 nm from 420 to 740 nm (7 x IO'photons/set/pm’). Duration of light was 0.25 see at each &‘aveiength and that of following intermkon of 3.75 sec. (C) Spectral response of Y;RB cell. Scanning was made in steps of 20 nm from 400 to 760 nm (2 x IO’ photons.sec!pm?). Duration of light was 0.X set at each wavelen_eth and that of following intermission was 575 set for the first and the third spectral scannings and was 7.25 set for the second one. Between the first and second spectral scannings, the retina was alternately illuminated with 440 and 630 nm lights of 3 set duration. OmV in the voltage calibration indicates the extracellular potential level. Iifumination was a circular spot 3 mm in diameter.
Normal s
BO2+
contolned
“E
WE g - -10
-’
I
mm
’
Fig. 6. Response of a L-cell to modulated light (thick bars on monitored trace) in normal and Ba’(2 mlM)-containing solutions. Stimulus light was modulated in white-noise fashion. intensity of its high level was 7 x 10-Y~cW!um’ and that of low level I x 10-vuWircmz. Stimuli &rd’ white light covering a circle of ‘4mm in diameter on the retina.
extinguished so that the resulting spectral response pattern became a monophasic hyperpolarization. Another divalent cation Sr’* was also tested, since ST’+ is known to have effects similar to Ba’+ on some retinal cells (Piccolino and Gerschenfeld, 1978, 1980; Gerschenfeld and Piccolino, 1980; Tachibana, 1981). 2 mM Sr’+ was found also to reduce the depolarizing red component of the R/G cell response, but it appeared less effective than Ba’+. When Ca” was reduced from Sr’+-containing medium, the depolarizing red component of R/G cell was consistently eliminated, though long exposure to the medium without Ca’+ extinguished the photoresponse. Since it turned out that Ba’+ depolarized the dark membrane potential of L cells and markedly fortified positive with a large their photoresponse white-noisewith experiments off-response, modulated light were carried out on L cells to study the properties of the entire system up to the L cell bathed in Ba”-containing medium. Figure 6 illustrates an example of this. In this experiment the average mean intensity of the modulated light used was 2 log unit lower than that of the brightest light in Figs 2, 3 and 4. Accordingly, the stimulus condition is comparable to (c) in
those figures, though the penetrated cell I, d!ti‘srcr,: Thus, when the modulated iipht \\as g:ien. [hi I. Eli responded with a DC component of :ltr<)ui -- 10 n?V and Ructuations around this it’vel &‘hen 2 mIi Ha’ _ was applied. the dark membrane potential &polarized gradually, and both DC and dynamic cornponents of the response to the same modulated light became much larger. It IS noted that ,i large off-response is observed when the moduiarsd light stimulus is turned off. In Fig. 7 the shift of the membrane potential and the plots of probability density functions of the responses in normal and Ba”-containmg solutions are shown. Comparing two distributions of the response amplitude, it is evident that the variance in Ba’+-containing solution is larger than that in normal solution. In addition, the amplitude distribution of the response in Ba’--containing solution appears departed from the normal distribution. Statistically, skewness values of the responses are 0.034 in normal solution and 0.522 with Ba’+ and the kurtosis values are 2.765 and 3.177, respectively. Thus, it is evident that the statistical deviation of the amplitude of response from the normal distribution arises in the symmetry rather than in the peakedness. In other cells, however, a meaningful deviation of kurtosis was also observed in Ba’--containing solution. The amplitude distribution of the modulated light can be regarded as the normal distribution Therefore. these results imply that the linearity of the system became worse in Ba’--containing solution. In other words, the input signal is not transmitted keeping its original waveform, and the resulting output response is dijtorted from the input signal. In Fig. 8(A) and (B), gain characteristics and cross-correlation functions obtained from the L cell‘s responses in Fig. 7 are illustrated. When plotting the gain characteristics, gain of the DC component in normal solution was taken as 0 dB. It is clear that application of Ba’+ increased the gain to a certain extent at low frequency region but lowered the corner A
3 Norma. . + _.--
Frequency
Membrane
potential
6~)
Fig. 7. Graphical representation of the probability density functions (distributions of amplitude) of the responses and the shift of dark membrane potential of the L cell illustrated in Fig. 6.
(Hz
I
Time
lsec)
Fig. 8. Dynamic properties of the L cell illustrated in Fig. 6. (A) Gain characteristics. Gain of DC component in normal solution was taken as a reference point (OdB). (B) Cross-correlation functions of responses (impulse responses) calculated directly from the input light signal and output L cell response. Top trace indicates the impulse response in normal medium and the middle that m Ba’--containing solution. They are superimposed for easy comparison (bottom).
Information
processm_e in the honzontal
frequency. This result indicates that the frequency of light to which the horizontal cell can respond is lowered and the amplitude of response is enhanced in Ba’- containing solution compared to those in normal solution. The cross-correlation functions between light signal and the resulting horizontal cell response (the impulse responses) were calculated in both normal and Ba’+-containing solution (Fig. SB). From two superimposed impulse responses. it is plain that the response in Ba“ -containing solution is larger and slower with longer latency as well as peak response time than that in normal solution. These effects of Ba” ion both on the spectral response curves and on the responses to white-noisemodulated light were consistently observed without failure. These results with Ba’+ will be discussed in relation to the interactions between cone photoreceptor cells and horizontal cells. DISCUSSION
Horizontal Iigh t
cell respotut’s
to ~l,llite-tloise-mo~iulnteti
The response of L horizontal cells to white-noisemodulated light fluctuated symmetrically around a mean value for dim light, and the amplitude distribution of the response showed the normal distribution. When the average mean intensity level of modulated light was increased, the fluctuation of response became larger and the amplitude distribution of the response departed from the normal distribution. As was already mentioned, the modulated light input can be regarded as a signal which has the normal distribution of amplitude. Therefore, the deviation of the amplitude distribution of the response from the normal indicates that some nonlinearity of the system from light to horizontal cell response was enhanced with increases in average mean intensity level of light. It is worth noting that the degree of nonlinearity of the system can be presented by using the amplitude distribution of the response without calculating the mean square error between the linear model response (first order kernel) and the experimental response as proposed by Naka et (11. (1975). Such a nonlinearity of response as mentioned above will be produced if the system has a logarithmic property. Therefore, one nonlinearity i? thought to be related to Wcber’s lava. If the amplitude distribution of data does not show the normal distribution, for example, in case of the modulated responses for brighter light, its nature cannot completely be described by its power spectrum alone. Therefore, higher order spectrum should be calculated. Our preliminary work on the third order spectrum (bispectrum) analysis revealed that the deviation from the normal distribution was accompanied by an increased interaction among frequency components. and in addition, such an interaction expanded toward higher frequencies with increases in light intensity.
cell
1235
Efeects of barium ion on the resporues q/‘ horizontal cells Neuron circuitry producing L and C type (R G and Y RB) horizontal cell responses was first proposed in turtle retina by Fuortes and Simon (1974). and in fish. Stell. Lightfoot and Wheeler (1975) presented a similar but a clearer model for goldfish retina Irom their histological studies. Their model IS as follows: the L cell receives signal input from red cones by sign-conserving transmission and sends signal output to ereen cones by sign-inverting transmission, and. similarly, the R G cell receives signals from green cones and sends signals to blue cones. Thus the R G cell’s depolarizing responses to longwavelength lights originate from L cells and the hyperpolarizing responses to short-wavelengths originate in green cones. The Y/RB cell receives signals only from blue cones, so that the responses of this cell to middle- and long-wavelength tights originating in the R/G cell reverse their signs again and the response to blue light remains hyperpolarizing. In carp retina. this model appears to have been substantially supported by several workers based on indirect experiments (Murakami et al., 1978; Watanabe er nl.. 1975; Shimoda, 1981: Toyoda et al., 1952). although the possibility of a direct pathway for the red component from L cell to R/G cell is also suggested (Watanabe, 1982). Our experiments with Ba”’ showed that the depolarizing red component of the spectral response of R/G cells and the hyperpolarizing deep red component of Y;RB cells were eliminated when the retina was superfused with Ba’+-containing solution. Sr’+ also indicated effects similar to Ba’- under a certain condition. From these results, it will be suggested that the action of barium or strontium ion resembles that of GABA antagonists such as picrotoxin and bicucullin on the C type horizontal cells of the carp reported by Shimoda and colleagues (1983). GABA was strongly suggested by Marc er (11. (1978) as a transmitter taken up and used by the L cell in goldfish retina. Since, houever, Ba’ ’ can remov’e the depolarizing responses to middle-wavelengths as vvell as the hyperpolarizing responses to tong-uav,elengths in some Y/RB cells. it can not be said to mimic the GABA antagonists, because the transmitter candid,ite of the R/G cell appears to be different from GABA (Marc e[ al., 1978; Toyoda and Fujimoto. 1982). Furthermore, when the spectral response pattern of R:G or Y/RB cell vvas changed by applying Ba’*, responses to deep red light still remained though small and reversing their polarity (Fig. 5A. B). This implies that, even if the sign-inverting transmission from L cell to green cones is completely btocked. there must be another pathway direct or indirect from red cones to the R/G cell than that via green cones, because the spectral sensitivity curve (Spekreijse. 1981) of the green cone is hardly thought to respond to light longer than 680 nm.
In some excitable
ceils Ba’- and Sr’ _ are known to channel as Ca’_ does and to facilitate the Ca” spikes (Reutsr. 1973: Hapiw-ara, 1974). and Ba” is also reported to block the K conductance of the cells (Werman er (11.. 1961: Sperelakis er al.. 1967). In turtle retina, Piccolino (1976) reported that Sr” was able to substitute for Ca” in the chemical transmission between cones and L cell, and Piccolino and Gerschenfeld (1980) found that Ba” as well as Sr’+ facilitated the Ca’- spikes of single cones evoked through the feedback from the L cell. According to Piccolino (1976), the postsynaptic effect of Sr’+ might cause the prolongation of the time-course of the hyperpolarizing onset in the L cell response to light. Marked sloudown of the rising phase of the L cell response by Ba” or Sr” observed in the present study might be accounted for by a similar mechanism. Tachibana (198 1) observed that in goldfish solitary horizontal cells cultured in a high K’ medium Ca’+-spikes were evoked by depolarizing current stimuli and were enhanced by Ba’+, Sr” and high Ca!+. In the present experiments, application of Ba” produced large positive off-responses in L ceils. Probably they are not regenerative responses of an all-or-none nature originating in the membrane of the L cell itself but rather are summated EPSPs of graded nature evoked by possible feedback spikes of cones like as those in turtle. and their nature is left for future investigation. Apparent blocking effects of Ba’+ and Sr’+ on a particular range of the spectral responses of C type cells are of special interest. However, such effects of these cations on chemical synapses have not been shown yet and their mechanisms are quite obscure. They remain also for future investigation. Existence of the negative feedback from L cell to red cones was first demonstrated in the turtle retina (Bayler et al., 1971) and some experimental results suggested similar feedback in the carp retina (Watanabe et al., 1978; Murakami et al., 1978; l_Jmino and Watanabe, 1981a). Lam er ~1. (1978) reported that in the catfish retina feedback from the horizontal cell to cones improved both frequency response and gain of the system at the same time. In general, however, the function of a negative feedback is known to improve the frequency response at the sacrifice of the gain. In the experiments with whitenoise-modulated light (Figs 6. 7 and S), it is evident that Ba’+ increases the static as well as dynamic components of the L cell response but produces a deviation from the normal distribution in its probability density function. In Fig. 8(A) the situation is more clearly shown, that is, upon Ba’+ application the frequency response is changed for the worse instead of the gain being improved. Thus the action of barium ion in this experiment is equivalent to disconnection of the negative feedback loop concerned. If so, this ion can be utilized as a good tool in studies on the properties of feedback from horizonpermeate
the
same
tal cell to cones. althouph
remains
the mechanism
i>!‘I!\ .iction
to be proved.
.~ci;no~~ird~emmrs--H’r wish to thank Professor :I. Ishida. Tokyo Medical and Dental Lmversity and Dr S. Kikkawa for valuable comments. 1lr 1’ btabama for help with rhc photography. and Mrs K. Koikc and 1~. Sakaur for technical assistance.
REFERENCES Baylor D. A., Fuortes M. G. F. and O’Bryan P. bl. (1971) Receptive fields of cones in the retina of the turtle. J. Physiol. 214, 265-291. Fuortes M. G. F. and Simon E. J. (1974) Interactions leading to horizontal cell responses in the turtle retina J. Phvsiol. 240, 177-198. Gerschenfeld H. M. and Piccolino IM. (1980) Sustained feedback effects of L-horizontal cells on turtle cones, Proc. R. Sot. Lend. B 206, 465-480. Hagiwara S. (1974) Ca-dependent action potential. In ,Clembranes-A Series qf’ Adcances (edited by G. Eisenman), Vol. 3. pp. 359-381. Dekker. New York. Kikkawa S., Yamaguchi T. and Suganara M. (1981) Some statistical properties of turbulence in the ascending aorta. Trans. lECE Japan MBESl-54, 5 l-58 (in Japanese). Lam D. M. K.. Lasater E. M. and Naka K. 1. (1978) ‘I-Aminobutyric acid: A neurotransmitter candidate for cone horizontal cells of the catfish retina, proc. ncltn. Acad. Sci. U.S.A. 75, 6310-6313. Marc R. E.. Stell W. K., Bok D. and Lam D. MM.K. (1973) GABA-ergic pathways in the goldfish retma. J. camp. ,Veurol. 182, 22 I-246. Murakami M., Shimoda Y. and Nakatam K. (1978) Effects of GABA on neuronal activities in the distal retina of the carp. Sensory Processes 2, 334-338. Naka K.-I., Marmarelis P. Z. and Chan R. Y. (1975) Morphological and functional identifications of catfish retinal neurons. 3. Functional identification. J. Neurophysiol. 38, 92- I3 I. Piccolino M. (1976) Strontium-calcium substitution in synaptic transmission in turtle retina. ,Valure, Lnnd. 261, 504-505. Piccolino M. and Gerschenfeld H. IM. ( 1978) Activation of a regenerative calcium conductance in turtle cones by peripheral stimulation. Proc. R. Sot. Lontl. B 201, 309-3 15. Piccolino M. and Gerschenfeld H. M. (1980) Characteristics and ionic responses induced in feedback spikes of turtle cones. Proc. R. Sac. Land. B 206, 439463. Reuter H. (1973) Divalent cations as charge carriers in excitable membranes. Prog. BiophFs. Molec. Biol. 26, l--13. Shimoda Y. (198 I ) Neural mechanisms of colour vision in the retina. J. Keio .\fed. SOL.. 58, I i-3 I (III Japanrsc). Shimoda Y., Uiyachi E.-I., Watanabe S.-I. and .Murakamt M. (1982) Effects of GABA antagonists on spectral responses of horizontal cells in the carp retina. Co/or Rcs Appl. 7, 149-151. Spekreijse H., Mooy J. E. M. and van den Berg T. .I. T. P. (I98 I) Photopigments and carp ganglion cell action spectra. Vision Res. 21, 1601-1604. Sperelakis N.. Schneider M. F. and Harris E. J. (1967) Decreased K*-conductance produced by Ba:- in frog sartorius fibers. J. gen. Physiul. 50, l565-1583. Stell W. K., Lightfoot D. 0. and Wheeler T. G. (1975) Goldfish retina: Functional polarization of conehorizontal cell dendrites and synapses. Science 190, 989.-990. Sturges H. A. (1926) The choice of class interval. J. .~uI. .-Icroc. 21. 65-66.
Information Tachtbana horizontal 321, Tomita
processmg in the horizontal
&I. (1951) Membrane properties of solitarq cells isolated from goldfish retina. J. PhFriol.
l-t-161. T. (1963)
Electrical acttvtt) J. opt. SW. .4n1. 53, 19-57.
in the carp retina. (in Japanese). Umino
m the vertebrate
retina
Toboda J. ( 1974) Frequency characteristics of retinal neurons tn the carp J. gen. PhJsml. 63. 211234. Toyoda I. and Fujimoto M. (1952) Effects of some chemicals on the interaction of second-order neurons in the retina. J. Phuiol. Sot. Japan 4, 421. Toq’oda J.. Kujiraoka T. and Fujimoto \I. (1982) The role of L-type horizontal cells in the opponent-color process. Color Ru. Appl. 7, 152-154. Trifonob Yu. A. and Byzov A. L. (1965) Response of cells-sources of S-potential of the tortoise retina-to current passed through the optic cup. Biojkika 10, 673-680 (in Russian). Umino 0. and Watanabe K. (19Sla) Receptive field characteristics of E-response in horizontal cells. ;Trcvrs. fECE Japan J&C, 4, 341-342 (in Japanese). Umino 0. and Watanabe K. (I981 b) Fundamental characteristics of E-response of photopic L-type horizontal cell
0.
1237
ceil
Sakaue
Trans. IECE Japun SlBE81-55. Y.. Kikkawa.
Analysis of horizontal m-sequence fashion.
S. and Watanabe
59-66
K. (1953)
cell response to light modulated Trans. IECE Japan J66-C,
in 1.
333-340 (in Japanese). Watanabe I;.. Katagiri Y. and Suds Y. (197s) Electricall\ evoked response (E-response) of L- and C-t)pe honzontai cells in the carp retina. Srnsor~ Procews 2. 326333. Watanabe K. and Umino 0. (1992) A double-input model of R.G type horizontal cell in the carp retina. J. Ph,rsid. Sot. Jnpm U. 3 16. Watanabe K. (1982) Electrically evoked responses (Eresponses) of the carp horizontal cells. T/w .S-p~~~rnricl[ (Edited by Drujan B. D. and Laufer M.). pp. 123-136. Alan R. Liss. New York. Wet-man R.. McCann F. V. and Grundfest H. (1961) Graded and all-or-none electrogenesis in arthropod muscle. I. The effects of alkali-earth cations on the neuromuscular system of Rott~trirrr microprc~m. J. ,qc~n.
Ph~sd
U, 979-995.
of Physiology.
Tokyo
Women’s
and K.
WAT.ASABE
Medical College. Japan
Kawada-cho.
Shinjuku-ku.
Tokyo
162.
Abstract-in
this paper. two types of experiments in the carp retina are presented. One is the analysis of responses to white-noise-modulated light at drfferent average mean intensity levels recorded from L-type horizontal cells. For weak light, the responses fluctuated symmetrically around mean values. For brighter hght. ho\\e\rr. the distribution of amplitude of the response departed from the normal distribution. The deviation from the normaI distribution was characterized in terms of symmetry rather than peakedness. The deviation also implies the existence of a nonlinearity in the system. The other type IS with barium Ion. Ba2- produced dramatic effects on the spectral response patterns of horizontal cells; the depolarizmg response IO long ua~elrngth lieht of the R G cell and the hypzrpolarizing response to deep red light of the Y RB cell were eliminated-in BaL”-containing solutions. The response of the L cell to whitr-noisrmodulated hght HIS also chanced by application of Ba?-. In Ba”-containing solution, the response became iar~er ;S well as slower than that in normal solution. The results of Ba” are discussed in relation IO the interactions
Carp retins
betwern
cone and horizontal
Horizontal ceil
B;$- iun
cells. White
INTRODUCTION
In a previous paper (Umino and Watanabe, 1981 b), considering a horizontal cell response to light as the response of an entire system composed of photoreceptors and a horizontal ceil, the system was divided into four subsystems, A, 8, C and D. The subsqstsm A relates principally to the first event of the system involving the absorption of photons, the subsystem B to the properties of the photoreceptor membrane, the subsystem C to those of synaptic transmission and the subsystem D to the horizontal cell itself. Among them. characterisiics of a system compused of the subsystems C and D were investigated by urihzing an electrically evoked response (E-response) of the horizontal cell as an impulse response of the sbstem to input from photoreceptor terminals (Umino and Watanabe, 1981b). since the E-response hits been thought to be the summated EPSP caused b) an increase in transmitter release evoked hg’ ;I tranricnt depolarization of the photorL’ccp:i): trrmlii;ii\ b> ;i rran~ctiiial electrical current pulse (Trifonov and Byzov. 1965; Watanabe ef nl., 197s). In the present paper. some properties of the response IO light of the entire system up to horizontal ceils \\erc e\;umincd by applying uhite-noisemodulated hght stimuli at dift‘erent average mean intensity levels. Probability density functions. also kno\\n as distribution of amplitudes of the response, were analyzed. and the etl-ect of barium ion on them was also examined. Although effects of barium ion on retinal cells have already been reported by several Lvorkers (Piccolino and Grrschenfeld, lY7S, 1980; I229
noise
Statistical
properties
Gerschenfeld and Piccolino, 1980; Tachibana. 1981). newly found dramatic effects of this ion on the spectral response patterns of horizontal cells will be presented and its influence to the probability density functions will be discussed. METHODS Prepmotions,
recording
cd
iighl
stinidi
Experiments were performed on the carp Cyprinrcs Recordings were obtained from the isolated retina, superfused with an artificial sahne solution. contained NaCl(iO0 mM), Normal solution NaHCO,(20 mM), KCI(2 mM), &Cl, (2 mM), MgCIz( 1 mM) and glucose (10 mM) and pH was adjusted to 7.8 by bubbling a gas mixture of 977,; O? and 3”,, CO?. Ions to be tested were added to the perfustng fluid without compensating for osmolarity. The horizontal cells were impaled with superfine microelectrodes (50-150 MQ) filled with 3 M KCI. Photic stimulation \vas provided by a t\vo-channel p~~t~)~titnul~t~r, both beam> soi to deli\e:r a spot of 4 mm diameter centered at the microelectrode tip. The unattenuated intensity of light was 7 x lO-‘~_~W/~rn’ for white light. In the experiments with white-noise stimuli, the light was modulated in m-sequence fashion (maximum-length linearly recurring sequence). According to Naka and Marmareiis (1975). the whitenoise approach nearly maximizes the rate of information gathering about the system behavior and results in a complete quantitative description. The whitenoise method employed was described in detail in a previous paper (Umino tr ul., lYS3). To obtain the carpio.
tar
Fig. I. Schematic diagram of the two-channel optical stimulator. Light source is a SOOW xenon arc lamp. Two light beams are symmetrically arranged except for a device placed in the test beam for modulating light in white-noise fashion. In each beam an electromagnetic shutter for conventional use is furnished. but it is not shown in the figure. To modulate the test light, a rotatory disk with twelve holes through which the light passes is directly driven by a stepping motor (PA-62 Nippon Pulse Motor Co.) which is controlled by the m-sequence (maximum-length linearly recurring sequence) signal. The sequence generator is a 5msec clock driven through I2 shift registers. Therefore, cycle of m-sequence employed is 20.475 set one [S msec x (2” - I)]. There were three neutral density filters (ND). ND filter I attenuates the test beam and ND filter II the background, so that the ratio of ND filter I to ND filter II determines the depth of modulation. ND tilter III serves to change the total light intensity and to determine the average mean intensity level keeping the modulation depth unchanged. The brightest modulated light (test beam + background) was composed of intensities 7 x lO-‘j~W/~rn~ for high level and I x IO-‘/rW/~m’ for low level (background light alone), and the ratio of the high level to the low one of modulated light is 7 in all experiments. The modulated light has a cutoff frequency around 90 Hz, and can be regarded as the analog signal which has the normal distribution for the retinal neuron system (see Methods). Inset shows a detailed sketch of the device for modulating the light stimuli. To improve the rise and fall times of light. a plate with an aperture stop smaller than one-tenth of the holes on the rotatory disk is placed behind the disk. Accordingly, rise and fall times are determined by the time the edge of holes on the disk passes across the aperture stop. Test light is monitored by a photomicrosensor composed of a photodiode and a detector (EE-SC Omron Tateisi Electronics Co.). (Produced by permission from Umino er 01.. 1983.)
white-noise-modulated light, a rotatory disk with I2 holes through which light passed was placed in the beam and driven by a stepping motor (PA-62, Nippon Pulse Motor Co.) vvhich was controlled by the m-sequence signal (Fig. I). The sequence generator was a 5 msec clock driven 12 shift registers. Therefore, one cycle of m-sequence employed was 20.475 set [5 msec x (2” - I)]. A neutral density filter designated ND filter I determined the intensity of test light and ND filter II placed in the second beam the intensity of background light. The ratio of ND filter I and ND filter II gave the depth of modulation, and ND filter III determined the total intensity of light under the condition of a constant modulation depth. The brightest light in the white-noise experiments was composed of intensities 7 x lO-‘~~Wi~~rn’ for high
level and I x !Om‘pU’ jlrn’
Some statistical functions were calculated to obtain a quantitative description of the response to whitenoise light. All calculations were made for the responses to the one cycle white-noise-modulated light. since the statistical properties of a part of the entire (one cycle) m-sequence is worse than those of the entire m-sequence (Umino er ul., 19S3). The probability density function describes the amplitude distribution of the response. It is useful to calculate this function to detect nonlinearity in the system, if it exists, since the deviation of the distribution of the response amplitude from the normal distribution implies its existence. The number of classes (NC) in the probability density function was calculated from the equation IV<= I + 3.3 x log ,t’ (IV: sample size) (Sturges. 1926). To describe the deviation from the normal distribution, skewness and kurtosis were calculated. Skeuness means lack of symmetry, and a measure of skewness shows how far the distribution departs from symmetry. Skewness ( fi) corresponds to the Third Moment and is defined \‘&
=/I,‘&
where N is the sample size, 11 is the mean value, x is the observed value, /lr is the Third Moment and CTis the standard deviation. For positively skewed distributions. .:/ii; :. I), for :I s~nlmi‘t~~z distribution ,_,x = 0 and for negative skewness, V/jy < 0. Kurtosis may be defined as “peakedness” and a measure of kurtosis serves to differentiate between a flat distribution curve and a sharply peaked curve. Kurtosis (IL) corresponds to the Fourth Moment and is defined
where pJ is the Fourth Moment. For a normal distribution kurtosis is equal to 3. If the distribution is more sharply peaked than the normal, kurtosis is
Informatton
processing
than 3, and if Ratter than the normal. kurtosis is less than 3. It is necessary to test these values. since the degree of freedom is different in each datum. The test adopted was reported by Kikkawa et al. (1981). The number of degrees of freedom (X-) is given by
m the horizontal
cell
vvhere T is the observation time and equivalent spectral width defined by
,
. \
I+‘, is the
where P(j) is the power spectrum. If we get the number of degrees of freedom from the data. we can calculate the level of significance in line with the traditional method (see Kikkawa et al., 1981). For skewness, a single-sided test was adopted, because Weber’s law exists in the biological system and it would be expected that the function might be distorted in one direction. For kurtosis. a two-sided test was adopted, since kurtosis might be changed in either direction. Cross-correlation was performed between the input (light) signal monitored (see Fig. I) and the resulting response. Gain characteristics were obtained by calculating Fourier transforms of the cross-correlation function
When the average mean intensity level of vvhitenoise-modulated light was increased. keeping the depth of modulation constant. both static (DC) and dynamic (fluctuating) components of the horizontal cell response became larger. Figure 2(A) shovvs an example of the Ructuating components of the responses to modulated white lights at different average mean intensity levels increased stepwise from (a) to (f), and the relation of the magnitudes of the DC component to the stimulus intensity IS illustrated in Fig. 2(B). For dim light, the response fluctuated around the mean level. For brighter light. however. upward fluctuations became larger than downward fluctuations. To describe this feature quantitatively. distributions of the amplitude (probability density function) of the responses were calculated and were plotted in Fig. 3. The average mean intensity level is increased from top to bottom. It is apparent that the probability density function of the response shows the normal distribution for the lowest average mean intensity level of light (a), but the function departed from the normal distribution with increases in the average mean intensity level. To evaluate the normality of the probability density function of the response, skewness and kurtosis were calculated (see Data analysis). Skewness means lack of symmetry and is 0 for the normal distribution. Skewness values of the responses at different intensity
(mV)
or :
L10
Light
intensity
(lag)
mV
5 set
Fig. 2. Responses of L-type horizontal cell to light modulated in white-now fashion (one cycle). The intensities7 x IO-‘pW!pm’ for high level and I x IO ktW)}lrn’ brightest light (0 log) was composed for low level. and the ratio of the high lebel to the low one was kept constant throughout the experiment. 1‘4) Dynamic (tluctuatmg) components of responses. Average mean intensity level of light is increased by 0.5 log untt steps from top (a) to bottom (f). Letters (a-f) correspond to those in Fig. Z(A) in the following tigurss. (B) R&non of the mean values of the response amplitudes to the average mean intensittos of the hght stimuli. Soltd circles indicate the mean value of responses and vertical poles standard devtatton of modulated response. Stimult were white light covet-in, 0 a circle of 4 mm tn dtametcr.
of
I
RESULTS
greater
,-
I23
-80
-70
-60
Msmbmnc
-so
potsnnol Imv)
Fig. 3. Probability density function of the responses (in other words, distribution of amplitude) of Fig. 2. Light intensity was increased from top (a) to bottom (f). Letters (a-f) correspond to those in Fig. 2. Solid triangle indicates the dark membrane potential of the cell.
normal distrtbution indxates that the nonhnexrr~ GL the system from light to honzontai cr!l r?:sp~-in% becomes stronger with increases in the average IIIZLLI: intensity level of the stimulus light. It is evident that variance ot the response becomes larger with increases in the average mean mtensity level of light except in the region of the saturation of the response. On the other hand, variance of the light signal input also increases proportionally to the increase in the light intensity with constant modulation of the depth. In comparison with the light signal inputs, the increase of the variance of the response is small, and the difference in variance between the input and output signals is magnified as the average mean intensity level of modulation hght is increased. The power spectra of horizontal cell responses to light showed Iowpass filter characteristics though not shown. Regarding this particular cell, the corner frequency of the power spectrum for the dimmest light was lower (5.4Hz) than those of ail other spectra, which showed no significant change in the corner frequency value (b, 6.5 Hz; c, 7.2 Hz:
A
1.0 levels of light are shown in Fig. 4(A). For the lowest average mean intensity level of light, the value of skewness is 0.047, indicating that the modulated response fluctuated symmetrically around the mean value. When the average mean intensity level of light is increased, however, the value of skewness became larger (b, 0.145; c, 0.017; d, 0.415; e, 0.510; f, 0.772). It is necessary to test these values, because the degree of freedom of each datum is different at different average mean intensity level of light. A dotted line in Fig 4(A) indicates a 50,; level of significance (see Data analysis). The test indicates that the measures of skewness at the three higher intensity levels depart from symmetry at the 57, level of significance. Kurtosis may be defined as “‘peakedness” and is 3 for the normal distribution. Kurtosis values are shown in Fig. 4(B) at different intensity levels of light (a, 3.08; b, 2.70; c, 2.97; d, 2.78; e, 2.84; f, 3.25). The test made for kurtosis indicates that none of the values deviate from the normai distribution (in terms of peakedness) at the 5% level of significance. We have cafculated the probability density functions of the responses at different intensity levels of light from I2 cells. All the data showed that the probability density function of the response departed from the normal distribution with increases in average mean intensity level of light, and the deviation was characterized in terms of symmetry rather than of peakedness. As already mentioned, the modulated light input can be regarded as the signal which has the normal distribution. Therefore, the deviation of the probability density function of the response from the
t
n” t
f3
.
0.5
0)
/!
l A
/
,,
0
*,‘A, __ -2.0
0
-1.0
Light mtensify (iogl
8
4.0
t _“-.-..
. . . . . . . .._._.............,.,,,,..,,.......
(,
.$ ,o s Y
_ _-
30
i 20
t
1
____ ,wK.z;d.:_ __ _
\ _ . . .. . ,.....-. - ___.___.
,..,,,_,,,.
/
-20
-to
__Lc
Light intensity (log) Fig. 4. Skewness (A) and kurtosis (B) of horizontal ceil responses. Cell is the same as in Figs 2 and 3. Skewness corresponds to the Third Moment in statistics and presents how far the distribution departs from symmetry. Kurtosis corresponds to the Fourth Moment and serves to differentiate a flat distribution curve from a sharply peaked one. Dotted lines indicate the 5% level of significance. Test indicates that the measures of skewness at the three higher levels departed from symmetry at the 5”,6 level of significance, and the kurtosis values did not deviate from the normal distribution (in terms of peakedness) at the 51,; level of significance (see Methods). Dashed line (B) indicates the kurtosis value for the normal distribution.
informarton processing in the honzontal d. 7.5 Hz: e. 6.3 Hz; f. 6.4 Hz). In some other celfs. however, the corner frequency shifted toward higher frequencies with increases in light intensity (see Umino er al.. 19833. E$LTIS of‘harium
ion on thr horixvmi
cell responses
It is cvell known
that. in the carp there are three types of photopic horizontal ceils classified by their spectral response patterns. i.e. L, R/G and Y, RB cells (Tomita, 1963). L cells produce hyperpolarizin~ responses to all monochromatic illuminations when the wavelength is scanned, R,‘G cells are characterized by a biphasic spectral response pattern. hyperpolarization in the short-wavelength range with depolarization in the long-wavelength range, and Y,‘RB cell has a triphasic spectral response pattern, hyperpolarization in blue, depolarization in orange and again hyperpolarization in deep red lights. In our recent study a dramatic effect of Ba’+ on the horizontal cell responses. especially on the spectral response patterns was found and was briefly reported (Watanabe and Umino. 1952). To confirm the previous results. experiments illustrated in Fig. 5 were carried out. Figure 5(A) shows an example of the spectral responses of a L cell scanned from 420 to 700 nm in JO nm sreps. After application of2 mM Ba’” the dark
cell
1133
membrane potential was gradually depolarized by about 15 mV. and the photoresponses became larger with little change in the spectral response pattern. Generally. each enhanced photoresponse in Ba”-containing medium is characterized by a slow rising phase and a large off response, the latter can be observed in Fig. 6. Application of Ba” depolarizes the dark membrane potential not only of the L cells but also of the chromatic type cells as seen in Fig. 5(B) and (C). Figure S(B) illustrates an example of the spectral responses of a R#G cell scanned from 420 to 740nm in 40 nm steps. It is noted that the depolarizing photoresponses to red light disappeared upon application of ?. mM Ba’+. In Fig. S(C) is shown an example of the effects of :! mM Ba’- on the spectral responses of a Y,‘RB cell. In this case. wavelength was scanned at the beginning from 400 to 760 nm in 20 nm steps. then alternate illuminations of 440 and 630 nm were given, and after the depolarization of the dark potential was brought about by Ba” application. the spectral response pattern was examined again. It is evident that, in Ba”-containing medium, the h~perpoiar~z~ng responses to deep red disappeared completely resulting tn a diphasic spectral response pattern. In some cases. the depolarizing responses to the middle-wavelength stimuli were also
A
B Normol
C Normol
Bo*+
contained
.-
Normol
Bo*+contained
)
Normal
Ba*+contoined
Normal
-+4-
Normal
Fig. 5. Etfrcts of Bit’- (2 mM) on photopic horizontal cells. Continuous application of Ba’- is indicated between arrows. (A) Spectral response of L cell. Scanning was made in steps of 40 nm from 420 (the left response III each scanning) to 700 nm with monochromatic lights adjusted to equal quanta (2 x IO’ photons:secipm’). Duration of light was 0.25 set at each wavelen_gth and that of followin_p intermission 3.73 sec. (B) Spectral response of R/G cell. Scanning was made III steps of 40 nm from 420 to 740 nm (7 x IO'photons/set/pm’). Duration of light was 0.25 see at each &‘aveiength and that of following intermkon of 3.75 sec. (C) Spectral response of Y;RB cell. Scanning was made in steps of 20 nm from 400 to 760 nm (2 x IO’ photons.sec!pm?). Duration of light was 0.X set at each wavelen_eth and that of following intermission was 575 set for the first and the third spectral scannings and was 7.25 set for the second one. Between the first and second spectral scannings, the retina was alternately illuminated with 440 and 630 nm lights of 3 set duration. OmV in the voltage calibration indicates the extracellular potential level. Iifumination was a circular spot 3 mm in diameter.
Normal s
BO2+
contolned
“E
WE g - -10
-’
I
mm
’
Fig. 6. Response of a L-cell to modulated light (thick bars on monitored trace) in normal and Ba’(2 mlM)-containing solutions. Stimulus light was modulated in white-noise fashion. intensity of its high level was 7 x 10-Y~cW!um’ and that of low level I x 10-vuWircmz. Stimuli &rd’ white light covering a circle of ‘4mm in diameter on the retina.
extinguished so that the resulting spectral response pattern became a monophasic hyperpolarization. Another divalent cation Sr’* was also tested, since ST’+ is known to have effects similar to Ba’+ on some retinal cells (Piccolino and Gerschenfeld, 1978, 1980; Gerschenfeld and Piccolino, 1980; Tachibana, 1981). 2 mM Sr’+ was found also to reduce the depolarizing red component of the R/G cell response, but it appeared less effective than Ba’+. When Ca” was reduced from Sr’+-containing medium, the depolarizing red component of R/G cell was consistently eliminated, though long exposure to the medium without Ca’+ extinguished the photoresponse. Since it turned out that Ba’+ depolarized the dark membrane potential of L cells and markedly fortified positive with a large their photoresponse white-noisewith experiments off-response, modulated light were carried out on L cells to study the properties of the entire system up to the L cell bathed in Ba”-containing medium. Figure 6 illustrates an example of this. In this experiment the average mean intensity of the modulated light used was 2 log unit lower than that of the brightest light in Figs 2, 3 and 4. Accordingly, the stimulus condition is comparable to (c) in
those figures, though the penetrated cell I, d!ti‘srcr,: Thus, when the modulated iipht \\as g:ien. [hi I. Eli responded with a DC component of :ltr<)ui -- 10 n?V and Ructuations around this it’vel &‘hen 2 mIi Ha’ _ was applied. the dark membrane potential &polarized gradually, and both DC and dynamic cornponents of the response to the same modulated light became much larger. It IS noted that ,i large off-response is observed when the moduiarsd light stimulus is turned off. In Fig. 7 the shift of the membrane potential and the plots of probability density functions of the responses in normal and Ba”-containmg solutions are shown. Comparing two distributions of the response amplitude, it is evident that the variance in Ba’+-containing solution is larger than that in normal solution. In addition, the amplitude distribution of the response in Ba’--containing solution appears departed from the normal distribution. Statistically, skewness values of the responses are 0.034 in normal solution and 0.522 with Ba’+ and the kurtosis values are 2.765 and 3.177, respectively. Thus, it is evident that the statistical deviation of the amplitude of response from the normal distribution arises in the symmetry rather than in the peakedness. In other cells, however, a meaningful deviation of kurtosis was also observed in Ba’--containing solution. The amplitude distribution of the modulated light can be regarded as the normal distribution Therefore. these results imply that the linearity of the system became worse in Ba’--containing solution. In other words, the input signal is not transmitted keeping its original waveform, and the resulting output response is dijtorted from the input signal. In Fig. 8(A) and (B), gain characteristics and cross-correlation functions obtained from the L cell‘s responses in Fig. 7 are illustrated. When plotting the gain characteristics, gain of the DC component in normal solution was taken as 0 dB. It is clear that application of Ba’+ increased the gain to a certain extent at low frequency region but lowered the corner A
3 Norma. . + _.--
Frequency
Membrane
potential
6~)
Fig. 7. Graphical representation of the probability density functions (distributions of amplitude) of the responses and the shift of dark membrane potential of the L cell illustrated in Fig. 6.
(Hz
I
Time
lsec)
Fig. 8. Dynamic properties of the L cell illustrated in Fig. 6. (A) Gain characteristics. Gain of DC component in normal solution was taken as a reference point (OdB). (B) Cross-correlation functions of responses (impulse responses) calculated directly from the input light signal and output L cell response. Top trace indicates the impulse response in normal medium and the middle that m Ba’--containing solution. They are superimposed for easy comparison (bottom).
Information
processm_e in the honzontal
frequency. This result indicates that the frequency of light to which the horizontal cell can respond is lowered and the amplitude of response is enhanced in Ba’- containing solution compared to those in normal solution. The cross-correlation functions between light signal and the resulting horizontal cell response (the impulse responses) were calculated in both normal and Ba’+-containing solution (Fig. SB). From two superimposed impulse responses. it is plain that the response in Ba“ -containing solution is larger and slower with longer latency as well as peak response time than that in normal solution. These effects of Ba” ion both on the spectral response curves and on the responses to white-noisemodulated light were consistently observed without failure. These results with Ba’+ will be discussed in relation to the interactions between cone photoreceptor cells and horizontal cells. DISCUSSION
Horizontal Iigh t
cell respotut’s
to ~l,llite-tloise-mo~iulnteti
The response of L horizontal cells to white-noisemodulated light fluctuated symmetrically around a mean value for dim light, and the amplitude distribution of the response showed the normal distribution. When the average mean intensity level of modulated light was increased, the fluctuation of response became larger and the amplitude distribution of the response departed from the normal distribution. As was already mentioned, the modulated light input can be regarded as a signal which has the normal distribution of amplitude. Therefore, the deviation of the amplitude distribution of the response from the normal indicates that some nonlinearity of the system from light to horizontal cell response was enhanced with increases in average mean intensity level of light. It is worth noting that the degree of nonlinearity of the system can be presented by using the amplitude distribution of the response without calculating the mean square error between the linear model response (first order kernel) and the experimental response as proposed by Naka et (11. (1975). Such a nonlinearity of response as mentioned above will be produced if the system has a logarithmic property. Therefore, one nonlinearity i? thought to be related to Wcber’s lava. If the amplitude distribution of data does not show the normal distribution, for example, in case of the modulated responses for brighter light, its nature cannot completely be described by its power spectrum alone. Therefore, higher order spectrum should be calculated. Our preliminary work on the third order spectrum (bispectrum) analysis revealed that the deviation from the normal distribution was accompanied by an increased interaction among frequency components. and in addition, such an interaction expanded toward higher frequencies with increases in light intensity.
cell
1235
Efeects of barium ion on the resporues q/‘ horizontal cells Neuron circuitry producing L and C type (R G and Y RB) horizontal cell responses was first proposed in turtle retina by Fuortes and Simon (1974). and in fish. Stell. Lightfoot and Wheeler (1975) presented a similar but a clearer model for goldfish retina Irom their histological studies. Their model IS as follows: the L cell receives signal input from red cones by sign-conserving transmission and sends signal output to ereen cones by sign-inverting transmission, and. similarly, the R G cell receives signals from green cones and sends signals to blue cones. Thus the R G cell’s depolarizing responses to longwavelength lights originate from L cells and the hyperpolarizing responses to short-wavelengths originate in green cones. The Y/RB cell receives signals only from blue cones, so that the responses of this cell to middle- and long-wavelength tights originating in the R/G cell reverse their signs again and the response to blue light remains hyperpolarizing. In carp retina. this model appears to have been substantially supported by several workers based on indirect experiments (Murakami et al., 1978; Watanabe er nl.. 1975; Shimoda, 1981: Toyoda et al., 1952). although the possibility of a direct pathway for the red component from L cell to R/G cell is also suggested (Watanabe, 1982). Our experiments with Ba”’ showed that the depolarizing red component of the spectral response of R/G cells and the hyperpolarizing deep red component of Y;RB cells were eliminated when the retina was superfused with Ba’+-containing solution. Sr’+ also indicated effects similar to Ba’- under a certain condition. From these results, it will be suggested that the action of barium or strontium ion resembles that of GABA antagonists such as picrotoxin and bicucullin on the C type horizontal cells of the carp reported by Shimoda and colleagues (1983). GABA was strongly suggested by Marc er (11. (1978) as a transmitter taken up and used by the L cell in goldfish retina. Since, houever, Ba’ ’ can remov’e the depolarizing responses to middle-wavelengths as vvell as the hyperpolarizing responses to tong-uav,elengths in some Y/RB cells. it can not be said to mimic the GABA antagonists, because the transmitter candid,ite of the R/G cell appears to be different from GABA (Marc e[ al., 1978; Toyoda and Fujimoto. 1982). Furthermore, when the spectral response pattern of R:G or Y/RB cell vvas changed by applying Ba’*, responses to deep red light still remained though small and reversing their polarity (Fig. 5A. B). This implies that, even if the sign-inverting transmission from L cell to green cones is completely btocked. there must be another pathway direct or indirect from red cones to the R/G cell than that via green cones, because the spectral sensitivity curve (Spekreijse. 1981) of the green cone is hardly thought to respond to light longer than 680 nm.
In some excitable
ceils Ba’- and Sr’ _ are known to channel as Ca’_ does and to facilitate the Ca” spikes (Reutsr. 1973: Hapiw-ara, 1974). and Ba” is also reported to block the K conductance of the cells (Werman er (11.. 1961: Sperelakis er al.. 1967). In turtle retina, Piccolino (1976) reported that Sr” was able to substitute for Ca” in the chemical transmission between cones and L cell, and Piccolino and Gerschenfeld (1980) found that Ba” as well as Sr’+ facilitated the Ca’- spikes of single cones evoked through the feedback from the L cell. According to Piccolino (1976), the postsynaptic effect of Sr’+ might cause the prolongation of the time-course of the hyperpolarizing onset in the L cell response to light. Marked sloudown of the rising phase of the L cell response by Ba” or Sr” observed in the present study might be accounted for by a similar mechanism. Tachibana (198 1) observed that in goldfish solitary horizontal cells cultured in a high K’ medium Ca’+-spikes were evoked by depolarizing current stimuli and were enhanced by Ba’+, Sr” and high Ca!+. In the present experiments, application of Ba” produced large positive off-responses in L ceils. Probably they are not regenerative responses of an all-or-none nature originating in the membrane of the L cell itself but rather are summated EPSPs of graded nature evoked by possible feedback spikes of cones like as those in turtle. and their nature is left for future investigation. Apparent blocking effects of Ba’+ and Sr’+ on a particular range of the spectral responses of C type cells are of special interest. However, such effects of these cations on chemical synapses have not been shown yet and their mechanisms are quite obscure. They remain also for future investigation. Existence of the negative feedback from L cell to red cones was first demonstrated in the turtle retina (Bayler et al., 1971) and some experimental results suggested similar feedback in the carp retina (Watanabe et al., 1978; Murakami et al., 1978; l_Jmino and Watanabe, 1981a). Lam er ~1. (1978) reported that in the catfish retina feedback from the horizontal cell to cones improved both frequency response and gain of the system at the same time. In general, however, the function of a negative feedback is known to improve the frequency response at the sacrifice of the gain. In the experiments with whitenoise-modulated light (Figs 6. 7 and S), it is evident that Ba’+ increases the static as well as dynamic components of the L cell response but produces a deviation from the normal distribution in its probability density function. In Fig. 8(A) the situation is more clearly shown, that is, upon Ba’+ application the frequency response is changed for the worse instead of the gain being improved. Thus the action of barium ion in this experiment is equivalent to disconnection of the negative feedback loop concerned. If so, this ion can be utilized as a good tool in studies on the properties of feedback from horizonpermeate
the
same
tal cell to cones. althouph
remains
the mechanism
i>!‘I!\ .iction
to be proved.
.~ci;no~~ird~emmrs--H’r wish to thank Professor :I. Ishida. Tokyo Medical and Dental Lmversity and Dr S. Kikkawa for valuable comments. 1lr 1’ btabama for help with rhc photography. and Mrs K. Koikc and 1~. Sakaur for technical assistance.
REFERENCES Baylor D. A., Fuortes M. G. F. and O’Bryan P. bl. (1971) Receptive fields of cones in the retina of the turtle. J. Physiol. 214, 265-291. Fuortes M. G. F. and Simon E. J. (1974) Interactions leading to horizontal cell responses in the turtle retina J. Phvsiol. 240, 177-198. Gerschenfeld H. M. and Piccolino IM. (1980) Sustained feedback effects of L-horizontal cells on turtle cones, Proc. R. Sot. Lend. B 206, 465-480. Hagiwara S. (1974) Ca-dependent action potential. In ,Clembranes-A Series qf’ Adcances (edited by G. Eisenman), Vol. 3. pp. 359-381. Dekker. New York. Kikkawa S., Yamaguchi T. and Suganara M. (1981) Some statistical properties of turbulence in the ascending aorta. Trans. lECE Japan MBESl-54, 5 l-58 (in Japanese). Lam D. M. K.. Lasater E. M. and Naka K. 1. (1978) ‘I-Aminobutyric acid: A neurotransmitter candidate for cone horizontal cells of the catfish retina, proc. ncltn. Acad. Sci. U.S.A. 75, 6310-6313. Marc R. E.. Stell W. K., Bok D. and Lam D. MM.K. (1973) GABA-ergic pathways in the goldfish retma. J. camp. ,Veurol. 182, 22 I-246. Murakami M., Shimoda Y. and Nakatam K. (1978) Effects of GABA on neuronal activities in the distal retina of the carp. Sensory Processes 2, 334-338. Naka K.-I., Marmarelis P. Z. and Chan R. Y. (1975) Morphological and functional identifications of catfish retinal neurons. 3. Functional identification. J. Neurophysiol. 38, 92- I3 I. Piccolino M. (1976) Strontium-calcium substitution in synaptic transmission in turtle retina. ,Valure, Lnnd. 261, 504-505. Piccolino M. and Gerschenfeld H. IM. ( 1978) Activation of a regenerative calcium conductance in turtle cones by peripheral stimulation. Proc. R. Sot. Lontl. B 201, 309-3 15. Piccolino M. and Gerschenfeld H. M. (1980) Characteristics and ionic responses induced in feedback spikes of turtle cones. Proc. R. Sac. Land. B 206, 439463. Reuter H. (1973) Divalent cations as charge carriers in excitable membranes. Prog. BiophFs. Molec. Biol. 26, l--13. Shimoda Y. (198 I ) Neural mechanisms of colour vision in the retina. J. Keio .\fed. SOL.. 58, I i-3 I (III Japanrsc). Shimoda Y., Uiyachi E.-I., Watanabe S.-I. and .Murakamt M. (1982) Effects of GABA antagonists on spectral responses of horizontal cells in the carp retina. Co/or Rcs Appl. 7, 149-151. Spekreijse H., Mooy J. E. M. and van den Berg T. .I. T. P. (I98 I) Photopigments and carp ganglion cell action spectra. Vision Res. 21, 1601-1604. Sperelakis N.. Schneider M. F. and Harris E. J. (1967) Decreased K*-conductance produced by Ba:- in frog sartorius fibers. J. gen. Physiul. 50, l565-1583. Stell W. K., Lightfoot D. 0. and Wheeler T. G. (1975) Goldfish retina: Functional polarization of conehorizontal cell dendrites and synapses. Science 190, 989.-990. Sturges H. A. (1926) The choice of class interval. J. .~uI. .-Icroc. 21. 65-66.
Information Tachtbana horizontal 321, Tomita
processmg in the horizontal
&I. (1951) Membrane properties of solitarq cells isolated from goldfish retina. J. PhFriol.
l-t-161. T. (1963)
Electrical acttvtt) J. opt. SW. .4n1. 53, 19-57.
in the carp retina. (in Japanese). Umino
m the vertebrate
retina
Toboda J. ( 1974) Frequency characteristics of retinal neurons tn the carp J. gen. PhJsml. 63. 211234. Toyoda I. and Fujimoto M. (1952) Effects of some chemicals on the interaction of second-order neurons in the retina. J. Phuiol. Sot. Japan 4, 421. Toq’oda J.. Kujiraoka T. and Fujimoto \I. (1982) The role of L-type horizontal cells in the opponent-color process. Color Ru. Appl. 7, 152-154. Trifonob Yu. A. and Byzov A. L. (1965) Response of cells-sources of S-potential of the tortoise retina-to current passed through the optic cup. Biojkika 10, 673-680 (in Russian). Umino 0. and Watanabe K. (19Sla) Receptive field characteristics of E-response in horizontal cells. ;Trcvrs. fECE Japan J&C, 4, 341-342 (in Japanese). Umino 0. and Watanabe K. (I981 b) Fundamental characteristics of E-response of photopic L-type horizontal cell
0.
1237
ceil
Sakaue
Trans. IECE Japun SlBE81-55. Y.. Kikkawa.
Analysis of horizontal m-sequence fashion.
S. and Watanabe
59-66
K. (1953)
cell response to light modulated Trans. IECE Japan J66-C,
in 1.
333-340 (in Japanese). Watanabe I;.. Katagiri Y. and Suds Y. (197s) Electricall\ evoked response (E-response) of L- and C-t)pe honzontai cells in the carp retina. Srnsor~ Procews 2. 326333. Watanabe K. and Umino 0. (1992) A double-input model of R.G type horizontal cell in the carp retina. J. Ph,rsid. Sot. Jnpm U. 3 16. Watanabe K. (1982) Electrically evoked responses (Eresponses) of the carp horizontal cells. T/w .S-p~~~rnricl[ (Edited by Drujan B. D. and Laufer M.). pp. 123-136. Alan R. Liss. New York. Wet-man R.. McCann F. V. and Grundfest H. (1961) Graded and all-or-none electrogenesis in arthropod muscle. I. The effects of alkali-earth cations on the neuromuscular system of Rott~trirrr microprc~m. J. ,qc~n.
Ph~sd
U, 979-995.