Retinoid therapy restores eye-specific cortical responses in adult mice with ret...

Effective treatment for childhood blinding diseases involving retinal dysfunction remains one of the major unmet needs in medicine. Over the last two decades, numerous strategies have emerged, including gene replacement therapy, gene editing approaches, optogenetics, retinal electrical implants, induced pluripotent stem cell (iPSC)-derived retinal cell replacements, and pharmacological agents., ,  These approaches can restore light-sensing ability at the retinal level and lead to improvements in visual perception in patients to varying extents. Studies of synthetic retinoid treatment in patients with RPE65 or LRAT mutations have led to improvements in visual field area and visual cortex activation, as seen on fMRI scans., ,  A separate study using intravitreally administered synthetic retinoid to treat RPE65-mutant dogs found that electroretinogram (ERG) recordings recovered a normal shape and response threshold following treatment, although the duration of this effect was limited to ten weeks. The treatment also improved vision-dependent behavioral tests at low light intensities. Clinical trials of gene therapy for recessive RPE65-associated retinopathies have also demonstrated behavioral and visual field area improvements and increases in visual cortex fMRI signals in treated patients and dogs.,, ,  Indeed, measures of visual cortex activity may provide an objective way to assess long-term treatment efficacy. Although some progress has been made,,, it still remains unclear the extent to which adult visual circuits can be restored to a fully functional state at the level of the visual cortex upon correction of the retinal defect.

It is widely appreciated that visual impairments in early critical periods of postnatal development can lead to lifelong deficits in central visual function, even when the initial abnormality is corrected. Recent studies have revealed exceptions to this critical period restriction. How the critical period applies to visual circuit remodeling in response to restoration of retinal function has not been defined. This question is crucial, as its answer would inform the optimal timing of visual restoration efforts.

We employed an established model of a childhood blinding disease, the lecithin:retinol acyltransferase knockout (Lrat−/−) mouse, to determine the extent of recovery of the central visual circuit following acute rescue of retinal light sensitivity. The Lrat−/− mouse has a metabolic blockade in the visual cycle pathway that produces the visual chromophore, 11-cis-retinal, necessary for the formation of photoreceptor visual pigments, thus recapitulating the primary pathology associated with Leber congenital amaurosis (LCA) type 14 (Figure 1A). This blockade can be circumvented pharmacologically by the administration of a retinoid, 9-cis-retinyl acetate (9-cis-R-Ac). 9-cis-R-Ac, a 9-cis-retinal prodrug, is currently under clinical investigation as a visual chromophore substitute.,, We confirmed that intraperitoneal administration of 9-cis-R-Ac (40 mg/kg in DMSO) to Lrat−/− mice for seven consecutive days led to substantial 9-cis-retinal accumulation in the eye (Figures S1A–S1C) and produced a rapid and prolonged improvement in light-induced ERG activity and optomotor reflexes (OMR) compared with vehicle (DMSO) control (Figures 1B–1H). Prior studies have shown that cone photoreceptors completely degenerate in Lrat−/− mice by postnatal day 42., Consistently, we observed no M-cone opsin immunostaining in the Lrat−/− mice used for our experiments, irrespective of retinoid treatment (Figure S2), which strongly indicated that the ERG and OMR improvements arose from restoration of rod photoreceptor function. We next employed widefield intrinsic signal optical imaging and two-photon calcium imaging to measure changes in primary visual cortex (V1) neuronal activity following retinoid treatment.

We first investigated the effect of 9-cis-R-Ac treatment on V1 activation in adult Lrat−/− mice using widefield intrinsic signal optical imaging (ISOI), which provides a bulk measure of visual activity (Figure 2A). Prior to treatment at baseline, Lrat−/− mice exhibited poor visual activation of V1 compared with wild-type mice (Figure 2B). Lrat−/− mice were administered either retinoid or vehicle daily for seven days (Figure 2C). We found that 9-cis-R-Ac treatment led to a significant increase in the amplitude of V1 activation (median of 28% increase when using stimuli with max. luminance of 33 cd/m2; Figures 2D–2G). The increased amplitude in retinoid-treated mice was sustained for at least 9 days post treatment and provided therapeutic effects at 27 days post treatment compared with the vehicle (Figures 2E–2G). Behavioral (OMR) improvements were sustained for at least 19–20 days post treatment (Figures 1F–1H). By contrast, vehicle-treated mice exhibited a progressive decrease in response amplitude over time (Figures 2D, 2E, and 2G). To investigate the nature of this progressive decline, we evaluated V1 activation in repeated ISOI imaging recordings in age-matched wild-type mice. Wild-type mice displayed a similar level of reduction in visual response amplitude with time, as we observed with vehicle-treated Lrat−/− mice (Figures 2E–2G). We speculate that these changes reflect an expected level of degradation in signal due in part to visual adaptation to repeated stimuli under awake recording conditions and likely do not indicate further decline of visual function in Lrat−/− mice over a relatively short period of time.

Since 9-cis-R-Ac partially restores photoreceptor function in Lrat−/− mice (Figures 1C–1E), the effect on visually evoked activity in V1 as observed using ISOI may be explained by improved retinal signals leading to (1) the activation of a greater number of V1 neurons, (2) increased response amplitude per neuron, or (3) both. To examine these possibilities, we measured neuronal activity in V1 at cellular resolution using two-photon calcium imaging. Adult Lrat−/− mice were injected with a syn-GCaMP6s virus into the binocular zone of V1, implanted with a cranial window, and treated with seven days of 9-cis-R-Ac. Typically, the same sets of V1 neurons were recorded at baseline, within one day following the end of treatment and nine days post treatment (Figure 3A). Neuronal activity was measured while mice viewed stimuli through both eyes, the contralateral (“contra”) eye or the ipsilateral (“ipsi”) eye, relative to the recorded hemisphere (Figure 3B).

Two-photon calcium imaging revealed that 9-cis-R-Ac treatment led to a significant increase in the response amplitude of individual V1 neurons (mean of 2-fold increase when using stimuli with max. luminance of 33 cd/m2) and that this effect reversed during the post-treatment period (Figures 3C and 3D). The number of neurons visually responsive to both- and contra-eye stimulation doubled after treatment, an effect that was largely reversed during the post period (Figure 3E). Strikingly, the number of neurons activated by ipsi-eye stimulation increased by nearly 5-fold after treatment, an increase that was sustained during the post period (Figure 3E). We confirmed using ISOI that the effect of treatment on V1 activation through the ipsilateral eye was observed at both low and high visual luminance levels (max. luminance of 33 and 330 cd/m2; Figure S3), indicating that even at high luminance levels, the cortical response to the ipsilateral eye stimulation is enhanced by the treatment.

Two-photon calcium imaging further revealed that V1 neurons displayed broader orientation tuning, preference for higher spatial frequencies (SFs) and larger SF bandwidth following treatment (Figure S4). The higher SF preference and broader SF tuning effects were transient, whereas the broader orientation tuning effect remained at post (Figure S4). These results indicate that Lrat−/− V1 increased in terms of response strength, number of activated neurons, and stimulus bandwidth. Although the enhancement of the contralateral-eye signal was transient, the ipsilateral eye signal underwent a particularly prominent and long-lasting improvement.

Given that 9-cis-R-Ac led to differential effects on Lrat−/− V1 responses depending on which eye was stimulated, we examined the effects on V1 neurons’ ocular dominance properties. Ocular dominance index (ODI) values near 1 indicate contra-eye dominance, values near −1 indicate ipsi-eye dominance, and values near zero indicate significant responses to both eyes (binocular). At baseline, V1 neurons in Lrat−/− mice displayed severely abnormal ODI with mostly contra-dominant cells (Figures 3F–3H). Immediately after retinoid treatment, ODI distributions shifted such that greater proportions of neurons displayed ipsi-dominant or binocular responses (Figures 3F–3H). However, ODI distributions in Lrat−/− mice did not fully shift to normal until the post period, becoming indistinguishable from those found in wild-type mice (Figures 2F–2H), indicating that binocular circuits in Lrat−/− mice take time to restore to normal following treatment.

One explanation for the increased ipsi-dominant population after treatment of the Lrat−/− mice is that newly ipsi-dominant neurons arise from a pool previously non-responsive to visual stimuli. Or, the newly ipsi-dominant neurons are originally contra-dominant or binocular neurons and switch their OD properties after treatment. To distinguish between these mechanisms, we tracked the activity of the same neurons over time. We designated neurons as “ipsi-destined,” “contra-destined,” or “binocular-destined” based on their post-period responses and compared them with their prior OD identity at baseline and after treatment. Although the majority of ipsi-destined neurons originated from a previously non-responsive pool of neurons (44/68 or 65%; Figure 3J), we found an unexpectedly large proportion of ipsi-destined neurons that were previously contra-dominant (18/68 or 26%; Figure 3I,J). This form of OD plasticity—monocular neurons switching OD to the other eye—is surprising, as it rarely occurs during juvenile development. The small remainder of the ipsi-destined population were either already ipsi-dominant neurons (4/68 or 6%) or neurons that switched OD status between baseline and treatment (“other,” 2/68 or 3%; Figure 3J).

Although some visual properties in V1 were restored by 9-cis-R-Ac treatment in adult Lrat−/− mice, it is unclear whether the recovery extends to functions involving multiple visual pathways. Arousal-mediated modulation of the visual system involves the coordination of multiple circuits and undergoes prolonged periods of development., To investigate arousal-mediated V1 modulation, we performed pupil tracking during two-photon calcium imaging of V1 neurons (Figure 4A). At baseline, Lrat−/− mice displayed large pupils with minimal size modulation compared with those found in wild-type mice, suggesting that their pupils are constantly enlarged with little arousal-mediated modulation (Figure 4B). Following retinoid treatment, mean pupil size was reduced but remained larger than that of wild-type mice under similar light conditions (Figures 4B and 4C). The amplitude of pupil size modulation was abnormally low in Lrat−/− mice, and following treatment, pupil size modulation became larger and was indistinguishable from wild-type values by the end of the post-treatment period (Figure 4D). However, we note that pupil size modulation may have been low at baseline in Lrat−/− mice partially due to ceiling effects of their abnormally large pupil sizes.

Within normal mice, V1 neuronal activity is tightly correlated with pupil size fluctuations., Accordingly, we found that the population calcium trace (average of individual traces) closely recapitulated the minute-to-minute pupil size fluctuations in wild-type mice (Figure 4E, the rightmost panel, top 2 traces). In wild-type mice, we found high cross-correlation between the pupil diameter trace and individual neurons’ calcium traces, with neural signals leading pupil fluctuations (negative pupil-neural lag; 4th panel in Figure 4F). By contrast, Lrat−/− mice displayed little correlation between pupil and neural signals before treatment (Figures 4E–4G, baseline). Following retinoid treatment, pupil and neural traces for Lrat−/− mice became more correlated such that correlation values became similar to wild-type values (Figure 4G). Cross-correlation values also became higher in negative pupil-neural lag windows compared with positive lag windows, indicating that neural signals led pupil traces in treated Lrat−/− mice (Figure 4F).

The extent of central vision recovery observed in our study is notable, given that Lrat−/− mice have dramatically reduced photoreceptor function, virtually no cone outer segments, and only ∼30% of the rod outer segments remaining by one month of age. We found that at baseline, adult Lrat−/− mice display poor activation of V1 by visual stimuli and ocular dominance distributions in V1 neurons that are abnormally skewed toward contralateral-eye dominance. Following 9-cis-R-Ac treatment, some properties such as visual response amplitude were increased immediately after treatment and reversed by post period. We speculate that some of these immediate and short-lasting effects may be due to circuit-wide changes in gain. In particular, the mean response amplitude in Lrat−/− mice immediately following treatment was numerically higher compared with responses in wild-type mice (Figure 3D and ISOI results in Figure 2F), although the differences were not statistically significant. In addition, Lrat−/− V1 neurons’ preferred spatial frequency and spatial frequency tuning bandwidth after treatment were higher and broader, respectively, compared with wild-type values (Figures S4B and S4C). We speculate that these transient overshoots may in part be due to the dark exposure during the treatment period shifting the excitatory-inhibitory balance in favor of excitation in cortical circuits, as previously shown in the visual cortex. In Lrat−/−mice, inhibitory circuits may be especially underdeveloped because of the prolonged lack of retinal signals. Immediately following treatment, stronger visual signals may reach V1 circuits due to the partially restored photoreceptor function, and inhibitory circuits may not be strong enough yet to counteract the excitation. The reduced inhibitory tone may also explain the broadening of orientation selectivity seen post treatment (Figure S4A). After some time (by post period), the excitatory-inhibitory balance may become partially re-established. Whether dark exposure during the treatment period plays a significant role in the over-activation of V1 during treatment remains to be determined. Intriguingly, in a clinical trial of 9-cis-R-Ac, some LCA patients reported photophobia, which may be consistent with V1 over-activation.

Visual function changes that take longer to manifest and are longer-lasting (e.g., increases in the number of ipsilateral eye responsive neurons and changes in ocular dominance distribution) may reflect new synapses being formed. With stronger input from retinae, new synapses may be added or existing synapses recruited to relay signals from the ipsilateral eye. Interestingly, we found that the majority of the newly ipsi-dominant neurons arose from previously non-responsive or contra-dominant neurons, which suggest not only a de novo cortical activity but a shift in visual tuning properties. The locus of these changes in the visual pathway (in thalamus or V1) and underlying mechanisms remain to be investigated in future studies. It also remains to be determined the extent to which the increase in light sensitivity in the retina after the treatment contributes to recruitment of a greater number of V1 neurons. Our immunohistochemistry results indicate that all visual responses in Lrat−/− mice at the adult age we examined, even in those that were treated, originate from rod photoreceptors. It remains to be explored whether a similar extent of central circuit restoration is possible in other models of retinal degeneration.

Our ISOI results show a sustained increase in V1 response that lasts until 9 days post-treatment, but two-photon calcium imaging results show a transient increase that is not sustained until 9 days post-treatment. These differences may reflect the nature of the two signals. ISOI measures oxygenated blood in a large volume of tissue resulting from a combination of neuronal firing and neuropil activity (input from other cortical layers and brain areas). In contrast, two-photon calcium imaging reports activity at the level of single neurons in the specific area and layers that were imaged (V1 L2/3 in our case). It may be that activity in circuits other than V1 L2/3 are restored in a sustained manner (e.g., deeper-layer neurons in V1, projection neurons in higher visual areas), whereas V1 L2/3 neurons show a transient increase, and ISOI amplitude may reflect increased inputs from these sources.

In our study, measures of subcortical visual functions were only partially rescued; however, many of the higher-level (V1) visual functions indicated restoration to normal levels. Previous studies that administered retinoid therapies to LCA patients reported that many experienced long-lasting improvements in visual acuity and visual field area., In these and other studies,,, no changes in ERG measurements were observed, despite significant subjective improvements in visual perception. Likewise in our study, improvements at the retina and those reflecting subcortical changes (e.g., ERG, pupil size, and OMR responses) indicated partial recovery, whereas many properties in the visual cortex were fully restored to normal levels (visual response amplitude, ocular dominance, and pupil-neural correlation). Together, these findings suggest that therapeutic interventions can lead to a much greater extent of central vision recovery than expected from retinal readings alone.

In summary, our findings indicate that a mouse model of retinal degeneration can undergo a significant restoration of vision when treated with retinoid replacement therapy, even in adulthood. V1 visual responses, both in terms of the number of visually responsive neurons and the strength of visual response per neuron, were strongly enhanced by the treatment. Ipsilateral eye visual processing, arousal-mediated modulation of neuronal activity in V1, and optomotor behavior showed marked improvements that lasted well beyond the end of the treatment. Although our finding of restored ocular dominance distribution suggests at least a partial restoration of binocular circuits, future studies using behavioral assays will be needed to fully assess whether binocular (depth) vision is restored by retinoid treatment in adulthood. Our results reveal an unexpected and promising plasticity of the visual circuit, suggesting that therapeutic strategies for childhood blinding diseases can be highly effective even if intervention is started in adulthood. Despite demonstrated efficacy, ease of oral administration, and excellent safety profile, retinoid therapies have remained relatively under-explored until recent years. Our findings showcase the potential of retinoid compounds as promising therapeutics for treating numerous retinal diseases involving defects in the visual cycle.