Lateral gain is impaired in macular degeneration and can be targeted to restore vision in mice

Macular degeneration is a leading cause of blindness. Treatments to rescue vision are currently limited. Here, we study how loss of central vision affects lateral feedback to spared areas of the human retina. We identify a cone-driven gain control mechanism that reduces visual function beyond the atrophic area in macular degeneration. This finding provides an insight into the negative effects of geographic atrophy on vision. Therefore, we develop a strategy to restore this feedback mechanism, through activation of laterally projecting cells. This results in improved vision in Cnga3−/− mice, which lack cone function, as well as a mouse model of geographic atrophy. Our work shows that a loss of lateral gain control contributes to the vision deficit in macular degeneration. Furthermore, in mouse models we show that lateral feedback can be harnessed to improve vision following retinal degeneration.


Mathematical models of different feedback mechanisms showing how they modulate the input-response relationship.
Schematic showing input-output functions at different gain values (indicated by corresponding color), when the gain is adaptive, implemented through a division of, or a subtraction from the input (left top and left bottom panels respectively) or suppressive, by the division of or subtraction from the response (right top and right bottom panels respectively). When the gain factor is applied to the input the shift/scale occurs along the x-axis (left panels), and when applied to the response it occurs along the y-axis (right panels). Implementing the gain through a divisive mechanism (top panels) scales the function, whereas implementation through a subtractive mechanism shifts the function without scaling (bottom panels). Note that different gain factor values were applied for the different computations for illustration purposes.

Surround luminance acts as an adaptive mechanism
A high luminance (64 cd/m 2 ) surround reduced contrast sensitivity similarly to a low luminance surround (n=9 normal vision subjects). Error bars denote standard deviation around a mean value. Source data are provided as a Source Data file.

Equiluminant
High luminance Surround-mediated effects do not require prolonged adaptation to surround luminance. Comparison of temporal contrast sensitivity with low luminance, equiluminant and high luminance surrounds in two normal vision subjects, each indicated with a different color. Contrast thresholds were measured using a 1° stimulus placed at a 6° eccentricity, with a 4 Hz temporal sinusoidal modulation. Closed circles indicate measurements made with the luminance of the background set throughout the test session. Very similar measurements were obtained (open circles) when the background luminance was set to equiluminant during the response period between stimuli presentations and only decreased (to black) or increased (to white) to coincide with stimulus onset. The similarity between the measurements obtained with these two different presentation parameters indicates that the effect of luminance in the surround is not due to slow adaptive mechanisms and that the surround provides fast feedback to the center. Source data are provided as a Source Data file. Comparable contrast threshold measurements obtained with and without eye tracking. Temporal contrast sensitivity in three normal vision subjects each indicated with a different color. Subjects were asked to fixate a spot placed in the center of the CRT monitor. Stimuli were 1 degree diameter, 4 Hz sinusoidal temporal modulation presented at 6° eccentricity on either side of the fixation spot. Subjects were asked to indicate on which side the stimulus was presented. Their ability to detect the stimulus was determined by their forced choice response. Contrast sensitivity was reduced by a low luminance (0.15 cd/m 2 ) vs luminance matched (27.7 cd/m 2 ) surround. Closed circles indicate threshold measurements made without the aid of a pupil-tracking device. Very similar threshold values were obtained (open circles) when measurements were made with the aid of a pupil tracking device that automatically discounted trials if saccades were made during the presentation of the stimulus (see Methods). Source data are provided as a Source Data file

Microperimetry data and Fundus Autofluorescence images for all tested Stargardt disease patients.
Left column: fixation stability (in cyan, see Methods) superimposed on a color fundus image of the eye. Middle column: sensitivity values obtained from microperimetry superimposed on color fundus photograph of the eye. Color bar indicates sensitivity in decibels. Red cross indicates average fixation spot for images in left and middle column. Superimposed white circles indicate stimulus size and locations used for psychophysical testing of contrast sensitivity proximal (central placement) and distal (peripheral placement) to the lesion. Right column: short wavelength (486 nm) Fundus Autofluorescence images. The area of atrophy presents with a decreased signal in all patients.
Patients with Stargardt disease do not have a generalized lack of lateral adaptation. a, Paired measurements in each patient (n=6 patients) show that an equiluminant surround imposes an improvement in vision (indicated by brackets) in locations distant from the scotoma similar to that observed in subjects with normal vision (cfr. Fig.1g, no scotoma). Locations of measurements were determined by retinal function as indicated by microperimetry measurements for each patient (see Methods). b, Same data as shown in panel a normalized to low luminance measurements within each patient to allow comparison of the data across the different retinal locations and with Fig.1g (n=6 patients). c, Data shown in panel a, expressed as a ratio of contrast sensitivity for equiluminant over low luminance surrounds at proximal and distal locations. Data from each patient is indicated by a different color (n=6 patients, p=0.0056; two-sided paired t-test). Error bars denote standard deviations around mean values. Source data are provided as a Source Data file.  Fig.2b,c). Source data are provided as a Source Data file.

Temporal contrast sensitivity functions for two patients with Congenital Stationary Night Blindness
For two patients, contrast sensitivity was improved by an equiluminant (grey traces) vs low luminance (black traces) surround. Comparative data for normal vision subjects is shown (grey and black markers, n=8 normal vision subjects). Values shown are mean averages, standard deviation indicated by grey shading, data as shown in Fig.1b). Source data are provided as a Source Data file.
Frequency (cycles/degree) Frequency (cycles/degree) WT CNGA3 -/-Contrast senstivity (1/c) Reduced contrast sensitivity in Cnga3 -/mouse model of achromatopsia. a, Contrast sensitivity was reduced in Cnga3 -/mice (n=9 mice) compared with wildtype mice (n=5 mice) measured using a standard optomotor behavior (0.128 cycles/degree p<0.0001; 0.25 cycles/degree p<0.0001; 0.381 cycles/degree, p=0.88, 0.45 cycles/degree, p=0.99; one-way ANOVA with Tukey's post hoc multiple comparison test). b, Contrast sensitivity functions could be obtained for wild-type (n=3 mice, black trace) and Cnga3 -/-(n=3 mice, blue trace) mice using a modified optomotor behavior paradigm driven by a 1° height sinusoidal stimulus. As expected, contrast sensitivity values were reduced compared to those obtained with full screen presentation of the same stimuli (panel a). c, The modified optomotor method allowed for assessment of different surround luminance values on contrast encoding in mice. Wild-type mice (black trace, n=9 mice) showed increased contrast sensitivity with an equiluminant surround compared to low-and high-luminance conditions, whereas Cnga3 -/mice (blue trace, n=4 mice) did not. Note that data for low and equiluminant surrounds in panel c is also shown in Figure 4. Error bars denote standard deviations around mean values. Source data are provided as a Source Data file Low and high luminance surrounds degrade contrast encoding over a wide retinal area. a, Expanding a black, low luminance (0.15 cd/m 2 ) annulus around the 1° stimulus lead to a cumulative degradation of contrast encoding compared to the scenario of no surround. (n=9, Annulus width: 0.05 degree, p=0.98; 0.1 degree, p=0.08; 0.25 degree, p=0.0028; 0.5 degree, p<0.0001; 1.5 degree p<0.0001; one-way ANOVA, with Dunnett's post hoc multiple comparison test). This effect of the surround decays in proportion to the distance from the center. b, Surrounding the stimulus area with a white, high luminance (64 cd/m 2 ) annulus reduced contrast sensitivity similarly to a low luminance annulus. The effect was also cumulative and a thin edge (0.05° annulus width) was not sufficient to reproduce the entire effect (n=9 normal vision subjects; Annulus widths: 0.05 degree, p=0.0013; 0.1 degree, p<0.0001; 0.25 degree, p<0.0001; 0.5 degree, p<0.0001; 1.5 degree p<0.0001; one-way ANOVA, with Dunnett's post hoc multiple comparison test