Dopaminergic modulation of photopic temporal transfer properties in goldfish retina investigated with the ERG

Vision Research 44 (2004) 2067–2081 www.elsevier.com/locate/visres

Dopaminergic modulation of photopic temporal transfer properties in goldfish retina investigated with the ERG Carlos Mora-Ferrer *, Konstantin Behrend Inst. Zoologie, Abt. III, J. Gutenberg Universit€at, Colonel Kleinmann Weg 2, SB II 55099 Mainz, Germany Received 4 July 2003; received in revised form 23 October 2003

Abstract The influence of dopamine (DA) through either D1- or D2-dopamine receptors (D1-/D2-R) onto temporal transfer properties of the retina has been investigated using the ERG. Single flash responses and flicker responses were measured in the vitreous under photopic illumination conditions after application of either D1-/D2-R agonists or antagonists. All DA-R drugs did change the single flash responses, but only blockade of D2-R or activation of D1-R also changed the temporal transfer properties. In the Bode plot the gain characteristic was changed and thereby the upper limit frequency reduced. The action of DA is discussed on the base of a membrane resonance model in the outer retina versus a feed-forward inhibition model in the inner retina. Ó 2004 Elsevier Ltd. All rights reserved.

1. Introduction Dopamine (DA) is a prominent retinal neuromodulator. It is involved in light-adaptive mechanisms such as the retinomotor movement of photoreceptors and pigment granules, control of synaptic plasticity, and response properties of retinal cells (Dearry & Burnside, 1989; Hedden & Dowling, 1978; Kolbinger, Kohler, Oetting, & Weiler, 1990; Negishi, Teranishi, & Kato, 1990; Vaquero, Pignatelli, Partida, & Ishida, 2001; Wagner, Behrens, Zaunreiter, & Douglas, 1992; Weiler, Kohler, Kirsch, & Wagner, 1988; Weiler et al., 1988; Yazulla, Studholme, Fan, & Mora-Ferrer, 2001). The DA induced changes in retinal processing can be measured in a quantitative manner in behavioral experiments. Blocking either D1- or D2-R induced changes in visual abilities such as perceived brightness, wavelength discrimination, motion perception and temporal resolution (Lin & Yazulla, 1994; Mora-Ferrer & Gangluff, 2000, 2002; Mora-Ferrer & Neumeyer, 1996). Blockade of retinal D2-R results in a 25% loss in flicker fusion frequency (FFF), elimination of DA lowers the FFF by

* Corresponding author. Tel.: +49-6131-392-4483; fax: +49-6131392-5443. E-mail address: [email protected] (C. Mora-Ferrer).

0042-6989/$ - see front matter Ó 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.visres.2003.11.028

about 50% while blockade of retinal D1-R had no effect on the FFF (Mora-Ferrer & Gangluff, 2002). The D1-R distribution in goldfish retina is well investigated (Mora-Ferrer, Yazulla, Studholme, & Haak-Frendscho, 1999). D2-R have been demonstrated on photoreceptors, on dopaminergic interplexiform cells (DA-IPC), and in both plexiform layers in all animals investigated (Rohrer & Stell, 1995; Wagner, Luo, Ariano, Sibley, & Stell, 1993; Wang, Harsanyi, & Mangel, 1997; Yazulla & Lin, 1995). Although D2-R have been reported in the inner plexiform layer it is unclear which cells express the D2-R. Some amacrine cells (Wagner & Behrens, 1993), and additionally M€ uller cells in amphibia (Muresan & Besharse, 1993), seem to express D2- or D2-like receptors. D2-R at DA-IPC seem to act as autoreceptors and thus to control DA release (Wang et al., 1997), D2-R on photoreceptors have been demonstrated to act on calcium currents, voltage-activated currents, and rod-cone coupling (Akopian & Witkovsky, 1996; Fan & Yazulla, 2002; Krizaj, Gabriel, Owen, & Witkovsky, 1998; Stella & Thoreson, 2000). However, the majority of experiments were performed in amphibia under scotopic or mesopic illumination conditions. This is also true for studies using the ERG. The aim of this study is to link the results from behavioral experiments on temporal resolution, performed under photopic illumination conditions using dopaminergic drugs (Mora-Ferrer & Gangluff, 2002),

2068

C. Mora-Ferrer, K. Behrend / Vision Research 44 (2004) 2067–2081

with electrophysiological data. To investigate the systems properties in a manner comparable to the behavioral experiments the ERG of the light-adapted goldfish eye was recorded and the behavioral experimental regimen was mimicked using both D1- and D2-R agonists and antagonists.

stimulus presentations, and stored on a personal computer (Macintosh 7600/132, scope Software, ADI). The reference electrode was placed into the contralateral nasal opening. The fish was grounded by a platinum wire placed with the respiration tubing in the month. 2.4. Data collection and stimulating procedure

2. Material and methods 2.1. Animals Adult goldfish (6–9 cm body length) were used in the experiments. The fish were obtained from a local fish dealer and kept in 100 l tanks at room temperature under a 12/12 h light/dark regime. 2.2. Surgery procedure Fish were anaesthetized with MS 222 (tricaine methanesulfonate, Sigma) and immobilized with Flaxedil (2% solution in fish ringer, 20–30 ll, Sigma) given intramuscularly. The fish were then placed in a moist rubber foam fish holder and the gills continuously perfused via a mouth tubing with an aerated water containing 0.005% MS 222 to maintain a slight anesthesia. All parts outside the fish holder, except the eye used for recording, were covered with paper tissue which, together with the surface of the eye, was constantly rinsed with water. The eye was then pierced with a syringe needle (Braun/Melsungen Gr. 20, Germany) at two positions: at the dorsal globe at about 1 o’clock, and at the margin of the scleral surface at about 9:30. Through the former a Hamilton syringe was inserted for drug administration, the latter to insert a glass micropipette for ERG recording. 2.3. Recording techniques With a Zeitz DM2 Universal Puller (Zeitz, Germany) two identical pipettes (Clark, GC 150 TF, UK) with a very long taper (
10 mm) and a tip diameter of 6–10 lm were pulled and filled with 3 M potassium chloride. An Ag–AgCl wire was inserted and the pipette sealed with a wax colophonium mixture. The electrode was mounted on a plastic holder and guided into the eye with a mechanical microdrive (Merzh€ auser, Germany) and a stepping motor guidance system (MS 21l 5, Lang Elektronik, Germany) allowing for 0.1 lm steps. The microdrive system allows to place the electrode at fixed positions in the eye. The signal was amplified by a differential amplifier (AM Systems 1700 Differential Amplifier or shop-build (University of Bonn)). The signal was band-passed (between 0.01 and 200 Hz), digitized (2 kHz sampling rate, MacLab/4s, AD Instruments), averaged over 16

The light stimulus was generated with a light emitting diode (Type WU-7-750 SWC, Wustlich, Opto Elektronik, Germany) in combination with a yellow filter (medium yellow no. 10, ROSCO, Germany). This resulted in a spectrum shown in the Appendix A (Fig. 1(A)). The LED was switched with the D/A output of the MacLab/4s, the time course of the stimulus recorded with a photo diode just besides the eye. Continuos background (for spectrum see Appendix A, Fig. 1(B)) light was used in order to keep the animal light adapted. The background light intensity was about 90 cd/m2 and stimulus intensity was about 2200 cd/m2 . All light intensities were measured at the plane of the eye using the probe of a photometer (Gossens Mavolux, Germany). Prior to a measurement, the Hamilton syringe was positioned in the vitreous body. Stimulating with 100 ms flashes, the recoding electrode was driven into the vitreous body using 100 lm steps until the ERG signal changed its time course, signaling that it touched the ganglion cell layer. Then the electrode was retraced to the position immediately before the ERG signal changed, ensuring a kind of ‘‘standardized’’ recording position. Two different stimulus and recording protocols were used in the experiments: (a) recording of ERG signals from the vitreous to single flashes of either 5 or 100 ms duration; (b) recording of a frequency response curve starting at 2 Hz and ending at 50 Hz. These two protocols were run before and after injection of any drug. Time course of drug action was controlled by single light flashes (100 ms) every 3–10 min. 2.5. Drugs Both D1- and D2-R agonists (SKF 38393 and Quinpirole) and antagonists (SCH 23390 and Sulpiride) were used (all drugs and chemicals were obtained from Sigma, Germany). Concentrations for drugs were, whenever possible, similar to drug concentrations used in behavioral experiments (in lM SCH 23390 (10), SKF 38393 (333), and Sulpiride (10) Lin & Yazulla, 1994; Mora-Ferrer & Gangluff, 2000, 2002; Mora-Ferrer & Neumeyer, 1996). The concentration of Quinpirole (1– 10 lm) was taken from Harsanyi and Mangel and Rashid and colleagues (Harsanyi & Mangel, 1992; Rashid, Baldridge, & Ball, 1993). To isolate the photoreceptor signal from the ERG, an injection of 3.55 M

C. Mora-Ferrer, K. Behrend / Vision Research 44 (2004) 2067–2081

2069

Fig. 1. ERG of untreated eye. (A) 100 ms stimulus; (B) 5 ms stimulus; (C)–(G) the stimulus is flickered with 2, 12, 22, 36 and 40 Hz respectively; H: Intensity-response function. In (A) the ERG-components are labeled: a: a-wave; b: b-wave; dc: dc-component and OFF: OFF-response. In (H) the intensity of the light flash (100 ms) was varied in the 9–33,000 cd/m2 range. The amplitudes of the a-wave, dc-component, OFF-component and for peak-peak are plotted as a function of log intensity. The experimentally used light intensity (about 2200 cd/m2 , black dot on X -axis) is within the light intensity range of linear increase. (I) and (J) Bode plot of the data obtained with flicker stimuli. Note the slight improvement of gain (I; n ¼ 17) towards the upper limit frequency (around 25–30 Hz) and the recovery of phase (J; n ¼ 8) at the same region pointing to a different mechanism than a pure low-pass. Data are given as means with the standard error of the mean. Note: because of the log scale, 0.01 is equivalent to zero gain.

2070

C. Mora-Ferrer, K. Behrend / Vision Research 44 (2004) 2067–2081

Aspartate was made (DeMarco & Powers, 1989). All drugs were dissolved in 5 mM HEPES buffer which has been demonstrated to have no effect on visual abilities as demonstrated in behavioral experiments nor on the temporal processing properties (this report). The effect of any of the used DA agonist or antagonist was similar whether administered in dark- or lightadapted animals (data not shown). The number of animals used for the different drugs is given in the figure legends (e.g. n ¼ 3).

3. Results Fig. 1(A) shows an ERG record from the vitreous body to a 100 ms stimulus which represents the standard for all subsequent experiments. It shows the typical components, i.e. a small negative going a-wave, followed by a positive going b-wave, the ON-bipolar cell response, the so called dc-component (Brown, 1968), and the d-wave as the OFF-bipolar cell response component (Dowling, 1987). When the stimulus is only a short pulse of 5 ms duration (Fig. 1(B)), the b-wave and OFFresponse become one indistinguishable wave following the a-wave, which seems not much altered in its time course and amplitude. Lengthening the stimulus duration to 250 ms (Fig. 1(C)) only stretches the dc-component, a- and b-wave staying unaffected. According to a study on primate ERG (Sieving, Murayama, & Naarendorp, 1994), the dc-component is the result of the interaction of the sustained part of the ON- and OFFbipolar cell responses. Although there is some variability in the 100 ms flash ERG-responses of untreated animals (e.g. Figs. 1(A), 2(A) etc.), the time course does not vary and differences are only in amplitude and amount of oscillations. When the stimulus was flickered with increasing frequency, the components of the ERG more and more disappeared and a more or less sinusoidal response of decreasing amplitude appeared, until at the fusion point no amplitude modulation was seen in the response (Fig. 1(C)–(G)). The intensity-response function of the ERG was also measured (Fig. 1(H)). Background light intensity was set to the intensity used throughout the ERG-recordings and the light-intensity of the light flash varied between 9 and 33,000 cd/m2 . ERG-recordings were obtained using a 100 ms light flash and the amplitude of the ERGcomponents determined. All ERG-components increased in amplitude with increasing light intensity. The largest increase was observed for the OFF-component and respectively for the peak-to-peak amplitude. All investigated components exhibited a linear increase in amplitude with the logarithm of intensity in the 700– 10,000 cd/m2 range. The light intensity used in the ERGrecordings presented here was within this range.

To investigate the dynamics of the system, the response amplitude was measured, scaled to its maximum which yields the gain, and plotted vs. the stimulus frequency on a double logarithmic scale (Fig. 1(I)). Above 2 Hz, peak-to-peak amplitude was taken after the response showed steady-state following to the stimulus to avoid the transients at stimulus onset. Similarly, the phase of response relative to stimulus onset was plotted vs. stimulus frequency on a semi-logarithmic scale (Fig. 1(J)). This is known in linear systems theory as the so called Bode plot. To mimic most closely the behavioral experimental situation, a square-pulse stimulus instead of a sine-wave modulation of the intensity was used. This might introduce an error when using the response amplitude directly for the gain, since higher harmonics contained in the stimulus wave form could have influenced it. This was estimated by Fourier transforming both stimulus and response and checking in the response the contribution of higher harmonics due to the stimulus higher harmonics. It turns out that only up to 8 Hz a considerable part in the response (about 40% at 2 Hz to about 20% at 6 Hz) contributed to the response amplitude. That means that at the very low frequency end the gain was slightly overestimated. Since the interesting properties of the gain plot appear at the high frequency end and since these properties would rather be more emphasized when filtering the response for the higher stimulus harmonics we have neglected this error. Fig. 1(I) shows the Bode plot prior to injection of a drug. The gain plot runs almost flat with a slight recovery towards the upper limit frequency which is about 25–30 Hz. The cut-off is rather steep having a slope of about 3, i.e. a third order low-pass, albeit the gain increase at the upper limit frequency suggests that it is not a pure lowpass. This is also suggested by the phase plot in which, instead of showing a continually increasing phase lag, at around 26–30 Hz there is a dramatic recovery of phase to increase again with increasing frequency (see Section 4). In the following only the gain part of the Bode diagrams are used to demonstrate drug action. 3.1. D1-R antagonist SCH 23390 Fig. 2 shows the same series of recordings as Fig. 1 after SCH 23390 injection. The most prominent change is a separation of the signal in transient peaks, emphasizing the components, and sharpening the peaks especially of the OFF-component. This was consistent for all animals tested. Despite of these changes the amplitude–frequency plot shows only a slight depression, especially the increase towards the corner frequency disappears. On the other hand, the fall off and therefore the corner frequency is the same as in the untreated eye, i.e. the temporal transfer properties seem not affected (Fig. 2(D)). It suggests, that the shape of the amplitude–

C. Mora-Ferrer, K. Behrend / Vision Research 44 (2004) 2067–2081

2071

0.020

(A) Pre

0.015

reponse stimulus

[mV]

0.010 0.005 0.000 -0.005 -0.010 0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

[S] 0.020 0.015

0.020

(B)

Post

0.015

0.010

0.010

0.005

0.005

0.000

0.000

-0.005

-0.005

-0.010

(C)

Post

-0.010 0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.0

0.1

0.2

0.3

0.4

[S]

0.5

0.6

0.7

[S]

1

gain

(D)

0.1

post pre 0.01 1

10

100

Frequency [Hz]

Fig. 2. (A)–(C) ERG before and after injection of SCH 23390 (D1-R antagonist, n ¼ 4). Here and in the subsequent figures the response to a 100 ms stimulus pre-injection (A), post-injection (B) and to a 5 ms flash also post-injection (C) are shown. Note in (B) the separation of the ERG-b-wave and dc-component as well as the sharpening of the OFF-response. (D) Gain plot. In comparison to the pre-injection data (open circles) the gain postinjection (filled circles) is scaled down a little. However, the corner frequency is unaffected. Axis and symbols as in Fig. 1. Note: last data point for post-injection data was shifted to 52 Hz for visibility reasons here and in following figures.

2072

C. Mora-Ferrer, K. Behrend / Vision Research 44 (2004) 2067–2081

frequency plot is dominated by the OFF-part of the ERG because this component seems the least affected by the drug. 3.2. D2-R antagonist Sulpiride Here also, the b-wave was influenced but, in contrast to the D1-R antagonist, instead of emphasizing the transients, the time course was considerably smoothed, the dc-part was higher in amplitude, and the OFFresponse was dampened considerably. Concerning the temporal transfer properties, the effect of Sulpiride was rather drastic. The corner frequency in the gain plot is shifted to a lower value and the curve takes the appearance of a more pure low-pass (Fig. 3(D)). 3.3. D1-R agonist SKF 38393 The most obvious effect of the drug was a decrease of the amplitude of the ON-response (Fig. 4(B)). The OFFresponse seemed not so much affected in amplitude, nevertheless, there were subtle changes in its time course. This was revealed when looking at the power spectra of the Fourier transform calculated from the respective OFF-responses (see Fig. 5(A)). Compared to the power spectra from the OFF-response for the D1antagonist or D2-agonist (Fig. 5(B)), both not changing the time transfer properties, the bulk of power was shifted towards lower frequencies, i.e. the transients of the response were dampened. Consequently, the gain plot (Fig. 4(D)) shows the same alterations as seen due to the D2 antagonist Sulpiride (having a similar power spectrum in the OFF-response Fig. 5(A)), although on the course of the ERG the effect of the respective drugs was quite different. 3.4. D2-R agonist quinpirole At concentrations of 1–10 lM Quinpirole no effect was seen in the shape of the ERG. For a 100 lM dose, a reduction of the dc-component was observed (Fig. 6(B)). However, the gain was not altered for any of the used concentrations (Fig. 6(D)). 3.5. Aspartate in combination with Sulpiride To rule out the possibility that DA already modulated photoreceptor activity in a way leading to the observed behavior, the transfer of receptor output to the second order neurons was interrupted with the glutamate substitute aspartate. As Fig. 6 shows, the ERG became a pure receptor response. The amplitude frequency plot is second to third order low-pass. Adding Sulpiride had no effect, which ruled out the photoreceptors as a source for the time transfer characteristics seen in the ERG as well as in the behavioral data.

4. Discussion As outlined in Section 1, the goal of this investigation was to show in the retinal signal processing pathway the influence of DA onto the temporal transfer properties. The behavioral data gave an indication that only the dopamine action via D2-R has significance on temporal resolution insofar as the flicker fusion point shifted to lower frequencies when D2-R were blocked with Sulpiride (Mora-Ferrer & Gangluff, 2002) and also that blocking the D1-R had no effect on the temporal transfer properties as judged by the FFF. Preparing the data of this study in a similar manner as here, a very good consistency of the gain plots for both behavior as well as the ERG for the action of Sulpiride and SCH 23390 can be seen (Fig. 8). Even the detail that prior to injection the gain shows a slight increase towards the corner frequency is present (Fig. 8(A) and (B)). Since the drugs were administered to the eye and their sole point of action could be the retina, this strongly suggests that temporal resolution ability of goldfish is completely determined by the retinal circuitry and not further shaped by central processing. On the other hand, because in the ERG most of the retinal circuitry is represented, the influence of the drugs on its components might reveal a more detailed insight as to which processing pathway of the many working in parallel plays the essential role. As already stated we will use the comparison of the Bode plots obtained for the different drugs used in this study. The interesting part of the gain plot of the untreated retina is the slight recovery of gain towards the upper limit frequency and the rather steep cut off. As already mentioned the pure low-pass characteristic of third order does not provide for the course of the gain. Especially the slight gain recovery cannot be fitted with such an assumption. A possibility is to introduce some kind of ‘‘resonance’’ into the circuitry. In our case it should be seen as the input being in phase with some ‘‘Eigenfrequency’’ of the processing chain. This ‘‘Eigenfreqeuncy’’ could–– unfortunately––be caused by several different mechanisms: (A) membrane properties of single cells, and (B) coupling properties between different cells. Discussing (A) first, one must look for cells carrying D2-R. Although, as pointed out already, D2-R exist in both the OPL and IPL, the explicit localization is not known except for the dopaminergic interplexiform cells (DAIPC) and photoreceptors. This makes the following discussion about membrane properties of single cells as source of the resonance purely hypothetical but useful for limiting the possibilities and as a guide for further experiments. The contacts of the DA-IPC are in the OPL mostly on bipolar cells (BC) and horizontal cells (HC), in the IPL mostly on amacrine cells (AC) (Dowling & Ehinger, 1978; Yazulla & Lin, 1995; Yazulla & Studholme, 1997;

C. Mora-Ferrer, K. Behrend / Vision Research 44 (2004) 2067–2081

2073

0.020

Pre

(A)

0.015

response stimulus

[mV]

0.010 0.005 0.000 -0.005 -0.010 0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7 [S]

0.020

0.020

(B)

Post

(C)

Pos t

0.015

0.015

0.010

0.010

0.005

0.005

0.000

0.000

-0.005

-0.005 -0.010

-0.010 0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.0

0.1

0.2

[S]

0.3

0.4

0.5

0.6

0.7 [S]

1

gain

(D)

0.1

post pre

0.01 1

10

100

Frequency [Hz] Fig. 3. ERG after injection of Sulpiride (D2-R antagonist, n ¼ 5). (A)–(D) Same as Fig. 2. Note in (B) the increase of the dc-component and the dampening and the prolongation of the OFF-component also seen in the ERG-response to the 5 ms pulse (C). In addition, the transients in the ERG are reduced. In (D) the gain is scaled down significantly and the corner frequency is shifted by a considerable amount. Also the course of the curve approaches one of a pure low-pass of second order.

2074

C. Mora-Ferrer, K. Behrend / Vision Research 44 (2004) 2067–2081 0.020

Pre

(A)

0.015 response stimulus

[mV]

0.010 0.005 0.000 -0.005 -0.010 0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7 [S]

0.020 0.015

0.020

Post

(B)

0.015

0.010

0.010

0.005

0.005

0.000

0.000

-0.005

-0.005

-0.010

Post

(C)

-0.010 0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.0

0.1

0.2

0.3

[S]

0.4

0.5

0.6

0.7 [S]

1

gain

(D)

0.1

post pre

0.01 1

10

100

Frequency [Hz] Fig. 4. ERG after injection of SKF 38393 (D1-R agonist, n ¼ 3). (A)–(D) Same as Fig. 2. Note in (B) the dramatic changes to the middle part (bwave and dc-component). The OFF-response is prolonged. Also the pulse-response (C) has changed with a prolonged OFF-response as seen in (B). The gain plot is almost identical to the one obtained after Sulpiride injection (Fig. 3(D)), although the shape of the ERG is vastly different for the two drugs.

C. Mora-Ferrer, K. Behrend / Vision Research 44 (2004) 2067–2081 400 350

(A)

D2-antagonist D1-agonist

Power

300 250 200 150 100 50 0 0

10

20

30

40

50

400 350

(B)

D2-agonist D1-antagonist

300 250 200 150 100 50 0 0

10

20

30

40

50

Frequency [Hz]

Fig. 5. Power spectra of the OFF-response after injection. The response was taken after stimulus OFF and Fourier transformed. The frequencies from 0 to 5 Hz are omitted. Note the shift of the bulk power towards lower frequencies with the drugs SKF 38393 (D1-R agonist) and Sulpiride (D2-R antagonist), the main peak being around 12–15 Hz, compared to the peak around 18–20 Hz with SCH 23390 (D1-R antagonist) and Quinpirole (D2-R agonist). X -axis: frequency in [Hz]; Y -axis: power.

Yazulla & Zucker, 1988). Photoreceptors can be ruled out as source for the resonance (see Fig. 7). Also, HC may be ruled out since they carry no D2-R. This leaves the BC dendritic membrane as the sole source for resonance in the OPL. Fig. 9(A) shows the gain plot fitted according to (A) with a resonance model (see Appendix A). The critical parameter of such a model is the damping in the resonance part. The Sulpiride effect as seen with this approach is merely the complete disappearance of the resonance, i.e. a high damping value. Burrone and Lagnado (1997) have investigated the membrane currents responsible for the damped oscillations around 30–40 Hz which occur in the response of ON-bipolar cells (depolarizing bipolar cells (DBC)), when a step of light is applied (Kaneko & Hashimoto, 1969). Their data show, that L-type calcium channels, calcium-activated potassium channels, and voltagedependent potassium channels shape the response leading to electrical resonance behavior of the DBC membrane potential. The resonance frequency is around 60 Hz but since they recorded isolated DBC’s the dif-

2075

ference to the oscillations observed by Kaneko and Hashimoto (Kaneko & Hashimoto, 1969) may be explained that the latter recorded in situ. These oscillations where also visible in our recordings on the DC part of the ERG and the frequency spectrum shows indeed a peak in the 60–70 Hz region. The reasoning would then be that the high end of the amplitude frequency plot is shaped by the membrane properties in the dendritic region of the DBC’s and that DA has an modulatory effect upon the particular ion channels involved in the resonance. Since Sulpiride, the D2-R antagonist, and SKF 38393, the D1-R agonist, influenced the high frequency end, the effect could be mediated by two mechanisms: (1) the D2-R are located in the dendritic region of the DBC and blocking the action of DA somehow jeopardizes the resonance properties. (2) blocking the D2-R of the DAIPC’s raises the DA level––somehow mimicked by the D1-R agonist––and this leads via D1-R at the dendritic membrane to the observed result. Now, when recording in the vitreous body the effect of Sulpiride was predominantly to suppress the transients in the ERG, especially obvious when looking at the pulse response. However the effect of damping both amplitude and time course was pre-dominantly on the OFF part of the ERG. Also, the so called DC part is raised in amplitude which is most probably due to a decrease of the OFFBC response (hyperpolarizing bipolar cells (HBC), (Sieving et al., 1994)).Taken together, this seems to rule out membrane properties of DBC’s dendrites as the site of Sulpiride action via D2-R. Indirect action via D1-R are still possible. Maguire and Werblin (1994) have shown that HBC’s of the tiger salamander retina show enhancement of response in the presence of DA mediated by D1-R. If true for the goldfish retina, the effect of blocking the D2-R of the DAIPC’s, and thereby raising the DA level, should have enhanced the OFF response instead of making it more sluggish. On the other hand, the data obtained with the D1-R agonist SKF 38393 seem to corroborate the finding of Maguire and Werblin (1994). Although a D1-R agonist certainly does not mimic an increased DA level, it forcibly activates the consequences of it. The visible effect on the shape of the ERG was a decrease in the dcpart which could be due to an increase in the HBC’s response amplitude to light. The effect upon the amplitude frequency function, however, is identical to blocking the D2-R and this seems to be due to a similar effect, namely the slowing down of the time course of the OFF response. This was indeed revealed in the respective power spectra of the OFF responses shown in Fig. 5, Although subtle, the shift towards lower frequencies in the spectrum is clearly visible. So far, resonance properties of HBC membrane have not been demonstrated (and are not visible in the recordings). Although the above reasoning has some uncertainties, there is more evidence speaking against the possibility that the time

2076

C. Mora-Ferrer, K. Behrend / Vision Research 44 (2004) 2067–2081 0.020 0.015

Pre

(A)

[mV]

0.010

response stimulus

0.005 0.000 -0.005 -0.010 0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7 [S]

0.020 0.015

0.020

(B)

Post

(C)

Post 0.015

0.010

0.010 signal stimulus

0.005

0.005

0.000

0.000

-0.005

-0.005

-0.010

-0.010 0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.0

0.1

0.2

0.3

[S]

0.4

0.5

0.6

0.7 [S]

1

gain

(D)

0.1

post pre 0.01 1

10

100

Frequency [Hz] Fig. 6. ERG after injection of Quinpirole (D2-R agonist, n ¼ 3). (A)–(D) Same as Fig. 2. Shown in (B) is the response for a 100 lM dose to demonstrate the only effect of Quinpirole. Note the separation of the b-wave from the dc-component together with a decrease in amplitude. The OFF-response is initially as fast as pre-injection but has a slightly changed fall-off. The 5 ms pulse response (C) has a rather small OFF-part compared to the reference measurement (Fig. 1(B)). Compare (B) to Fig. 2(B) where the same separation of b-wave and dc-component can be seen. The gain plot (D) shows the data of all animals and concentrations used.

C. Mora-Ferrer, K. Behrend / Vision Research 44 (2004) 2067–2081

2077

0.020

Pre

0.015

(A) response stimulus

[mV]

0.010 0.005 0.000 -0.005 -0.010 0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7 [S]

0.020

[mV]

0.015

0.020

Post

0.015

(B)

0.010

0.010

0.005

0.005

0.000

0.000

-0.005

-0.005

-0.010

Post

(C)

-0.010 0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.0

0.1

0.2

0.3

[S]

0.4

0.5

0.6

0.7 [S]

1

gain

(D)

0.1

pre post aspartate post aspartate + sulpiride

0.01 1

10

100

Frequency [Hz] Fig. 7. ERG after injection of Aspartate and additional Sulpiride ðn ¼ 6Þ. (A)–(D) Same as Fig. 2. Note the ERG changes to a pure receptor potential after injection of Aspartate. The additional injection of Sulpiride has no effect, i.e. the ERG remains a pure receptor response. The gain plot (D) becomes, with the injection of Aspartate, a pure low-pass of second order with corner frequency around 10 Hz and this frequency characteristic is not changed with the addition of Sulpiride.

2078

C. Mora-Ferrer, K. Behrend / Vision Research 44 (2004) 2067–2081

log normalized choice frequency [%]

100

(A)

(B)

90 85

SCH 23390

Sulpiride 80 75 70 65 60 pre post

55

pre post

50 10

20

30

40

50

1

10

20

30

40

50

1 (C)

gain

(D)

gain

Sulpiride

0.1

SCH 23390

0.1

post

post pre

pre

0.01

0.01 1

0

100

1

10

100

Frequency [Hz] Fig. 8. Comparison of behavioral measurements ((A) and (B); see Figs. 2 and 3 in Mora-Ferrer & Gangluff, 2002) and ERG data (C (Fig. 3(D)) and D (Fig. 2(D))). The gain–frequency plots for the two data sets show a remarkable consistency for the untreated (open symbols) and the treated animals (filled symbols). Note: 50% choice frequency is equivalent to zero amplitude of ERG-wave. (A)–(D) X -axis: log stimulus frequency [Hz]; (A) and (B): Y -axis: normalized choice frequency [%], (C) and (D) gain. Number of animals for (A) and (B): n ¼ 4.

transfer properties are shaped by membrane properties such as described for DBC’s in the OPL circuitry. Looking at the IPL, the possibilities are quite different to the OPL. First of all, as already mentioned above, the bulk of DA-IPC innervation is on AC. Second, the AC are activated by the BC and provide inhibition fed via GABA-R upon both DBC’s and HBC’s. This speaks for the resonance implemented by coupling properties between different cells (see p. 13, possibility B). We formulate this model as the output of a second order low-pass fed forward as inhibition with a fixed delay (see

Fig. 9 of Dong & Werblin, 1998, which shows the elements involved). The AC functions as the delay element, the BC as the low-pass. There are two critical parameters: the delay introduced and the inhibitory strength. Fig. 9(B) shows the gain plot, Fig. 9(C) the phase plot, fitted according to this model. The delay was chosen to be just half the period of the upper limit frequency of the low-pass (30 Hz) which is about 15 ms. The output of the delay element will then just be in phase with the original signal, in this way counteracting the damping of gain introduced by the low-pass the signal has gone

C. Mora-Ferrer, K. Behrend / Vision Research 44 (2004) 2067–2081

gain

1

(A)

0.1

pre Resonance-model (De=0.24) Resonance-model (De=2)

0.01 1

10

100

Frequency [Hz] 1

gain

(B)

0.1

pre Inhibitory strength: 0.1 0.35 4

0.01 1

10

100

50

(C) 0 -50

Phase [°]

-100 -150 -200 pre

-250

Inhibitory strength: 0.1 0.35 4

-300 -350

1

10

100

Frequency [Hz]

Fig. 9. Bode plot of the models calculated. (A) Low-pass combined with resonance; (B) feed-forward inhibition with fixed delay. (A) The pre-injection data can be reasonably fitted with this model. The parameters chosen were fto: 35 Hz, fro: 19 Hz and damping De: 0.24. When the damping De is increased to 2 the gain-plot appears very similar to the gain plots obtained for Sulpiride and SKF 38393. (B) The pre-injection data are fitted when the inhibitory strength in this model is set to an intermediate level (0.35). A reduction of inhibitory strength (0.1) results in a gain-plot similar to the ones for Sulpiride and SKF 38393. Increasing inhibitory strength (4) results in a gain plot similar to the one obtained with SCH 23390 and quinpirole, especially the phase plot (bottom) is very good fitted (details for the calculation see Appendix A).

through already. In all the studies on GABAergic inhibition upon BC-terminals, the GABAc-R was the predominant source ((Dong & Werblin, 1994, 1998; Shen & Slaughter, 2001) in the tiger salamander; (Shields, Tran, Wong, & Lukasiewicz, 2000) in the ferret; (KapoustaBruneau, 2000) in the rat). And in all studies the effect of the AC feedback was to sharpen the time course of the

2079

bipolar output by making the otherwise sustained signal more transient. Again one could discuss direct vs. indirect influence of blocking DA action. An interesting finding in this connection is by Dong and Werblin (1994) who demonstrated that DA, acting via D1-R, had an reducing effect upon the GABAc elicited current. Since the D1-R antagonist rather improved the OFFpart in the pulse response, and did not impair the high end of the amplitude frequency plot, as was the case with the D2-R agonist, it is suggestive to attribute the action of Sulpiride to a raising of DA which, according to the above finding would dampen the GABAc inhibition and as a consequence making the circuitry’s time transfer behavior more low-pass like. The result obtained with the D1-R agonist only hints in the same direction. In summary, higher DA levels tend to dampen the time transfer properties via action on D1-R and the site of action is a decreased inhibition via GABAc receptors at the axonal terminal of BC. Blocking the D1-R should then increase the inhibition, which leads to the demonstrated effects upon the gain plot (Fig. 2(D)). In Fig. 9(B) and (C) these effects could be mimicked at least qualitatively by weakening or strengthening the inhibition leading to the results predicted by the above reasoning. As already stated, a discussion of circuit details on the base of the ERG is highly hypothetical. On the other hand, if the results of modeling with linear systems tools are taken as a means to reveal possibilities hitherto hidden, as is the case with a gain increase as a result of properly phased inhibition, then its usefulness is certainly demonstrated and the results are helpful for understanding retinal operations. One may wonder why under photopic conditions, when the DA-level is raised, the temporal transfer properties should be impaired as the data suggest. One action of DA is the decoupling of the HC network with subsequently improved local spatial resolution and a speeding up of light responses. In the detection of motion spatial resolution and temporal transfer properties are coupled in the sense that change of location over time of a visual object has to be detected and there has to be always a trade off between both spatial resolution and velocity resolution. The ‘‘normal’’ DA level might just keep this delicate balance in so far as in the OPL it serves the spatial resolution by decoupling horizontal cells thereby making possible contrast enhancement, whereas in the IPL it fine tunes the inhibition needed for overcoming the sluggishness at the bipolar output but controlling the inhibitory strength as not to wipe out everything. Concerning the temporal resolution in the goldfish, the high end of the gain plot is equivalent to an object passing with a velocity of about 0.2–0.4 m/s enabling the fish a tolerable resolution even during his occasional outbursts of swimming speed.

2080

C. Mora-Ferrer, K. Behrend / Vision Research 44 (2004) 2067–2081

Acknowledgements

3x10

3x10

2x10

[mW]

We wish to thank Drs. Yazulla, W€ assle and Neumeyer for their critical comments on earlier versions of the manuscript and Dr. Scharstein for his help with the models. This work was supported by the Forschungsfond of the University of Mainz.

2x10

1x10

Appendix A

-3

-3

(A)

Stimulus

-3

-3

-3

-6

To simulate the gain curve assuming a low pass with resonance a third order low-pass was combined with a second order low-pass with resonance (Cruse, 1996). Their amplitudes depending on input frequency are given by: 1 A ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiord f2 1 þ fto 2

ð1Þ

500x10

0 400

500

600

700

-6

5x10

(B)

Background -6

4x10

-6

3x10

2

fro A ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2 2 ðfro  f 2 Þ þ 4De2 f 2 fro2

ð2Þ -6

2x10

with f ¼ input frequency, fto, fro ¼ corner frequency of (1) and (2) respectively, De ¼ damping coefficient. The two were combined and the corner frequencies were chosen according to the gain plot prior to injection for the resonance part and to the D2-R antagonist curve for the low-pass. The damping was adjusted to fit the curve in the manner shown in Fig. 9(A). Increasing the damping would approximate more and more the lowpass characteristic seen in the D2-R antagonist plot. The second simulation has been calculated according to the following model: the output of a second order low-pass is fed into two branches, one attributed with no delay, the other with a fixed delay chosen such that at the corner frequency the delay is just 180°. The delayed signal, because of its inhibitory nature, is subtracted from the non-delayed signal. The amplitude of the output after the subtraction is given by qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi   1 þ he2  2he cos pf fo 0:2 A¼f ð3Þ qffiffiffiffiffiffiffiffiffiffiffiffi2ffiord f 1 þ fo2 and the phase by   arctan pf fo U ¼ 360 p  180 arctan

-6

1x10

0 400

500

600

700

[nm]

Fig. 10. Spectra of the LED used for stimulation (A) and background (B) illumination. (A) LED with yellow filter driven at stimulus intensity setting. (B) Background illumination at experimental setting. Y axis: radiant power [mW]; X -axis: wavelength [nm].

the signal despite the increase at the very low frequencies (2–6 Hz) due to the effect of higher harmonics in the rectangular stimulus (see Section 3). The corner frequency of the low-pass was set to 25 Hz a value obtained from the gain after injection of Sulpiride. In Fig. 9(B) (gain) and C (phase) the effects of varying the inhibitory strength (he) are shown (Fig. 10). The spectra of both the LED and the background illumination light source was measured by replacing the fish eye with the spectrometer probe (Spectro 100, Instrument Systems, Germany). References

he sin

 pf  fo

1 þ he cos

 pf  fo

! 

1 p

ð4Þ

with f ¼ input frequency, fo ¼ corner frequency, he ¼ inhibitory strength. The additional term f 0:2 was introduced to account for the error introduced by measuring the amplitude of

Akopian, A., & Witkovsky, P. (1996). D2 dopamine receptor-mediated inhibition of a hyperpolarization-activated current in rod photoreceptors. Journal of Neurophysiology, 76(3), 1828–1835. Brown, K. T. (1968). The electroretinogram: its components and their origins. Vision Research, 8(6), 633–677. Burrone, J., & Lagnado, L. (1997). Electrical resonance and Ca2þ influx in the synaptic terminal of depolarizing bipolar cells from the goldfish retina. Journal of Physiology, 505(Pt 3), 571–584.

C. Mora-Ferrer, K. Behrend / Vision Research 44 (2004) 2067–2081 Cruse, H. (1996). Neural networks as cybernetic systems. Stuttgart, New York: Thieme, pp. 1–167. Dearry, A., & Burnside, B. (1989). Light-induced dopamine release from teleost retinas acts as a light-adaptive signal to the retinal pigment epithelium. Journal of Neurochemistry, 53(3), 870– 878. DeMarco, P. J., Jr., & Powers, M. K. (1989). Sensitivity of ERG components from dark-adapted goldfish retinas treated with APB. Brain Research, 482(2), 317–323. Dong, C. J., & Werblin, F. S. (1994). Dopamine modulation of GABAC receptor function in an isolated retinal neuron. Journal of Neurophysiology, 71(3), 1258–1260. Dong, C. J., & Werblin, F. S. (1998). Temporal contrast enhancement via GABAC feedback at bipolar terminals in the tiger salamander retina. Journal of Neurophysiology, 79(4), 2171–2180. Dowling, J. E. (1987). The retina: an approachable part of the brain. Cambridge, London: Harvard University Press, pp. 1–282. Dowling, J. E., & Ehinger, B. (1978). The interplexiform cell system. I. Synapses of the dopaminergic neurons of the goldfish retina. Proceedings of the Royal Society of London Series B––Biological Sciences, 201(1142), 7–26. Fan, S. F., & Yazulla, S. (2002). Interactions of cannabinoid CB1 receptor and dopamine D2 receptor agonist on voltage–currents in goldfish cones. Ft. Lauderdale: ARVO, IOVS. Harsanyi, K., & Mangel, S. C. (1992). Activation of a D2 receptor increases electrical coupling between retinal horizontal cells by inhibiting dopamine release. Proceedings of the National Academy of Sciences of the United States of America-Physical Sciences, 89(19), 9220–9224. Hedden, W. L., Jr., & Dowling, J. E. (1978). The interplexiform cell system. II. Effects of dopamine on goldfish retinal neurones. Proceedings of the Royal Society of London Series B––Biological Sciences, 201(1142), 27–55. Kaneko, A., & Hashimoto, H. (1969). Electrophysiological study of single neurons in the inner nuclear layer of the carp retina. Vision Research, 9(1), 37–55. Kapousta-Bruneau, N. V. (2000). Opposite effects of GABA(A) and GABA(C) receptor antagonists on the b-wave of ERG recorded from the isolated rat retina. Vision Research, 40(13), 1653–1665. Kolbinger, W., Kohler, K., Oetting, H., & Weiler, R. (1990). Endogenous dopamine and cyclic events in the fish retina, I: HPLC assay of total content, release, and metabolic turnover during different light/dark cycles. Visual Neuroscience, 5(2), 143– 149. Krizaj, D., Gabriel, R., Owen, W. G., & Witkovsky, P. (1998). Dopamine D2 receptor-mediated modulation of rod–cone coupling in the Xenopus retina. Journal of Comparative Neurology, 398(4), 529–538. Lin, Z. S., & Yazulla, S. (1994). Depletion of retinal dopamine increases brightness perception in goldfish. Visual Neuroscience, 11(4), 683–693. Maguire, G., & Werblin, F. (1994). Dopamine enhances a glutamategated ionic current in OFF bipolar cells of the tiger salamander retina. Journal of Neuroscience, 14(10), 6094–6101. Mora-Ferrer, C., & Gangluff, V. (2000). D2-dopamine receptor blockade impairs motion detection in goldfish. Visual Neuroscience, 17(2), 177–186. Mora-Ferrer, C., & Gangluff, V. (2002). D2-dopamine receptor blockade modulates temporal resolution in goldfish. Visual Neuroscience, 19, 807–815. Mora-Ferrer, C., & Neumeyer, C. (1996). Reduction of red–green discrimination by dopamine D1 receptor antagonists and retinal dopamine depletion. Vision Research, 36(24), 4035–4044. Mora-Ferrer, C., Yazulla, S., Studholme, K. M., & Haak-Frendscho, M. (1999). Dopamine D1-receptor immunolocalization in goldfish retina. Journal of Comparative Neurology, 411(4), 705–714.

2081

Muresan, Z., & Besharse, J. C. (1993). D2-like receptors in amphibian retina: localization with flourescent ligands. Journal of Comparative Neurology, 331(2), 149–160. Negishi, K., Teranishi, T., & Kato, S. (1990). The dopamine system of the teleost fish retina. Progress in Retinal Research, 9, 1–48. Rashid, K., Baldridge, W. H., & Ball, A. K. (1993). Evidence for D2 receptor regulation of dopamine release in the goldfish retina. Journal of Neurochemistry, 61(6), 2025–2033. Rohrer, B., & Stell, W. K. (1995). Localization of putative dopamine D2-like receptors in the chick retina, using in situ hybridization and immunocytochemistry. Brain Research, 695(2), 110–116. Shen, W., & Slaughter, M. M. (2001). Multireceptor GABAergic regulation of synaptic communication in amphibian retina. Journal of Physiology, 530(Pt 1), 55–67. Shields, C. R., Tran, M. N., Wong, R. O., & Lukasiewicz, P. D. (2000). Distinct ionotropic GABA receptors mediate presynaptic and postsynaptic inhibition in retinal bipolar cells. Journal of Neuroscience, 20(7), 2673–2682. Sieving, P. A., Murayama, K., & Naarendorp, F. (1994). Push-pull model of the primate photopic electroretinogram: a role for hyperpolarizing neurons in shaping the b-wave. Visual Neuroscience, 11(3), 519–532. Stella, S. L., Jr., & Thoreson, W. B. (2000). Differential modulation of rod and cone calcium currents in tiger salamander retina by D2 dopamine receptors and cAMP. European Journal of Neuroscience, 12(10), 3537–3548. Vaquero, C. F., Pignatelli, A., Partida, G. J., & Ishida, A. T. (2001). A dopamine- and protein kinase A-dependent mechanism for network adaptation in retinal ganglion cells. Journal of Neurosciene, 21(21), 8624–8635. Wagner, H. J., & Behrens, U. D. (1993). Microanatomy of the dopaminergic system in the rainbow trout retina. Vision Research, 33(10), 1345–1358. Wagner, H. J., Behrens, U. D., Zaunreiter, M., & Douglas, R. H. (1992). The circadian component of spinule dynamics in teleost retinal horizontal cells is dependent on the dopaminergic system. Visual Neuroscience, 9(3–4), 345–351. Wagner, H. J., Luo, B. G., Ariano, M. A., Sibley, D. R., & Stell, W. K. (1993). Localization of D2 dopamine receptors in vertebrate retinae with anti-peptide antibodies. Journal of Comparative Neurology, 331(4), 469–481. Wang, Y., Harsanyi, K., & Mangel, S. C. (1997). Endogenous activation of dopamine D2 receptors regulates dopamine release in the fish retina. Journal of Neurophysiology, 78(1), 439–449. Weiler, R., Kohler, K., Kirsch, M., & Wagner, H. J. (1988). Glutamate and dopamine modulate synaptic plasticity in horizontal cell dendrites of fish retina. Neuroscience Letters, 87(3), 205–209. Weiler, R., Kohler, K., Kolbinger, W., Wolburg, H., Kurz-Isler, G., & Wagner, H. J. (1988). Dopaminergic neuromodulation in the retinas of lower vertebrates. Neuroscience Research, 8(Suppl.), S183–S196. Yazulla, S., & Lin, Z. S. (1995). Differential effects of dopamine depletion on the distribution of [3H]SCH 23390 and [3H]spiperone binding sites in the goldfish retina. Vision Research, 35(17), 2409– 2414. Yazulla, S., & Studholme, K. M. (1997). Differential reinnervation of retinal bipolar cell dendrites and axon terminals by dopamine interplexiform cells following dopamine depletion with 6-OHDA. Journal of Comparative Neurology, 382(4), 535–545. Yazulla, S., Studholme, K. M., Fan, S. F., & Mora-Ferrer, C. (2001). Neuromodulation of voltage-dependent K+ channels in bipolar cells: immunocytochemical and electrophysiological studies. Progress in Brain Research, 131, 201–213. Yazulla, S., & Zucker, C. L. (1988). Synaptic organization of dopaminergic interplexiform cells in the goldfish retina. Visual Neuroscience, 1(1), 13–29.