|Home | About | Journals | Submit | Contact Us | Français|
The GUCY2D gene encodes retinal membrane guanylyl cyclase (RetGC1), a key component of the phototransduction machinery in photoreceptors. Mutations in GUCY2D cause Leber congenital amaurosis type 1 (LCA1), an autosomal recessive human retinal blinding disease. The effects of RetGC1 deficiency on human rod and cone photoreceptor structure and function are currently unknown. To move LCA1 closer to clinical trials, we characterized a cohort of patients (ages 6 months—37 years) with GUCY2D mutations. In vivo analyses of retinal architecture indicated intact rod photoreceptors in all patients but abnormalities in foveal cones. By functional phenotype, there were patients with and those without detectable cone vision. Rod vision could be retained and did not correlate with the extent of cone vision or age. In patients without cone vision, rod vision functioned unsaturated under bright ambient illumination. In vitro analyses of the mutant alleles showed that in addition to the major truncation of the essential catalytic domain in RetGC1, some missense mutations in LCA1 patients result in a severe loss of function by inactivating its catalytic activity and/or ability to interact with the activator proteins, GCAPs. The differences in rod sensitivities among patients were not explained by the biochemical properties of the mutants. However, the RetGC1 mutant alleles with remaining biochemical activity in vitro were associated with retained cone vision in vivo. We postulate a relationship between the level of RetGC1 activity and the degree of cone vision abnormality, and argue for cone function being the efficacy outcome in clinical trials of gene augmentation therapy in LCA1.
Light absorbed by vertebrate photoreceptors triggers electrical signals by the process of phototransduction, the complex G-protein-mediated biochemical cascade in rods and cones (reviewed in 1–4). The discovery of key genes/proteins in phototransduction activation, amplification and deactivation was followed by finding associations between human retinal diseases and the mutant phototransduction-pathway genes (reviewed in 5,6).
Leber congenital amaurosis (LCA) is a group of early-onset human retinal blinding diseases, which includes at least 16 molecular causes with different pathogeneses (7,8). LCA1, the first LCA locus identified (9), is caused by recessive mutations in GUCY2D (10), which encodes retinal membrane guanylyl cyclase (RetGC1), a key enzyme along the phototransduction pathway in photoreceptors (11–15). Mutations in GUCY2D account for ~6–12% of all genotyped LCA (7,8). In the human retina, RetGC1 is detected mainly in the outer segments of rod and cone photoreceptors (11,16).
What do we know about the effects of RetGC1 deficiency in human rod and cone photoreceptors? The RetGC1 deficiency state, LCA1, has been described mainly in clinical terms—such as inability to follow light, photophobia, preference for bright lights, preference for dim lights, hyperopia, severely reduced visual acuities, nystagmus, normal-appearing fundus or salt-and-pepper appearance or early macular and retinal degeneration, non-recordable electroretinograms (ERGs), and no measurable visual fields (8,17–21). There are discordant histopathological observations in the literature: a 26-week-old abortus from a family with LCA1 that had severe retinal degeneration on post-mortem retinal examination (22); a 12-year-old LCA patient with a GUCY2D mutation showing photoreceptor degeneration (23); and non-invasive cross-sectional retinal imaging of a 31-year-old patient described as showing normal macular thickness (20) and two patients in the third and sixth decade of life reported as similar to normal in retinal architecture (24).
We analyzed retinal structure and function in a cohort of LCA1 patients harboring different mutations in the GUCY2D gene and studied the biochemical effects of these mutations on RetGC1 activity in vitro. Key similarities as well as differences were found among affected individuals, and in vitro studies helped explain some of these differences. The results provide the first opportunity to determine whether human results are being faithfully modeled by naturally occurring avian mutants and genetically engineered mice (25–30).The importance of such a comparison at this time is that the animals are currently being used for proof-of-concept studies with the intent to advance to gene augmentation therapy in LCA1 clinical trials (31–35).
Clinical features of our LCA1 cohort (6 months to 37 years of age; n = 11; Table 1) were similar to those previously reported (8,17–21). All patients had nystagmus with visual impairment noted in the first year of life. Ophthalmoscopic findings at the ages examined included retinal vessel attenuation and a granular appearance to the peripheral fundus. P11 had macular pigmentary disturbances. Visual acuity was abnormal and ranged from 20/100 to light perception. Visual fields by kinetic perimetry were detectable in only four patients: P4, age 11, and P8, age 22 showed detectable vision to large bright targets (V-4e) across much of the field, whereas P7, age 19, and P9, age 22, had small central or peripheral islands of vision, respectively, to this target.
En face views of retinal pigment epithelium (RPE) health using autofluorescence imaging are shown for a representative normal subject and three individuals affected with LCA1 (Fig. 1A; Table 1). The image of P8 approximates that of the normal and indicates relatively well-preserved RPE. P7 and P11 have central retinal RPE abnormalities to different degrees: a ‘bull's eye’ appearance in P7 and a wider region of central RPE lipofuscin disturbance in P11.
Cross-sectional optical images along the horizontal meridian illustrate the laminar architecture of the otherwise-transparent retina (Fig. 1B). Normal cross-sections show discernible laminae: cellular layers are of low reflectivity and intervening higher-reflectivity laminae represent synaptic and nerve-fiber connections. Deep in the retina is a series of hyperreflective signals that can be subdivided into components representing photoreceptor inner and outer segments, RPE, and anterior choroid (36–40). The retinal laminar architecture of P7 shows very little difference to that of the normal section. The photoreceptor nuclear layer (or outer nuclear layer, ONL) peaks centrally and declines with distance from the fovea. In contrast, the cross-sectional image of P11 is abnormal, with loss of structural integrity of the ONL in the very central retina but preserved ONL to either side of this foveal region. The quantitation of the photoreceptor nuclear layer across central, temporal and nasal retina in the 11 LCA1 patients is shown (Fig. 1C). Only in the central (foveal) retina were there abnormal reductions in ONL thickness and these were dramatic in P10 and P11 but borderline in P6 and P5. Photoreceptor-loss profiles were plotted to define better the extent of the defect in those patients with abnormalities (Fig. 1D). There is a tendency in all LCA1 patients to show a central depression of ONL relative to the ONL fraction at extracentral locations. To estimate whether the central retinal defect was an exclusive loss of cone photoreceptor nuclei, which have the highest density in primate foveas (41), the ratio of rod to cone densities is also plotted (Fig. 1E). The central cell loss is mainly that of cone nuclei, but as the dimension of the lesion increases, rod nuclei may also be lost to the degenerative process. Of importance is that in the rod-dominated retina outside of the fovea, all ONL thickness measurements are within normal limits. The latter measurement, however, could not determine whether extra-central cone cells were lost.
Localization studies place the GUCY2D gene product in the outer segments of human rods and cones (11–13,42). Considering the severity of visual dysfunction in LCA1, we asked whether there was loss of the outer segments of rods and cones in these retinas (Fig. 1F). Rod outer segment (ROS) laminae were present and thickness was within normal limits in an extracentral region of high rod density (2–7 mm temporal retina). A similar measurement but in the central 1 mm of retina would quantify cone outer segment (COS) thickness. This measure was within normal limits, at the lower limit of the normal range in P9, or definitely abnormal in P11, who also showed the major loss of foveal photoreceptor nuclei. Laminar architecture of the central retina in all LCA1 patients, however, was abnormal. The hyperreflectivity attributed normally to COS tips was not discernible in any of the patient scans.
In summary, in vivo measures of photoreceptor cell layer thickness in LCA1 were within normal limits with the exception of some individuals who showed foveal losses. The interface between COSs and RPE, however, was not normal in any patient.
Visual function in LCA1 was assayed with psychophysics, ERGs, pupillometry and performance on a mobility course. Full-field sensitivity testing (FST) was performed in nine patients (Fig. 2A). Chromatic measurements showed that all nine patients detected blue stimuli using rods. Sensitivity to the blue stimulus was abnormally reduced in all but one patient, P3, but there were different degrees of abnormality. P2, P11, P4 and P9 showed 5.6, 5.5, 5.2 and 5.2 log10 units (l.u.) sensitivity, respectively, compared with the normal average of 6.6 l.u. (±2 SD, ±0.6), indicating ~1–1.5 l.u. of rod sensitivity loss. This is in contrast to P6, P7, P8 and P5 whose average sensitivities were 3.2, 2.3, 2.1 and 1.7, respectively, indicating ~3.5 to 5 l.u. of rod sensitivity loss. It can be concluded that substantial rod function is detectable by full-field psychophysics in GUCY2D-LCA patients; there was no clear relationship of patient age to the presence of residual rod function (Fig. 2A).
The presence of considerable psychophysically measured rod function in P3, P11, P9 and P4 was confirmed objectively with ERGs (Fig. 2B). White flashes of light of increasing intensity in the dark-adapted state in normal subjects elicited ERG waveforms that increased in the amplitude of the b-wave (positive component) and the a-wave (negative component) with brighter stimuli. Waveforms in P3 were not detectable at the lowest intensity shown for the normal subject but there was a response to all other intensities, indicating a threshold elevation of ~0.5–1.0 l.u. compared with normal (based on blue light intensity series; not shown). P9, P4 and P11 (not shown) had no detectable waveforms for intensities below −1.24 log10 (phot) cd.s.m−2; thresholds are ~2.5 l.u. elevated compared with normal. The maximum waveform is similar in these three patients and there are both b- and a-wave components. P7, representative of all other patients in this cohort (Table 1), had no detectable waveforms even to the brightest stimuli. Cone ERGs to single white flashes on a background or to 29 Hz flickering white light are shown for the normal subject (Fig. 2B). None of the patients in this study had detectable cone ERGs to these stimuli.
The transient pupillary light reflex (TPLR) was used to evaluate objectively the transmission of visual information from the retina to higher visual centers (43). Families of TPLR responses elicited with brief stimuli in a representative normal subject are shown for comparison with those from two patients with GUCY2D-LCA (Supplementary Material, Fig. S1). The TPLR threshold of P11 was near −4.7 log10 scot-cd.m−2 (Supplementary Material, Fig. S1B) and not substantially different from normal (Supplementary Material, Fig. S1A). In P10 on the other hand, there was no detectable TPLR response throughout the tested luminance range (Supplementary Material, Fig. S1C). Seven other patients (P2-P4, P6-P9) showed a spectrum of TPLR sensitivity losses and there was good correlation between TPLR and FST estimates (Supplementary Material, Fig. S1D).
The regional retinal variation of the detectable rod function by FST was studied with chromatic thresholds in the dark-adapted state across the visual field in five of the patients (Fig. 2C; 44). P9 and P11 had detectable rod sensitivity at most loci with an average reduction of ~2.5 l.u. The rod sensitivity map for P7 showed no detectable perception, indicating ~5 l.u. of sensitivity loss (FST has a greater dynamic range of stimulus intensities, Fig. 2A). Unlike the cone visual field map of normal subjects, there was no detectable cone sensitivity in most patients, except for some central field loci in P8 and P4 (not shown) and a single central locus in P7, all three of whom were among those with the best acuities measured in this cohort (Table 1).
We also asked whether these psychophysical, ERGs and pupillometric test results had any relation to the ability of subjects to navigate an obstacle course under different ambient illuminations (Fig. 2D). P2, P3, P4 and P9, who had considerable rod function, showed a mobility performance with no incidents for the lower two ambient illuminations (0.2 and 0.6 lux) and between 0 and 1 incidents at the higher illumination levels (2, 4 and 100 lux). P6 and P10 had greater rod function losses and found the task difficult at all light levels.
To study the consequences of GUCY2D mutations on visual adaptation to ambient illumination (45), we developed a testing paradigm applicable to the low levels of vision in LCA1. Full-field chromatic stimuli (46,47) on full-field achromatic backgrounds were used to examine light adaptation properties of rod- and cone-based vision with psychophysical increment thresholds adapted from our previous studies (48–50). Sensitivity of the rod- (and cone-) mediated systems, demonstrating adaptation across ~7 l.u.-wide range of background light intensity in normal subjects, is shown relative to absolute dark-adapted (and cone-plateau) sensitivity (Fig. 3A, upper left). Normal rod vision usually starts showing sensitivity changes near −3 log10 scot-Td background, and adaptation continues along a linear trajectory on log-log coordinates similar to previously published results (48,49,51–54). At backgrounds higher than 1 log10 scot-Td, cone-mediated sensitivities take over and show adaptation along a linear trajectory, also (48,49,55). Saturation of the rod adaptation is normally hidden by cone-mediated vision but can be demonstrated under special experimental conditions or in patients lacking cone vision (Supplementary Material, Fig. S2A).
Two patterns of results were present in the 7 LCA1 patients who were studied for light adaptation (Fig. 3). The first pattern (illustrated in data from P3, P4 and P7) showed detectable cone-mediated sensitivity and different degrees of abnormality in rod-mediated sensitivity. Rod-mediated sensitivities of P3, P4 and P7 were reduced by 0.8, 1.7 and 4.2 l.u., respectively, compared with normal (Fig. 3A, lower panels). Equal shifts of the normal light adaptation function up and to the right could explain the results of P3 and P4; whereas for P7, the right-shift of the curve (1.5 l.u.) was substantially less than the sensitivity loss. Cone-mediated sensitivities were also detectable in P3, P4 and P7 but were reduced compared with normal (Fig. 3A, upper panels). At high backgrounds, there was evidence of light adaptation of the cones in all three patients. The right-shift of the normal curves (by 0.5, 1 and 1.3 l.u., respectively) was less than the sensitivity losses. Results in P8 (not shown) also supported abnormal but detectable rod- and cone-mediated function, but formal light adaptation testing was not performed.
Four GUCY2D-LCA patients (P2, P6, P9 and P11) showed detectable rod function but no detectable cone function; the sensitivities in these four patients were mediated by the rod system even with the highest background light intensities (Fig. 3B). P11 had only 1.1 l.u. loss of sensitivity and thus represented the patient closest to mean normal in this category. Upon a 0.7 l.u. horizontal shift, lower range light adaptation of P11's rod system could be described with the normal function. The exception was at the higher background intensities, where the normal switch over to the cone system was lacking; chromatic stimuli confirmed continued rod-mediation. Also, the expected saturation of the rod system was lacking. Instead, the rod system continued to adapt along a linear trajectory from the lower adaptation levels supporting Weber behavior in an ambient light regime normally involving saturation of rods (see Supplementary Material, Fig. S2B). P2 and P9 had 1.7–1.8 l.u. losses of rod sensitivity and the majority of their light adaptation was consistent with the normal curve horizontally shifted by 0.7 and 1.0 l.u., respectively. At the highest background intensities, both patients also appear to deviate from the expected normal saturation in a manner similar to that of P11. An apparent expansion of the range of rod light adaptation is difficult to reconcile with the hypothesis that normal modulation of guanylyl cyclase activity is an important contributor to the light adaptation of rod photoreceptors. To find out whether this extended Weber behavior is intrinsic to RetGC1-deficient rods, we carried out suction-pipette recordings from single rods of the RetGC1−/− mouse. However, we did not detect such extended Weber behavior (Supplementary Material, Fig. S2B, large open and red triangles). Finally, P6 had a large (4.3 l.u.) loss of rod sensitivity, and no evidence of cone function. Light adaptation of P6 could be explained by the right-shift of the normal curve by 2.5 l.u.
RetGC1 is a membrane form of guanylyl cyclase (11,56,57), with a single transmembrane region separating the extracellular domain (ECD) of RetGC1 from the cytoplasmic portion of the protein that contains a kinase-homology domain (KHD) and the catalytic domain (CAT), both connected via a short dimerization domain (DD) required for RetGC1 to form a functional homodimer [58; Fig. 4]. Mutations detected in the GUCY2D gene of our cohort of LCA patients can be divided into two major groups based on their projected functional consequences (Fig. 4). Some of them a priori destroy the enzymatic activity of RetGC1, by eliminating the vital portion of the enzyme. The Leu42del6 deletion mutation eliminates a portion of the N-terminal signal peptide in RetGC1 (56,57), required for proper synthesis of RetGC1 and its integration into the membrane, thus rendering photoreceptors unable to express functional RetGC1. The M476ins10 insertion causes a frame-shift after Gly481, which ends the protein sequence 77 amino acid residues later, soon after the transmembrane domain. Similar to that, the Gln545x nonsense mutation also truncates a major portion of the intracellular segment of the cyclase, thus eliminating its entire function.
Four mutations likely cause deletion in the CAT per se (Fig. 4). The Leu865del1 and Ser981del1 mutations generate a frame-shift with the subsequent termination of the primary structure after 13 and 38 out-of frame amino acid residues, respectively. The +1 splice site change of G > C in one allele of P7 has an uncertain consequence (some splice variants only mis-splice some of the time); it may affect a splicing signal and truncate the CAT after Ser1076. In preliminary experiments, we established that the elimination of the last 21 amino acid residues from the C-terminus of RetGC1 by a frame-shift mutation is sufficient to inactivate it (data not shown). Therefore, two of the three mutations a priori render the cyclase inactive while the effect of the splice variant remains speculative. However, the Arg1091x deletes a shorter 13 amino acid residues long fragment from the C-terminus of the protein. Hence, the effect of that naturally occurring non-sense mutation could not be predicted based on the previously available biochemical data.
There were also several missense mutations in this cohort of LCA1 patients: the Ser248Trp located in the ECD portion of RetGC1, Arg768Trp in the KHD domain and the His980Leu in the CAT. The recently biochemically characterized Arg768Trp mutation (14) inactivates RetGC1 complex with the activator protein, GCAP1. However, RetGC1 in living photoreceptors is regulated not only by GCAP1 but also by GCAP2 (59), and it remains unclear whether this mutation can truly create a RetGC1 null condition in vivo.
The Arg822Pro mutation is located in the DD, and could affect the catalytic activity of the cyclase, which is active only as a homodimer (58,60). The His980Leu mutation in the CAT domain does not directly affect amino acid residues that form the active site of the CAT required for the conversion of GTP to cGMP (60,61). However, its close proximity to the active site argues that this mutation could adversely affect the enzymatic activity.
The potential effect of Ser248Trp substitution also remained to be tested, because although catalytic activity and the ability to bind GCAPs do not require the ECD portion of the cyclase (14,62), we cannot a priori exclude that altered structure of the ECD segment can indirectly affect the cytoplasmic portion of the enzyme dimer. Therefore, we analyzed the GCAP-dependent activity of the five RetGC1 mutants (Ser248Trp, Arg768Trp, Arg822Pro, His980Leu and Arg1091x) in vitro.
We expressed wild-type RetGC1 and the five mutants in HEK293 cells and reconstituted the membranes containing recombinant RetGC1 (Fig. 5A) with purified GCAP1 and GCAP2 expressed in E. coli at saturating free Mg2+ concentrations in the presence of EGTA, which is the optimal condition for the RetGC1 activity (15,63).
The activity of the recombinant Ser248Trp RetGC1 was slightly above that of the wild-type in the presence of GCAP1 (Student's t-test, P< 0.0001) and GCAP2 (P = 0.007) (Fig. 5B–D), but the difference was rather small (by ~23–25%). In contrast, the Arg822Pro, Arg768Trp and His980Leu all showed little or no activity in the presence of either GCAP1 isoform (30- to 100-fold reduction when compared with the wild-type) or GCAP2 (P< 0.0005) (Fig. 5B and D). There was also a substantial, 4–5-fold, reduction in activity in the Arg1091x RetGC1 with both GCAP1 (P< 0.0001) and GCAP2 (P = 0.0003) (Fig. 5B and C). The apparent affinities for GCAP1 in Ser248Trp and Arg1091x mutants for GCAP1 were similar to that of the wild-type RetGC1 (1.3, 1.1 and 0.7 µM, respectively) (Fig. 5C), well below the estimated in vivo concentrations of GCAPs (15,64). These results argue that the low activity of the Arg1091x mutant was caused by suppression of its catalytic activity, rather than its affinity for the activator protein. Ca2+ sensitivity of the Ser248Trp and Arg1091x mutants remained normal (Supplementary Material, Fig. S3).
GUCY2D-LCA patients have normal photoreceptor laminar architecture except for foveal COS abnormalities and, in some patients, foveal photoreceptor cell losses. This phenotype is unlike any other molecular subtype of LCA studied in detail to date (37,38,65–70). Whereas the degree of visual disturbance with conventional clinical testing approximates the severity of, for example, AIPL1-LCA and CEP290-LCA, the former shows major loss of photoreceptors across wide expanses of retina and the latter, in distinct contrast, retains mainly foveal cone cells (69–71). The measurements of retinal structure in the current cohort of GUCY2D-LCA extend previous qualitative observations of normal-appearing OCT scans in three such patients (20,24). The extent of foveal degeneration, when analyzed for its relationship with known human rod:cone ratios (41, Fig. 1), suggested that rod photoreceptors are also included at the edges of the central lesion. Ideally, cone (and rod) cell counting, as can occur with certain adaptive optics techniques, may be helpful to understand the level of cone disease outside the fovea in GUCY2D-LCA (72).
There were two discernible LCA1-disease expressions based on cone function. The majority (7/11) of the patients had no detectable cone-mediated visual function by psychophysical and electrophysiological methods, a life-long lack of perception of colors and a severe loss of visual acuity. The retinal structural abnormalities we observed in the fovea would not be predicted to lead to such profound cone dysfunction. This suggests a primary functional deficit in human cones due to the GUCY2D mutations. The remaining (4/11) patients had detectable (but abnormal) cone function by psychophysical methods, central fixation, perception of colors and relatively retained (but abnormal) visual acuity. This level of cone function may likely be more closely aligned with the minor retinal structural abnormalities observed.
The majority of GUCY2D-LCA patients had detectable but abnormal rod function; under dark-adapted conditions, rod sensitivity could be from 0.5 to 5 l.u. reduced compared with normal. The degree of retained rod function was not correlated with that of cone function. In a subset of patients, we tested the hypothesis that their rod function was actually useful for vision by quantifying behavior in a mobility course. There was definite avoidance of obstacles at low levels of ambient illumination in patients with relatively retained rod function and this was not the case in patients with more impaired rod vision. Pupillometric sensitivity in the dark closely followed the perceptual estimates of sensitivity and provided objective evidence of rod function. Substantially retained rod photoreceptor function in GUCY2D-LCA patients has not been previously reported, but it would be consistent with rod ERG recordings in RetGC1−/− mice (27,30,59).
Rod vision of GUCY2D-LCA patients showed clear evidence of adaptation to scotopic ambient conditions (−3 to 0 log10 scot-Td). Adaptation to this range of light levels is thought to be normally dominated by post-receptoral mechanisms (51,52,73) and our results suggest a lack of substantial contribution from RetGC1 to this mechanism even though some studies have indicated possible presence of RetGC1 in synaptic layers of the retina (16,74). To date, there is no compelling evidence that RetGC functions outside the photoreceptor outer segment. Demonstration of continued light adaptation of GUCY2D-LCA rod vision along the Weber-law behavior through mesopic and low photopic ambient conditions (up to 2 log Td) suggests that the activity of RetGC2 (or RetGC1-unrelated molecular mechanisms) within human rods is sufficient for maintaining near-normal adaptation of rod vision to low photopic lights (27,30,75–77).
Also surprising was the apparent lack of saturation of rod vision in GUCY2D-LCA patients for ambient conditions in the high photopic range as bright as 4.7 log Td; these results were only uncovered in the subset of patients with no detectable cone function. Single cell recordings of light adaptation at least in RetGC1−/− mouse rods argued against the possibility that this apparent extension of the dynamic range of light adaptation had resulted from the specific lack of RetGC1, because, if anything, the Weber range seems to be shortened in this genotype (Supplementary Material, Fig. S2B), which is not entirely surprising because RetGC1 partakes in the calcium feedback underlying adaptation. Instead, one possible explanation is that this extended Weber-range arises from the lack of normal inhibition by cones (i.e. the removal of normal rod–cone interactions; 78) due to non-functional cones in these patients. Future studies evaluating the full range of light adaptation across all retinal layers of RetGC1−/− mice (73) may provide greater insight into our observations in human GUCY2D-LCA.
With a few notable exceptions, the mutations found in this LCA1 cohort were predicted to render RetGC1 inactive (Table 1, Fig. 4). Although some of the nonsense or frame-shift mutations in the group result in truncation of the essential catalytic portion of the enzyme, the Arg1091x mutation (one allele in P8, a compound heterozygote; Table 1) demonstrates detectable residual activity that is only 4–5-fold lower than the wild-type (Fig. 5). Although a reduction of this magnitude may cause reduced function of guanylyl cyclase per se, it would unlikely explain a complete loss-of-function. However, the proline-rich carboxy-terminal portion of RetGC1 is specific for photoreceptor cyclases and has little homology with the peptide hormone receptor guanylyl cyclases GUCY2A and GUCY2B. Therefore, this region may participate in other interactions, such as proper compartmentalization of the cyclase. It would not be unreasonable to suggest that in addition to the reduction of RetGC1 catalytic activity the truncation of the last 13 amino acid residues may cause other adverse effects when produced in retinal rods and cones. This question cannot be resolved using recombinant cyclase expressed in HEK293 cells but could be addressed in vivo using living photoreceptors.
The enzymatic properties of the Ser248Trp missense mutation (one allele in P7, a compound heterozygote; Table 1) also do not appear to be consistent with complete loss-of-function of RetGC1. RetGC1 harboring this mutation does not demonstrate any loss of function; if anything, there is a small (~25%) increase in maximally stimulated Ser248Trp RetGC activity in the conditions of the in vitro assay (Fig. 5). This would not come as a major surprise considering that the entire ECD domain of the cyclase can be removed without the elimination of RetGC1 catalytic activity or regulation by GCAP1 and GCAP2 (14,62). We cannot exclude that such a mutation may affect RetGC1 folding, expression or targeting to the proper compartment in a living photoreceptor. It is less likely that inactivation of photoreceptor function resulted from the slight hyperfunction of the Ser248Trp mutant observed in vitro. In the light, phosphodiesterase effectively negates the accumulation of cGMP in photoreceptors (79), while high concentrations of Ca2+ in dark-adapted photoreceptors suppress RetGC activity. Hyperfunction of RetGC1 is known to produce a dominant phenotype of cone–rod dystrophy (CORD6), because it drastically reduces Ca2+ sensitivity of the RetGC1/GCAP1 complex and renders the cyclase active at high Ca2+ concentrations (12,58,63,80,81).The Ser248Trp RetGC1 does not show this behavior and becomes suppressed by Ca2+ within a normal range (Supplementary Material, Fig. S3). In summary, P7 is the only subject with two novel and different mutant alleles and further studies are warranted to try to clarify mechanism leading to disease. The clinical phenotype of P7, however, is definitely LCA, and the finding of measurable cone vision in this subject is shared with three others (P3, P4 and P8) who have better understood mutations.
In contrast to Ser248Trp, the biochemical properties of other missense mutations that do not cause a frame-shift—Arg768Trp, Arg822Pro and His980Leu (Table 1)—were consistent with abolished cyclase activity, capable of producing a functional RetGC1 null phenotype in human photoreceptors similar to the effect of Gucy2e gene disruption in a mouse model (27). None of these mutants displayed activity that resembles the activity of the wild-type (Fig. 5).
Of interest, there was only a trace amount of catalytic activity detectable in the Arg822Pro mutant, even in the conditions providing maximal stimulation of RetGC1 by either GCAP1 or GCAP2 (Fig. 5B and D). This mutation affected the predicted DD (57) of the cyclase, and the lack of activity can likely be caused by inability of the cyclase to properly form a functional homodimer (60). The molecular mechanism that can affect dimerization by this mutation is not immediately apparent, because Arg822 is not located in one of the key positions of the heptades forming the coiled-coil structure contacts between two alpha-helixes of the DDs (58). It is more likely that the replacement of the Arg with Pro destabilizes the general alpha-helical structures of the DD instead.
Within the sensitivity of the method, we were unable to detect any activity above the background in Arg768Trp and His980Leu RetGC1. The Arg768Trp mutation shows a high frequency of occurrence—8 out of 11 LCA1 patients of various ethnic descents carried this mutation in one or both alleles. Recent observations (14) demonstrate that the Arg768Trp mutation completely blocks interaction of the cyclase with GCAP.
In the case of the His980Leu mutation, found to be homozygous in P5, we were unable to detect any catalytic activity, which has a likely explanation based on the molecular structure of the RetGC1 catalytic core (Supplementary Material, Fig. S4). Although this mutation does not represent any of the side chains that directly bind Mg2+GTP substrate in the active site of RetGC1 (60), it should nevertheless have a strong impact on the active site. The His980 is predicted to form, through Thr1002, a close contact with the alpha-helix, Asp1001–Thr1012, which contains several key residues coordinating GTP and Mg2+. Substitution with Leu980 must cause rearrangement of this contact from Thr1002 to Ala1006, which can directly interfere with either GTP or Mg2+ binding.
In summary, the two mutant alleles with biochemical evidence of some RetGC1 activity, Arg1091x and Ser248Trp, are in patients P7 and P8, with measurable cone visual function (Fig. 3A). The other alleles in these compound heterozygotes show no biochemical evidence of RetGC1 activity. The two siblings with the Arg822Pro allele, P3 and P4, also had measurable cone vision, and this allele showed a trace amount of catalytic activity. The second allele, the common Arg768Trp, is considered a null (14). Notable is that these four patients had very different extents of rod vision (Fig. 2). We postulate that the in vitro biochemical RetGC1 phenotype of these single alleles in four compound heterozygotes with LCA1 may be related to the presence of the in vivo phenotype of measurable cone vision.
Knowledge of the photoreceptor disease in humans with LCA caused by GUCY2D mutations prompts questions about which of the currently available animal models most faithfully mimics the human retinopathy. The question is important as there are already published proof-of-concept data for gene augmentation treatment of some of the models and initiation of such therapy in human LCA1 is being discussed. Until now, it has been assumed that progressive retinal degeneration is a part of human LCA1 and models with photoreceptor degeneration have been considered better replicas of human disease than ones without (13). To mimic the phenotype in our cohort of LCA1, disease models should have normal rod morphology but measurable rod dysfunction, severe cone dysfunction and progressive foveal (cone) degeneration. The avian rd (retinal degeneration) model has a null mutation in the guanylyl cyclase gene (26) and, like the human retina in many of our GUCY2D-LCA patients, there is normal retinal morphology, but this persists for only the first 7–10 days post-hatching. The retina is non-functional by ERG at this early stage when retinal morphology is normal (25,31). Thereafter, progressive photoreceptor degeneration occurs (25,36). Gene-targeting in the zebrafish of photoreceptor-specific Gucy2d leads to dysfunction but also to photoreceptor degeneration (82).The progressive retinal degeneration in the avian and zebrafish models, both of which are diurnal and have both rods and cones (83,84), makes them unlike the cohort of human LCA1 we studied.
Gene disruption of RetGC1 in the mouse (specifically, a GC-E or Gucy2e null genotype) results in morphologically normal rods but an abnormal scotopic ERG (13,27,59); cone photoreceptors do not respond by ERG and degenerate over months (24). GC-E is detectable in murine rods and cones, as it is in man. There are thus similarities to human LCA1 of cone dysfunction and cone degeneration (assuming the foveal degeneration is mainly cones, see above) in the presence of measurable but abnormal rod function in morphologically normal rods. There is also evidence of a range of rod functional abnormalities, as shown with ERG data (13,27,59). A double knockout of GC-E and GC-F (Gucy2f), the latter being present only in rods, has been more recently engineered on the assumption that progressive degeneration of rods and cones may better mimic the human disease (13,35). There is no evidence to date that both cyclase isozymes are affected in any of the reported human LCA1 cases.
Based on the human results in the current study, the disease in the mouse knockout of Gucy2e represents the closest phenotype to that of human LCA1. Further, it is important to note that there are already three studies of gene augmentation therapy in this model (32–34). AAV2/5 and AAV2/8 vectors restored cone ERG function with or without increased rod function. These studies are worthy as proof-of-concept that gene augmentation is potentially a valuable treatment modality for LCA1.
It will be important to decide how to determine the outcome of therapy. Conventional clinical measures of vision, such as visual acuity and visual fields (85), will need to be supplemented with measures specifically suited to this severe disease with profound cone visual loss at baseline and a range of rod dysfunction. A full-field ERG outcome would not be favored assuming treatments will be focal subretinal injection(s) of vector-gene in a Phase 1 trial. Most patients have no recordable ERGs, making safety monitoring of negative changes untenable, and most patients have nystagmus, making ERG recordings difficult and potentially artefactual. Chromatic FST was an informative method in the present study leading to our finding of previously unrecognized rod-driven vision in LCA1; this method has also proven valuable in our LCA2 trial of gene augmentation therapy (43). Although FST uses a full-field stimulus, it detects the most sensitive locus in the retina (46) and would therefore detect any improvement in rod sensitivity; negative changes in a safety trial, however, may not be detectable unless the therapy specifically affected loci of greatest sensitivity. We previously defined the limits of variability for FST in patients with retinal degenerations including LCA (46,47,86), and it would be the most logical candidate outcome for measuring focal therapeutic benefit to rod function in a trial of patients with GUCY2D-LCA.
Expectations for increases in rod vision from gene augmentation in LCA1 are uncertain and will depend on baseline rod vision as well as increased understanding of how much of the dysfunction is at the photoreceptor level (and related to RetGC1 deficiency) and how much is post-receptoral. It may be advantageous to begin trials by including patients with lower levels of rod function and advance to those with higher levels of rod function in later cohorts, depending on the outcome of the initial cohort. More proof-of-concept studies specifically directed at rod functional improvement in the GC1 knockout model will also need to be performed (32,33). It is of note that GC1 knockout mice have been reported to show a wide range of remnant rod function from nearly normal to nearly undetectable (59) and thus model the range of results in human patients.
GUCY2D-LCA patients can have near normal foveal architecture but severe visual acuity abnormalities, making the central retina a target for treatment. Sophisticated monitors of central vision, such as microperimetry, that control for nystagmus and have published repeatability measures would be ideal to complement standard visual acuity as outcome measures (87). Individuals with no detectable foveal photoreceptors should be excluded from subfoveal injection but para- and peri-central retina may well benefit from treatment and an eccentric fixation could develop, as has been demonstrated to occur in our LCA2 clinical trial (88). For safety monitoring of foveal subretinal injections, non-invasive cross-sectional OCT imaging should be performed. The slow return of outer segment structure after foveal detachment, such as we previously documented (43), could impact a potential recovery of cone function from therapy. The expectations for an increase in central cone vision from gene augmentation in GUCY2D-LCA with retained foveal architecture are unknown. The role of amblyopia and cone degeneration outside the fovea in this severe form of LCA remains uninvestigated. Amblyopia, in this case bilateral, could have a significant effect on outcome despite correction of a foveal cone photoreceptor guanylyl cyclase defect. Further, it would be important in those with the more severe visual disturbance and normal or near normal retinal architecture to have magnetic resonance imaging studies to confirm the integrity of the visual pathways (69,89,90).
Eleven patients (ages 6 months to 37 years) with known GUCY2D mutations were included (Table 1). Sequencing of parental samples was possible in 9 of the 10 families and confirmed that the two GUCY2D mutations were on separate alleles. Informed consent or assent was obtained; procedures followed the Declaration of Helsinki and had institutional review board approval. All patients underwent a complete eye examination as well as specialized tests of retinal structure and visual function.
The health of the RPE was estimated by imaging the lipofuscin fluorophores with a reduced-illuminance autofluorescence imaging method using short-wavelength excitation (SW-RAFI) as previously published (91). Cross-sectional images of the retina were obtained with a spectral-domain optical coherence tomography (SD-OCT) instrument (RTVue-100, Optovue, Inc., Fremont, CA, USA). Overlapping, non-averaged, 4.5 or 9 mm length scans were used to study the retina along the horizontal meridian through the fovea. All scans used for digital montage included the foveal depression for unambiguous retinal localization in patients with a limited control of eye position. Each scan, formed by a series of longitudinal reflectivity profiles, was analyzed with custom programs (MatLab 6.5; MathWorks, Natick, MA, USA). For the outer retinal sublaminae, signal peak assignments were based on our previously published work (39,40,92,93) consistent with hypotheses on the correspondence between OCT signals and histologically defined layers (36,94). Specifically, outer retinal sublaminae were assumed to include: (i) a hyperreflective layer at the outer plexiform layer; (ii) a hyporeflective layer at the ONL; (iii) a hyperreflective layer at the external limiting membrane; (iv) a hyperreflective layer near the junction of inner and outer segments possibly corresponding to the ellipsoid region of inner segments; (v) a hyperreflective layer near the interface between COS and RPE apical processes; (vi) a hyperreflective layer near the interface between ROS and RPE apical processes and (vii) a hyperreflective layer near the interface between basal RPE and Bruch membrane. ONL thickness was defined as the distance from (i) to (iii), COS+ thickness from (iv) to (v) and ROS+ thickness from (iv) to (vi).
Full-field ERGs were performed with bipolar Burian-Allen contact lens electrodes and a standard protocol using an Espion system (Diagnosys, Lowell, MA, USA) with methodology previously described (95). Psychophysical thresholds were measured in the dark-adapted state with FST. FST was performed using an LED-based ganzfeld stimulator (Colordome; Diagnosys, Lowell, MA, USA) as previously described (46,47). Rod sensitivity was determined with blue light results for patients showing rod or mixed rod-and-cone mediation; cone sensitivity was determined using red light for patients showing cone or mixed rod-and-cone mediation. Normal dark-adapted cone sensitivity was obtained during the cone plateau phase of dark adaptation. Light adaptation of vision was quantified with FST on steady white backgrounds ranging from −1.7 to 4.7 log10 scot-Td in a full-field variant of our methods previously described (48–50). Chromatic FST sensitivities were obtained at each background to allow the identification of the rod or cone mechanisms mediating detection (46,47).
The direct TPLR was elicited and recorded as previously described (43,96). Luminance-response functions were derived from TPLR amplitude at fixed time (0.9 s) to increasing intensities (from −6.6 to 2.3 log10 scot-cd.m−2) of green stimuli with short duration (0.1 s) presented monocularly in the dark-adapted state. TPLR luminance-response functions were analyzed to estimate the threshold intensity to produce a criterion response (0.3 mm constriction).
Dark- and light-adapted chromatic static threshold perimetry (500 and 650 nm stimuli, dark-adapted and 600 nm, light-adapted; 200 ms duration, 1.7°-diameter target; 72 loci on a 12° grid) was performed with a modified computerized perimeter (HFA-750i analyzer, Zeiss-Humphrey, Dublin, CA, USA). Location-specific sensitivities for rods were obtained using the 500 nm target in the dark-adapted state for rod-mediated locations (determined by the difference between results with the 500 and 650 nm stimuli, dark adapted). Cone sensitivities were measured with the 600 nm target on a 10 cd.m−2 white background. Our methods for data collection and analyses have been published (44,97).
A mobility performance task was used to quantify the ability of the patients to move through an indoor obstacle course as previously published (43). In short, a configurable path delimited by large styrofoam panels suspended from the ceiling of a 4.6 × 2.7 m room was traveled by the subjects using five different ambient illumination levels (0.2, 0.6, 2, 4 and 100 lux). Floor-level obstacles consisting of light-colored 0.6 m high objects were included in the course at randomized positions. The patients were initially dark-adapted and instructed to reach the far end of the course without touching the walls or obstacles and avoiding unnecessary delays. The task was performed binocularly. At least three traversals of the course (trials) were performed for each illumination level, proceeding from dimmest to brightest. The number of incidents (bumps to obstacles and walls; reorientations) per trial was recorded and the average calculated for each illumination level, by dividing the total number of incidents by the number of trials.
The R768W RetGC1 mutant was produced as described previously (15). Other mutations were introduced in wild-type human RetGC1 cDNA by PCR using the ‘splice by overlap extension’ technique (98) using high-fidelity Phusion Flash DNApolymerase (Finnzymes/Thermo Scientific). Prior to the mutagenesis, the pRCCMV vector (Invitrogen) containing wild-type RetGC1 cDNA (62) was modified by the removal of a XhoI/XhoI fragment and self-ligated. For the S248W, the first round of PCR generated two fragments using 5′-TGGTGATGCACTGGGTGCTGCTGG GTGGC/5′-TGTCTAGCAGCACGAATGGGGGCTCCTC (r1) and 5′-CACCCAGCAGCACCCAGTGCATCACCA/5′-ACGCCCTGCTTCGCGCATTCGGCT (f1) pairs of primers, and the resultant fragments were spliced by the second round of amplification using the r1 and the f1 primers. The spliced product was purified using the Promega Wizard PCR purification kit and ligated into the BssHI/AvrII sites of the modified pRCCMV-RetGC1 plasmid, thus replacing the corresponding wild-type fragment. The R822P mutant was generated using the 5′-TACTGCTCCAGCATCGGAAGCATCGAGTCAAT/5′-GGATCTCCTGCAGGCTGATGCCGTGGTT (r3) and 5′-ATTGACTCGATGCTTCC GATGCTGGA GCAGTA/5′-ACAGAAGGTGCTACCGGAGCCTCCCAGAG (f1) pairs of primers, spliced by second round of amplification using 5′-GCTGCTTAGGGACCCAGC CCTGGAG (f4) and 5′-ATCTCCTGCAGGCTGATGCCGTGGTTGCT (r4) primers and ligated into the plasmid harboring RetGC1 cDNA, replacing the KflI/SbfI fragment of the wild-type.
The H980L mutant was generated using the 5′-ACGCATGGACCCGAGAGCAGGCCT ATG/r3 and 5′-CATAGGCCTGCTCTCGGGTCCATGCGT/f1 pairs of primers, spliced by second round of amplification using f4 and r4 primers and ligated into the plasmid harboring RetGC1 cDNA, replacing the KflI/SbfI fragment of the wild-type.
The R1091x mutant was generated by PCR amplification using 5′ -ATCAGCCTGCAGGAG ATCCCACCCGAGCGGtGACGGAAG and 5′-TTATTTCTAGACAGGAATTCCGGAAGA CAGCATGCCTTTATTTC primers and ligated in the SbfI/XbaI sites to replace the corresponding wild-type fragment. All mutant DNA constructs were sequenced on both strands to confirm the proper outcome of the mutagenesis. Endotoxin-free plasmid preparations isolated from E-Cloni 10G Elite E. coli strain (Lucigen) using a Promega PureYield Plasmid purification reagent were used for transfection of the HEK293 cells by the calcium–phosphate precipitation technique as previously described (14,63), except that 20 µg of the plasmid DNA were used per each 100 mm culture dish. The membrane fraction from HEK293 cells was isolated as previously described (63).
Membrane fractions isolated from RetGC1-expressing transfected cells were subjected to electrophoresis in 7% SDS–PAGE and transferred to PVDF membrane by electroblotting using iBlot setup (Invitrogen). The blots were probed by rabbit polyclonal anti-RetGC1 antibody (62), developed using peroxidase-conjugated goat anti-rabbit polyclonal IgG (Cappel/MP Biomedical) and visualized using Pierce SuperSignal chemiluminescent substrate (Thermo Scientific). The images were acquired using a Fotodyne Luminous FX imager.
RetGC1 activity was assayed as described (63) with minor modifications. Briefly, the assay mixture (25 μL) contained 30 mm MOPS-KOH (pH 7.2), 60 mm KCl, 4 mm NaCl, 1 mm DTT, 2 mm EGTA, 10 mm free Mg2+, 0.3 mm ATP, 4 mm cGMP, 1 mm GTP, 10 mm creatine phosphate, 0.5 unit of creatine phosphokinase, 1 μCi of [α-32P]GTP, 0.1 μCi of [8-3H]cGMP (Perkin Elmer), phosphodiesterase inhibitors zaprinast and dipyridamole and variable concentrations of GCAPs. The reaction was incubated for 20 min at 30° and quenched by heating at 95° for 2 min. The [32P]cGMP product of the reaction was isolated using thin-layer chromatography on fluorescent-back polyethyleneimine cellulose plates (Merk), eluted in 2 m LiCl solution and quantified by liquid scintillation counting using ScintiSafe scintillation cocktail (Fisher Scientific) containing 20% ethanol.
Overnight-dark-adapted mice were anesthetized by Avertin and euthanized by cervical dislocation in accord with the Institutional Animal Care and Use Committee of the Johns Hopkins University. The eyes were enucleated and hemisected under infrared illumination. The retina was chopped into small pieces, placed in the experimental chamber and viewed on the microscope with infrared optics. Bath perfusate was bicarbonate-buffered, 35–37°C Locke's solution (in mm): 112.5 NaCl, 3.6 KCl, 2.4 MgCl2, 1.2 CaCl2, 10 HEPES, 0.02 EGTA, 20 NaHCO3, 3 Na2-succinate, 0.5 Na-glutamate, 10 glucose, pH 7.4. Two Xenon arc lamps provided 500 nm illumination (10 nm bandwidth) for background light and incremental test flashes, respectively. The solution in the recording suction-pipette contained (in mm): 140 NaCl, 3.6 KCl, 2.4 MgCl2, 1.2 CaCl2, 3 HEPES, 0.02 EGTA, 10 glucose, pH 7.4. Membrane current was recorded from a rod outer segment protruding from a retinal piece, digitized at 10 kHz and low-pass filtered at 20 Hz (8-pole Bessel). Data were analyzed with Originpro 7.5 and presented as mean ± SEM.
Conflict of Interest statement. None declared.
This work was supported by grants from the Foundation Fighting Blindness-Clinical Research Institute; National Eye Institute (EY11522—A.M.D.); Hope for Vision; Macula Vision Research Foundation; The Chatlos Foundation and Pennsylvania Department of Health (A.M.D.). A.V.C. is a Research to Prevent Blindness Senior Scientific Investigator. A.M.D. is Hafter Chair Professor of Pharmacology.