Impact of Second-Generation PDE 5 Inhibitor, Avanafil, on Retinal Function: Studies From Ex Vivo ERG

Abstract

Purpose

This study investigates whether avanafil, a second-generation phosphodiesterase 5 (PDE5) inhibitor, exhibits reduced off-target effects on retinal function compared to first-generation inhibitors, by quantifying its impact on photoreceptor and bipolar cell signaling using transretinal electroretinography (tERG).

Methods

We conducted ex vivo tERG using wild-type C57BL/6J and Gnat^−/− mice. The dark-adapted isolated retinas were stimulated with 530-nm full-field flashes of light while perfused with controlled avanafil concentrations at 0.1, 0.3, 1, 3, and 10 µM. The inhibition constant of avanafil for light-activated phosphodiesterase 6 (PDE6) was determined from flash responses for rods and cones. The effects of avanafil on bipolar cell signaling were also assessed.

Results

Avanafil exhibited dose-dependent inhibition of rod and cone phototransduction, characterized by slower response kinetics and reduced amplitude of dim flash responses. The inhibition constants for light-activated PDE6 were determined to be 1.74 µM for rods and 6.3 µM for cones. This study demonstrated that avanafil does not inhibit spontaneous PDE6 activity, and it has a lower inhibitory effect on light-activated PDE6 compared to other PDE5 inhibitors like sildenafil and zaprinast. Additionally, we conclude that avanafil primarily impacts photoreceptor cells, with no significant direct effect on rod bipolar cell signaling.

Conclusions

This study provides quantitative insights into avanafil's impact on retinal function, supporting the hypothesis that it has reduced off-target effects on PDE6 and retinal signaling.

Phosphodiesterases (PDEs) are crucial in cellular signaling by regulating the level of intracellular cyclic nucleotides, which act as vital second messengers in various biochemical pathways.¹ These enzymes hydrolyze cAMP and cGMP, contrasting with their synthesis by adenylyl and guanylyl cyclases.² Given their central role in cellular function, PDEs are seen as promising therapeutic targets across a spectrum of diseases.^3,4

The PDE superfamily comprises 11 distinct gene groups (PDEs 1–11), encompassing approximately 100 isoforms distinguished by their substrate specificity and regulatory mechanisms.⁵ PDEs are broadly classified as cAMP-specific, cGMP-specific (including PDE5 and PDE6), or dual-specificity enzymes.⁵ PDE5 is widely expressed in vascular smooth muscle, notably the corpus cavernosum, and in various tissues, including the kidney, pancreas, heart, lung, liver, brain, placenta, and gastrointestinal tract,⁶ while PDE6 plays a unique and vital role in retinal photoreceptor cells.⁷ The rod PDE6 consists of two catalytic subunits, α and β, alongside two inhibitory γ-subunits, whereas the cone PDE6 is made up of two catalytic α′-subunits and two inhibitory γ′-subunits.⁸

Research into PDE5 inhibitors continues to be an attractive area of study, given their broad potential applications in treating conditions ranging from erectile dysfunction and benign prostatic hyperplasia to pulmonary arterial hypertension, Alzheimer disease, long-term memory deficits, and tuberculosis ^4,9–15 The first-generation PDE5 inhibitors, such as sildenafil, vardenafil, and tadalafil, have proven to be pharmacologically effective in a variety of therapeutic conditions^16,17 However, the first-generation PDE5 inhibitors have been reported to cause visual disturbances.^18,19 PDE5 closely resembles PDE6 in terms of amino acid patterns as well as biochemical and pharmacologic properties, more than with any other PDE family.²⁰ This may be the reason for the insufficient selectivity of the first-generation PDE5 inhibitors that can cause inhibition of photoreceptor PDE6.^1,7,21 Sildenafil and vardenafil exhibit relatively low selectivity, being only 10 to 15 times more specific for PDE5 than for PDE6.²²

Phototransduction begins with light-activating rhodopsin or cone opsins, which subsequently activate the G-protein transducin.^18,23 Activated transducin binds to and disinhibits PDE6,²⁴ accelerating cGMP hydrolysis by PDE6 and leading to a decrease in the intracellular cGMP concentration and, consequently, to the closure of cGMP-gated ion channels in the photoreceptor plasma membrane.^25,26 Dysfunction of cGMP metabolism can severely impact vision, even causing photoreceptor cell death.²⁷

Avanafil, a second-generation PDE5 inhibitor, is specifically designed for rapid action, short half-life, and enhanced selectivity.^28,29 It has been identified as a promising therapeutic option for erectile dysfunction.³⁰ Avanafil was reported to express the lowest incidence of side effects, and it exhibited 100-fold higher specificity for PDE5 than for PDE6 among PDE5 inhibitors such as sildenafil and vardenafil.^18,20,31 In vivo ERG recordings on humans suggest only weak effects of avanafil on retinal function.^28,32,33 However, quantitative studies on PDE6 inhibition by avanafil in mammalian photoreceptor cells are lacking. Although avanafil appears to have fewer adverse effects than first-generation PDE5 inhibitors, it is not clear whether this is caused by a higher inhibition constant for PDE6 or less effective passage to the photoreceptor cells from blood circulation. Our working hypothesis was that avanafil exhibits a higher inhibition constant for PDE6 than the first-generation PDE5 inhibitors, resulting in lower adverse retinal effects. To test this hypothesis, we employed ex vivo ERG to evaluate the inhibitory effect of avanafil on PDE6 in both rod and cone photoreceptor cells of isolated mouse retinas and, further, to assess its impact on rod ON-bipolar cells. Our results demonstrate that, while avanafil exhibits lower inhibitory potency toward PDE6 compared to other PDE5 inhibitors, it can impact both rod and cone phototransduction. Additionally, our findings suggest that avanafil’s effects are limited to photoreceptor cells and do not appreciably affect signal mechanisms within rod ON-bipolar cells.

Methods

Ethical Approval

The use of animals was carried out according to the Finnish Act on Animal Experimentation of 2013 and the Project Authorization Board of Finland guidelines.

Animal Subjects, Perfusion, and Drug Preparation

Retinas from wild-type C57BL/6J and Gnat^−/− mice of both genders, ranging in age from 8 to 10 weeks, were employed. Before the experiment, the mice were subjected to dark adaptation overnight for at least 6 hours, followed by euthanasia through CO₂ inhalation and cervical dislocation.

The recordings of photoreceptor cell responses were conducted using HEPES buffered solution containing (in mM): Na⁺ 133.4, K⁺ 3.3, Mg²⁺ 2.0, Ca²⁺ 1.0, Cl⁻ 142.7, glucose 10, EDTA 0.01, HEPES 12.0, and Leibovitz culture medium L-15 (0.72 mg/mL). The medium’s pH was regulated to 7.5 with either NaOH or HCl. To suppress K⁺ currents in Müller glial cells, 100 µM BaCl₂ was included. The b-wave was eliminated using 40 µM D,L-2-amino-4-phosphonobutyric acid, an agonist of metabotropic glutamate receptors.

A liter of Ames’ solution was prepared using one bottle of Ames’ media in powdered form (obtained from Sigma-Aldrich, St. Louis, MO, USA) and 1.9 g NaHCO₃. Throughout the experiment, the medium was bubbled with carbogen (5% CO₂/95% O₂). Penicillin-streptomycin with 10,000 units of penicillin and 10 mg streptomycin/mL (Sigma-Aldrich) was used to inhibit bacterial contamination in longer experiments.³⁴ The perfusate osmolality was set to 0.279 to 0.280 osmol/kg.

Avanafil (purchased from ChemScene, Monmouth Junction, NJ, USA) was dissolved into dimethyl sulfoxide (DMSO) (Sigma-Aldrich) to create a 50-mg/mL stock solution, and a sonic bath was employed to agitate particles within the solution.³⁵ The stock solution was added to the perfusate to produce concentrations of 0.1, 0.3, 1, 3, and 10 µM. The control solution contained the same concentration of DMSO as the avanafil-containing solution. The DMSO concentration in the perfusate was 0.0096 v/v%, much below the value of 0.1 v/v% recommended as the safe level for intravitreal injections.³⁶

The eyes were enucleated and incised along the equator under subdued red light (long-pass filtered at λ > 700 nm). The retina was gently separated from the pigment epithelium with forceps, and the extracted retina was positioned in a specimen holder on filter paper with photoreceptors facing upward. The retinas were subjected to a steady perfusion rate (4–5 mL/min). The specimen holder was on a heat exchanger that kept the solution temperature at 32 ± 0.5°C. The temperature of the retina was monitored with a thermistor (Tewa Termico TT5-10KC3-72, Tewa Temperature Sensors Ltd., Lublin, Poland, diameter 0.5 mm) inside the specimen holder.

Light Stimulation and Ex Vivo ERG

The electroretinogram recording was done using two silver/silver chloride pellet electrodes (EP2, World Precision Instruments, Hitchin, UK) on both sides of the retina connected to the perfusate. The signal was amplified (1000×) and underwent low-pass filtering using the Model 950 Bessel 8-pole filter (Frequency Devices, Ottawa, IL, USA), with a cutoff frequency of 500 Hz, and was sampled at a rate of 5 kHz (PCIe-6351; National Instruments, Austin, TX, USA). The dark-adapted retina was stimulated with uniform full-field green light flashes. A fiber-coated LED (M530F2; Thorlabs, Newton, NJ, USA) emitting at a nominal wavelength of 530 nm was used to produce 1-ms flashes, controlled by a driver (DC2200; Thorlabs). A camera-based beam profiler (Model SP503U; Ophir-Spiricon, Logan, UT, USA) was used to verify the beam uniformity. An optical power meter (PM100D; Thorlabs Sweden AB, Mölndal, Sweden) measured the light’s intensity at the location. The intensity of light reaching the retina was also regulated using computer-operated neutral density filters. Custom-made LabVIEW (National Instruments) software managed both data collection and stimulation controls. Rod photoresponses were recorded in both the control solution and in one to five different avanafil concentrations. Cone photoresponses recorded up to three different avanafil concentrations in each experiment. The avanafil test protocol always consisted of perfusion with control solution, solution with avanafil, and washout with control solution, repeated for different concentrations. The retinas were always allowed to stabilize for at least 20 minutes in each solution.

Data Analysis and Determination of Inhibition Constant for Light-Activated PDE6

Linear baseline correction was applied to the recorded data in MATLAB (MathWorks, Natick, MA, USA). Subsequently, a second-order notch filter was used to attenuate power-line interference, and the flash response traces recorded with the same stimulus strength under similar experimental conditions were averaged.

The number of photoisomerized rhodopsin molecules per rod cells (R*rod⁻¹) and activated pigment molecule per cone (P*rod⁻¹) was calculated using the visual pigment template by Govardovskii et al.,³⁷ incorporating the collecting area, LED emission spectrum, and optical power meter sensitivity, as described in Heikkinen et al.³⁸ The rod and cone collecting areas were calculated to be 0.6 and 0.078 µm², as explained in Saeid et al.³⁴ The stimulus–response curves for rod and cone photoreceptors were generated by applying the Hill equation

$$\begin{array}{c}\hfill \frac{R}{R_m}=\frac{1}{1+{\left(\frac{I_{1/2}}{\varphi }\right)}^n}\end{array}$$ (1)

to the response amplitude (R) and the corresponding flash strengths (ϕ). In this equation, (I_½) refers to the stimulus strength required to produce a half-maximal response, n represents the Hill coefficient, and R_m is the saturated response amplitude.

Statistical analysis was performed using one-way ANOVA, followed by Holm–Bonferroni post hoc testing. OriginPro 2024 (OriginLab Corporation, Northampton, MA, USA) was utilized for graphical representation, while MATLAB was employed for analysis of response characteristics and fitting.

When a PDE inhibitor drug compound is introduced to the retina, there is a decrease in the amplification of phototransduction due to the changes in cGMP hydrolysis activity.

The light-activated PDE6 enzyme inhibition constant (K_i,Light) determination was conducted as described in Turunen and Koskelainen.³⁹ Shortly, if a drug inhibits the catalytic activity of PDE6, phototransduction gain is reduced. The change in the gain can be obtained by fitting the Lamb–Pugh activation model²⁶

$$\begin{array}{c}\hfill r\left(t\right)=r_{\max }\left(1-exp\left[-1/2\varphi A{\left(t-t_d\right)}^2\right]\right)\end{array}$$ (2)

to the early phase of flash responses. In this equation, r(t) represents the response waveform and ϕ the flash strength, r_max denotes the saturated response amplitude, and t_d accounts for the delays in both phototransduction reactions and in the measurement equipment. We fitted the model to the first 40 ms with rod responses and 15 ms with cone responses. With cones, the experiments were conducted on Gnat^−/− mice to reduce the Ca²⁺-mediated negative feedback to phototransduction. We used a constant delay to all responses in each experiment (rods 4–7 ms, cones 4 ms). The alteration in the amplification constant is related to the inhibition constant and the concentration of the inhibitor, and it is given by the following equation:

$$\begin{array}{c}\hfill \frac{A_{Control}}{A_{Inhibitor}}=\frac{\left[I\right]}{K_{i,Light}}+1.\end{array}$$ (3)

Here, A_Control and A _Inhibitor represent the amplification constants derived from the transretinal electroretinography (tERG) responses recorded in control and avanafil solutions at different concentrations, respectively. In this study, A_Control was defined as the average of the amplification constants derived in the control solution before and after avanafil application. A MATLAB-based optimization to fit our experimental data to the model of Equation (3) was used to obtain the K_i,Light.

Results

Avanafil Shows a Dose-Dependent Inhibition of Rod Phototransduction

Before starting the tERG recordings, the wild-type (WT) mouse retinas were perfused for 60 minutes to achieve stable responses. Figure 1 compares the characteristics of recorded rod flash responses in control solution and in the presence of varying avanafil concentrations (0.1–10 µM).

Figure 1.

Effect of avanafil on rod flash photoresponses. (a–c) Representative response families recorded from a C57BL/6J mouse retina at 32.0 ± 0.5°C in solution with 100 µM BaCl₂and 40 µM D,L-2-amino-4-phosphonobutyric acid (APB): (a) control solution (black traces), (b) 1 µM avanafil (red traces), and (c) after washout with control solution (blue traces). Flash strengths: 10.5, 18.7, 33.3, 59.2, 105, 187, 333, 592, 1053, 1873 R*rod⁻¹. (d) Comparison of responses from a–c. (e) Stimulus strength–response amplitude curves in control solution (black line), 1 µM avanafil (red line), and after washout with control solution (blue line). Hill equation fitted to the data points yields ϕ_1/2 = 67, 112, and 62 R*rod⁻¹ in control solution, 1 µM avanafil, and washout control solution, respectively. (f) Mean normalized saturated amplitudes, (g) mean time to peak values of dim flash responses (18.7 R*rod⁻¹), and (h) mean normalized inverse light sensitivity of rods with 0.1 (red bar), 0.3 (blue bar), 1 (green bar), 3 (purple bar), and 10 (orange bar) µM avanafil and without avanafil (gray bars) administration. (c) Data in f–h are from six retinas. One-way ANOVA Holm–Bonferroni test. Significance comparisons shown are relative to control values. Error bars show ± SEM (DF:23 for all comparisons, F values: normalized time to peak: 15.66, normalized saturated response amplitude: 25, normalized light sensitivity [ϕ_1/2]: 159). *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

A set of three rod response families recorded in control solution and in the presence of 1 µM avanafil is illustrated in Figures 1a–c. Figure 1d compares the kinetics of rod responses (flash strengths: 18.7, 105, 1873 R*rod⁻¹) from the same response families, showing that 1 µM avanafil only slightly decreased the saturated response amplitude in this retina, while the small-stimulus response amplitude was more substantially reduced. The dependence of the response amplitude on flash strength in this experiment is illustrated in Figure 1e, showing a decrease in rod light sensitivity in the presence of avanafil. The flash strengths needed to achieve half-maximal response (ϕ_1/2) were 67, 112, and 62 R*rod⁻¹ in control, 1 µM avanafil, and washout control solution, respectively. Panels a to e demonstrate that rod responses show full recovery from 20 minutes treatment with 1 µM avanafil after washout solution application.

Panels f to h show mean data collected from six retinas. On average, the saturated response amplitude was not affected by low concentrations of avanafil, while only the highest concentration studied (10 µM) caused a modest 20% drop in the amplitude (Fig. 1f).

The time to peak of dim flash responses (Fig. 1h) grew in a dose-dependent way, reflecting slowing of the flash response kinetics. The increase, however, was significant only at concentrations of 1 µM or higher. The stimulus strength needed to produce half-maximal responses (ϕ_1/2) was also increased in a dose-dependent way, showing that 10 µM avanafil decreased the rod light sensitivity by a factor of more than 4 (Fig. 1h). In all experiments, a complete recovery of rod flash responses was observed up to a concentration of 3 µM, while at 10 µM avanafil, partial recovery happened.

Avanafil Shows a Dose-Dependent Inhibition of Cone Phototransduction

Compared to rods, the higher energy demands of cone photoreceptor cells⁴⁰ can make preserving their light responses during prolonged experiments more difficult. To address this, individual experiments testing the effect of avanafil on cones were conducted using only two to three avanafil concentrations. To isolate the cone photoresponses, we used transducin α-subunit knockout (Gnat^−/−) mouse retinas that lack responding rods. Figure 2 illustrates the characteristics of cone flash responses recorded in control solution and in the presence of varying concentrations of avanafil (0.1–10 µM). Panels a to c show a representative set of three cone response families recorded in both control conditions and with 1 µM avanafil. Panel d compares the kinetics of cone responses (flash strengths: 992, 14, 314, 138, 625 P*cone⁻¹) from the same experiment. The relationship between response amplitude and flash strength is presented in Figure 2e. The flash strengths needed to achieve a half-maximal response were 3577, 5232, and 3749 P*cone⁻1 in control, 1 µM avanafil, and washout control solution, respectively. These results demonstrate that the light sensitivity of cones decreased in the presence of 1 µM avanafil.

Figure 2.

Effect of avanafil on cone flash responses. (a–c) Representative response families recorded from a Gnat^−/− mouse retina at 32.0 ± 0.5°C in solution with 100 µM BaCl₂ and 40 µM APB: (a) control solution (black traces), (b) 1 µM avanafil (red traces), and (c) after washout with control solution (blue traces). Flash strengths: 400, 992, 5781, 14,314, 55,989, and 138,625 P*cone⁻¹. (d) Comparison of responses from a–c. (e) Stimulus strength–response curves in control solution (black line), in 1 µM avanafil (red line), and after washout in control solution (blue line). Hill equation fitted to the data points yields ϕ_1/2 = 3577, 5232, and 3749 P*cone⁻¹ in control solution, in 1 µM avanafil, and after washout in control solution, respectively. (f) Normalized mean saturated amplitudes, (g) mean time to peak values of dim flash responses, and (h) mean inverse light sensitivity of cones with 0.1 (red bar), 0.3 (blue bar), 1 (green bar), 3 (purple bar), and 10 (orange bar) µM avanafil and without avanafil (gray bars) administration. Data in f–i are from eight retinas. One-way ANOVA Holm–Bonferroni test. Significance comparisons shown are relative to control values. Error bars show ± SEM (DF:17 for all comparisons, F values: normalized time to peak: 21.1, normalized saturated response amplitude: 6.3, normalized light sensitivity [ϕ_1/2]: 54).

Panels f to h display the mean effects of different avanafil concentrations (two to three concentrations per retina) on the normalized saturated response amplitudes, time to peak of small-stimulus responses, and inverse light sensitivity (ϕ_1/2) across all (n = 8) Gnat^−/− mice retinas studied. Data are normalized to the average control value (control and washout control solution) for each concentration. The normalized saturated amplitude (Fig. 2f) showed a significant reduction only at 10 µM avanafil. In panel g, significant increases in normalized time to peak values were present at 1- to 10-µM concentrations of avanafil. Finally, the stimulus strength needed to produce responses with half-maximal amplitude (ϕ_1/2, Fig. 2h) was significantly increased at 1 µM, 3 µM, and 10 µM. All retinas showed full cone recovery up to a concentration of 3 µM.

Quantifying Avanafil's Impact on Phototransduction Gain

The molecular gain of phototransduction plays a critical role in determining the absolute sensitivity of vision. To quantify avanafil’s effect on molecular amplification of phototransduction in both rods and cones, we employed the method previously used by our group.³⁹ The Lamb–Pugh activation model²⁶ (Equation 2) was fitted to the initial phase of subsaturated responses, spanning the first 0 to 40 ms for rod responses (Fig. 3a) and 0 to 15 ms for cone responses (Fig. 3b). In the fitting, we used a delay of 4 to 7 ms (constant for each retina) for rod responses and 4 ms for cone responses. This analysis was performed on responses from C57BL/6J (n = 6) and Gnat^−/− (n = 8) mouse retinas under control conditions and across five avanafil concentrations. (Equation 3) was then used to calculate avanafil’s inhibition constant. Figure 3c demonstrates a linear relationship between relative amplification constant and inhibitor concentration, as predicted by Equation (3). The data point corresponding to the highest avanafil concentration was omitted, because it was well below the fitted line. This result coincides well with the earlier observation that in both rods and cones, full recovery of phototransduction was not achieved after the introduction of 10 µM avanafil. The calculated inhibition constants for avanafil were 1.74 ± 0.06 µM for light-activated rod PDE6 and 6.3 ± 0.6 µM for light-activated cone PDE6. The relative standard error for avanafil was 3.5% (rod PDE6) and 9.4% (cone PDE6).

Figure 3.

Fitting the Lamb–Pugh activation model (rod, blue; cone, green) to the initial phase of subsaturated (a) rod and cone (b) flash responses. (c) Determination of the avanafil inhibition constant for light-activated PDE6. The slopes of the linear fit to the data are 0.0574 and 0.157 µM, corresponding to the inhibition constants K_i,light of 1.74 ± 0.06 and 6.3 ± 0.6 µM for rod and cone light-activated PDE6, respectively. Data were collected from C57BL/6J (n = 6) and Gnat^−/− (n = 8) mouse retinas at 32.0 ± 0.5°C.

Effect of Avanafil on ON-Bipolar Cell Responses

Previous research suggests that the effects of PDE inhibitors may not be purely confined to photoreceptor segments, potentially impacting the inner retina as well.^19,41,42 This is supported by the observations that PDE5 is localized in bipolar cells, ganglion cells, and retinal blood vessel endothelium.^43–45 Furthermore, cGMP has been demonstrated to modulate bipolar cell signaling, and bipolar cells are concluded to contain PDE that regulates the cGMP level inside the cells (for a discussion, see, e.g., Snellman et al.⁴⁶).

To assess the potential effects of avanafil on rod ON-bipolar cells, we analyzed the amplitude and the kinetics of the b-wave recorded from C57BL/6J mouse retinas. Figure 4 shows exemplary b-wave flash response families in control solution (a), in the presence of 3 µM avanafil (b), and after washout by control solution (c). Panels j to k depict the normalized mean amplitudes and normalized time to peak values of rod bipolar cell b-waves from all recordings (n = 6). K⁺ currents in glial cells were suppressed using 100 µM BaCl₂ in the perfusate. Panels d to f illustrate dim flash responses with 3 µM and without avanafil and panels g to i with 10 µM avanafil. Panels j to k show that a 0.3-µM avanafil concentration significantly decreases the mean dim flash b-wave amplitude, while a significant increase in b-wave time to peak is observed at avanafil concentrations of 1 µM and higher. These results demonstrate a dose-dependent reduction in the b-wave amplitude and a slower time to peak value in the presence of avanafil. However, a comparison of the mean a-wave and b-wave amplitudes (Fig. 4l) suggests that the reduction in amplitude and increased delay observed in the b-wave originate in the photoreceptor cells and are associated with the inhibition of light-activated PDE6 in rods. The slower photoreceptor response kinetics may lead to a delayed reduction in glutamate release in the synaptic cleft.

Figure 4.

Effect of avanafil on rod b-wave from C57BL/6J mouse retinas. Flash response families recorded from one retina in control solution (a), in the presence of 3 µM avanafil (b), and after washout in control solution (c). (d–i) Effects of 3 and 10 µM avanafil on the in-dim flash (18.7 R*rod⁻¹) response amplitude and kinetics in two separate experiments. (j) Normalized mean amplitudes and (k) normalized time to peak values of the rod b-wave with 0.1 (red bar), 0.3 (blue bar), 1 (green bar), 3 (purple bar), and 10 (orange bar) µM avanafil and without avanafil (gray bars) at 32.0 ± 0.5°C. (l) Comparison of dim stimulus a-wave (red bars) and b-wave amplitude (blue bars). Data in j–l are from six retinas. A total of 100 µM BaCl₂ is present in all experiments. One-way ANOVA Holm–Bonferroni test. Significance comparisons shown are relative to control values. Error bars show ±SEM (DF:20 for all comparisons, F values: normalized b-wave amplitude: 129, and normalized b-wave time to peak: 21.5).

Discussion

PDE6’s close structural and functional similarity to PDE5 raises concerns about off-target PDE6 inhibition by PDE5 inhibitors.⁴⁷ This may lead to a reduced rate of cGMP hydrolysis by PDE6 and, in turn, affect the kinetics of light responses and the sensitivity of rods and cones. A previous ERG study on anesthetized dogs suggests avanafil’s greater selectivity for PDE5 compared to sildenafil.³³ However, despite reports suggesting avanafil’s superior isoform selectivity,^15,28,29 investigations directly assessing the drug’s impact on retinal function at precisely controlled concentrations are lacking. In this study, we used tERG to evaluate avanafil’s influence on rods and cones as well as on rod ON-bipolar cell signaling. Since direct retinal avanafil concentrations have not been measured, we estimated retinal concentrations based on reported plasma and brain tissue levels. Considering human plasma concentrations (∼2.8–12.1 µM) and rat brain data (∼240 ng/g at peak, corresponding to roughly 2 µM retinal concentration due to the fourfold higher permeability of the blood–retinal barrier compared to the blood–brain barrier, we approximate human retinal concentrations after oral dosing of 50 to 200 mg to fall within ∼1–5 µM.^48–50

We found that avanafil decelerates flash response kinetics and decreases sensitivity in both rods and cones at concentrations of 1 µM or higher, while we did not observe any direct effect of avanafil on bipolar cell signal mechanisms.

The cGMP concentration in photoreceptor cells is tightly controlled by cGMP synthesis by guanylate cyclases and cGMP degradation by PDE6. The cGMP hydrolysis activity of PDE6 is regulated by the inhibitory PDE6 γ-subunits that, when PDE6 is not activated, cover the two catalytic binding sites for cGMP.⁵¹ When light activates the phototransduction cascade, the transducin α-subunits can bind to PDE6, which leads to uncovering of the catalytic sites. The adverse effects of PDE5 inhibitors on PDE6 are believed to be caused by binding of the inhibitor molecules to the catalytic binding sites.^25,52,53

In dark-adapted photoreceptor cells, the slower photoresponse kinetics and reduced sensitivity in the presence of PDE inhibitors typically arise from two major factors: binding of the inhibitor to the catalytic sites reduces the effective number of activatable PDE subunits. The other factor can come from elevated intracellular cGMP concentration, if the inhibitor molecules are able to bind to the catalytic sites without light stimulation, that is, if the PDE inhibitor can suppress the spontaneous (dark-adapted) PDE6 activity.^39,54 Our results that avanafil can decelerate the flash response kinetics and reduce sensitivity (i.e., reduce the amplification factor of phototransduction) in both rods and cones without increasing saturated response amplitude suggest that avanafil can bind to the catalytic site of cGMP hydrolysis. Further, the study by Liu and others⁵⁵ demonstrated that both the α- and β-subunits of PDE6 contain catalytic binding sites that are structurally homologous to those in PDE5 and can bind the PDE5 inhibitor vardenafil. However, the observation that avanafil did not increase the saturated response amplitude in rods or cones is completely different from that with other common PDE inhibitors like 3-isobutyl-1-methylxanthine (IBMX), vardenafil, and sildenafil.³⁹ This result demonstrates that avanafil cannot suppress spontaneous PDE6 activity in mouse rods and cones. The likely reason for this is avanafil’s bulkier structure, which prevents its access to the catalytic binding pocket when it is covered by the PDE6 γ-subunits.⁵⁶ This is consistent with the earlier observations that the small nonspecific PDE inhibitor IBMX can very rapidly and effectively block the spontaneous PDE6 activity in frog and mouse rods,^39,57 while the somewhat larger molecules vardenafil and sildenafil can only more slowly inhibit the spontaneous PDE6 in mouse rods.³⁹ This slower effect by vardenafil and sildenafil, however, may at least partly be explained by their lower permeability through the rod outer segment plasma membrane. The milder side effects of avanafil compared to first-generation PDE5 inhibitors are unlikely to result from differences in half-life. The reported half-life of avanafil in rat plasma is approximately 4.87 hours,⁴⁸ which is sufficiently long to suggest that half-life differences do not play a role in our results.

A recent study reported that sildenafil may induce oxidative stress in the retina and acidification in the subretinal space.⁵⁸ The oxidative stress and acidification effects were hypothesized to result from sustained cGMP elevation, leading to prolonged opening of cyclic nucleotide–gated channels, increased circulating current, and increased mitochondrial activity. The increased mitochondrial activity results in the production of acidic metabolic by-products that accumulate in the subretinal space, leading to a reduction in pH. This triggers the activation of water transport mechanisms in the RPE, which facilitates water removal and causes dehydration and thinning of the external limiting membrane (ELM)-RPE layer. Additionally, the heightened mitochondrial activity generates reactive oxygen species, and when these exceed the retina’s antioxidant capacity, they lead to localized oxidative stress.⁵⁸

In our experiments, unlike with sildenafil, avanafil did not increase saturated response amplitudes, indicating that the circulating current in photoreceptor cells did not increase. This suggests that the mechanisms causing oxidative stress and acidification with sildenafil are unlikely to occur with avanafil. Further, in our recording geometry, the perfusate in contact with the photoreceptor side of the retina is changed more than 10 times per second, effectively preventing possible acidification.

The observation that avanafil does not increase the saturated response amplitude, indicating that the circulating current through the outer segment plasma membrane does not rise, may also have clinical importance. Since Ca²⁺ ions can pass the cGMP-gated channels, an increase in the circulating current would lead to an elevated level of intracellular calcium, which might be harmful to the photoreceptor cells. With avanafil, this does not take place. Instead, at higher concentrations of avanafil, the saturated response amplitude shows a small but significant decrease. The mechanism behind this decrease is not known, and we did not study it further.

Avanafil is considered a more specific inhibitor for PDE5 over PDE6 than the first-generation PDE5 inhibitors. Our inhibition constant determinations for light-activated mouse rod PDE6 using the same method as was used with vardenafil and sildenafil³⁹ indeed gave an inhibition constant for avanafil that is 3.1 times higher than for sildenafil and 1.8-fold for vardenafil. However, the pharmacologically determined IC₅₀ (half maximal inhibitory concentration) value of avanafil for PDE5 inhibition is about two times higher than for sildenafil and five times higher than for vardenafil,¹ suggesting that the specificity of avanafil for PDE5 over PDE6 might not differ much from the first-generation PDE5 inhibitors. According to our results, the weaker adverse effects might be a consequence of avanafil’s inability to inhibit spontaneous PDE6 activity rather than higher specificity for PDE5 over PDE6.

Finally, our observed reduction in b-wave amplitude and increase in time to peak value suggest alterations in signal transduction. However, these effects may be attributed to the inhibition of PDE6 in the photoreceptor cells rather than a direct impact on rod bipolar cells.

Gnat ^{− / −} mice have been reported to show mildly elevated oxidative stress and altered outer retina energy metabolism.^59,60 Our study design focused on comparing responses within the same retina under controlled conditions (control, avanafil, and washout). Although Gnat^−/− retinas may experience higher metabolic load due to sustained circulating current in rods, this largely corresponds to the dark-adapted state in WT mice. We therefore consider any minor oxidative or metabolic differences unlikely to affect our interpretations.

Finally, it should be noted that transretinal ERG recording is an indirect method of assessing PDE6 inhibition and cGMP channel activity, and it cannot directly demonstrate whether avanafil binds to and inhibits PDE6. Moreover, avanafil exposure times in our study were short, and the measurements reflect acute responses rather than cumulative effects. Thus, conclusions based on these assays should be interpreted accordingly. Future studies integrating additional biochemical and imaging techniques, such as OCT-based assessment of ELM-RPE thickness, could provide complementary spatial or metabolic information and further support functional interpretations based on tERG data.

Conclusions

This study utilized the ex vivo tERG method to explore the effects of avanafil, a second-generation PDE5 inhibitor, on retinal function. Our findings indicate that avanafil causes dose-dependent inhibition of phototransduction in rod and cone photoreceptors, with inhibition constants of 1.74 µM for rods and 6.3 µM for cones, showing lower PDE6 inhibition potency compared to other PDE5 inhibitors, such as sildenafil and vardenafil. Notably, avanafil’s inability to suppress spontaneous PDE6 activity, likely due to its bulkier structure, differentiates it from first-generation PDE5 inhibitors. While avanafil slowed photoresponse kinetics and reduced photoreceptor sensitivity at concentrations of 1 µM or higher, its effects were fully reversible up to 3 µM for rod and cone photoreceptors, suggesting minimal risk of retinal damage at lower drug concentrations. These findings emphasize the need for comprehensive preclinical evaluations of PDE5 inhibitors to better understand their impact on retinal health.