Effects of Hypoxemia on the a- and b-Waves of the Electroretinogram in the Cat Retina

Abstract

purpose. Slow components of the electroretinogram (ERG) are sensitive to even mild hypoxemia (60 < P_a o ₂ < 100 mm Hg) in the cat eye. However, the electrical responses of the inner retina remain unchanged until P_a o ₂ is below 40 mm Hg. In this study, the effects of hypoxemia on photoreceptors, on which both slow ERG components and inner retinal activity depend, were examined by recording the a-wave of the ERG.

methods. The ERG of dark-adapted, anesthetized cats was recorded between an Ag-AgCl electrode in the vitreous humor and a reference electrode near the eye. Responses to bright flashes of diffuse white light were recorded at 3-minute intervals during hypoxemic episodes lasting 15 minutes to 2 hours.

results. The cat a-wave was well described by the Lamb and Pugh a-wave model during normoxia and hypoxemia. During mild hypoxemia (P_a o ₂ of 50–60 mm Hg), small changes in a-wave amplitude were detected but did not become greater during severe hypoxemia. The mean decrease in the a-wave amplitude during severe hypoxemia (P_a o ₂ of 20–30 mm Hg) was 8.9% from the mean amplitude during air breathing. The effects of hypoxemia were more severe on the b-wave amplitude. The mean decrease in the b-wave was 35% at P_a o ₂ of 20–30 mm Hg.

conclusions. The a-wave is more resistant to severe hypoxemia than the b-wave. This implies that photoreceptor transduction works almost normally during hypoxemia and that failure of inner retinal Po ₂ regulation causes the decrease in the b-wave. Previously observed changes in the amplitudes of slow ERG components during hypoxemia may result from changes in the ionic environment, rather than a failure of photoreceptor energy metabolism.

Previous studies of the effects of hypoxemia (decreased blood Po ₂) on retinal electrical signals can be divided into two categories. The first category concerns hypoxemic effects on the inner (proximal) retina, which have been studied by evaluation of the amplitudes of the electroretinogram (ERG) b-wave, scotopic threshold responses (STR), and ganglion cell sensitivity in the cat retina. All these measures are resistant to hypoxemia until P_a o ₂ is close to 40 mm Hg. ^{1 2 3 4} When the ERG b-wave was examined during severe hypoxemia (P_a o ₂ below 40 mm Hg) a large decrease in amplitude was observed. It is believed that the inner retina is unaffected during hypoxemia, at least for P_a o ₂ above 40 mm Hg, due to the efficient regulation of inner retinal Po ₂ by the retinal circulation (e.g., References 1, 5, 6). The second category concerns hypoxemic effects on the outer (distal) retina, which have been studied by recording the slow ERG components that arise from the retinal pigment epithelium (RPE). In cats, the slow ERG components, including the c-wave, fast oscillation, and light peak are sensitive to hypoxemia beginning at P_a o ₂ as high as 60 mm Hg. ^{2 3} The photoreceptor, which initiates the RPE responses, was believed to be the primary site of hypoxemic changes based on measurements of [K⁺] _o in the subretinal space ⁷ and of photoreceptor oxygen consumption. ⁶  

It is unclear why the outer retina is sensitive to hypoxemia, but the inner retina is resistant, because the inner retinal responses are dependent on signals from the photoreceptors. A key element missing from previous studies was direct evaluation of photoreceptor electrophysiology. For more than 60 years, it has been known that the leading edge of the a-wave is associated with photoreceptor activity. ^{8 9} However, the a-wave could not be used in a quantitative way to reveal receptor response properties until Hood and Birch ¹⁰ and Lamb and Pugh ¹¹ proposed a computational model of the rod response. Breton et al. ¹² demonstrated that the a-wave exhibits the same kinetics and amplification as the photocurrents of single primate rods. The Lamb and Pugh ¹¹ model fits the a-wave reasonably well, giving the fraction of circulating current (i.e., photocurrent) as a function of time in terms of an amplification constant, an effective delay time, and the number of photoisomerizations. 

The purpose of this study was to examine the effects of hypoxemia on dark-adapted photoreceptors in the cat retina by recording the a-wave in response to bright flashes. The a-wave model parameters were analyzed to examine phototransduction during normoxia and hypoxemia. The a-wave was also compared with the b-wave to simultaneously evaluate the effects of hypoxemia on the outer and inner retina. 

Methods

Animal Preparation

All experimental procedures were in accordance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. Adult cats were anesthetized initially either with thiopental (5% solution, 0.35 ml/kg intravenously) or with ketamine/acepromazine (25 mg/kg; 0.12 mg/kg intramuscularly) if the cat was difficult to handle. Atropine (0.3 mg subcutaneously) was administered after the induction. Urethane (200 mg/kg loading dose followed by 20 to 40 mg/kg per hour) was used throughout the experiment as the long-term anesthetic. The surgical preparation consisted of inserting two venous cannulae for drug infusion, a femoral arterial cannula for sampling blood and monitoring arterial pressure, and a tracheal cannula for respiration. The temporal side of the right eye was exposed by removing bone and cartilage. The eye was mounted on a stainless steel eye ring by attaching the conjunctiva. Topical atropine (1%) and phenylephrine (1%) were administered to dilate the pupil. Topical flurbiprofen sodium ophthalmic solution USP (0.03%; a prostaglandin inhibitor; Ocufen; Allergan, Irvine, Ca) was administered every 5 to 6 hours to prevent pupil constriction. The cat was then paralyzed with pancuronium bromide (0.2 mg/kg per hour; Sigma, St Louis, MO) and was artificially ventilated. The temperature was monitored by a rectal temperature probe and kept at 38°C to 39°C using a heating pad. The electrocardiogram and arterial blood pressure were monitored throughout the experiment. The arterial pH, Pco ₂, Po ₂, blood glucose, and blood lactate were monitored by a blood gas analyzer (model 860; Chiron, Norwood, MA). The stroke volume of the respirator was adjusted so that during air breathing, pH was between 7.35 and 7.45, Pco ₂ was approximately 30 mm Hg, and Po ₂ was higher than 90 mm Hg. The blood glucose was controlled with a combination of intravenous porcine insulin (2.5 mU/ml; Iletin; Eli Lilly, Indianapolis, IN) and glucose (40%, Sigma; St. Louis, MO) to achieve normoglycemia (range, 90–120 mg/dl). Hypoxemia was induced by inspiration of a mixture of N₂ and air. 

Visual Stimulation

Bright flashes of diffuse white light, produced by a photoflash (Model 283; Vivitar, Santa Monica, CA) were used to elicit the a- and b-waves. Stimuli were reflected to the eye by a ganzfeld hemisphere 30 cm in diameter with an interior surface coated with reflective white paint. Flash intensity was varied over 6 log units by means of neutral density filters (Eastman Kodak, Rochester, NY) available in steps of 0.5 log units. At 0 log units, the intensity of the flash at the cornea was 28,057 scotopic candelas (cd)-seconds per square meter (measured with a photometer; model 40X; UDT, Hawthorne, CA) in photopic units and converted to scotopic units by assuming the spectral distribution of the xenon-filled flash tube). ¹³ The conversion from the intensity of the flash at the cornea to the number of photoisomerizations per rod per flash (Φ), which was used in the a-wave model, was performed according to Breton et al. (see Reference ¹² , equation 13) with a few corrections for cat eye. The Φ is a product of Q, the retinal illuminance in scotopic troland-seconds (scot td-sec) and K, the overall conversion factor in photoisomerizations/rod per scot td-sec. The pupil area was 113 mm², and retinal illuminance Q in our study was therefore 3.17 × 10⁶ scot td-sec (or 12.2 log q/deg²). The conversion factor K was adjusted for the cat eye by considering the following three factors: 1) the smaller loss in media and tapetal reflection, a difference of 0.24 log units from human ¹⁴ ; 2) the smaller posterior nodal distance in cat, differing 0.26 log units from human ¹⁴ ; and 3) the smaller area of the rod outer segment, assumed to have an average diameter of 1.3 μm in cat. ¹⁵ The value of K was then 27.13 photoisomerizations/rod per scot td-sec. Our maximum unattenuated retinal illuminance corresponded to 7.93 log photoisomerizations per rod per flash. To test for a cone contribution, responses were obtained to short-wavelength flashes (Wratten 47B filter; Eastman Kodak, Rochester, NY) and long-wavelength flashes (Wratten 26 filter). 

Recording

The ERG was measured between a vitreal electrode and a reference electrode behind the eye. The vitreal electrode was a chlorided silver wire in a 0.5-mm glass tube. A 20-gauge needle was inserted in the eye to hold the Ag-AgCl vitreal electrode in place. The reference electrode was a chlorided silver plate, which was sewn into tissue near the eye. Potentials were amplified (DC to 3 kHz, M4A Amplifier; WPI, Sarasota, FL) and displayed on an oscilloscope. The responses were then digitized at a rate of 5 kHz by an analog-to-digital board in a computer. The flashes were triggered by the computer, and the data were digitized by use of commercial software (Epic-XL; LKC, Gaithersburg, MD). After each flash, the eye was allowed to dark adapt for at least 3 minutes. All responses shown were to single flashes, not averages. 

a-Wave Analysis

The responses to all flash energies were fitted to equation 1 based on the Lamb and Pugh model: ¹¹  

\[F(t){=}\mathrm{exp}{[}\mathrm{-}\ \frac{1}{2}\ {\Phi}\ A(t-t_{\mathrm{eff}})^{2}{]}\]

with t > t _eff and where F(t) is the cGMP-activated current expressed as a fraction of its dark value, Φ is the number of photoisomerizations per rod, t _eff is a brief delay, and A is an amplification constant. When applied to the a-waves, F(t) is equivalent to the normalized a-wave from flash onset to the peak of the a-wave:  

\[a(t){=}{[}1-F(t){]}a_{\mathrm{max}}\]

where a(t) is the a-wave response, and a _max is the a-wave amplitude in response to a flash that saturates the amplitude. All the fitting was performed on computer (Photran software; LKC). Equation 2 was fitted to a group of responses to obtain values of A and t _eff that produced the best fit. The measured a _max as well as the fitted parameters A and t _eff were compared to determine the effects of hypoxemia on the photoreceptors. 

Double-Flash Experiments

Two photoflashes (Vivitar) were used to elicit two a-wave responses to test recovery from a bright flash during hypoxemia (e.g., References 16, 17). Both stimulus flashes were reflected to the eye by the ganzfeld. Flash intensity was varied by means of neutral-density filters. A test flash (5.93 log photoisomerizations per rod per flash) was given at time 0, and a probe flash followed at a predetermined interstimulus interval (ISI). The probe flash (6.93 log photoisomerizations per rod per flash) was used to produce amplitude saturation of the a-wave. The photoflashes were triggered by computer (LabTech Notebook software running within a Visual Basic program; Microsoft, Redmond, WA). The ISI was controlled by the computer. The ISI used during hypoxemia was determined in a preliminary experiment in which the probe flash was given at varying times after the test flash, usually ranging from 500 to 3000 msec. The responses were digitized at a rate of 500 Hz by an analog-to-digital board. After each pair of responses, the eye was allowed to dark adapt for 3 minutes. 

Results

a-Wave during Normoxia

Figure 1A shows the a- and b-waves of the ERG of a normal, dark-adapted cat retina. The a-wave amplitude was measured from the baseline to the a-wave trough, and the b-wave amplitude was measured from the a-wave trough to the peak of the b-wave. The figure illustrates graded increases in the amplitude and the rate of rise of the a-wave as the flash attenuation was varied from 2.93 to 7.93 log photoisomerizations per rod per flash. The a-wave amplitude saturation was observed at 6.93 log photoisomerizations per rod per flash for all cats, whereas the b-wave saturated at approximately 4.93 log photoisomerizations per rod per flash. Oscillatory potentials can be observed on the rising phase of the b-wave, but were not analyzed. 

Figure 1B shows the a-wave responses, illustrated in Figure 1A , normalized to the saturated a-wave amplitude, a _max which was obtained at 6.93 log photoisomerizations per rod per flash. Equation 2 was applied to the data simultaneously from 2.93 to 4.93 log photoisomerizations per rod per flash and is shown as dashed lines. The values for A and t _eff were 0.83 seconds⁻² and 4.13 msec, respectively. The responses at the highest intensities were not well described with the parameters predicted from this simultaneous fit. Instead, the responses to the three highest intensities were fitted individually (for parameter values, see figure legend). When analyzing the biochemical cascade, it was expected that A would be independent of Φ, but at the higher intensities, A and t _eff declined in human ^{11 12} and rat. ¹⁸ In cats, A also declined at higher intensities (Fig. 2) . 

Table 1 shows the a-wave model parameters for eight cats during normoxia at the fixed intensity used to track the responses during hypoxemia. The mean maximum a-wave amplitude produced by the saturating flash was −550 ± 111 μV. The mean values of A and t _eff for 4.93 log photoisomerizations per rod per flash were 0.4 ± 0.2 seconds⁻² and 2.30 ± 0.9 msec, respectively. For 5.93 log photoisomerizations per rod per flash, A and t _eff were 0.21 ± 0.04 seconds⁻² and 2.16 ± 0.2 msec, respectively. Because A and t _eff decreased at high intensity, the values of A and t _eff presented in Table 1 were averaged at each intensity level in the last two rows of the table. 

Cone Contribution

The data presented in this study are responses to white light. At least in humans, a cone contribution adds to the signals from rods so that the cone component of the responses must be subtracted to study rods alone. ¹⁰ Cone intrusion was examined by obtaining amplitude versus intensity data for blue, red, and white flashes. These are plotted in Figure 3 for one cat. Each curve was fitted to the Hill equation, \(R(I){=}\ \frac{R_{\mathrm{max}}I^{n}}{I^{n}{+}{\varsigma}^{n}}\) where R(I) is a-wave peak amplitude, I is intensity, R _max is the maximum a-wave amplitude, and ς is half saturation. The Naka–Rushton equation (n = 1) did not fit the peak amplitude data well, although it fit when the a-wave was measured at a fixed time after the flash (not shown). The average Hill equation parameter ς was 4 ± 0.3 log units attenuation for white light (for three cats), 2 ± 0.2 log units attenuation for blue light (for three cats), and 1 log unit attenuation for red light (for one cat). The responses to white, blue, and red light were separated from each other by a fixed amount over the entire range, suggesting that there was no Purkinje shift and no significant cone contribution. When a response to red light at 0 log units attenuation was subtracted from a response to blue light at 1 log unit attenuation, the subtracted response was virtually zero giving more evidence that there was little or no cone intrusion. The inset to Figure 3 shows the superimposed responses to blue (1 log unit) and red (0 log units) lights. These data indicate that the responses recorded were the rod ERG uninfluenced by a cone component. 

Hypoxemia Experiments

For the hypoxemia experiments, the a-wave was measured at a fixed intensity of either 4.93 or 5.93 log photoisomerizations per rod per flash to allow comparison between control and hypoxic a- and b-waves. Values of P_a o ₂ higher than approximately 90 mm Hg were considered to be normal. The protocol in most of these experiments was to reduce inspired oxygen gradually, maintaining P_a o ₂ at each level for 15 minutes. Figure 4 shows the a- and b-wave amplitudes plotted over time for cat 214. Each symbol represents a level of hypoxia or control. The a-wave amplitude is plotted as positive for easier comparison with the b-wave amplitude. There was often a small decrease in the a-wave at the onset of hypoxia, as illustrated in Figure 4A . However, the a-wave was then stable with decreasing P_a o ₂ until the inspired O₂ was reduced to 3% in this example (P_a o ₂ of 29 mm Hg). The a-wave amplitude recovered to near normal after hypoxic episodes. The b-wave amplitude (Fig. 4B) was also stable with decreasing P_a o ₂ until 4% O₂ (P_a o ₂ of 31 mm Hg) was reached. However, the hypoxic effect on the b-wave was much more dramatic than that on the a-wave (Fig. 4A , inset). The b-wave amplitude did not reach steady state at very low P_a o ₂, in contrast to the a-wave, whose amplitude stabilized. The b-wave amplitude also recovered to near normal values after hypoxic episodes. 

Figure 5 shows the relative a- and b-wave amplitudes as a function of P_a o ₂. Each point represents the mean a- or b-wave in one episode of hypoxemia from eight different cats. The amplitudes were normalized to the maximum mean amplitude during air breathing. As noted in the example of Figure 4 , small changes in the a-wave were sometimes detected at P_a o ₂ in the 40- to 60-mm Hg range, but these did not generally become greater at lower P_a o ₂ (Fig. 5A) . A paired t-test was performed to determine whether these responses during mild hypoxemia (P_a o ₂ of 50–60 mm Hg) were different from the normoxic mean responses, and there was a small but significant difference between the two levels (P = 0.04, n = 4). The a-wave responses during severe hypoxemia (P_a o ₂ of 20–30 mm Hg) decreased an average of 8.9% ± 6.8% (paired t-test, P = 0.007, n = 7). The largest decrease in the a-wave amplitude during severe hypoxemia was 14.7% from the amplitude during air breathing. The b-wave was resistant to hypoxemia until P_a o ₂ was less than 40 mm Hg, confirming previous work (Fig. 5B) . ^{2 3 4} The effects of hypoxemia were more severe on the b-wave amplitude at low P_a o ₂ than on the a-wave amplitude. The b-waves shown in Figure 5B are averages of the five responses obtained during hypoxemia, and at more severe levels of hypoxemia, where the b-wave was not stable, this underestimates the changes. The average b-wave always decreased by at least 40% during the most severe hypoxemia. 

In some cases it was possible to obtain a complete intensity–response function during hypoxia. The a-wave amplitudes as a function of flash attenuation were fitted to the Hill equation before and during hypoxemia. As expected from the data in Figure 5 , there was a small decrease in the a-wave amplitude during hypoxemia, which was observed at all intensities. Table 2 summarizes the fits of the Hill equation to the a-wave amplitude as a function of intensity before and during hypoxemia for five cats. Paired t-tests showed that R _max decreased during hypoxemia (P = 0.02, n = 5 cats), whereas n and ς were not significantly different. 

To determine whether there might be a delayed effect of hypoxemia that was not detected with 15-minute steps of decreased oxygen, the P_a o ₂ was held at a moderate level (42 mm Hg) for 2 hours in one cat. The initial decrease in the a-wave amplitude (∼14%) was observed a few minutes after the onset of moderate hypoxemia, but the a-wave was then stable over time, implying that 15-minute episodes of hypoxemia allowed a good estimate of the a-wave during hypoxia and that the photoreceptor can function almost normally during moderate hypoxemia for long periods. The b-wave was also stable at this level of hypoxemia. This indicates that the changes in the a- or b-wave are directly related to the P_a o ₂, and not the duration of the hypoxemic episode. 

The Lamb and Pugh Model in Hypoxemia

The Lamb and Pugh model ¹¹ was applied to determine the values of A and t _eff during hypoxemia. Table 2 shows the parameters of the Lamb and Pugh model for five cats. At each level of hypoxemia, five responses to a flash at a fixed attenuation (5.93 log photoisomerizations per rod per flash) were compared with the control (normoxic) a-wave fits except in cat 205 (three responses to 4.93 log photoisomerizations per rod per flash) and cat 209 (five responses to 4.93 log photoisomerizations per rod per flash). Both the normoxic and hypoxic a-waves were well fitted by the Lamb and Pugh model. The overall amplitudes decreased significantly as the percentage of oxygen decreased; however, the change was small enough that the values of A and t _eff were not different from the normoxic a-wave fits. The rate of rise of the a-wave did not change with lowering of the percentage of oxygen. The time to peak and the shape of the a-wave also did not change during severe hypoxemia, whereas the b-wave was essentially gone in some cats (Fig. 4A , inset). 

Double-Flash Experiments

To study the full time-course of the rod flash response, the double-flash or paired-flash method was used to examine the recovery kinetics of the rod ^{16 17} during hypoxemia. Figure 6 (inset) shows the a-wave responses to the probe flash at various ISIs. When the ISI was less than 1500 msec, almost no probe a-wave was observed. For the hypoxemia experiments, an ISI of 2500 msec was used. As in the hypoxemia experiments described earlier, 15-minute steps of hypoxemia were used, but paired flashes rather than single flashes were issued at 3-minute intervals. 

As indicated by Figure 3 , the test responses are rod a-waves with no cone contribution. However, due to the cone’s faster recovery after strong light, the probe responses were compared with the single flash responses to check for any cone contribution. There were no differences between the waveform of the probe responses in the double-flash procedure and the single-flash responses, suggesting that cone responses were absent. Figure 6 shows both the test and probe a-wave amplitudes as a function of P_a o ₂, each normalized to its own normoxic amplitude. The amplitude of the test a-wave decreased by less than 10% over the P_a o ₂ range, consistent with the result in Figure 5 . The a-wave in response to the probe flash tended to increase in amplitude at lower levels of P_a o ₂, but the changes were generally not significant when compared with the normoxic probe response (t-test). The probe b-wave behaved similarly to the test b-wave during hypoxemia, in that the amplitudes decreased with the severity of hypoxemia. 

Discussion

Normoxia

The present work applied the Lamb and Pugh ¹¹ a-wave model to characterize the a-wave of the cat retina and to examine the effects of hypoxemia. The cat a-wave is rod-dominated and does not require subtracting a cone component. The present work is the first to examine the cat a-wave using the Lamb and Pugh model and can be compared with a study by Breton et al. ¹² on the human a-wave. The leading edge of the a-wave in cat appears to fit the model as well as the human a-wave. When the feline intensity series were compared (Fig. 1B) to the human a-wave intensity series, similar a-wave characterization was observed, but there were a few differences in the a-wave model fits. The cat pupil is approximately 12 mm in diameter, whereas the maximum pupil diameter in humans is approximately 7 mm ¹² ; therefore, for the same intensity at the cornea, the cat has a higher retinal illuminance (Q). In addition, the overall conversion factor K (Breton et al., ¹² equation 13), which converts Q to units of photoisomerizations per rod per flash, needed to be adjusted from the human conversion factor K, as described in the Methods section. The maximum Φ in the Breton et al. ¹² work was 6.08 log photoisomerizations per rod per flash, whereas in the present work the maximum Φ was 7.93 log photoisomerizations per rod per flash. 

The amplification (A) is constant at low intensities, but at high intensities it declines because of saturation in the rate of rise of the photocurrent in human ^{11 12} and rat. ¹⁸ A similar result was obtained in cat (Fig. 2) . The maximum value of A was 5 to 10 seconds⁻² in humans, ¹² whereas in rats the maximum was 2 seconds⁻². ¹⁸ The value of A for monkey was 4 to 7 seconds⁻² (Pugh and Lamb ¹⁹ estimated from data of Baylor et al. ²⁰ ). The maximum value of A in our cat data was approximately 0.7 seconds⁻². Pugh and Lamb ¹⁹ reported that A for mammalian rods at body temperature (37°) was 4 to 10 seconds⁻² and that of amphibian rods at room temperature (22°) was 0.04 to 0.1 seconds⁻². There could be several reasons for the difference between our value for A and that of Pugh and Lamb. ¹⁹ First, when Pugh and Lamb ¹⁹ estimated A for mammals (using experiments from other studies) they assumed t _eff to be 5 msec. Breton et al. ¹² and our calculations showed that t _eff was between 2 and 3 msec for both cat and human. The higher t _eff used by Pugh and Lamb ¹⁹ could explain the higher A (see equation 1 ). Second, the value of A that we obtained represents the intact cat eye, whereas in the other mammalian work (monkey ²⁰ and cat ²¹ ) Pugh and Lamb estimated the amplification constant from isolated retina. Finally, the conversion factor K was corrected for cat in this study, but it is not clear that the correction was made in the Pugh and Lamb ¹⁹ summary. It is also possible that there may be some error in our conversion factor K, although we used the best values available. 

Hypoxemia

Previous studies of hypoxic effects on the retina did not examine the photoreceptor activity directly. The data presented here provide the first information on the electrical activity of photoreceptors and on phototransduction during graded hypoxia. The present work showed that there was a small change, a decrease on average of approximately 9%, in the a-wave amplitude at P_a o ₂ between 50 and 60 mm Hg. At this level there was no change in the b-wave. For P_a o ₂ between 20 and 30 mm Hg, a large change in the b-wave amplitude was observed, but the a-wave amplitude never decreased by more than 14.7%, indicating that the a-wave is more resistant to hypoxemia than the b-wave. The Lamb and Pugh model was applied to the hypoxic a-wave and even at very low P_a o ₂, the parameters were not different from those in normoxia (Table 2) except for a small decrease in a _max. The intensity–response functions showed little change at any intensity. The preservation of a-wave amplitude and the parameters of the Lamb and Pugh model suggest that transduction works almost normally during severe hypoxemia. 

The recovery of the photoreceptor response after an intense flash was examined by using a double-flash technique. We hypothesized that, although hypoxemia had little effect on the flash response, it might impair the ability of the photoreceptor to recover from a flash. However, rather than being impaired, the probe a-wave amplitude was also resistant to hypoxemia. 

There are a few concerns that should be addressed at this point. The first is that the b-wave amplitude may be overestimated in Figure 5 , because it often did not reach steady state at the most severe level of hypoxemia studied. However, the a-wave was not overestimated, because its amplitude was relatively stable even at low P_a o ₂. Thus, the difference between the a- and b-waves was even greater than suggested by Figure 5 . It would be useful to know at what Po ₂ the a-wave deteriorates, but the cats could not tolerate more severe hypoxemia. Second, the a- and b-waves were stable over long periods (2 hours) during moderate hypoxemia, indicating that any changes in the amplitudes were due to the changes in oxygen level and not the duration of the hypoxemic episode. Third, it is possible that there is a regional failure of photoreceptors during hypoxemia, which may contribute to a decrease in the maximum amplitude. The a _max may have decreased because some photoreceptors are affected more by hypoxemia than others or because each photoreceptor was affected a little by hypoxemia. Currently, there is no way to evaluate the regional loss of the photoreceptors. Fourth, there may be a concern that the a-wave appeared resistant to hypoxemia, because flashes nearly saturated its amplitude, but intensity–response data (Table 2) show that the effect of hypoxemia was small at all intensities. Finally, there was a concern that a reduction in the photoreceptor response might actually be more substantial than the a-wave indicates, because the a-wave change in hypoxemia might be obscured by a reduction in the b-wave. Rodieck ²² proposed this as one explanation for failure to see a change in the a-wave during anoxia. This cannot explain the current results because in response to intense flashes the a-wave peaks well before the b-wave starts ²³ and there is no evidence that the b-wave obscures the a-wave peak. Instead, the results are consistent with those of Noell ²⁴ and Granit ⁸ who found that even during anoxia there was relatively good preservation of the a-wave. 

Possible Mechanism

Both the a- and b-waves were resistant to hypoxemia, but the b-wave was more affected by low P_a o ₂. It can now be suggested that when the b-wave decreases, it does so not because of a failure of the photoreceptor signals that initiate it, but more likely from a failure of the retinal circulation to maintain oxygenation of the inner retina. It is possible that the failure of the b-wave does not result directly from an effect of hypoxemia on Müller cells or bipolar cells but from a failure of synaptic transmission in the outer plexiform layer. There are currently no data that would allow us to distinguish between these possibilities. In either case, the decrease in the b-wave would be tied to changes in retinal circulation, because the inner nuclear and outer plexiform layers both rely on retinal circulation. Linsenmeier and Braun ⁶ and Enroth–Cugell et al. ¹ showed that the inner retinal Po ₂ during normoxia was low (18 ± 12 mm Hg), but that it was well regulated during hypoxemia. Mean inner retinal Po ₂ was significantly affected only at P_a o ₂ below 45 mm Hg, which corresponded with the point at which the b-wave and ganglion cell sensitivity begin to break down in this and earlier studies. ^{1 4 25}  

Perhaps the most puzzling aspect of the current findings is that we know that photoreceptor oxidative metabolism decreases, ⁶ and slow ERG components, which are dependent on the photoreceptors, change over a range of P_a o ₂ when there is little change in the a-wave. ³ Linsenmeier and Steinberg ⁷ suggested that slowing of the Na⁺-K⁺ pump was responsible for changes in potassium activity during hypoxemia, which eventually led to changes in the slow ERG components. It now seems necessary to postulate that when oxidative metabolism is affected, the photoreceptors switch more to glycolysis for energy production. Linsenmeier and Braun ⁶ calculated that the increase in adenosine triphosphate (ATP) production from glycolysis could potentially compensate for the decrease in oxidatively derived ATP production. 

The present data may not require a fundamental alteration in our understanding of the mechanism of changes in the slow ERG components during hypoxemia. It is possible that the changes in[ K⁺] _o are due to changes in the mechanism of ATP production during hypoxemia, rather than the amount of ATP production, as was previously thought. Yamamoto and Steinberg ²⁶ showed that systemic hypoxia further acidifies the extracellular space outside rods in dark-adapted cat retina, a region that is already acidic in normoxia. Yamamoto et al. ²⁷ suggested that the acidity found outside rods is from rod glycolysis. This compensatory mechanism, an increase in ATP production by glycolysis that is needed to maintain rod function, may be responsible for keeping the a-wave unchanged when the slow ERG components are affected during hypoxemia. The slow ERG components may change because of acidosis accompanying hypoxemia, rather than because of hypoxemia per se. Respiratory acidosis alone is known to cause similar changes in the standing potential, c-wave, and light peak to those caused by hypoxemia. ^{3 28} Perhaps the change in[ K⁺] _o during hypoxia occurs not because of a change in the pump rate but rather because of other changes that occur in the subretinal space during hypoxemia. One possibility is that there may be ionic redistribution that could occur because of the H⁺ change. The exact mechanism of ion redistribution is unknown. 

A more likely possibility is that the compensation for the loss of oxidative metabolism is not quite complete, allowing a small change in[ K⁺] _o . The increase in[ K⁺] _o during hypoxemia may account for the small changes in the a-wave amplitude during mild hypoxemia. When the fractional changes in the photoreceptor light response in hypoxia (due to changes in K⁺) were calculated, there was a small change in rod voltage responses (for calculation, see Appendix). There was an approximately 5% change in the calculated membrane potential, which could account for the small changes (average of 3.4% ± 2.8%) in the a-wave amplitude during mild hypoxemia. The b-wave might not be altered by this small change in rod voltage responses, because the amplification at rod synapses is large, and the full b-wave can be produced with only a small rod hyperpolarization. 

To test the idea that glycolysis compensates for the loss of oxidative metabolism, the a-wave was recorded during hypoglycemia, which will be discussed in a future article. 

Appendix 1

Linsenmeier and Steinberg ⁷ suggested that the slowing of the Na⁺-K⁺ pump was responsible for changes in potassium activity during hypoxemia. Herein we show that a small change in[ K⁺] _o during hypoxemia may account for the small changes in the a-wave amplitude during mild hypoxemia. 

Let \(z{=}\ \frac{g_{\mathrm{Na}}}{g_{\mathrm{K}}}\) in the dark. To a first approximation, rod membrane potential in the dark can be expressed by the chord conductance equation ²⁹  

\[V_{\mathrm{m,dark}}{=}\ \frac{z\ E_{\mathrm{Na}}{+}E_{\mathrm{K}}}{1{+}z}\]

where g _Na and g _K are conductances to Na⁺ and K⁺, and E _Na and E _K are the equilibrium potentials. 

During strong illumination, all Na⁺ channels close so z = 0, and the voltage difference between light and dark is  

\[{\Delta}V_{\mathrm{m}}{=}V_{\mathrm{m,light}}-V_{\mathrm{m,dark}}{=}E_{\mathrm{K}}-\ \frac{z\ E_{\mathrm{Na}}{+}E_{\mathrm{K}}}{1{+}z}{=}\ \frac{z}{1{+}z}\ (E_{\mathrm{K}}-E_{\mathrm{Na}})\]

The fractional change in the light responses due to hypoxia is  

\[\mathrm{fractional}\ {\Delta}{=}\ \frac{{\Delta}V_{\mathrm{m,H}}-{\Delta}V_{\mathrm{m,N}}}{{\Delta}V_{\mathrm{m,N}}}\]

where H designates hypoxemia and N designates normoxia. 

Assume that in the dark z does not change during hypoxia so that the \(\frac{z}{1{+}z}\) factor drops out. Then,  

\[\mathrm{fractional}\ {\Delta}{=}\ \frac{E_{\mathrm{K,H}}-E_{\mathrm{Na,H}}-E_{\mathrm{K,N}}{+}E_{\mathrm{Na,N}}}{E_{\mathrm{K,N}}-E_{\mathrm{Na,N}}}{=}\ \frac{\mathrm{log\ }\ \frac{\mathrm{K}_{o,\mathrm{H}}}{\mathrm{K}_{i,\mathrm{H}}}-\mathrm{log\ }\ \frac{\mathrm{Na}_{o,\mathrm{H}}}{\mathrm{Na}_{i,\mathrm{H}}}-\mathrm{log\ }\ \frac{\mathrm{K}_{o,\mathrm{N}}}{\mathrm{K}_{i,\mathrm{N}}}{+}\mathrm{log\ }\ \frac{\mathrm{Na}_{o,\mathrm{N}}}{\mathrm{N}_{i,\mathrm{N}}}}{\mathrm{log\ }\ \frac{\mathrm{K}_{o,\mathrm{N}}}{\mathrm{K}_{i,\mathrm{N}}}-\mathrm{log\ }\ \frac{\mathrm{Na}_{o,\mathrm{N}}}{\mathrm{Na}_{i,\mathrm{N}}}}{=}\ \frac{\mathrm{log\ }\left(\frac{\mathrm{K}_{o,\mathrm{H}}}{\mathrm{K}_{i,\mathrm{H}}}\right)\ \left(\frac{\mathrm{K}_{i,\mathrm{N}}}{\mathrm{K}_{o,\mathrm{N}}}\right)\ -\mathrm{log\ }\left(\frac{\mathrm{Na}_{o,\mathrm{H}}}{\mathrm{Na}_{i,\mathrm{H}}}\right)\ \left(\frac{\mathrm{Na}_{i,\mathrm{N}}}{\mathrm{Na}_{o,\mathrm{N}}}\right)}{\mathrm{log\ }\ \frac{\mathrm{K}_{o,\mathrm{N}}}{\mathrm{K}_{i,\mathrm{N}}}-\mathrm{log\ }\ \frac{\mathrm{Na}_{o,\mathrm{N}}}{\mathrm{Na}_{i,\mathrm{N}}}}\]

We assume that the bigger percentage changes occur in[ K⁺] _o and[ Na⁺] _i so that we can ignore changes in [K⁺] _i and [Na⁺] _o . Then  

\[\mathrm{fractional}\ {\Delta}{=}\ \frac{\mathrm{log\ }\left(\frac{\mathrm{K}_{\mathit{o,}\mathrm{H}}}{\mathrm{K}_{o,\mathrm{N}}}\right)\ \left(\frac{\mathrm{Na}_{i,\mathrm{H}}}{\mathrm{Na}_{i,\mathrm{N}}}\right)}{\mathrm{log\ }\left(\frac{\mathrm{K}_{o,\mathrm{N}}}{\mathrm{K}_{i,\mathrm{N}}}\right)\ \left(\frac{\mathrm{Na}_{i,\mathrm{N}}}{\mathrm{Na}_{o,\mathrm{N}}}\right)}\]

For illustration, let K _{o,N =} 5 mM; K _{i,N =} 100 mM; K _o,H= 6 mM; Na _{o,N =} 100 mM; Na _{i,N =} 10 mM; Na _i,H = 11 mM. The change in[ K⁺] _o during hypoxia is approximately 1 mM. ⁷ Similarly, we assume that the change in [Na⁺] _i during hypoxia is approximately 1 mM. Substitution of these values gives a fractional change in the expected voltage response of rods of 5.2%. The mean a-wave amplitude decrease at P_a o ₂ in the 50 to 60 mm Hg range was 3.4%. Thus, the small change in the a-wave amplitude during mild hypoxia may be due to the change in the membrane potential due to the increase in[ K⁺] _o during hypoxemia, rather than any change in transduction. 

 Supported by National Institutes of Health Grant EY 05034.

 Submitted for publication December 28, 1999; revised May 22, 2000; accepted June 8, 2000.

 Commercial relationships policy: N.

 

Figure 1.

 

The ERG of a normal, dark-adapted cat retina. (A) The a- and b-waves of the ERG. The flash intensity (Φ) varied from 2.93 to 7.93 log photoisomerizations per rod per flash. The a-wave amplitude saturated at −673 μV (cat 209). (B) The a-wave responses, illustrated in (A), normalized to the saturated a-wave amplitude, −673 μV. The Lamb and Pugh ¹¹ model, applied to the data, is shown as dashed lines. Responses to 2.93 to 4.93 log photoisomerizations per rod per flash were fitted simultaneously. A and t _eff were 0.83 seconds⁻² and 4.13 msec, respectively (cat 209). Responses to three brighter flashes (intensity of 5.93–7.93) were fitted individually. A and t _eff were 0.17 seconds⁻² and 2.23 msec, respectively, for 5.93 log photoisomerizations per rod per flash; A = 0.04 seconds⁻², t _eff= 1.75 msec for 6.93 log photoisomerizations per rod per flash; A = 0.02 seconds⁻², t _eff = 1.73 msec for 7.93 log photoisomerizations per rod per flash.

Figure 2.

 

The parameter A plotted as a function of Φ, the flash intensity in photoisomerizations per rod per flash for five cats. Values of A were obtained from fits to the Lamb and Pugh. ¹¹ Both A and t _eff were varied to obtain the best fits. Each symbol represents a different cat.

Table 1.

 

The a-Wave Model Parameters during Normoxia

Cat	a max (μV)	log10 Φ	A (sec−2)	t eff (msec)
202	−678	4.93	0.22	1.02
205	−602	4.93	0.55	3.02
208	−363	4.93	0.6	2.56
209	−673	4.93	0.24	2.59
213	−429	5.93	0.24	2.35
214	−551	5.93	0.26	2.36
218	−594	5.93	0.18	2.00
223	−513	5.93	0.17	1.92
Mean± SD	−550± 111 (n = 8)	4.93	0.40 ± 0.2 (n = 4)	2.30 ± 0.9 (n = 4)
Mean± SD		5.93	0.21 ± 0.04 (n = 4)	2.16 ± 0.2 (n = 4)

 The maximum amplitude of the a-wave was obtained by using the amplitude saturating flash (log Φ = 6.93). Column 3 shows the flash intensity in log₁₀ Φ (photoisomerizations per rod per flash) at which the a-wave model fitting was performed. For each fit, five responses were used, except in cat 202 (three responses) and cat 205 (four responses).

Figure 3.

 

An intensity series to white light (•) varied from 2.93 to 7.93 log photoisomerizations per rod per flash (cat 236; 5–0 log units attenuation). An intensity series to blue light (▪) varied from 5 to 0 log units attenuation. An intensity series to red light (▴) varied from 5 to 0 log units attenuation. The Hill equation parameters for white light were: R _max = 729 μV, n = 0.53, ς = 4.2 log units attenuation. The Hill equation parameters for blue light were: R _max = 760 μV, n = 0.58, ς = 2 log units attenuation. The Hill equation parameters for red light were: R _max = 721 μV, n = 0.62, ς = 1 log unit attenuation. The inset shows superimposed responses to blue (1 log unit) and red (0 log units). Both blue and red light intensities are expressed in terms of log units, because the values were not converted to photoisomerizations per rod per flash.

Figure 4.

 

(A) The a-wave amplitude as a function of time during prehypoxemia, hypoxemia, and posthypoxemia (cat 214). At each oxygen level (shown by different symbols, and corresponding to the oxygen levels in B), five responses were recorded at intervals of 3 minutes. The absolute value of the a-wave amplitude is plotted. The inset shows the actual a- and b-wave waveforms (one response at each level of oxygen). (B) The b-wave amplitude as a function of time during prehypoxemia, hypoxemia, and posthypoxemia. At each oxygen level, five responses were recorded. The solid line and right ordinate show P_a o ₂ as a function of time.

Figure 5.

 

Relative a- or b-waves as a function of P_a o ₂. (A) Relative a-wave as a function of P_a o ₂. Each symbol represents the mean a-wave in one episode of hypoxemia from eight cats. (B) Relative b-wave as a function of P_a o ₂. The a- and b-wave amplitudes were normalized to the maximum mean amplitude for each cat during air breathing.

Table 2.

 

Parameters from the Hill Equation Fits and the a-Wave Model Fits before and during Hypoxemia

Cat	Pa o 2 (mm Hg)	Hill Equation Fits					a-Wave Model Fits
				R max (μV)	n	ς (log Φ)	a max (μV)	A (sec−2)	t eff (msec)
205	97	679	0.50	4.28	−602	0.55	3.02
	32	660	0.41	4.31	−556	0.41	2.68
209	107	717	0.52	4.78	−673	0.24	2.59
	22	631	0.44	5.03	−549	0.25	2.23
213	126	490	0.41	4.98	−429	0.24	2.35
	22	450	0.53	4.97	−412	0.22	2.22
214	120	576	0.60	4.66	−551	0.26	2.36
	29	461	0.66	4.64	−452	0.27	2.45
218	94	592	0.63	4.52	−594	0.18	2.00
	29	574	0.62	4.52	−564	0.15	1.85

 Column 2 shows normoxic and severely hypoxemic P_a o ₂ for each cat. Column 3 shows the maximum amplitude obtained by the Hill equation fits. Column 6 shows the maximum a-wave amplitude (in response to 6.93 log photoisomerizations per rod per flash) used to fit the a-wave model at each level of O₂. At each level of O₂, A and t _eff were determined from five responses to 5.93 log photoisomerizations per rod per flash except in cat 205 (three responses to 4.93 log photoisomerizations per rod per flash) and cat 209 (five responses to 4.93 log photoisomerizations per rod per flash).

Figure 6.

 

The test and probe a-wave amplitude plotted as a function of P_a o ₂. The responses are normalized to their own (either test or probe) normoxic amplitudes. For the hypoxemia experiments, an ISI of 2500 msec was used. The test flash intensity was 5.93 log photoisomerizations per rod per flash, and the probe flash intensity was 6.93 log photoisomerizations per rod per flash. Inset: a-Wave responses to the probe flashes at various ISIs.

The authors thank Michael Breton for the use of the Photran program and his A/D board, Neal Peachey and Kenneth Alexander for helpful comments on technical setups, David Pepperberg for helpful comments on double-flash experiments, and John J. K. Derwent, Christina Enroth–Cugell, Yun Kim, Monique McRipley, and Lissa Padnick–Silver for their assistance.