Probing the functional impact of sub-retinal prosthesis

  1. Sébastien Roux  Is a corresponding author
  2. Frédéric Matonti
  3. Florent Dupont
  4. Louis Hoffart
  5. Sylvain Takerkart
  6. Serge Picaud
  7. Pascale Pham
  8. Frédéric Chavane  Is a corresponding author
  1. CNRS, Aix-Marseille Université, France
  2. Aix Marseille Université, Hôpital Nord,Hôpital de la Timone, France
  3. CEA-LETI, France
  4. Université Grenoble Alpes, France
  5. Inserm, UMRS-986, Institut de la vision, France
12 figures

Figures

Experiment design.

(A) Schematic view of the experimental setup with the camera and the visual pathway from the retina to V1 activated with normal visual stimuli (left) or with sub-retinal electrical stimulation using a MEA (right). Retinal ganglion cells’ (RGCs) axons leaving the retina and projecting to V1 via the LGN are schematized (black when non activated; in red for direct activation; in orange for activation of en passant fibers that could occur in the electrical stimulation case). Blue curves on the top of V1 schematize the expected spatial profile of cortical activation (symmetric for visual stimulation and asymmetric for electrical stimulation if axons en passant are activated). (B) Clear optical access through thinned bone over V1; scale bar: 2 mm. L: lateral; M: medial; A: anterior; P: posterior. (C) Image of the eye fundus with the 9 electrodes (top left) and the 17 electrodes (top right) MEA; scale bars: 500 microns. Note that the use of an additional magnifying lens induced optical artifacts (halos of light). Retinal OCT B-scan (bottom) of an implanted animal showing the MEA and intact retina (PE: pigmetary epithelium; *: shadowing of the external reference surrounding the 17 electrodes MEA).

https://doi.org/10.7554/eLife.12687.003
Figure 2 with 1 supplement
Position.

The visual (A&E), retinal (C&G) and cortical (B,D&F,H) expected position and size of the MEA are compared to their corresponding visual stimuli. (A) Schematic view of the visual field showing the 20 positions (grid) of the visual stimuli used for retinotopic mapping. Optic disk: asterix; MEA: colored circles; nearest visual stimulus: white square (stands for all Figures). (B) V1 retinotopic polar map for azimuth (left) and elevation (right). Color hue and brightness code respectively for the retinotopic position and the strength of the response. Scale bar: 2 mm. (C) Image of the eye fundus with the implant. (D) Extent and center-of-mass (contour and circle respectively) of V1 activations generated by visual (white, right map) and stimulation at ± 150 µA wMEA (red, left map). The activation amplitude depicted in the colorbar is expressed both in Z-score and in DI/I (see Materials and methods). The solid line indicates the activation contour of the corresponding map and the superimposed dashed line corresponds to the compared condition; scale bar: 2 mm. (E-H) Same as in (A-D) in a different animal for 2 SE stimulations (cyan and purple; dashed circle: MEA position) at ± 200 µA; see Figure 6A for an example at low intensity in the same animal. (I) Retino-cortical magnification factor for azimuth computed over 20 animals (n = 177 displacements). Boxplots represent the median and interquartile range; whiskers represent ± 2.7σ or 99.3 coverage if data are normally distributed (any points outside are considered as outliers). (J) Retinotopic based positional cortical error for visual and electrical activations (right: mm; left: equivalent degrees of visual angle). Black dots correspond to visual counterparts of electrical activations. One-sided two-sample Wilcoxon rank sum test for paired data: pSE vs. wMEA=0.056, nwMEA = 21, nSE = 82, NwMEA = 6, NSE = 7. Wilcoxon rank sum test for paired data: *p=0.025, n = 13, N = 4 and ***p=5.35 10–5, n = 80, N = 6 (*p<0.05, **p<0.01, ***p<0.001, n = number of sample, N = number of rats).

https://doi.org/10.7554/eLife.12687.004
Figure 2—figure supplement 1
Raw maps.

Raw z-score maps without overlaid contours corresponding to a sample of figures presented in the manuscript (see subtitles). Colorbars are expressed both in z-score as well as in DI/I.

https://doi.org/10.7554/eLife.12687.005
Size.

(A) V1 activation generated by visual stimuli of increasing size (value indicated above the maps). Center-of-mass: white circle; extent: white contour; equivalent ellipse orientation: white cross; scale bar: 2 mm. (B) Extent of cortical activation as a function of visual stimulus size pooled over 7 rats (linear fit: dashed black line). (C) Extent of cortical activations generated in 2 animals by SE (blue) and wMEA (red) stimulation (top maps) at a high current intensity (± 200 and 150 µA respectively) and their corresponding 20° visual stimulus (white, bottom maps). Centers of mass of the activation: colored circles; scale bar: 2 mm. (D) Size of cortical activation generated by visual (gray, N = 20 rats); SE (blue, N = 8 rats) and wMEA (red, N = 10 rats) stimulation. Wilcoxon rank sum test for paired data: **p=0.0019 (n = 54, N = 8), ***p=1.83 10–4 (n = 15, N = 10). An alternative ordinate of equivalent visual counterpart is given on the right. Gray thin lines link paired electrical stimulation to visual activation. Solid horizontal black line indicates the size of a 20° visual stimulus estimated from the fit in B. Note that we could not reveal any effect of the electrode-to-counter-electrode distance on activation size in the various SE configurations (not shown).

https://doi.org/10.7554/eLife.12687.006
Figure 4 with 2 supplements
Shape.

(A) Shape of cortical activations generated in 2 animals by SE (blue) and wMEA (red) stimulation at high current intensity (top) and their corresponding 20° visual stimulus (white, bottom). (B) Aspect ratio (AR) of cortical activations (Wilcoxon rank sum test for paired data, ***p=1.06 10–4, n = 44, N = 7). (C) Predictions of the elongation of electrical activations as a function of the contribution of axons en passant and the distance to the optic disk. Insets correspond to a model of retinal activation due to direct isotropic activation plus passive electrical diffusion and anisotropic activation due to axons en passant recruitment for 3 different electrode sizes. The brightness codes the strength of the response. Center of the white dashed target: position of the optic disk; black circle: position and size of the MEA active surface; gray lines: 'shadow cone' angle sustained by the MEA active surface respective to the optic disk location; colored contour: size and shape of the global retinal activation for an axons en passant contribution of 1 (alpha, see Materials and methods). (D) Elongation of electrical activations relative to their corresponding visual activations (AR electrical/AR visual) as a function of the 'shadow cone' angle. (E) Cortical radial organization of prosthetic activations. Solid segments: orientation of cortical activations; dashed segments: optimal radial orientation towards the black disk; segment crossing; geometrical center; red dot: center-of-mass of cortical activations; Dark disk: cortical position that optimized radial organization; gray disk: median position of the optic disk. The blue lines connect the center-of-mass to the geometrical center of activations. Scale bar: 0.5 mm. Inset: distribution of median angular deviation expected by chance compare to our observation: blue segment. (F) Top: centered and reoriented deviations of the center-of-mass (blue disks) to the geometrical center (center of the representation), horizontal dashed axis corresponds to the orientation of the radial organization. Bottom: averaged, centered and reoriented SE (with AR > and < than 1.6, left and middle respectively) and wMEA maps (right). White circle: center-of-mass; additional dashed contour corresponds to a Z-score of −4.5.

https://doi.org/10.7554/eLife.12687.007
Figure 4—figure supplement 1
Model of retinal anisotropic activation.

Model of retinal anisotropic activation due to axons en passant for 3 different electrode sizes (rows in B&C). (A) Modeling steps used to compute the activation of axons en passant (see Materials and methods). The brightness codes the strength of the response, all representations are scaled between 0 and 1. (B) Direct activation with passive electrical diffusion (left column). Center of the white dashed lines: position of the optic disk; black circles: position and size of the MEA active surface. Anisotropic activation due to axons en passant recruitment (middle column). Size and shape of the global retinal activation (colored contours, right column) for an axons en passant contribution of 0.5 (alpha, see Materials and methods). (C) same as in A for an axons en passant contribution of 1.

https://doi.org/10.7554/eLife.12687.008
Figure 4—figure supplement 2
Model of retino-cortical transformation.

To check whether retino-cortical transformation can introduce further bias to our predictions (Figure 4—figure supplement 1), we implemented a retino-cortical transformation based on the magnification measured in our data (A). This model simply generates a transformation using the formula RCM = 1 /(aR+b), where RCM is the retino-cortical magnification factor (mm/deg), R the retinal eccentricity (deg), a and b constants (along the horizontal meridian : a = 0.7; b = 30 & a = 0.4, b = 40 for the vertical dimension). In the Figure A, the resulting transformation is shown for a retinal pattern with circles of different diameter and eccentricity. (B) We show for 2 cone angles (blue and red) and ratio of axon-en-passant (same as Figure 4—figure supplement 1), what such transformation does. (C) From these maps, we calculated the cortical activation aspect ratio, normalized to the shape of activation without axon-en-passant (ratio = 0 not shown), plotted with the same convention as in Figure 4C.

https://doi.org/10.7554/eLife.12687.009
Intensity.

(A) V1 activation generated by visual stimuli of increasing luminance (value indicated above maps). Center-of-mass: white circle; extent: white contour; equivalent ellipse orientation and size of activation contour: white cross; scale bar: 2 mm. (B) V1 activation generated by SE (top) and wMEA (bottom) in 2 different animals at high current intensity. Center-of-mass of visual and electrical activations are indicated with white and colored circles respectively and extent of electrical activation in colored contours; scale bar: 2 mm. (C) Amplitude of cortical activation as a function of visual stimulus luminance computed over 8 rats (population fit: dashed black line; individual fits: gray thin lines). (D) Amplitude of cortical activations generated by wMEA stimulation at high current intensity. Population fit: dark red dashed line; individual fits: orange thin lines (N = 10). (E) Amplitude of cortical activations generated by SE stimulation at high current intensity. Population fit: dark blue dashed line (N = 25). Note that individual fits (cyan thin lines) were only performed on N=6 animals (tested with 7 different levels of intensity) and not on the others (N=19, tested with only 2 different levels: 50 and 200 µA). (F) Constant of semi-saturation (c50) for visual (gray in cd/m2) and electrical activations (wMEA: red and SE: blue, in µA). Two-sample Wilcoxon rank sum test: **p=0.0017, nSE = NSE = 6, nwMEA = NwMEA = 10. (G) Exponent of the naka-rushton fits for visual (gray) and electrical activations (wMEA: red and SE: blue). Two-sample Wilcoxon rank sum test: ***pSE vs. visual=6.66 10–4 (nSE = NSE = 6, nvisual = Nvisual = 8); **pwMEA vs. SE=0.0075 (nSE = NSE = 6, nwMEA = NwMEA = 10); pwMEA vs. visual=0.896 (nwMEA = NwMEA = 10, nvisual = Nvisual = 8). (H) Amplitude-based correspondence for all electrical activations to their visual counterpart (wMEA: red and SE: blue) as a function of current intensity level (in equivalent cd/m2).

https://doi.org/10.7554/eLife.12687.010
Focalization.

(A) SE activations generated by square pulses of different polarity (anodic first: blue; cathodic first: green) at 2 intensity levels (50 and 200 µA). Scale bar: 2 mm. (B) Effect of polarity and intensity on cortical extent for SE (anodic first: blue; cathodic first: green, N = 10). One-sided Wilcoxon rank sum test for paired data, p=0.3802 (n = 18, N = 10) and 0.0615 (n = 11, N = 10) for 50 and 200 µA respectively. (C) Individual example of asymmetrical (green) and symmetrical (blue) square pulses, same animal as in (A). (D) Effect of square pulse asymmetry (green) for two polarities on cortical extent for SE (N = 10). One-sided Wilcoxon rank sum test for paired data: **pCathoSym vs. CathoAsy=0.0052 (n = 13, N = 10); *pAnoAsy vs. CathoAsy=0.0469 (n = 6, N = 10).

https://doi.org/10.7554/eLife.12687.011
Figure 7 with 1 supplement
Impedance spectroscopy adaptation.

(A) Example of electrode-tissue interface filtering on pulse shape (at 500 mVpp) with (green) and without (blue) IS based adaptation. Top: injected pulse; bottom: pulse shape reaching the tissue. (B) Individual example of IS adaptation (green, top maps) and reference (blue, bottom maps) for 2 voltage levels; scale bar: 2 mm. Note that for this protocol only, we switched from current to voltage injection. (C) Effect of IS adapted pulses on cortical extent for SE at 2 intensity levels (5 and 12 Vpp in voltage injection mode). IS adapted pulses: green; reference: blue, N = 4. One-sided Wilcoxon rank sum test for paired data: *p=0.0117 (n = 8, N = 4), **p=0.0078 (n = 7, N = 4).

https://doi.org/10.7554/eLife.12687.012
Figure 7—figure supplement 1
Principle of IS adaptation.

(A) The stimuli processing platform’s architecture integrates an Impedance Spectroscopy (IS) recording module; an Identification Algorithm module fitting IS data to an Electronic Equivalent Circuit (EEC); a Transfer Function Computing module and an Adapted Stimuli Shaping module computing adapted stimuli emitted by the Stimuli Generator. (B) Linear EEC used for this study. I: current; V: voltage; CPE: constant phase element (capacitance of the electrical double layer formed at the interface between a metallic electrode and an ionic solution); C: retinal membrane capacitance; R: resistance; bulk: subretinal fluid. (C) Mean duration (top) and frequency content (bottom) of reference (left) and IS adapted (right) pulse shapes recorded in the tissue (reference pulses: cathodic first symmetrical biphasic squared pulses). Duration: full width at half maximum (FWHM) of each recorded transients for the reference pulses (corresponding to the 3 abrupt pulse transitions) and of the 2 sustained phases for the adapted ones. Frequency: power of the Fourier transformed computed on the pulses temporal profiles.

https://doi.org/10.7554/eLife.12687.013
Differential effect of SE patterns.

Averaged cortical size and aspect ratio (± sem: black error bars) elicited by visual stimuli, by the wMEA and by the different SE stimulation patterns delivered at high intensity levels. Gray circles: 10 and 20 deg visual stimuli (n = 7 and 75, respectively); cyan and red squares: high intensity symmetrical stimulation patterns for SE (n = 57) and wMEA (n = 10) respectively (both polarities); green and cyan circles: respectively IS adapted and non-adapted pulses delivered at 5 Vpp (n = 7, same order of magnitude as the other pulses delivered at high current intensity); green and cyan up triangles: respectively asymmetrical (n = 7) and symmetrical (n = 21) anodic first pulses delivered at ± 200 µA; green and cyan down triangles: respectively asymmetrical (n = 8) and symmetrical (n = 24) cathodic first pulses delivered at ± 200 µA. Lines link comparable conditions. The different SE patterns evoke activations that exhibit strong correlation between size and aspect ratio (gray shaded area), except for IS adapted stimuli which converge towards visual responses.

https://doi.org/10.7554/eLife.12687.014
Author response image 1
Predictions of the distance between the center-of-mass (COM) and the geometrical center of electrical activations(normalized to implant size) as a function of the contribution of axons en passant and the distance to the optic disk (from the same model presented in Figure 4C).

Values >0 correspond to distances away from the BS, when compared to the geometrical center (<0 is towards the BS).

https://doi.org/10.7554/eLife.12687.015
Author response image 2
V1 cortical activation measured with voltage sensitive dye imaging in non-human primate, elicited by local drifting gratings in awake (left) and anesthetized (right, midazolam) condition.

Red and white contours represent 75 and 50% of maximal activation; scale bar: 1 mm. Drowsy state of the animal (right) was induced by intranasal administration of 0.2 ml midazolam (7-chloro-imidazo, 0.1 mg/kg). Midazolam is a short-acting hypnotic-sedative drug with anxiolytic and amnestic properties. From the benzodiazepines class of tranquilizer drugs, the pharmacologic effects of midazolam are mediated through a majority of GABAA receptors (see Sigel 2002 for a review).

https://doi.org/10.7554/eLife.12687.016
Author response image 3

(A) Simulated deformation of cortical activations (right) to circular retinal regions located at various eccentricities and angles (left), the center of the gray pattern corresponding to the optic disk location. For this simulation, we used our estimation of the retino-cortical magnification factor. (B) Orientations of visual activations on the cortical surface. Length of the segments is proportional to the aspect ratio of the activation. Gray disk represents the median blind spot representation, scale bar: 0.5mm.

https://doi.org/10.7554/eLife.12687.017
Author response image 4
Correlation between DI/I and Z-score values.

(A) Histogram of highly significant*** r2 correlation coefficients between all pixels of Z-score vs DI/I maps (median r2= 0.81, N=9 rats, n=225 maps, all pval= 1.40e-45). (B) Representative examples of DI/I maps (same as the one presented in Figure 2 D, H and Figure 2—figure supplement 1D, H).

https://doi.org/10.7554/eLife.12687.018

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Sébastien Roux
  2. Frédéric Matonti
  3. Florent Dupont
  4. Louis Hoffart
  5. Sylvain Takerkart
  6. Serge Picaud
  7. Pascale Pham
  8. Frédéric Chavane
(2016)
Probing the functional impact of sub-retinal prosthesis
eLife 5:e12687.
https://doi.org/10.7554/eLife.12687