ERG assessment of altered retinal function in canine models of retinitis pigmentosa and monitoring of response to translatable gene augmentation therapy

Abstract

Purpose:

To analyze ERG responses from two dog models of retinitis pigmentosa, one due to a PDE6A mutation and the other a CNGB1 mutation, both to assess the effect of these mutations on retinal function and the ability of gene augmentation therapy to restore normal function.

Methods:

Scotopic and photopic ERGs from young affected and normal control dogs and affected dogs following AAV-mediated gene augmentation therapy were analyzed. Parameters reflecting rod and cone function were collected by modeling the descending slope of the a-wave to measure receptor response and sensitivity. Rod-driven responses were further assessed by Naka-Rushton fitting of the first limb of the scotopic b-wave luminance:response plot.

Results:

PDE6A^−/− dogs showed a dramatic decrease in rod-driven responses with very reduced rod maximal responses and sensitivity. There was a minor reduction in the amplitude of maximal cone responses. In contrast, CNGB1^−/− dogs had some residual rod responses with reduced amplitude and sensitivity and normal cone responses. Following gene augmentation therapy rod parameters were substantially improved in both models with restoration of sensitivity parameters log S and log K and a large increase in log R_max in keeping with rescue of normal rod phototransduction in the treated retinal regions.

Conclusions:

Modeling of rod and cone a-waves and the luminance:response function of the scotopic b-wave characterized the loss of rod photoreceptor function in two dog models of retinitis pigmentosa and showed the effectiveness of gene augmentation therapy in restoring normal functional parameters.

Introduction

The electroretinogram (ERG) is an invaluable tool in characterizing the pathophysiological changes that occur in inherited retinal diseases, for monitoring disease progression and more recently for assessing recovery of function as treatments for these blinding conditions are developed. Dogs with spontaneous gene mutations have proven to be a source of valuable large animal models of a variety of human inherited retinal diseases such as the retinitis pigmentosas (RPs) [1–3] and are commonly used to study these conditions and develop translational therapies such as gene augmentation [1, 4–7]. The presence of a retinal region of high photoreceptor density with a cone-rich center, the area centralis, in canine eyes make them particularly valuable because of the similarity to human retina which has a macula with central fovea [8, 9]. Accurate characterization of canine retinal function and dysfunction is critical, both in the characterization of potential dog models of human retinal disease as well as for the assessment of translational therapies.

Two canine models of retinitis pigmentosa that have been utilized for assessment of translatable gene augmentation therapy are dogs with a mutation in rod cyclic GMP phosphodiesterase alpha subunit gene (PDE6A) and those with a mutation in the cyclic nucleotide gated channel beta 1 subunit gene (CNGB1) [10, 11]. The PDE6A mutation is a frameshift mutation resulting in a lack of PDE6 formation and complete failure in rod phototransduction resulting in a rapid loss of rod photoreceptors leaving the young animal with cone only function. At the age of retinal maturation (7 to 8 weeks of age) cone ERG a-waves are slightly reduced, while the cone b-wave has normal amplitude [10]. Studies in younger PDE6A^−/− dogs gives the opportunity to investigate cone function in the dog without interaction from rods. The mutant dogs do progressively lose cones secondary to loss of rods (similarly to RP patients). This is reflected in a progressive loss of cone ERG amplitudes over the first 12 months of age, and so ERGs must be recorded at a young age to assess cone-only function [10]. Given the early and rapid loss of rod photoreceptors, gene augmentation therapy to rescue rod function in PDE6A^−/− dogs must be performed at a young age. Successful gene augmentation therapy has been achieved using adeno-associated virus (AAV) vectors delivered by subretinal injection to express a normal copy of the gene in the remaining rod photoreceptors. This treatment restores rod function and preserves both rod and cone photoreceptors [5].

The CNGB1^−/− dog model of retinitis pigmentosa results from a truncating mutation in CNGB1 due to a complex mutation in exon 26 that leads to exon skipping and a premature stop codon. This results in the absence of full-length CNGB1a protein [11]. In the normal retina CNGB1a combines with CNGA1 in a 1 to 3 ratio to form the CNG channel that is critical for the rod response to light. In the absence of CNGB1a the trafficking of CNGA1 to the outer segment is severely impaired [11, 12]. The CNGB1^−/− dog lacks normal rod function, although there is evidence of residual desensitized rod function. There is no recordable ERG response to the ISCEV rod stimulus and the scotopic ERG has an elevated response threshold. However, stronger stimuli result in a combination of cone plus desensitized rod responses leading to a larger b-wave amplitude than results from cone only responses. The residual rod response is slower than the normal rod response and slowly diminishes with age reflecting the progressive rod loss [11]. In this model, rod photoreceptor loss is slower than that of the PDE6A^−/− dog. Secondary loss of cones follows rod loss but occurs more gradually than in the PDE6A^−/− dog. CNGB1a AAV-mediated gene augmentation therapy restores dark-adapted vision and results in long-term preservation of rod function [1].

The purpose of this paper is to describe and compare electroretinographic findings in normal dogs and the two dog models of retinitis pigmentosa both prior to and after gene augmentation therapy using previously established mathematical models that extract values for photoreceptor function and sensitivity from the ERG waveforms. Although these methods have been used extensively in humans and other primates, they have seen limited use in analysis of dog ERGs [1, 13, 14]. Calculation of these quantifiable parameters enhances the comparison of the pathophysiologic changes seen in inherited retinal disease and the response to gene augmentation therapy in dogs.

MATERIALS AND METHODS

Ethics Statement

All procedures were performed in accordance with the ARVO statement for the Use of Animals in Ophthalmic and Vision Research and approved by the Michigan State University Institutional Animal Care and Use Committee.

Animals

All dogs used for these studies were bred onto a laboratory beagle background. Purpose bred PDE6A^−/−, CNGB1^−/−, and unaffected dogs maintained in a colony at Michigan State University were investigated in this study. They were housed under 12hr:12hr light:dark cycles. PDE6A^−/− dogs included both 10 untreated animals and 6 young dogs who received subretinal injections of an adeno-associated viral vector serotype 2/8 delivering human PDE6A cDNA under control of a human rhodopsin promoter (AAV8-hRHO-hPDE6A) [5]. Untreated dogs were tested at a young age (1–3 months), which is while they have well preserved cone function and little to no rod response. CNGB1^−/− dogs included both 11 untreated animals and 4 young dogs who received subretinal injections of the recombinant AAV2/5 vector with the canine CNGB1 cDNA controlled by a human G protein–coupled receptor kinase 1 (GRK1) promoter (AAV5-hGRK1-cCngb1) [1]. Untreated dogs were tested at a young age (1–3 months), which is prior to loss of cone function and at which time there is a desensitized residual rod response. Eleven phenotypically normal beagle and beagle crosses were included in the study. Normal control dogs were tested as adults ranging from 6 months to 3 years of age. Gene augmentation therapy treated dogs were reported in previous publications and detailed methods of vector production and administration are included in those publications [1, 5]. The dogs that had gene augmentation therapy underwent ERGs on up to 9 occasions between 1 and 24 months post-treatment. The results from each visit are shown as separate points in the figures and the statistical analysis (see below) adapted to account for repeated measures.

Anesthesia

Adult dogs were premedicated (0.03 mg/kg intramuscular acepromazine) at least 15 minutes prior to anesthetic induction. Induction of anesthesia was achieved by isoflurane administered by mask for puppies and by 4–6 mg/kg intravenous propofol for older animals. Dogs were subsequently intubated and maintained under anesthesia with isoflurane (IsoFlo, Abbott Laboratories, North Chicago, IL, USA) [between 2–3.5% in a 1–2L/min oxygen flow via a rebreathing circle system for dog over 10 kg and via a Bain system for dog under 10 kg].

Electroretinography (ERG)

Electroretinograms were recorded as previously described [15]. Specific apparatuses are noted below. Briefly, dogs were dark-adapted for one hour prior to anesthetic induction and pupils dilated with tropicamide (Tropicamide Ophthalmic Solution UPS 1%, Falcon Pharmaceuticals Ltd., Fort Worth, TX, USA). An ERG-Jet electrode (Fabrinal Eye Care, La Chaux-De-Fonds, CH) which is a contact lens with inbuilt gold-leaf ring electrode, was used and for reference and grounding platinum needle skin electrodes (Grass Technologies, Warwick, RI, USA) were placed 5mm lateral to the lateral canthus and over the occiput respectively. ERGs were recorded using an Espion E² Electrophysiology system with ColorDome Ganzfeld (Diagnosys LLC, Lowell, MA).

White Flash ERG

For low strength stimuli, each flash was presented at one second intervals, and repeated to generate an averaged response detectable against background electrical noise. As stimulus strength was increased the time between flashes was lengthened to prevent light-adaptation of rod photopigment. After completion of the scotopic flashes, the animal was light-adapted (exposed to continuous, bright white light at 30 cd/m²) for 10 minutes to suppress rod responses, and the trials repeated in the light-adapted eye.

Mathematical Models

Rod-driven a-wave

To determine parameters for the rod-driven a-wave, we restricted our initial analysis to flashes ranging in strength from −1.6 to 1.4 log cd.s/m². For our calculations, the parameters were calculated after subtracting photopically matched ERG waveforms [16]. We fit the following equation described by Birch & Hood to the leading edge of the rod a-wave [17, 18]:

$$R(I,t)=(\mathit{1}-\exp [-I·S·{(t-t_d)}^2])·R_{max}fort>t_d$$ (Equation 1)

The amplitude R is a function of the retinal luminance I and time t after the flash and t_d is a brief delay. S is a sensitivity factor and R_max is the maximum amplitude of the response.

Cone-driven a-wave

To determine parameters for the cone-driven a-wave, we restricted our initial analysis to photopic ERG flashes ranging in strength from 0 to 1.4 log cd.s/m². We fit the following equation described by Birch & Hood [19] to the leading edge of the cone a-wave:

$$R(I,t)=(I·S_c·{(t-t_d)}^3)/(I·S_c·{(t-t_d)}^3+\mathit{1})·R_{m{p}_3}fort>t_d$$ (Equation 2)

The amplitude R is a function of the retinal luminance I and time t after the flash and t_d is a brief delay. S_c is a sensitivity factor and $R_{m{p}_3}$ is the maximum amplitude of the response.

Naka-Rushton function

We used an older and unrelated equation, the Naka-Rushton function, to fit the first limb of the b-wave luminance response plot [20]:

$$R/V_m=L^n/(L^n+K^n)$$ (Equation 3)

V_m represents the maximum response amplitude of the first limb of the b-wave luminance:response plot, K is a semi-saturation constant considered a measure of retinal sensitivity, and n is dependent on of the slope of the plot at the position of K, which may reflect retinal homogeneity [21–23].

Curve fitting

We calculated parameters using the lmfit curve-fitting program in the Python 3.6 environment [24], using the Levenberg-Marquardt algorithm to calculate optimal parameter values via least squares minimization [25]:

$$f({\mathit{X}}_i,\mathbf{\beta }+\mathbf{\delta })\approx f({\mathit{X}}_i,\mathbf{\beta })+{\mathbf{J}}_i\mathbf{\delta }$$

Where J_i is the gradient of f with respect to β. Successive calculation of the parameter δ that minimizes the sum of square of the residuals S is performed computationally until final model parameters are obtained [26, 27].

We determined model goodness-of-fit with the least-squares parameter, with values less than 0.25 considered a good fit [18]:

$$lsq=\frac{{\sum }_{i=1}^n{\left(y_i-f\left(X_I,\beta \right)\right)}^2}{{\sum }_{i=1}^n{\left(y_i-mean(y)\right)}^2}$$

Statistical analysis

All variables were tested for homoscedasticity using the Breusch-Pagan test, and for normality using the Shapiro-Wilks test, prior to calculation of the F-statistic using the F-test of the linear mixed effects model (LME).

To account for the repeated measures in the data, an LME model was employed utilizing the Statsmodels package in Python to examine statistical significance of serial ERGs performed in dogs following gene augmentation therapy, as well correlations between model parameters in all dogs, fitting the following equation:

$$Y_{ij}={\beta }_0+{\beta }_1{X}_{ij}+{\gamma }_i+{\epsilon }_{ij}$$

Y_ij is the j^th measured response for subject i, 𝑋_ij is the covariate for this response, 𝛾_i is the random effects parameter for subject i, and 𝜀_ij is the error parameter for this response. β₀ and β₁ are fixed effect parameters for all subjects, corresponding to intercept and slope, respectively, and are fit according to the restricted maximum likelihood (REML) optimized with the Broyden-Fletcher-Goldfarb-Shanno (BFGS) algorithm [28].

RESULTS

Characterization of the rod-driven a-wave

To fit equation 1 to the leading edge of the rod a-wave, we found that using four flash stimuli – −0.4, 0, 0.4, and 0.86 log cd.s/m² – yielded excellent fits in the phenotypically normal adult dogs (see representative normal dog in Figure 1). We calculated average parameters for untreated PDE6A^−/− dogs and PDE6A^−/− dogs following AAV gene augmentation therapy (representative tracings in Figure 2A and B; composite results in Figure 2C). Additionally, we calculated parameters for untreated CNGB1^−/− dogs and CNGB1^−/− dogs following AAV gene augmentation therapy (representative tracings in Figure 3A and B; composite results in Figure 3C). Average parameters are shown in Table 1.

Fig 1. Rod a-wave in normal dogs.

Rod-driven a-wave modeling of a normal dog with 4 different flash stimuli, ranging in strength from −0.4 to 0.86 log cd.s/m². Model parameters are provided in the inset

Fig 2. Changes in the scotopic a-wave of PDE6A^−/− dogs before and following gene augmentation therapy.

A. Rod-driven a-wave modeling of a PDE6A^−/− dog before treatment using 4 different flash stimuli, ranging in strength from −0.4 to 0.86 log cd.s/m². Model parameters are provided in the inset

B. Rod-driven a-wave modeling of the same PDE6A^−/− dog 3 months following gene augmentation therapy using 4 different flash stimuli, ranging in strength from −0.4 to 0.86 log cd.s/m². Model parameters are provided in the inset

C. Log R_max (amplitude) vs. log S (sensitivity) of untreated PDE6A^−/− dogs, PDE6A^−/− dogs following gene augmentation therapy, and normal dogs. This demonstrated a significant reduction in both parameters in the affected dogs and a significant increase in all eyes following gene augmentation therapy

Fig 3. Changes in the scotopic a-wave of CNGB1^−/− dogs before and following gene augmentation therapy.

A. Rod-driven a-wave modeling of a CNGB1^−/− dog before treatment using 4 different flash stimuli, ranging in strength from −0.4 to 0.86 log cd.s/m². Model parameters are provided in the inset

B. Rod-driven a-wave modeling of the same CNGB1^−/− dog 9 months following gene augmentation therapy using 4 different flash stimuli, ranging in strength from −0.4 to 0.86 log cd.s/m². Model parameters are provided in the inset

C. Log R_max (amplitude) vs. log S (sensitivity) of untreated CNGB1^−/− dogs (n = 14), CNGB1^−/− dogs following gene augmentation therapy (n = 4), and normal dogs (n = 12). This demonstrated a significant reduction in both parameters in the affected dogs and a significant increase in all treated eyes following gene augmentation therapy

Table 1.

Rod-driven a-wave parameters in normal dogs and dog models of retinitis pigmentosa. Shown as mean ± SD.

Rod a-wave Parameters	Normal (n = 12 eyes of 6 dogs)	PDE6A−/− untreated (n = 17 eyes of 10 dogs)	PDE6A−/− gene therapy (n = 6 eyes of 6 dogs)	CNGB1−/− untreated (n = 14 eyes of 7 dogs)	CNGB1−/− gene therapy (n = 4 eyes of 3 dogs)
log Rmax (μV)	1.969 ± 0.139	0.471 ± 0.164	1.02 ± 0.122	0.881 ± 0.288	1.539 ± 0.26
log S 1/(cd/m2 s3)	1.972 ± 0.105	1.547 ± 0.291	1.875 ± 0.284	1.507 ± 0.297	1.939 ± 0.099
td (mSec)	1.46 ± 0.317	1.784 ± 1.539	1.599 ± 1.066	2.169 ± 1.427	2.054 ± 0.89

In the PDE6A^−/− dogs, both log R_max (F=18.84, p=3.67×10⁻⁷⁹) and log S (F=4.32, p=1.54×10⁻⁵) values were significantly smaller in PDE6A^−/− untreated dogs compared to normal controls. A comparison of model parameters in pre-treatment to ERGs performed between 1- and 24-months post-treatment showed a significant increase in both R_max and S parameters following treatment. Furthermore, both log R_max (F=8.70, p=3.35×10⁻¹⁸) and log S (F=3.49, p=4.82×10⁻⁴) values were significantly increased post-treatment compared to untreated eyes. Compared to normal dogs, treated dogs had a lower mean value for log S but this difference was not statistically significant (F=1.23, p=0.22), although the log R_max was still significantly reduced (F=14.29, p=2.73×10⁻⁴⁶).

In the CNGB1^−/− dogs, both log R_max (F=8.86, p=7.79×10⁻¹⁹) and log S (F=5.01, p=5.56×10⁻⁷) values were significantly smaller in CNGB1^−/− untreated dogs compared to normal controls. Comparison of model parameters in untreated dogs to ERGs performed between 1- and 9-months post-treatment showed a significant increase in both R_max and S parameters following treatment. Furthermore, both log R_max (F=6.04, p=1.54×10⁻⁹) and log S (F=9.28, p=1.66×10⁻²⁰) values were significantly increased following gene augmentation therapy compared to untreated controls. Compared to normal dogs, the treated dogs had a lower mean log S value but this difference was not statistically significant (F=0.32, p=0.75), the mean log R_max value although substantially improved was still significantly lower than that of the normal control dogs (F=4.06, p=4.99×10⁻⁵).

Characterization of the cone-driven a-wave.

To fit equation 2 to the leading edge of the cone a-wave, we found that using four stimuli – 0, 0.4, and 0.86, and 1.4 log cd.s/m² – yielded excellent fits in the phenotypically normal adult dogs (see representative normal dog in Figure 4A). We further calculated model parameters young (< 3 months of age) untreated PDE6A^−/− dogs (representative tracing in Figure 4B) as well as young (< 3 months of age) untreated CNGB1^−/− dogs (representative tracing in Figure 4C); average parameter values are shown in Table 2. The only significant difference between normal and untreated affected dogs was a smaller log R_mp3 value in PDE6A^−/− animals (F=2.43, p=0.008) (Figure 4D). This improved in PDE6A^−/− dogs following AAV gene augmentation therapy, restoring the amplitude to statistical equivalence with normal dogs (F=0.98, p=0.33). However, there was no significant difference in the mean between treated and untreated dogs (F=1.82, p=0.07).

Fig 4. Cone a-wave in normal dogs and dog models of retinitis pigmentosa.

Cone-driven a-wave modeling using 4 different flash stimuli, ranging in strength from 0 to 1.4 log cd.s/m² of A. a normal dog. B. an untreated PDE6A^−/− dog. Model parameters were similar in both affected and normal dogs, albeit with a slight reduction in the amplitude parameter log R_mp3 and C. an untreated CNGB1^−/− dog. Model parameters were similar in both affected and normal dogs. D. Log R_mp3 (amplitude) vs. log S_c (sensitivity) of normal (n = 10), CNGB1^−/− (n = 8), and PDE6A^−/− (n = 10) dogs. In A-C model parameters are provided in the insets.

Table 2.

Cone-driven a-wave parameters in normal dogs and dog models of retinitis pigmentosa. Shown as mean ± SD.

Cone a-wave Parameters	Normal (n = 12 eyes of 6 dogs)	PDE6A−/− untreated (n = 10 eyes of 6 dogs)	PDE6A−/− gene therapy (n = 6 eyes of 6 dogs)	CNGB1−/− untreated (n = 14 eyes of 8 dogs)
log Rmp3 (μV)	1.039 ± 0.140	0.744 ± 0.133	0.878 ± 0.153	0.886 ± 0.236
log Sc 1/(cd/m2 s3)	3.510 ± 0.128	3.548 ± 0.247	3.561 ± 0.105	3.582 ± 0.183
teff (mSec)	0.062 ± 0.145	0.233 ± 0.555	0.161 ± 0.450	0.003 ± 0.012

Characterization of the first limb of the scotopic b-wave

After measurement of scotopic ERG b-wave amplitudes, we fit the Naka-Rushton function (equation 3) to the luminance:response scotopic ERGs recorded in the phenotypically normal adult dogs (representative response in Figure 5A and B). Following successful application of the Naka-Rushton function to scotopic ERGs in normal dogs, we calculated average parameters in untreated PDE6A^−/− dogs and PDE6A^−/− dogs following AAV gene augmentation therapy (representative responses in Figure 5C and D). Additionally, we fit the function to untreated CNGB1^−/− dogs and CNGB1^−/− dogs following AAV gene augmentation therapy (Table 3; representative responses in Figure 5E and F).

Fig 5. The scotopic b-wave in normal dogs and dog models of retinitis pigmentosa.

A. Luminance:response curve of the scotopic b-wave of a normal dog. Note the two ascending limbs, separated by an inflection point, which follow semi-saturation kinetics

Naka-Rushton (Michaelis-Menten) modeling of the luminance:response curve of the scotopic b-wave of: B. a normal dog (same dog as in A). C.an untreated PDE6A^−/− dog. The reduced amplitude and sensitivity are reflected in a reduction in the V_m parameter and an increase in the K parameter, respectively and D. a PDE6A^−/− dog 4 months following gene augmentation therapy. E. an untreated CNGB1^−/− dog. The reduced amplitude and sensitivity are reflected in a reduction in the V_m parameter and an increase in the K parameter, respectively. F. a CNGB1^−/− dog 3 months following gene augmentation therapy. In B – F the model parameters are provided in the insets.

Table 3.

Naka-Rushton equation parameters in normal dogs and dog models of retinitis pigmentosa. Shown as mean ± SD.

Naka-Rushton Parameters	Normal (n = 12 eyes of 6 dogs)	PDE6A−/− untreated (n = 17 eyes of 10 dogs)	PDE6A−/− gene therapy (n = 6 eyes of 6 dogs)	CNGB1−/− untreated (n = 11 eyes of 7 dogs)	CNGB1−/− gene therapy (n = 4 eyes of 3 dogs)
log Vm (μV)	2.159 ± 0.193	1.283 ± 0.21	1.659 ± 0.211	1.869 ± 0.29	1.820 ± 0.211
log K (cd.s/m2)	−2.4 ± 0.17	−0.57 ± 0.591	−2.333 ± 0.419	0.063 ± 0.262	−2.184 ± 0.296
n	1.355 ± 0.194	1.054 ± 0.232	1.475 ± 0.356	1.162 ± 0.258	1.327 ± 0.213

In PDE6A^−/− dogs, all three parameters log V_m (F=9.61, p=7.25×10⁻²²), log K (F=9.6, p=8.02×10⁻²²), and n (F=3.05, p=0.002) values were significantly different in affected dogs compared to normal controls, with significantly decreased log V_m and n values and a substantially increased log K parameter. Furthermore, significant increases in log V_m and n and a decrease in log K were seen in treated PDE6A^−/− dogs compared to untreated dogs, with log V_m (F=4.23, p=2.32×10⁻⁵), log K (F=11.73, p=9.28×10⁻³²), and n (F t=2.0, p=0.046). Compared to normal dogs, treated dogs demonstrated statistically equivalent log K (F=1.32, p=0.19) and n (F=1.67, p=0.09) values, although the log V_m was still significantly reduced (F=6.78, p=1.18×10⁻¹¹).

In CNGB1^−/− dogs, all three parameters log V_m (F=2.32, p=0.022), log K (F=19.83, p=5.7×10⁻⁸⁷), and n (F=2.41, p=0.016) values were significantly different in affected dogs compared to normal controls. Comparing treated CNGB1^−/− dogs to untreated controls, only the log K (F=13.0, p=1.24×10⁻³⁸) was significantly increased post-treatment. Compared to normal dogs, treated dogs demonstrated statistically equivalent log K (F=1.23, p=0.22) and n (F=0.41, p=0.68) values, although the log V_m was still significantly reduced (F=3.17, p=0.002).

Correlations between ERG waveforms

Within each group, we determined Pearson r values for all pairs of amplitude and sensitivity parameters calculated previously. The strongest, and only consistent, correlation was found between the rod-driven a-wave parameter R_max and the Naka-Rushton value V_m. Correlations between parameters were tested within groups and plotted with a line of best fit (Figure 6).

Fig 6. Correlations of amplitude parameters in normal dogs, untreated dog models of retinitis pigmentosa, and dogs following gene augmentation therapy.

Correlation of rod-driven b-wave amplitude parameter log V_m and rod-driven a-wave amplitude parameter log R_max. The r value for each goodness of fit is included. Inset key indicates genotype and treatment status and the significance of the fit to the dotted line.

Although correlations were significant for each group, there was an increase in the Pearson r value in both groups of dogs following gene augmentation therapy, with a slight increase in CNGB1^−/− dogs and substantial increase in PDE6A^−/− dogs. This finding fits with results from the previous sections – untreated PDE6A^−/− dogs have cone-driven dark-adapted a- and b-waves due to the absence of functional rods, whereas the scotopic ERG following gene augmentation has a substantial rod contribution. In contrast, untreated CNGB1^−/− dogs do have rod activity, albeit with altered kinetics and much lower sensitivity compared to that of normal controls. In these dogs, the scotopic b-wave is rod-driven before and after gene augmentation therapy and thus there is a smaller change in the correlation following treatment.

DISCUSSION

In this study, we utilized two established dog models of retinitis pigmentosa caused by rod-specific loss-of-function gene mutations. We assessed abnormal function as well as the effectiveness of gene augmentation therapy in normalizing retinal function. The leading edge of rod and cone a-waves were fit with mathematical models to measure photoreceptor response and sensitivity. The direct assessment of rod phototransduction response by a-wave modeling was also compared with Naka-Rushton fitting of the first limb of the luminance:response function of the rod b-wave which is used as a measure of retinal sensitivity. All the tested models provided excellent fits in both normal dogs and the two dog models of retinitis pigmentosa showing the applicability of these models for use in the dog.

Functional and morphological changes have previously been characterized in both dog models of retinitis pigmentosa [10, 11]. PDE6A^−/− dogs lack PDE6 in the rod outer segments [10]. This phosphodiesterase is a critical component of the rod phototransduction cascade and lack of PDE6 activity results in accumulation of cGMP in rods. This occurs during photoreceptor maturation in animal models with mutations that results in a lack of PDE6 activity resulting in a failure of rod outer segment elongation. In the PDE6A^−/− dogs rods initially have stunted outer segments and then undergo a relatively rapid degeneration triggered by cGMP accumulation [10]. Although cones do not express PDE6A, they are impacted by the abnormalities in the surrounding rods and do not develop outer segments of normal length during retinal maturation [10]. Dogs were used in this study at an age before substantial rod loss and prior to the secondary loss of cones that occurs later as the disease progresses. Rod a-wave modeling and Naka-Rushton fitting of the b-wave luminance:response plot showed a profound decrease in the scotopic log R_max and b-wave V_m parameters, which reflect the failure of rod phototransduction as would be expected with the lack of rod phosphodiesterase activity in this model [10]. There was also a decrease in the residual scotopic log S parameter, and substantial increase in the b-wave log K parameter, which reflects the decreased sensitivity of dark-adapted responses. The residual dark-adapted a-wave remaining following subtraction of the matched photopic response was very small and although it could possibly indicate a very small rod response it is most likely due to a difference between the response of dark- versus light-adapted cones. Finally, there was a slight decrease in the cone-driven log $R_{m{p}_3}$ parameter, although the sensitivity parameters were unchanged. The decrease in log $R_{m{p}_3}$ is likely to be explained by the stunting of cone outer segments which is apparent early in the disease process [10].

CNGB1^−/− dogs have a lack of CNGB1 in the rod outer segments and marked reduction in the other CNG channel subunit (CNGA1). For normal CNG channel function both subunits are required [12]. The normal light-induced closure of rod CNG channels interrupts the “dark current” and hyperpolarizes the rod. This generates the rod PIII response (fast PIII) which is a major generator of the rod a-wave. In contrast to PDE6A^−/− dogs who have profound early loss of rod function, the scotopic ERG in CNGB1^−/− dogs demonstrates evidence of residual, but markedly desensitized, rod function and relatively normal photopic responses [unpublished findings]. CNGB1^−/− dogs have decreased values for rod-driven log R_max and b-wave V_m parameters compared to normal dogs but the decrease is not so marked as in PDE6A^−/− dogs. This decrease in rod driven measures suggests either a reduction in the response from all rods, or a reduced population of rod photoreceptors capable of driving the scotopic response. The ‘n’ parameter derived by Naka-Rushton fitting of the b-wave luminance:response series is suggested to be an indication of retinal homogeneity and as that was comparable to that of normal dog, this suggests that there is likely a reduced response from all rods. However as discussed below, studies in mice with a similar mutation in CNGB1 suggested that only a low proportion of rods were functional [12]. Single cell recording in dogs would be needed to further investigate this. The CNGB1^−/− dogs do have a profound decrease in rod-driven log S parameters and a substantial increase in the b-wave log K parameter.

Taken together these findings support the presence in young CNGB1^−/− dogs of some residual rod function but with a marked loss of sensitivity. CNGB1^−/− dogs have low but detectable levels of CNGA1 in rod outer segments (unpublished observations). It is possible that the residual rod activity is accounted for by the presence of CNGA1 homomeric channels. In vitro studies have shown the CNGA1 subunits can form a functional channel but that there are profoundly altered dynamics compared to the native rod channel which consists of CNGA1: CNGB1a at a 3:1 ratio (see Kaupp and Seifert 2002 for a review [29]). Single rod cell recording from CNGB1^−/− mice which have a deletion of exon 26, and thus a very similar mutation to that of the dogs which have skipping of exon 26 and a premature stop codon (1, 2), showed residual light-induced response from a subset of rods [12]. The photopic a-wave parameters in the young CNGB1^−/− dogs examined in this study indicates that there is initially normal cone function. With the progressive slow rod loss in CNGB1^−/− dogs, eventually cone function secondarily declines [1].

We analyzed the effect of gene augmentation therapy on the ERG parameters in both dog models of RP. This is delivered by subretinal injection of recombinant AAV and transgene expression is effectively restricted to the retinal areas detached during the injection [30]. This means that the post-treatment recordings will consist of contributions from the treated retinal areas (mainly central) as well as from the retinal areas that did not receive a subretinal therapeutic injection. Despite not having the entire retina treated we found both scotopic sensitivity parameters (the rod-driven log S and b-wave log K) were restored to levels comparable to normal dogs in treated PDE6A^−/− and CNGB1^−/− dogs. These findings suggest that normal rod phototransduction is restored in the treated retinal regions. We additionally found significant increases in the rod-driven log R_max and (in treated PDE6A^−/− dogs) b-wave V_m amplitude parameters compared to untreated controls. These measures did not reach the values recorded from normal dogs, most likely reflecting the fact that only a part of the retina (approximately one-third) received therapy.

Furthermore, in gene augmentation therapy treated PDE6A^−/− dogs we found a slight increase in the cone-driven log $R_{m{p}_3}$ parameter to levels comparable to that in normal dogs (although it was not significantly increased compared to untreated controls, which may indicate a type II statistical error due to the relatively low number of treated eyes in the study). Histological examination of treated PDE6A^−/− dogs showed that cones in the treated region had longer normal appearing outer segments compared to cones in the untreated retinal regions [5]. Future investigation should examine the correlation between percentage of retina treated and the improvement in these measures of retinal function [10, 11]. The gene augmentation therapy in both models was performed at a young age, prior to substantial loss of photoreceptors [1, 5]. Assessment of the rescue of the ERG parameters of dogs treated at later disease stages with less remaining “rescuable” rods will be of interest.

We also looked to see if there was a correlation between amplitude and sensitivity parameters across the tested models in all dogs included in this study. In normal dogs the only significant correlation was between the amplitude parameters from modeling the rod a-wave (R_max) and the first limb of the scotopic b-wave luminance:response plot (V_m). Although we also found this correlation in untreated CNGB1^−/− and PDE6A^−/− dogs, the strength of the relationship increased following gene augmentation therapy. This change was small in CNGB1^−/− dogs, while the improvement was much greater in the PDE6A^−/− dogs, likely reflecting the substantial increases in rod responses compared to that of the untreated eyes. As untreated CNGB1^−/− dogs have rod-driven dark-adapted responses, albeit with markedly reduced sensitivity, the most substantial increase was in rod sensitivity, and the proportionally lower increase in the correlation of amplitude parameters mirrors the smaller increase in maximal rod response. Finally, the lack of correlation of the sensitivity parameters S and K from rod a-wave modeling and Naka-Rushton fitting, respectively, may be due to differences in model design – however this negative finding was in all dogs and indicates that the slope of the a-wave (reflecting the speed of the photoreceptor-driven response) does not have a direct relationship with the semi-saturation constant of the b-wave (the threshold of the half-maximal bipolar cell response). This may be because multiple rods synapse with each bipolar cell and the sensitivity of the rod input does not have a linear correlation with bipolar cell generated electrical currents.

There are several potential limitations to the results of modeling of the ERG performed in this study. It is well recognized that postreceptoral contributions to the photopic a-wave are significant in some species such as primates – this brings into question the interpretation of the a-wave model to represent cone phototransduction parameters [31, 32]. However in species such as mice and rats [33, 34] as well as the dog [Petersen-Jones et al unpublished ERG drug dissection studies] postreceptoral cells provide a less substantial contribution to the cone a-wave than in primates. Thus, in these species it is likely that photopic a-wave modeling more accurately represents cone phototransduction. Low-amplitude ERGs in the disease models make it more difficult to accurately assess sensitivity of responses as the low signal is more difficult to detect above background noise. There may be utility for frequency-based approaches in such instances, such as the relatively recent development of wavelet analysis in signal processing to examine time-sensitive components of the ERG signal [35, 36].

In summary, we investigated the photoreceptor amplitude and sensitivity parameters and rod b-wave parameters derived from the ERG waveforms of normal dogs and two dog models of retinitis pigmentosa at early stages in the disease process as well as in dogs following gene augmentation therapy. These results supported previously described ERG changes and correlated to physiologic and morphologic retinal findings. Specifically, we found significant decreases in rod amplitude and sensitivity parameters in both dog models. In the PDE6A^−/− dog model this appeared to reflect a profound loss of rod responses while in the CNGB1^−/− model there was evidence of some residual but markedly desensitized rod function. Following gene augmentation therapy, rod sensitivity was restored and response amplitude increased but not to the level of normal dogs. This suggests that normal rod phototransduction is restored in the treated retinal regions but because a smaller than normal retinal area is contributing to the normal ERG the overall amplitude of response does not meet that of normal dogs. Finally, we showed that established models of the rod- and cone-driven a-waves, and the ON-bipolar cell driven b-wave, are suitable for use in fitting the canine ERG and should be considered in future ERG studies to characterize and corroborate pathophysiologic changes in dog models of retinitis pigmentosa.

Availability of data and material:

the datasets generated and analyzed during the current study are available from the corresponding author on reasonable request