A p97/valosin-containing protein inhibitor drug CB-5083 has a potent but reversible off-target effect on phosphodiesterase-6.

CB-5083 is an inhibitor of p97/valosin-containing protein (VCP), for which Phase I trials for cancer were terminated due to adverse effects on vision, such as photophobia and dyschromatopsia. Lower dose CB-5083 could combat inclusion body myopathy with early-onset Paget disease; and frontotemporal dementia or multisystem proteinopathy, caused by gain-of-function mutations in VCP. We hypothesized that the visual impairment in the cancer trial was due to CB-5083's inhibition of phosphodiesterase-6 (PDE6), which mediates signal transduction in photoreceptors. To test our hypothesis, we used in vivo and ex vivo electroretinography (ERG) in mice, and a PDE6-activity assay of bovine rod outer segment (ROS) extracts. Additionally, histology and optical coherence tomography were used to assess CB-5083's long-term ocular toxicity. A single administration of CB-5083 led to robust ERG signal deterioration, specifically in photoresponse kinetics. Similar recordings with known PDE inhibitors, sildenafil, tadalafil, vardenafil, and zaprinast, showed that only vardenafil had as strong an effect on the ERG signal in vivo as did CB-5083. In the biochemical assay, CB-5083 inhibited PDE6 activity with a potency higher than sildenafil but lower than that of vardenafil. Ex vivo ERG revealed a PDE6-inhibition constant of 80 nM for CB-5083, which is 7-fold smaller than for sildenafil. Finally, we showed that the inhibitory effect of CB-5083 on visual function is reversible, and its chronic administration does not cause permanent retinal anomalies in aged VCP disease-model mice. Our results warrant re-evaluation of CB-5083 as a clinical therapeutic agent. We recommend preclinical ERG recordings as a routine drug safety screen. Significance Statement This report supports the use of a VCP inhibitor drug, CB-5083, for the treatment of neuromuscular VCP disease, despite CB-5083's initial clinical failure for cancer treatment due to side effects on vision. Our data show that CB-5083 displays a dose-dependent, but reversible, inhibitory action on phosphodiesterase-6, an essential enzyme in retinal photoreceptor function, but no long-term consequences on retinal function or structure.


INTRODUCTION
VCP is a member of the AAA (ATPases associated with a variety of cellular activities) protein family and it is a highly abundant protein accounting for ~1 % of total cellular protein content (Song, Wang, and Li 2003). Many of VCP's physiological actions are related to the ubiquitin-proteasome degradation pathway. Missense mutations in the gene encoding the valosin-containing protein (VCP) are causative for VCP disease that is characterized by a complex set of clinical manifestations, such as musculoskeletal and neurodegenerative conditions associated with Limb-Girdle Inclusion Body Myopathy, Paget's disease of the bone, Frontotemporal Dementia and Amyotrophic Lateral Sclerosis (Watts et al. 2004). In vitro assays of VCP mutants have shown moderately enhanced ATPase activity, increased cofactor binding and reduced mitofusin levels, providing evidence for a gain-of-function mechanism of the disease (Zhang et al. 2017;Blythe et al. 2017). Early generation VCP inhibitors, NMS873 and ML240, were shown to ameliorate mitochondrial dysfunction and cell death in a drosophila model of VCP disease (Zhang et al. 2017). Of these inhibitors, one is allosteric and one is ATP-competitive, demonstrating that different VCP-inhibition modalities drive a similar beneficial response.
This finding suggests that direct inhibition of VCP activity is the basis for the therapeutic effects. Unfortunately, both NMS873 and ML240 lack suitable pharmaceutical properties for clinical development. Instead, a newer generation VCP inhibitor, CB-5083, is an orally bioavailable compound (Le Moigne et al. 2017) that has previously reached Phase 1 clinical trials for cancer (ClinicalTrials.gov Identifiers: NCT02223598 and NCT02243917).
Many types of cancer in humans are associated with elevated VCP expression (Yamamoto et al. 2005;Tsujimoto et al. 2004). CB-5083 is a reversible and competitive inhibitor that is highly specific for the D2 ATPase domain of the AAA ATPase p97/VCP, making CB-5083 ideal for probing the specific therapeutic benefit of normalizing the upregulated VCP activity in patients (Le Moigne et al. 2017). Promising preclinical data in cellular and rodent cancer models prompted the developer of CB-5083, Cleave Biosciences, to launch a Phase 1 trial in 84 patients with solid tumors and multiple myeloma. CB-5083 was documented as clinically safe and generally well tolerated.
However, CB-5083 was not successful in achieving the desired endpoints in the cancer trial because it was halted due to suspected off-target activity against PDE6. The PDE6 inhibition ostensibly led to visual adverse events consisting of photophobia and dyschromatopsia that were mild, episodic and self-resolving at lower doses; but these symptoms increased in frequency and severity at increasingly higher doses, ultimately ending the dose escalation study. Similar side effects have been associated with other drugs in the phosphodiesterase (PDE)-inhibitor family, such as sildenafil and vardenafil that are primarily indicated for the treatment of erectile dysfunction (Yafi, Sharlip, and Becher 2018;Moschos and Nitoda 2016). Indeed, the chemical structure of CB-5083 is closely related to those of sildenafil and vardenafil (https://pubchem.ncbi.nlm.nih.gov/).
Notably, drug doses required to normalize aberrant over-activity of VCP mutants could potentially be much lower than the doses required to drive a cytotoxic response in cancer cells. We postulate that CB-5083 could be repurposed for the treatment of VCP disease after careful dose-titration. However, its adverse effects on vision must be investigated at a mechanistic level before other major research efforts concerning therapy of VCP disease are warranted.
Previously, adverse effects on vision associated with clinical administration of PDE5 inhibitors were attributed to potential inhibition of PDE6, an enzyme required for phototransduction in retinal photoreceptors. Since CB-5083 in the previous Phase 1 cancer trial caused side effects analogous to those of known PDE5-inhibitors, we hypothesized that these visual impairments could be due to off-target deactivation of PDE6. In the current study, we present a comprehensive analysis of the impact of CB-5083 on retinal phototransduction to evaluate the potential of this off-target effect to impede clinical use of the drug in the VCP-disease setting. From the perspective of drug safety our data indicate that repurposing CB-5083 for the treatment of VCP disease or other applicable conditions is a viable option.

Animals, compounds and drug regimen
We used mice as a primary model organism. In acute repeated measures experiments, 2-4-month-old C57BL6/J wild type (WT) mice were utilized (n=3 per drug). In addition, heterozygous VCP R155H knock-in mice between 12 and 18 months of age were used in a chronic ocular toxicity trial n=4;vehicle,n=5). In the terminal phase of this trial, 18-month-old C57BL6/J mice (n=3) were used as a WT control. An independent ERG recording was performed also in non-treated 24-month-old VCP R155H (n=5) and 19-monthold WT (n=7) mice to further investigate if VCP R155H knock-in mutation causes alterations to retinal function. VCP R155H mice (later referred to as VCP model mice) recapitulate many clinical features of VCP disease, including muscle, bone, brain and spinal cord pathology and demonstrate progressive muscle weakness most noticeably from the age of 9 months onwards (Badadani et al. 2010;Nalbandian et al. 2012;Yin et al. 2012 solution to a final concentration of 3 mg/ml. Methyl cellulose solution was used as vehicle control in the chronic CB-5083 trial, wherein the mice were treated with daily oral gavage. Twelve-month-old VCP model mice were treated with 15 mg/kg CB-5083 or vehicle for 6 months. In acute CB-5083 trials, WT mice were treated with vehicle, 15 mg/kg CB-5083, or 30 mg/kg CB-5083 by a single gavage administration 45 min prior to initiation of ERG recordings. These doses were chosen based on previous studies where a minimum gavage dose of 30 mg/kg CB-5083 was used for cancer therapy in mice using a 4 days on/3 days off regimen (Le Moigne et al. 2017;Anderson et al. 2015).
Sildenafil, tadalafil, vardenafil and zaprinast were all purchased from Cayman Chemical Company (Ann Arbor, MI). Drug stocks were first dissolved in DMSO at a 10 mg/ml concentration and dosed at 50 µl volume via intraperitoneal injection 45 min prior to ERG recording.

In vivo electroretinography (ERG)
Before scotopic ERG recordings, the mice were dark-adapted overnight and animal handling before recording was performed under dim red light. The ERG was performed using a Diagnosys Celeris rodent ERG device (Diagnosys, Lowell, MA), as described previously (Orban et al. 2018). Briefly, mice were anesthetized with ketamine (100 mg/kg, KetaVed®, Bioniche Teoranta, Inverin Co., Galway, Ireland) and xylazine (10 mg/mg, Rompun®, Bayer, Shawnee Mission, KS) by intraperitoneal injection, and their pupils were dilated with 1% tropicamide (Tropicamide Ophthalmic Solution USP 1%, Akorn, Lake Forest, IL) and thereafter kept moist with 0.3 % hypromellose gel (GenTeal, Alcon, Fort Worth, TX). An additional 3 min were allowed in full dark adaptation before recordings. Light stimulation was produced by an in-house scripted stimulation series in Espion software (version 6; Diagnosys). The eyes were stimulated with a green LED (peak 544 nm, bandwidth 160 nm) using a 6-step ascending flash strength series in 1log-unit increments between 0.0005 and 50 cd·s/m 2 . After scotopic ERG, the mouse eyes were adapted to a rod-suppressing green background light at 20 cd/m 2 for 1 min and this background was maintained during the subsequent photopic ERG recordings. Here, stimulation was performed with a UV LED (peak emission 370 nm, bandwidth ∼50 nm) at light strength increments of 0.1, 1.0, and 10.0 cd·s/m 2 , and separately with a green LED (peak emission 544 nm, bandwidth ∼160 nm) at 0.3, 3.0 and 30.0 cd·s/m 2 . As the short-wavelength cone opsin (S-opsin) and medium-wavelength cone opsin (M-opsin) sensitivities peak at 360 and 508 nm, respectively, in mice (Nikonov et al. 2006), UV and This article has not been copyedited and formatted. The final version may differ from this version. green LEDs stimulated mouse S-and M-opsins relatively selectively. LED light emission spectra were measured with a Specbos 1211UV spectroradiometer (JETI Technische Instrumente GmbH, Jena, Germany). The ERG signal was acquired at 2 kHz and filtered with a low-frequency cutoff at 0.25 Hz and a high-frequency cutoff at 300 Hz. Espion software automatically detected the ERG a-wave (first negative ERG component) and bwave (first positive ERG component) implicit times (latencies) and amplitudes; a-wave amplitude was measured from the signal baseline, whereas b-wave amplitude was measured as the difference between the negative trough (a-wave) and the highest positive peak.

Mouse serum and tissue collection for drug level measurements
Three-month-old C57BL/6J mice (n=10) were used for these experiments. Mice were terminally anesthetized by 3-4-fold overdose of ketamine-xylazine cocktail. Once fully unresponsive, the mouse chest was incised open and a small hole was punctured to the heart's right atrium and ~0.5 ml of blood was quickly collected and transferred on ice. A perfusion needle was quickly inserted into the left ventricle and vasculature was perfused with ice-cold saline for 3 min using a peristaltic pump. After perfusion, the retinas were quickly excised by performing three incisions starting from the optic nerve head and cutting towards the ora serrata, which allowed easy and quick separation of the retina from the rest of the eye cup. We also collected a 10-20 mg biopsy of frontal cortex for retina-brain comparative analysis. Blood was centrifuged at 4 ºC and 13 500 rpm for 20 min and thereafter the supernatant was collected as a serum sample. Tissue and serum samples were stored at -20 ºC until analysis.

Liquid chromatography-mass spectrometry (LC-MS)
Each serum sample (100 µL) was precipitated with 400 µL of pre-cooled methanol, and centrifuged at 17,000g for 15 min at 4 °C. The supernatant was carefully transferred to a SpinX centrifuge tube filter with a 0.45 µm cellulose acetate membrane (Costar, Salt Lake City, UT), and centrifuged at 7,000 g for 2 min. Filtered samples were dried under vacuum, reconstituted in 100 μL 50% methanol/water, and centrifuged at 17,000g for 15 min at 4 °C. The resulting supernatants were ready for LC-MS analyses. Retina and brain biopsy samples were homogenized in methanol (2 x 800 µL). The resulting mixture was centrifuged at 17,000g for 15 min at 4 °C. The supernatant was dried under vacuum, reconstituted in 100 μl 50% methanol/water, and centrifuged at 17,000g for 15 min at 4 °C. Ten microliters of the supernatant extracted from serum or tissue samples was injected into an Ultimate 3000 HPLC system coupled with LXQ mass spectrometer (ThermoFisher Scientific, Waltham, MA) with an electrospray ionization unit. The separation was performed on a Proshell EC-18 column (2.7 μm, 3.0 x 150 mm, Agilent, Santa Clara, CA) using a mobile phase consisting of 0.1% aqueous formic acid (A) and acetonitrile (B) at a flow rate of 300 µL·min -1 and the mobile phase gradients and time course were as follows: 0-3 min, 80% A / 20% B; 3-13 min, 80%-10% A / 20%-90% B.
The signals were detected in the selected reaction monitoring (SRM) mode, with the m/z values of the parent and the daughter ions set at 414.2 and 397.2, respectively. Serum, retina or brain samples from mice without drug administration were spiked separately with 0, 1, 3, 10, 30, 100, or 300 nmol authentic CB-5083 compound, and then extracted and analyzed, as described above, to generate the standard curves representing the relationship between the amounts of CB-5083 and the areas under the corresponding chromatographic peaks for specific tissue samples.

PDE6 activity assay in bovine rod outer segment (ROS) extracts
Each cGMP molecule acted upon by PDE6 releases a free proton when converted to GMP. The accumulation of protons in solution through PDE6 activity was monitored with the pH-sensitive fluorophore 5-(and-6)-Carboxy SNARF-1 (Invitrogen, Carlsbad, California), using a previously described method (Baker and  containing only buffer and cGMP/SNARF-1 was subtracted from all samples; and group background from wells containing the highest concentration of each inhibitor in buffer plus cGMP/SNARF-1 in the absence of protein was also subtracted. All data was first normalized so that the reading at t = 0 min was set to 0 A.U., and then the data from 580 nm was averaged with the absolute value of the data at 640 nm at each time point (they have in inverse relationship to decreasing pH). The average of the two emissions was used to limit fluorescence measurement artifacts. The data points were then normalized as a percentage of the saturated value for the reaction without inhibitors (black markers in Fig. 2). Three technical replicates were utilized for each condition and the data are presented as mean ± SD.

Ex vivo ERG and assessment of PDE6 inhibition constant
Female and male C57BL/6J mice at the age of 8-10 weeks were utilized in the ex vivo ERG experiments (n=10 mice). The mice were dark-adapted overnight and sacrificed with CO2 aspiration and cervical dislocation. Dissection of the eyes was executed under dim red light in a nutrition solution. The nutrition solution consisted of (in mM) Na + , 133.4; K + , 3.3; Mg 2+ , 2.0; Ca 2+ , 1.0; Cl − , 142.7; glucose, 10.0; EDTA, 0.01; HEPES, 12.0; and Leibovitz culture medium L-15 (0.72 mg/ml). Immediately after sacrificing, the eyes were enucleated and cut open along the equator. The retina was segregated from the sclera.
Subsequently, the retina was placed in a specimen holder (modified from (Donner, Hemila, and Koskelainen 1988)) on a filter paper with photoreceptors directed upwards.
A continuous flow (5-6 ml/min) of nutrition solution perfused the retina on the This article has not been copyedited and formatted. The final version may differ from this version. photoreceptor side. The temperature of the solution was kept at 36.0 ± 0.5 °C with a heat exchanger in contact with the specimen holder and a thermistor fixed near the retina. K + currents of Müller glial cells were obstructed by utilizing 50 μM BaCl2 (Bolnick, Walter, and Sillman 1979;Nymark et al. 2005). Moreover, 20 μΜ of D,L-2-amino-4phosphonobutyric acid (APB), a group III metabotropic glutamate receptor agonist, was added into the solution to isolate the photoreceptor component of the ERG response (Vinberg, Kolesnikov, and Kefalov 2014). CB-5083 was dissolved in dimethyl sulfoxide (DMSO) and added to the solution to achieve final concentrations of 31.3, 62.5, 93.8, or 125 nM. The control solution had the same concentration of DMSO as the solutions above, but no CB-5083. The pH was set to 7.5 with NaOH. CB-5083 was obtained from

MedChemExpress (Sollentuna, Sweden) and all other chemicals from Merck
Group/Sigma-Aldrich (Darmstadt, Germany). In previous studies, DMSO has been demonstrated to have a detrimental effect on retinal function (Tsai, Bui, and Vingrys 2009;Galvao et al. 2014). In our experiments, however, the concentration of DMSO in the perfusion solution was 0.00026 v/v%, which is well below the safe level proposed for intravitreal injections in rats (0.1 v/v%) (Tsai, Bui, and Vingrys 2009). Fig. S6 shows the behavior of response latency in photoreceptoral ex vivo ERG. The addition of DMSO to the perfusion did not cause an observable change in the response kinetics.
The isolated dark-adapted retina was stimulated with homogenous full-field flashes of light with 1 ms exposure time from the photoreceptor side. A fiber-coupled LED (M530F2, Thorlabs Inc, New Jersey, USA) with a nominal wavelength of 530 nm was employed to generate flashes of varying light intensity using a driver (DC2200, Thorlabs Inc, New Jersey, USA). The homogeneity of the beam was confirmed with a camera-based beam profiler (Model SP503U, Ophir-Spiricon Inc., Logan, UT, USA). The dimmest stimulus strength was selected based on its ability to induce ~10% response amplitude compared to the saturation level. The subsequent stimuli in the series increased by 0.5log-unit increments. The light power incident on the retina was measured with an optical power meter (PM100D, Thorlabs Sweden AB, Mölndal, Sweden). Light stimulus strengths were defined by calculating the number of photoisomerized rhodopsins per rod cell (R*rod -1 ), using the pigment template by (Govardovskii et al. 2000), the emission spectrum of the LED, and the sensitivity spectrum of the optical power meter, as described before (Heikkinen, Nymark, and Koskelainen 2008). The stimulus strengths used in the experiments ranged from 6.0 to 2900 R*rod -1 . Transretinal mass potential was recorded by two silver/silver chloride pellet electrodes (EP2, World Precision Instruments, Hitchin, UK) placed inside the solution-filled channels above and below the retina. The The method for determination of a compound's inhibition constant for lightactivated phosphodiesterase-6 (PDE6) via ex vivo ERG has been described in detail elsewhere (Turunen and Koskelainen 2018). Briefly, the introduction of the PDE6 inhibitor compound to the retina leads to a decrease in the amplification of the phototransduction.
This effect can be quantified by modeling the initial phase of the ERG responses with the activation model developed and described in detail by Pugh and Lamb (Pugh and Lamb 2000). Fitting the activation model equation (Turunen and Koskelainen 2018, Equation 6) to the response traces normalized to the saturation amplitude (

Histology and optical coherence tomography (OCT)
Histology and OCT were performed as described earlier (Leinonen et al. 2019;Leinonen 2020 4, panels L and M, were taken at the dorsal retina centered at ~800 µm from the ONH.
IS/OS and ONL layer and whole retina thicknesses were measured at 500 µm distance from ONH both at the superior and inferior sides and the superior-inferior average was used in Fig. 4N. Histology image morphometry was performed using ImageJ 1.52a software.

Statistical analysis
This article has not been copyedited and formatted. The final version may differ from this version. were compared using one-way ANOVA. All ANOVAs that found a significant betweensubjects or within-subjects/between-subjects interaction effect were followed by Bonferroni's post hoc tests. The data are presented as mean ± SD and the level of statistical significance was set at P<0.05.

CB-5083 suppresses mouse retinal function in vivo
PDE6 is an essential enzyme for energy transfer from the photons of visible light to electrical signals in our retinas, known as phototransduction. These electrical signals can be readily recorded with a distant electrode inserted close to the eyelids (in humans) or in contact with the eye´s cornea (in humans and animals) using a minimally invasive ERG method (Leinonen and Tanila 2018). Any systemically administered experimental compound that crosses the physical barrier of the eye, the blood-retina barrier, and inhibits PDE6 may reduce the phototransduction amplification that is easily apparent in the ERG signal. Therefore, we began our investigation by recording scotopic (rod- photoreceptor dominant) and photopic (cone-photoreceptor specific) ERGs first after vehicle treatment (baseline) and then after acute CB-5083 administration in 1-week intervals using the same mice. These experiments were performed in healthy WT mice starting 45 min after treatments.
CB-5083 decreased retinal sensitivity prominently at doses of 15 mg/kg and 30 mg/kg per mouse body weight (Fig. 1). Specifically, both dosing regimens completely abolished the ERG response to dim 0.0005 cd·s/m 2 light stimuli that cause a robust response under baseline, non-treated, conditions (Fig. 1A). With 1-2 log-units higher light stimuli (0.005 -0.05 cd·s/m 2 ), the ERG response started to recover characteristic waveform appearance after administration of 15 mg/kg CB-5083, but it was still significantly distorted after the 30 mg/kg dose. Quantitative measurements of ERG awave latency (Fig. 1B), amplitude (Fig. 1C), and b-wave amplitude (Fig. 1D) confirmed the significance of the effect. Among these parameters, the a-wave latency most reliably defines the drug effect on phototransduction. The modified scotopic ERG waveform caused by impaired phototransduction renders amplitude comparisons to the healthy/normal waveform less reliable. Recordings in the photopic condition ( Fig. 1E-H where flash stimuli is superimposed on a steady rod-suppressing background, revealed that CB-5083 might desensitize cone function, especially the UV-selective cone function in mice (Fig. 1F, H), even more so than the rod-function.  (Fig. S4).

CB-5083 reaches a high concentration in the mouse retina after an acute gavage administration
We next determined to what extend CB-5083 reaches the retina and what is the molar ratio between the circulating and retinal levels of the drug. We performed CB-5083 administration by gavage into mice similarly as before the acute in vivo ERG experiments.
To match the timing to what would be roughly halfway through the ERG experiment, we collected blood and tissue samples 1 h after drug administration. LC-MS analysis showed high drug concentrations in both the mouse serum and retinae (Fig. 1I). In fact, the molar CB-5083 concentrations in the mouse serum and in the perfused retinae were nearly identical, indicating that CB-5083 may freely cross the blood-retina barrier. To inspect CB-5083 levels in another neuronal tissue, we collected and analyzed frontal cortex biopsies. CB-5083 concentration in this brain region was ~ 3-fold lower compared to serum and retinae (Fig. 1I).

CB-5083 is a potent inhibitor of PDE6 in bovine ROS extracts
To test the hypothesis that CB-5083 targets PDE6, we performed a previously characterized PDE6-activity assay in bovine ROS (Baker and Palczewski 2011).
Consistent with our observations from acute in vivo ERG experiments (Figs. 1, S1, S2), we found that CB-5083's inhibitory potency on PDE6 is between those of sildenafil and vardenafil (Fig. 2). From the dosages we tested, sildenafil started to inhibit cGMP hydrolysis at 3 µM, whereas CB-5083 and vardenafil showed some inhibitory effect already at 0.03 µM concentration (Fig 2A). At 3 µM concentration, vardenafil strongly inhibited cGMP hydrolysis whereas CB-5083 had a moderate effect (Fig. 2C). With the largest dose tested, 30 µM (Fig 2D), vardenafil and CB-5083, to a slightly lesser extent, almost completely blocked cGMP hydrolysis during the 50 min experiment. Our assay using ROS extracts rather than purified PDE6 did not allow inspection of precise doseresponse relationships or determination of IC50 values.  To quantify the potency of CB-5083, we utilized a recently established ex vivo ERG-based method, enabling the determination of a compound's inhibition constant (Ki) against PDE6 in native conditions (Turunen and Koskelainen 2018). In this method, the inhibitor compound is introduced to the retina, causing a decrease in PDE6 catalytic activity and, thus, a reduction in the molecular amplification of phototransduction. The effect is quantified by analyzing the initial phase of the ERG signal in the theoretical framework of the phototransduction activation model developed by Pugh and Lamb (Lamb and Pugh 1992). We fitted the activation model to the leading edges of photoresponses recorded from 17 retinas in four different CB-5083 concentrations and in the control solution (Fig. 3B). The relative amplification constant was directly proportional to inhibitor concentration (Fig. 3C), which is consistent with Equation 1, and illustrates the dose-dependent effect of CB-5083 on photoreceptor function. Accordingly, the inhibition constant, Ki, against PDE6 was calculated as the reciprocal of the slope of the linear fit of the data points, leading to a value of Ki = 80 nM, which is 7-fold smaller compared to the Ki obtained for sildenafil previously, using the same method (Turunen and Koskelainen 2018).

Retinal function recovers after treatment cessation and no ocular abnormalities are observed after chronic CB-5083 treatment in middle-aged VCP-disease model mice
The strong dependence of phototransduction suppression on CB-5083 concentration indicates that its clinical application in VCP disease could be feasible after dose-titration, regardless of off-target effects on PDE6. However, this assertion requires that CB-5083´s negative effects on visual function are reversible, and that function will recover to normal, if unacceptable side effects occur. In addition, chronic retinal toxicity needs to be ruled out.
To assess the long-term ocular safety of chronic CB-5083 usage, we utilized middle-aged heterozygous VCP R155H knock-in mice. We dosed the mice via daily gavage (CB-5083, 15 mg/kg) for 6 months, over the period from 12 months of age to 18 months.
An equal number of the mice were dosed in parallel with the same volume of vehicle to serve as controls. After the 6-month treatment, we recorded ERGs after a 24-h period from the last administration to avoid the acute effects of CB-5083 on phototransduction.
At 18 months of age, both A-and B-wave amplitudes in vehicle-treated VCP mice were slightly smaller compared to age-matched, non-treated WT mice (Fig. 4B, D; 2-way ANOVA between-subjects effects: A-wave, F1.6=6.10, P<0.05; B-wave, F1.6=12.4, P<0.05). This could be an effect of long-term vehicle-treatment or related to mouse genetic background (not littermates), as we later recorded non-treated 24-month-old heterozygous VCP and slightly younger WT mice (19-month-old) and did not find statistically significant differences between them (Fig. S7). Therefore, it seems unlikely that the heterozygous VCP R155H knock-in mutation deleteriously affects visual function.
Importantly, our results from drug-treated mice show that the negative effects of CB-5083 on phototransduction are reversible, and visual function in mice recovers virtually to WT levels already after a 24-h washout, despite many months of continuous treatment ( Fig.   4A-F). After drug washout, we did detect a small but statistically significant increase in awave amplitude in CB-5083-treated versus vehicle-treated VCP mice ( Fig. 4B; 2 be either a type 1 error, or it could be a neural rebound effect after the 6-month-long partial phototransduction suppression by PDE6 inhibition. However, further investigation of this phenomenon was beyond the scope of the current study. Since CB-5083 has an off-target inhibitory effect on PDE6 that is most abundantly expressed in the photoreceptors, they would be the locus where long-term ocular toxicity would most likely occur. The thickness of the outer retina (photoreceptor nuclei and photoreceptor ciliary segments) is an established measure of photoreceptor toxicity and can be reproducibly measured with noninvasive OCT imaging (Marmor 2012). Therefore, we performed OCT after the 6-month-long trial in WT and VCP-disease mice. We did not detect any obvious abnormalities in the OCT images ( Fig. 4G-I) nor did we find differences in photoreceptor layer thickness among the study groups (Fig. 4J). Furthermore, a more high-resolution inspection from histological retinal sections did not reveal abnormalities to retinal structure after chronic CB-5083 treatment (Fig. 4L-M) or alterations in photoreceptor inner + outer segment, photoreceptor nuclei layer or whole retina thickness ( Fig. 4N), consistent with the recovered retinal function.

DISCUSSION
VCP disease is a rare genetic disorder caused by missense gain-of-function mutations in the VCP gene (Watts et al. 2004). One report estimated its prevalence in the UK at 0.66/million (Figueroa-Bonaparte et al. 2016). Nevertheless, the disease is devastating for the patients and their families. No curative therapies exist for these conditions, and due to their rarity, the major pharmaceutical industry is unlikely to pursue extensive research efforts. Typically, with respect to rare genetic disorders such as VCP disease, it is up to the academic community, patient advocate groups, small pharmaceutical/biotech This article has not been copyedited and formatted. The final version may differ from this version. companies, individual venture capitalists and/or philanthropists to make the effort and investment to discover prospective pharmacotherapies. One potential pharmacotherapy is the CB-5083 compound, originally developed to drive cytotoxic cellular responses in cancer (Le Moigne et al. 2017). Given the gain-of-function mechanisms of VCP-diseasecausing mutations, CB-5083 could potentially be repurposed for the treatment of neurodegenerative VCP disorders where it could be used to normalize aberrantly elevated activity. The option of drug repurposing is attractive, as major preclinical evaluations of CB-5083 have already been conducted and its general safety was documented in the recent Phase 1 clinical trial. However, the clinical trial in cancer patients was halted due to CB-5083's adverse effects on vision.
Since the documented side effects of CB-5083 closely resemble those observed in association with usage of erectile dysfunction drugs (Yafi, Sharlip, and Becher 2018; Moschos and Nitoda 2016), we hypothesized that they share the same mechanism. The erectile dysfunction drugs vardenafil, sildenafil and tadalafil are designed to selectively inhibit the PDE5 enzyme; however, especially vardenafil and sildenafil also inhibit PDE6 as an off-target effect (Hatzimouratidis et al. 2016;Zhang, Feng, and Cote 2005). As PDE6 is an essential enzyme in phototransduction in our photoreceptors, the severity of the side effect on vision depends on the drug's distribution properties from the blood circulation to the eye and on its inhibitory potency for PDE6, which can be described by the inhibition constant. Our investigation of CB-5083 concentration from mouse serum and retinae revealed that the drug crosses the blood-retina barrier essentially without resistance, so that the retinal drug level and the circulatory drug level were nearly the same after acute administration (Fig. 1I). Our ERG recordings in WT mice after acute CB-This article has not been copyedited and formatted. The final version may differ from this version. 5083 dosing showed that the drug significantly desensitized retinal function (Fig. 1A-H).
It particularly delayed the photoreceptor response (increased scotopic a-wave latency), consistent with the hypothesis of direct PDE6 inhibition (Turunen and Koskelainen 2018).
We recorded similar ERGs after acute administration of vardenafil (Fig. S1), sildenafil ( Fig. S2) and tadalafil (Fig. S3); only vardenafil affected retinal function as much, or more, than CB-5083 (Figs. 1, S1). Sildenafil and tadalafil had only a moderate phototransduction-suppressive effect with the relatively large dose of 30 mg/kg per mouse body weight (Figs. S2, S3). Interestingly, an experimental compound zaprinast, a relatively selective PDE6 inhibitor with Ki against PDE6 of 30 nM (10-fold selectivity for PDE6 compared to closely related PDE5) (Zhang, Feng, and Cote 2005), did not have any significant effects on the ERG signal (Fig. S4). This result indicates that zaprinast may not enter the eye from the circulation and reach a sufficient concentration to affect photoreceptor function.
To directly assess CB-5083's inhibitory action on PDE6, we utilized a wellcharacterized PDE6-activity assay (Baker and Palczewski 2011), using bovine ROS extracts. We found that treating the ROS extracts with CB-5083, sildenafil, or vardenafil significantly slowed the rate of cGMP hydrolysis, a reaction driven by PDE6 in this system. under its native conditions without biochemical manipulations, using a novel electrophysiology-based method (Turunen and Koskelainen 2018). We used the same method and setup to investigate the inhibitory potency of CB-5083. Our Ki value of 80 nM for CB-5083 (Fig. 3) shows it to be a 7-fold more potent inhibitor of light-activated PDE6 than sildenafil. CB-5083's dose-response relationship for phototransduction suppression followed the relation: (a non-linearized form of Eq. 1), with concentrations up to 125 nM in intact mouse retinas (Fig. 3). These results indicate that the clinical side effects of CB-5083 on vision are likely to be directly dose-dependent in a regular and predictable fashion.
Typically, the drug doses needed to suppress tumor growth in cancer therapy are significantly higher than needed for therapeutic efficacy in other indications, such as in autoimmune diseases (Liu et al. 2011). Accordingly, we hypothesize that a lower dose of  (Kimonis et al. 2008), and these patients would likely constitute the population in prospective clinical trials. After a 24-h washout period following 6-monthlong daily CB-5083 administration, the ERG signal practically recovers to normal (Fig. 4).
We also did not observe any abnormalities in retinal macrostructure by OCT imaging or standard histology. Retinal ONL thickness, an established and sensitive marker of photoreceptor toxicity (Marmor 2012), was the same among all study groups as assessed by two distinct methods. Taken together, our data show that CB-5083 dose-dependently inhibits PDE6, causing a reduction of phototransduction amplification that results in visual dysfunction. However, this effect is reversible and disappears shortly after drug use is ended.
Finally, even though the visual side effects of vardenafil and sildenafil are relatively common, millions of people have used them safely for decades (Yafi, Sharlip, and Becher 2018). The use of these drugs in erectile dysfunction is typically intermittent, but low dose  Zoumalan et al. 2009). However, as always with pharmacotherapy, the potential risks still need to be weighed with the benefits.
In conclusion, our data suggest that repurposing CB-5083 for the treatment of VCP disease or other applicable conditions is a viable option with respect to drug safety, as lowering drug doses would decrease the occurrence and severity of visual side effects.
In addition, according to our data from mice, these side effects are transient and do not cause permanent ocular toxicity. Finally, regarding drug research and development in general, we recommend adding in vivo and/or ex vivo ERG recordings as an economic and efficient method to screen drug effects on PDEs and neurotransmission.