Temporal transfer and nonlinearity properties of turtle erg: Tuning by temperature, pharmacology, and light intensity

Temporal transfer and nonlinearity properties of turtle erg: Tuning by temperature, pharmacology, and light intensity

Ksion f&-s. Vol. 3. No. 4. pp. 403-492, 1985 Printed m Great Britain. All rights reserved 0022-6989 ‘85 s3.Qo + 0.00 C 1985 Pergamon Press Ltd Copyr...

1MB Sizes 57 Downloads 27 Views

Ksion f&-s. Vol. 3. No. 4. pp. 403-492, 1985 Printed m Great Britain. All rights reserved

0022-6989 ‘85 s3.Qo + 0.00 C 1985 Pergamon Press Ltd

Copyright

TEMPORAL TRANSFER AND NONLINEARITY PROPERTIES OF TURTLE ERG: TUNING BY TEMPERATURE, PHARMACOLOGY, AND LIGHT INTENSITY ALAN R. ADOLPH Eye Research Institute of Retina Foundation and Department of Ophthalmology, School. 20 Staniford Street, Boston, MA 02114, U.S.A.

Harvard Medical

(Received 18 Mqv 1984~in re&sedform 16 November 1984) Abstract-ERG impulse response, amplitude and phase temporal spectral transfer functions. and nonlinearities were measured in turtle retina under different retinal temperatures, pharmacological treatments, and light intensities. &Wave amplitude is strongly temperature dependent; amplitudes of the a-wave and slow P-III are less sensitive. Their time courses all slow markedly as temperature decreases. ERG amplitude transfer function is bimodal bandpass with narrow low-frequency peak (below I Hz) and broader mid-frequency peak (5 Hz at 8°C). Both peaks broaden and their frequency increases (low, 1 Hz mid, 15 Hz, at 23°C) as temperature increases. Phase transfer function slope decreases (from -7O”/Hz at 8°C to -25”IHz at 23°C) as temperature increases. Nonlinear properties of ERG at high input intensity are modelled by a quadratic nonlinearity, low-pass prefilter with cutoff above 12.4Hz. and low-pass postfdter broadly peaked at 4-10 Hz with cutoff above 20 Hz. For low input intensity, ERG exhibits linear properties with low-pass filtering sharply cut off above 6 Hz. o-Wave and slow P-III were isolated by aspartate treatment; depolarizing bipolar cell activity was examined using ethanol/GABA treatment of retina. High-frequency components, including broad mid-frequency peak, were attenuated and lowfrequency components were enhanced with aspartate. Transfer function narrows and peaks at a lower frequency with ethanol/GABA. Temperature effects ERG Turtle retina

Pharmacology

Noise analysis

INTRODUmION The retina is a neural network exquisitely temperature. The elements comprising

sensitive

to

this retinal network, the retinal neurons and glial cells, and their functional interactions via synapses, tight junctions, ionic fluctuations, etc. are differentially sensitive to the effects of variations in temperature (Negishi and Svaetichin, 1966; Schellart ef ai., 1974; Lamb, 1984; Armington and Adolph, 1984). Differential thermal sensitivity produces changes in the functional properties of the various intraretinal pathways. This thermal “tuning” of the network properties is reflected in field potentials corresponding to intraretinal pathway activity, e.g. the electroretinogram (ERG). Although many of the nonmammalian vertebrates used in retinal research are poikilothe~ic, temperature has been largely ignored as a factor influencing vertebrate retinal function, in retina research paradigms. One obvious exception is in the studies of visual pigment kinetics and the first stages of visual transduction in retinal photoreceptors (e.g. photon quantum response fluctuations, etc.). Perhaps the use of temperature-stabilized environments in retina experiments in uirro or the natural thermal stability assumed for mammalian neural tissue in uivo has contributed to the neglect of temperature as an extremely effective tool in retina research. Granit

Transfer functions

Nonlinearity

(1947) and more recently Armington and Adolph (1984) have used temperature variation and its effects on the ERG as a means of studying the ERG component processes and their possible interaction. In the experiments reported in this paper I have used temperature variation, as well as pharmacological manipulation, to examine the turtIe retina ERG, its components, and their interactions. In addition to the traditional flash stimulus ERG, white noise and sum-of-sinusoids stimuli were also used to efficiently assess temporal amplitude and phase transfer properties and any nonlinearities associated with the genesis of the ERG, and the thermal sensitivity of these effects. EXPERIMENTAL .METHODs

An eyecup or isolated retina preparation from the pond turtle (Pseudemys scripta elegans) was used in the experiments. For an eyecup, the globe was hemisected with a razor at a level slightly posterior to the ora serrata. Most vitreous was removed by vitrectomy and tissue planchet drainage, although a thin layer usually remained and was left in place. Hemisection of the eyecup and inversion of the nonoptic nerve-containing segment onto a filter paper disk was the next step in preparing the isolated retina. This was followed by peeling the sclera/pigment epithelium

away from the retina. which was leii adhering to the filter paper with its receptors facing upward. The eyecup or isolated retina vvas held in place in a circular experimental chamber whose bottom was formed by a Peltier device (Cambridge Thermionic Corp.. Cambridge, MA) covered by a thin, anodized ai~rn~nurn plate. The chamber sides were a plenum that carried moist 100”; oxygen and discharged it continuously over the ryecup. The chamber also contained a ring of chlorided silver wire as an indifferent electrode, a thermistor sensor for the bipolar temperature controtter, and a second bead thermistor for a digital temperature indicator. Stable physiological conditions could be maintained in such a preparation for Z or 3 hr. The optical system combines the images of two apertures, one illuminated by a steady incandescent source, the other by a glow modulator source whose intensity can be varied by any appropriate modulating signal, e.g. white noise, sine waves, sums of sinusoids. square waves, etc. In the experiments reported in this paper, fuull-field stimuli (LOmm diameter or greater) were used to illuminate the entire eyecup or isolated retina. The individual beams may also be interrupted by electromechanical shutters, in the conventional manner. The maximum irradiance, at the retinal Iocation in the optical system, produced by the unattenuated beams is 1.2 p W/cm’ for the glow modulator tube source, and 87 pW/cm? for the tungsten-halogen source (UDT Model 61 Optometer). The light source used for ail spectral analyses, i.e. amplitude (XFR) and phase (4) transfer functions, coherence function (CF), and sum-of-sines {S/S), was the glow modulator tube with attenuation by neutrai density filters, as indicated in the figure legends. The ERG flash responses (Fig. 1) were to the unattenuated tungsten-halogen source using a full-field stimulus. A sample of the optical input stimulus to the retina was used as the “input” signal to the FFT spectrum analyzer in transfer function measurements. A chlorided-silver wire dipped in the eyecup vitreous or touching the isofated retinal surface was used as the recording electrode. Electrical activity was amplified by conventional electtophysiological electronic instruments, displayed on a CR0 screen, recorded on a chart recorder (d.c.40 Hz FS.), and recorded on an instrumentation tape recorder (d,c.-2.5 kHz, Ilm. mode). Retinal “output” spectral analysis and transfer function spectral analysis were performed by a dual-channel, reai-time spectrum analyzer that incorporates digitai filtering techniques and an FFT-based spectral analysis implemented via a dedicated microprocessor (Hewlett-Packard 3582A). The unit can measure amplitude and phase spectra of two signais, or the gain and phase of their transfer function (128 equispaced points on the frequency scale). in addition, it computes the coherence function {normalized, input-output cross power spectrum) which gives a measure of power in the output

ERG vf

Temperatura

Fig. 1. ERG flash response as a function of retina temperature. Responses to IO&macetiashts at 28°C (top), l8’C (second top}, 13.3Q.Z(third top), and 7.8”C (bottom). Calibrations are 100JoV and 0.4 see in top three traces; 50 p V and 0.4~~ in bottom trace. (Amplitude calibration in gVV/division in each trace varies.) Unattenuated halogen source. Full-field stimulus.

tungsten-

ERG transfer function and nonlinearity tuning

485

Table 1. 909; confidence limits on CF. XFR. d measurements Measured coherence function (CF) values 0.2 CF XFR (dB) 4 fdeg)

+5.2 - 14.6 +54

0.3 +4.2 -8.4 f38

0.4 f0.19 -0.25 1-3.5 -6.0 k30

0.5 f0.17 -0.25 + 3.0 -4.5 224

0.6 +0.14 -0.24 c2.s - 3.5 119

0.7 +O.ll -0.20 +2.1 -2.7 515

.-_I___ 0.8 +o.os -0.15 il.6 -2.0 2 I2

0.9 +0.04 -0.09 + 1.1 -1.3 +8

The spectrum measurements presented in the paper are the r.m.s. averages of 16 spectrum measurements under each condition. The accuracy, or expected range of variability, may be described in terms of the “90?~ confidence limits” of the averaged spectrum data, i.e. the range of values within which 90:, of the measurements will fall. For example.suppose the measured values of XFR and d are -20dB and -40’. respectively, at some frequency, and the corresponding CF is 0.7. The 90% confidence limits for CF. XFR. and @Iare (0.5. 0.81). ( - 22.7 dB. - 17.9 dB), and (-55’. -25”). respectively (based on Bendar and Piersol. 1971).

signal due to the input signal {low coherence regions of the spectrum indicate noise and nonlinearity generated signals arising within the system under test). Spectral spans of either O-25 Hz (200mHz resolution) or O-50 Hz (400 mHz resolution) were used in the turtle retina signal analysis. The output and transfer function spectra of the ERG and, in other experiments (Adolph, l983a, b), intracellular to below noise levels at responses, attenuated frequencies well below 50 Hz. White-noise was used to drive the glow modulator tube (for making transfer function measurements). The noise is a digitally synthesized, flat-amplitude, frequency comb with spectral components at each of the (128) spectrum analysis points. In the present experiments, output spectra in response to sum-ofsinusoid inputs are presented as a qualitative assessment of retinal nonIinearities. A microcomputer (Apple II+) was used to synthesize the sum of four sinusoids with appropriate frequencies and phase relationships (Victor and Shapley, 1980). The frequencies of the four sines used in these experiments are 2.8, 6.0, 12.4, and 25.2 Hz (based on a relation of the form k(2” - I), where k =0.4andn = 3-6). Such a set of frequencies has no coincidences with its set of second- and third-order sums, differences, and harmonics. Coincidences of frequencies begin to appear between second and fourth order, and first and fifth order, components. These coincidences may be removed by shifting the relative phases of the primary components by half a cycle, according to an algorithm developed by Victor and Shapley (1980). For our set, 2.8 and 25.2 Hz are not shifted: 6.0 and 12.4 Hz are shifted. The absence of coincident frequency components permits one to observe the production of nonlinear components at expected frequencies when the sum-of-sinusoid signal is used as an input to a putatitively nonlinear system, such as the retina under appropriate conditions. Coherence functions were measured, in many instances, in addition to amplitude and phase spectra of the various transfer functions, and output amplitude spectra in response to sum-of-sinusoid input signals. The coherence function, CF, is the fraction of total output power (amplitude squared) due to the input of a system. It ranges from unity to zero and

is a function of frequency. As noise and nonlinear components become a larger fraction of output power, at any frequency, CF becomes smaller. CF has two other useful features: knowledge of its quantitative value at any frequency translates to confidence limits on the measurements of amplitude and phase transfer functions (Bendat and Piersot, 1971); CF is equivalent to the squared, input-output cross-correlation coefficient. Table 1 indicates the expected ranges of variability (confidence limits) of various spectrum measurements under the reported experimental conditions. Electrophysiological data must be stable (quasistationary) during at least the first 42 set, which is the time period required by the spectrum analyzer to input, compute, and average sixteen O-50 Hz spectra. At the conclusion of this time, the averaged spectral data are frozen in storage in the analyzer, prior to the onset of the remainder of the processing, display, and storage, etc. performed by the microcomputer. Each experimental step or condition requires about 1.5 min minimum for data analysis and storage, plus any other intervening time required, e.g. to change and stabilize temperature, apply pharmacological agents, modify stimulus configuration, etc. This has not turned out to be a problem in ERG experiments. RESULTS

Temperature dependence of the ERG

The ERG response to a 100 msec flash with the retina at a range of temperatures from 28’ to 7.8”C is shown in Fig. 1. The amplitude calibration at 28°C is greater than at 18” and 13.3”C, which are the same, and these in turn are greater than at 7.8”C. The b-wave amplitude is strongly temperature dependent; it goes down as temperature is decreased. a-Wave amplitude is relatively insensitive to temperature, as is the amplitude of slow-PHI. The time-course of all three components (a, 6, and slow-PIII) are quite temperature sensitive; they slow significantly as temperature is decreased. There are some minor oscitlatory potentials superimposed on the three major waves, and their frequency appears to slow as temperature drops. Similar findings have been reported by Armington and Adolph (1984).

ERG

Transfer

Prc~srties

vs Temoeroture

23.4’

16.5’

8.0’

Fig. 2. ERG transfer function and coherence function spectra as functions of retina temperature. Left amplitude transfer function (XFR) and coherence function (CF) spectra at 23.4”C, IdS”C. If.S”C, and 8.O”C (top to bottom). Right column, phase (4) and amplitude (XFR) transfer function spectra, at same indicated temperatures. Amplitude calibrations for all (XFR) spectra, lOdB/div. For CF spectra, 1.0 full scale. For q5spectra, 0” center and tiO”/div. Spectral spans for all spectra (XFR. CF, 4) are 0 to 25 Hz (linear). &attenuated glow modular tube source. Full-field stimulus.

column,

ERG transfer function and nonlinearity tuning Thermal sensitivity of ERG transfer properties

ERG amplitude transfer function has a bimodal shape consisting of a relatively narrow peak at low frequencies and a much broader mid-frequency peak (Fig. 2). At the lowest temperatures (8’C in Fig. 2), the low-frequency peak is so narrow that it is essentially a delta function at D.C., indicating that a net D.C. component remains in ERG even at low temperatures. The relatively broad mid-frequency “bandpass” peak becomes broader as temperature increases. as does the narrow low-frequency peak. High significance values of coherence function parallel the spectral distribution of the bandpass peak. The slope of the phase transfer function increases as temperature decreases, i.e. rate of change of phase is related to broadness of amplitude transfer function peak. The 180” phase shift points, which occur at the discontinuities in the phase spectrum plots, decrease in frequency as temperature decreases. ERG response linearity If the presence of nonlinear distortion products is used as a measure of the extent of nonlinearity, then the linearity of ERG response is dependent on stimulus intensity. A sum-of-sines input to the retina at nominal 0 log attenuation, in the experiment shown in Fig. 3, produces several significant distortion peaks within the passband of the ERG amplitude transfer function (XFR). Attenuating stimulus intensity by I log unit has, under these conditions, attenuated the response to the two highest frequency input sines as well as the nonlinear responses in the corresponding range of frequencies. The responses to the two lowest frequency input sines are still significantly higher than the background noise spectrum. The sum-of-sines (S/S) input signal consists of sine waves at 2.8, 6.0, 12.4, and 25.2 Hz with equaf amplitudes. The frequency spectra of the S/S inputs at 0 log and - I log attenuation of the light intensity are shown in the lower two graphs in Fig. 3, marked S/S,,. These spectra are measured from the output of a photomultiplier photometer that samples the intensity of the stimulus light beam projected onto the retina. The output (ERG) spectrum in response to the 0 log attenuated S/S input (Fig. 3, center, graph) contains several peaks at frequencies other than the fundamental frequencies. A table of the 2” possible first-order distortion products (sums, differences, and second harmonics)

*Both output spectra (0 log, - 1 log) also contain a peak at 0.8 Hz, which is not at a fundamental, harmonic, or sum or difference frequency of the fundamentals. The constant amplitude of this peak in the spectra for differing input intensity levels and its frequency unrelated to input frequencies for both input conditions, indicate that this peak reflects an artifact such as a periodic low-frequency noise. The sharp dip in coherence at 0.8 Hz for the white noise XFR (Fig. 3, top) is further support for this conclusion.

387

and the n fundamental below

frequencies (n = 4) is shown

Fundamentals

2.8, 6.0, 12.4, 25.2

Sums:

8.8, 15.2, 18.4, 28.0, 31.2, 37.6

Differences: Second harmonics:

3.2, 6.4, 9.6, 12.8, 19.2, 22.4 5.6, 12.0, 24.8, 50.4

The measured frequencies and amplitudes of the various peaks in the 0 log S/S output spectrum are: Frequency Amplitude

(Hz): (dB):

2.8 6.0 8.8 9.6 12.4 15.2 18.4 23 33 23 14 23 14 13

The amplitudes are relative to the baseline, which decreases exponentially with frequency.* Distortion peaks at frequencies within 0.4 Hz of a fundamental peak may be “merged” with the spectral display of the fundamental since the equivalent analysis filter bandwidth of the digitally synthesized FFT

(Al

(61

Ll""""l

I





1

S/S in -

Fig. 3. ERG nonlinearly properties. Top, ERG amplitude transfer function (XFR) and coherence function (CF) to white noise input. (Mean stimulus intensity, 0 log attenuated glow tube source.) Middle, input (S/S,) and output (S/S,,,) amplitude spectra for sum-of-sines input with 0 long attenuation. Bottom, S/S, and S/S,,, sum-of-sines input with - I log attenuation. Temperature, 18.2”C for all spectra, and O-JO Hz span (linear). Full-field stimulus.

filter is 0.4 Hz (i.e. center frquency iO.2 Hz at I 2 poser points). Using this assumption. 3.2 and 6.4 Hz

difference frequencies and 5.6 and 12.0 Hz second harmonics would be merged with nearest fundamentals. The 25.2 Hz fundamental and the various distortion products generated by interactions with 25.2 Hz. i.e. 12.8. 19.2, 22.4. 28.0, 31.2, 37.6, and 50.4 Hz, are not detectable. Although the 12.8 Hz distortion peak might be considered “merged” with the 12.4 Hz fundamental, a more general hypothesis that explains the absence of the peaks associated with 25.2 Hz interactions is that the input to the assumed nonlinearity does not contain any 25.2 Hz energy, i.e. the S:S input has been low-pass filtered through a retinal “prefilter” that cuts off at a frequency below 25.2 hz. The nature of the retinal nonlinearity and its possible physiological basis. the post-nonlinearity shaping of the S S output spectrum, and the basis for the spectral characteristics in response to an S/S input at - I log attenuation, are considered in the Discussion section. Pharmacological modtjica~ion of ERG properties

Several experiments were performed that were designed to isolate pharmacologically a specific ERG component or a specific retinal functional unit that contributes uniquely to the ERG, and to measure relevant temporal transfer properties. In one case, sodium aspartate was applied by nebuhzer spray to the isolated retina (estimated aspartate concentration at the retinal surface, ca 250 rt M). Figure 4 illustrates the changes observed in the ERG Rash responses and amplitude spectra. The response shown at right is an ERG record in an eyecup preparation from another experiment, for comparison with the ERG recorded from isolated retina before (center) and 4min after (Ieft) aspartate treatment. The ERG components of the isolated retina are inverted relative to those recorded from an eyecup, since the isolated retina is inverted relative to the active electrode. The u-wave is larger than the eyecup ERG a-wave, although both h-waves are approximately equal in amplitude. This probably reflects the greater proximity of the receptors to the active eiectrode. In both the eyecup and control isolated retina, the ERG b-wave exhibits a slight degree of oscillatory behavior on the rising phase and peak. The time courses of all ERG components are similar for the ERG in both types of retinal preparation at the same temperature (ca. 21 ,C). The amplitude spectrum (XFR) of the control isolated retina ERG (Fig. 4, center) exhibits some characteristics typical of an eyecup ERG transfer function (e.g. Fig. 2, top, or Fig. 3, top). It has a broad peak in the region of 1O-1 5 Hz, a slow fall-off at high frequencies, and some hint of a minor peak at low frequencies (ca 2-3 Hz). After aspartate treatment, the b-wave has been completely abolished from the ERG, possibly unmasking underlying oscillatory potentials. The a-wave/slow P-III complex remains, reflecting receptor activity in the retina (mainly red

cones. Fuortes et ui.. 1973). Under these condition,. the broad mid-frequency peak seen in the cnntrol XFR is attenuated. whereas the response in the lowest frequency range may be relatively enh.mced. Dick and Miller (1978) and Dick (19781 hare shown that ethanol (EtOH) sefectiveiy depresses the responses of hyperpolarizing bipoiar cells and ganglion cells, while enhancing depolarizing bipolar ce!t retina. In earlier inresponse in mudpuppy vestigations, Bernhard and Skogiund t.194 1) showed that EtOH diminished the d-wave in frog retina; Negishi and Svaetichin i 1966) demonstrated that EtOH causes horizontal cells to hyperpofarize and lose light sensitivity. Dick (1978) confirmed the action of EtOH on mudpuppy horizontal cells. Gammaaminobutyric acid (GABA) attenuated amacrine cell and ganglion cell activity, in addition to selectively affecting depolarizing bipolar cells (DPBC) compared to hyperpoiarizing bipolar cells (HPBC) (Dick and Miller, 1978). Miller ei ni. (1981) Glickman c~fiti. (1982). and Ariel and Adolph (1984) all showed the response inhibiting action of CABA on ganglion cells. GABA and EtOH, applied together, act synergistically to selectively attenuate proximal extracellular [K+lo relative to distal [K’],. Under these pharmacological conditions, the ERG h-wave, which reflects [K’]O changes in distal and proximal retina via Muller cell responses (Miller and Dowling, 1970) mainly signals distal &+I0 changes dependent on depolarizing bipolar cefi activity (Dick and Miller. 1978). Figure 5 illustrates the results of applying Ringer’s solution containing EtOH (10%) and GABA (20mM) directly to the isolated retina. The control ERG [Fig. 5(Al)] has large u- and b-waves and a slow P-III which is truncated by light-offset and the appearance of a large d-wave composed of an initial positive transient and a slower, more sustained negative component. Early and fate oscillatory potentials appear on the b-wave and slow P-III, respectively. The ERG transfer function [Fig. 5(BI)] is broadly peaked between IO and 20 Hz, with a slow roil-off at high frequencies. No secondary low-frequency peak is apparent. Two minutes after applying the EtOH + GABA spray [Fig. 5(At)f. the h-wave amplitude is attenuated, as is the sustained. negative d-wave component, apparently unmasking a larger u-wave and a positive, light-off response whose leading edge may form the d-wave positive transient. The transfer function [Fig. 5(B2)] has narrowed and peaks at a lower frequency (5-10 Hz). Nine minutes after treatment [Fig. 5(A3)]. the ERG has nearly regained its control appearance, the oscillatory components have not yet reappeared, and the transfer function Fig. 5(B3)] has broadened and shifted its peak to higher frequencies. DISCUSSION

ERG a-wave and slow P-III, which reflect receptor activity, are less temperature sensitive than The

ERG transfer function and nonlinearity tuning

u

f

5

d

>a

t”

:: -..-,

: w” ,.. .I

._

.. .

-...-...-- ~

.._ ._. _ 1. ..

...

u a

5

a

1:

. ............

.+..-g

_. .....

1 ..-.-a

*

389

-

_I-

_.

q--N I 1I-’ I I-

.,

.1-_ - . -. .

1”’

:‘i

; I 1

iI

:

I

2

i I :

Fig. 5, ERG and spectral response properties of retina treated with ethanol (EtGH) plus gammaaminobutyric acid (GABA). Left column, ERG response to 1 SC flash before (Al), 2 min after (A2), and 9 min after (AJ) application of EtOW + OABA (details in text). Right column, amplitude and coherence transfer functions before (Bl), during maximal action (BZ), and after recovery from (B3) EtGH + GABA application. Temperature, 20°C. Calibrations, 5OpV and 0.4 set far all ERG responses, indicated in A3. Spectra amplitude, IO dljdiv, and coherence, 1.0 full-scale. Frequency span, O-50 Hz. Stimulus intensity, 0 log attenuated glow source for all ERG and XFR. Full-Eeld stimulus.

is the b-wave, which reflects Mulltr cell and, indirectly, DPBC activity (Dick and Miller, 1978). These

effects are similar to those seen in isolated carp retina by Armington and Adolph (1984). Studies of the temperature sensitivity of toad rod dark noise (Matthews, 1984) and fiash response kinetics (Lamb, 1984) may suggest the basis, at the sir&e photoreceptor leveil, For the temperature sensitivity of the

receptor-linked ERG components paper. Both dark noise (spontaneous

reported in this thermal isomer-

izations of photopigment) and flash response kinetics (rise, peak duration, and decay times) are highly temperature dependent. Temperature changes evoked only a shift of time scale, not response waveform. Although the receptors contributing a major share of the n-wavefsiow P-111 components are, in these experiments, probably (red) cones (the eyecup is generally functioning under mesopic-phoIopic conditions), the toad rod resuhs may offer a qualitative explanation of the present experimental findings.

ERG transfer function and nonlinearity tuning

b-Wave amplitude and time-course are quite sensitive to temperaK~re. as is time-course of the receptor-linked components. b-wave, for example, is a wavelet af much shorter duration than is slow P-111. and by inference, changes in b-wave might be more strongly reflected in mid- to high-frequency retions of the ERG XFR, whereas slow P-XII changes might influence low- and mid-frequency regions. These inferred changes could explain the variation in ERG XFR shapes seen in Fig. 2. The relatively narrow, low-frequency peak narrows further and attenuates in amplitude as temperature decreases. These changes parallel the slowing of a-and P-IX1 waves. The broad mid-frequency peak narrows as temperature decreases: b-wave attenuates and slows. Perhaps one might even distinguish the contributions to XFR of each receptor-linked component; an XFR for the aspart~te-iso~t~ u-wave was measured and presented in Fig. 4. The isolated a-wave XFR (cone XFR?) is relatively flat from d.c. to ca 10 Hz and falls monotonically as frequency increases above 10 Hz. Toy&a (1974) measured XFRs for carp intracetlufar cone responses as well as slow P-III isolated functionally by means of a coaxial electrode technique. His receptor XFRs have classic low-pass characteristics (Lorentzian shape), Le. flat from d.c. to upper cutoff frequency, which in carp in his experiments was ca 3-5 Hz. The relatively higher cutoff frequency for turtle than for carp is analogous to comparable findings for amacrine celf responses (Adolph, 1983a,b). The ERG transfer function shape of goldfish, measured in rko. was found to be insensitive to temperature by Schellart ef al, (1974). Only the frequency location of a constant-shape XFR shifted to lower values as temperature decreased. This is obviously a disparity from the results presented here. Although Schellart er al. did not present flash ERGS vs temperature, it is unlikely that the changes seen in turtle do not occur to some extent in goldfish..The major difference between their experimental condition and the present one is that they measured the ERG in cico in unanaesthetjzed goldfish. The turtle ERG was recorded in an isolated, oxygenated eyecup. Nature of tire nonlirzearitq

Nonlinearity, defined as the presence in the amp& tude spectrum of distortion product frequencies (i.e. sums, differences, harmonics, etc.) appears to be a function of stimulus intensity. The transfer function also exhibits a measure of intensity dependence as inferred from the envelope of the discrete peaks, both fundamental and high-order, which changes its spectral configuration between the high-intensity (0 log) and Iow-intensity (- I log) stimulus conditions. If the total signal path Bows through a single nonlinearity within the retina, then the nonlinearity cannot be a purely “power-law” function y = Ax”, where b 3 2. There are outputs measured at the fundamental frequencies, 2.8, 6.0, and I2.4Wz. The output from a pure “power-law” nonlinearity wouId

491

contain on& sum and digerence frequencies and “b th order” harmonics. A single non~~nea~ty would have to be of the form .V= Ax + Bx’ in order to have outputs of the type found at 0 log attenuation. Alte~ativeIy~ the signal path co&d split into (at least) two paraflel paths, one of which is linear and the other containing a “power-law” nonlinearity. The latter configuration is unlikely as the functional information path from photoreceptor input to ganglion cell output since, in general, aft ~nfo~ation passing through the outer retinal elements (primarily linear} must encounter inner retinal elements that are essentially non&near. However, for the components of the ERG no such constraints exist. The b-wave, for example, may simuitaneo~sly reflect the activity of distal retina (linear elements) and proximal retina (nonlinear elements} via Nulier cell responses to distal and proximal [K’;t, changes. The parallel (linear/nonli~ear) pathway model may explain the response to - 1 log attenuation of the input. There, only the two lowest frequency fundamentals (2.8 and 6.0 Hz) are passed, and the outputs have equal amplitudes. If the attenuated input response is primarily reflected by activity in the distal retina, the linear elements, then we might expect ERG component only at the fundamental frequencies. The cquaf amptitudes of the outputs at the two lowest frequencies suggest a pre- or postnonlinearity filtering that is flat up to some frequency between 6.0 and 12.4 Hz, and attenuates sharply at higher frequencies, thereby eliminating the output at the two highest fundamental frequencies. The experimental findings concerning the temperature ~nsitivity of the turtle retina as reflected by changes in ERG properties, and their imp~jcatio~s about the functional stability of various intraretinal pathways under varying temperature conditions, are intriguing. As a pojki~otherm, the turtle’s retina, and indeed its entire nervous system, is subject to environmental temperature ~uct~ations. One cannot escape concluding that the processing of a specific pattern of visual stimuli by the retina, the further processing of this retinal message by the visual CNS, and any subsequent responses initiated and effected by the CNS vary with temperature. Perhaps the visual CNS compensates for this in order to maintain perceptual stability in a thermally varying environment. Alternatively, although intuitively unlikely, perceptual constancy may not be an important aspect to poikiiotherm existence. Acknowledgemenr-Supported 03383.

by NE1 grant ROI EY

REFERENCES

Adolph A. R. (1983a) Temporal transfer properties &turtle retina: Tuning by temperature and stimulus intensity. lnvesf Oph. thufmoi. visual Sci., AR VO Suppl. 24, 220.

Adolph A. R. (1983b) Temporal transfer properties of intraretinal pathways in turtle: Effects of temperature, spat size, and intensity. Sot. Neurusci. Abstr. 9, 686.

Ariel .Lt. and Adolph A. R. ( 1YS-t) Synaptic pharmacology of directionally sensttite ganglion cells in turtle retina. inwsr. Ophdd. ~isuui Sri.. .-1RIZ) Suppl. 2% I IO. Armington 1. C. and ,-\dolph A. R. (198-t) Temperature effects on the el~troretinogram of the isolated carp retina. .-fcru ~ph~~~~.62, 498-509. Bendat J. S. and Picrroi A. G. (1571) &r&m Data: ilnaiws anti .Lleaszrremern Prace&es. Wiley-Interscience, New York.

Bernhard C. G. and Skoglund C. R. (1941) Selective suppression with ethyi alcohol of inhibition in the optic nerve and the negative component PI11 of the electroretinogrdm. .4cfu physiol. swnd. 2, IO-2 1, Dick E. (lY79) Light- and dark-dependent extracellular Kactitity modulations in the vertebrate retina: origins of el~troretinographic components. p. 310. Ph.D. thesis. SUNY Buffalo. SY. Dick E. and Miller R. F. (19X) Light-evoked potassium activity in mudpuppy retina: Its relationship-to the bwave of the electroretinoeram. Brtrin Res. 1.54. 388-394. Fuortes M. G. F., Schwartz E. A. and Simon E. J. (1973) Colour-dependence of cone responses in the turtle retina. J. Plykd. 234, 199-2 16. Glickman R. D., Adolph A. R. and Dowhng J. E. (1982) Inner plexiform circuits in the carp retina: Effects of cholinergic agonists, GABA, and substance P on the ganglion cells. Bririn Res. 231, 81-99. Granit R. (1947) Sen,so~~ .Weclmrri.tnr.r o/ tire Rerinn. Oxford Univ. Press.

Lamb T. D. I IYN) EtTects of temperature .:hdng.-‘\ ‘I: :.: tII rod photocurrents. J, Ph_tsid. 346, SE-‘--S’s, Matthews G. (193-4~Dark noise in the outc: \cgrn-r:i membrane current of green rod photoreceptors in m ?;).I;: retina. J. Physioi. 349, hi)?--618 Miller R. F. and Dealing J. E. [ iy_V) Iniractllt~t~r responses of the Muller (ghal) cells of mudpuppy r:tma. their relation to the h-vvave of the rlectroretinogrsm. J. ,Veuroph,vsioi. 33, !23-3-t I. [Miller R. F.. Frumkes T. E.. Slaughter >l. and Dacheaw R. F. (1981) Physiological and pharmacological basis of GABA and glycine action on neurons of mudpuppy retina. Il. Amacrine and esnplion cells. J. .~re’errroah~,sir,;. _ 45, 763-752. Negishi K. and Svactichin G. ( 1966) Eti’ecrs of alcohais and vofatiie anesthetics on S-potential producing cells an3 on neurons. P$ugers .-1rch. ges. Physioi. 292, 21%2%. Negishi K. and Svaetichin G. (1966) Elects of temperature on S-potential producing ceils and on neurons. !‘flfli~~erv Arch. yes. Physiol. 292, 206-217. Schellart N. A. 41.. Spekreijse H. dnd van den Berg T. J. T. P. (1974) Influence of temperature on retinal ganglion ceil response and E. R. G. of goldfish. .I Ph_~~-ird. 238. 25 l-267. Toyoda J.-I. (1974) Frequency characteristics of ret1m11 neurons in the carp. f.