Mitochondria in cone photoreceptors act as microlenses to enhance photon delivery and confer directional sensitivity to light

Mammalian photoreceptors aggregate numerous mitochondria, organelles chiefly for energy production, in the ellipsoid region immediately adjacent to the light-sensitive outer segment to support the high metabolic demands of phototransduction. However, these complex, lipid-rich organelles are also poised to affect light passage into the outer segment. Here, we show, via live imaging and simulations, that despite this risk of light scattering or absorption, these tightly packed mitochondria “focus” light for entry into the outer segment and that mitochondrial remodeling affects such light concentration. This “microlens”-like feature of cone mitochondria delivers light with an angular dependence akin to the Stiles-Crawford effect (SCE), providing a simple explanation for this essential visual phenomenon that improves resolution. This new insight into the optical role of mitochondria is relevant for the interpretation of clinical ophthalmological imaging, lending support for the use of SCE as an early diagnostic tool in retinal disease.


The PDF file includes:
Figs. S1 to S8 Legend for movie S1 Table S1 Other Supplementary Material for this manuscript includes the following: Movie S1 Each group of panels contains images from a sectioned retinal sample during live experiments and in stained images following immunolabeling. Live high-resolution images labeled i, ii, iii are marked in low-resolution images (10x) where available as well as in matching immunolabeled images. Attempts have been made to rotate images to facilitate comparison. Live samples were labeled with TMRE and MitoTracker Green; fixed samples were labeled with TOMM20 (mitochondrial marker), DAPI (nuclei); some samples were also stained for RCVRN (cone marker).
A) Example post-fixation labeled sample highlighting the isolation of mitochondria-bearing ellipsoids. Compare to the cartoons in Fig. 1. Orthogonal projection corresponds approximately to the region enclosed with a dashed box. Note the different color scheme used for this panel compared to the remainder of this figure.

B-D)
Example samples labeled for verification. In particular, in panel B, example i, the bracketed region corresponds to the cluster of cells highlighted in panel A.

B)
Quantification of peak light concentration and focal length (distance from putative IS-OS border) for individual cones (small dots) and the average within each sample (large red circle; lines indicate the mean ± standard deviation).

C)
Maximum intensity projections of the samples in A immunolabeled following live imaging. These samples possessed clear, intact cell bodies with nuclei. Focusing deeper into the sample during imaging (not shown) revealed a lack of nuclear (DAPI) or mitochondrial (TOMM20) staining, indicating that those inner retinal layers were removed during vibratome sectioning. Images have been rotated and annotations included to facilitate comparisons among images. Cone arrestin (ARR3) is specific to cone photoreceptors; in particular, in examples 2 and 3, arrows indicate ARR3-negative rod photoreceptors, which also concentrated light but at focal lengths that differed from their cone neighbors, as the apparent focal points were often not in the same plane. In orthogonal projections from immunolabeled samples, note the inconsistent presence vs. absence of outer segments. For each reconstructed cone, the following images or data are presented, from left-to-right: Matching orthogonal cross-sections from original SBEM data and the corresponding dielectric structure from FDTD simulations; 3D model of segmented mitochondria; equivalent skeletonized mitochondrial reconstructions; and heatmaps of mitochondrial alignment analysis. This analysis is described in the Methods, see also Fig. 2D. Here, histograms of mitochondria branch alignment are displayed as a function of cone height (z). Lines indicate the median (red) and lower and upper quartiles (white) for each z value. Color scale encodes the percentage of all mitochondria branches for that cone (scale at bottom of figure applies for all heatmaps). Note that for all cones, mitochondrial orientations are most closely aligned near the center (z ≈ 0); however, this ordered alignment is much more apparent for model cones from active than hibernating GS. Z-axes on alignment heatmaps also serve as scalebars for 3D reconstructions. Mesh cleaning, orientation and discretization of 3D dielectric structure Import and FDTD simulation of light energy propagation using Maxwell's equations using MEEP: Energy export, image processing and quantification of light concentration 3D reconstruction of cone mitochondria and cell membrane from SBEM images At each Z step: Pixellation exaggerated for illustration (structure discretized at 10 nm resolution) For further details, see Methods.
A) 3D reconstructions of GS cone mitochondria from SBEM image datasets.

B)
Preconditioning of 3D models followed by discretization of the dielectric grid. Grid locations intersecting membranes were encoded with the normal vector of the mesh triangle intersecting that point as well as the volumetric fill fractions inside vs. outside the mesh at that point.

C)
Assignment of dielectric relative permittivity assignment in the simulation grid and injection of a linearly polarized current source. For the sample cross-section, note the difference in scaling for each image, which is necessary to show the full range at each time step due to the increasing concentration of light energy over time. The X value for each image indicates the ratio of the maximum energy value in that image to the background light intensity (i.e., 1x indicates no light concentration).

D)
Conversion of E-M energy density in the simulation volume to a two-channel Z-stack image and the subsequent light concentration analysis.  Simulations in this figure were performed in one of two modes: "Simple" or "Cristae". "Simple" simulations were as described elsewhere in the present study; mitochondria were modeled with uniform average refractive indices (n mito ) separated from the cytosol by a membrane envelope with refractive index n memb . "Cristae" simulations were modified by dividing mitochondria into separate inner matrix and inter-membrane compartments with refractive indices n m and n i , respectively, separated by a second internal membrane (n memb = 1.46, as elsewhere in the present study). For Simple simulations, the average n mito is indicated, whereas for Cristae simulations, refractive indices are instead indicated as n m /n i .

A)
Images from examples depicting top-down (horizontal) and vertical cross-sections of simulations of an active GS cone under three different refractive configurations; note that higher refractive indices are encoded by brighter colors.

B) Example comparison of internal mitochondrial structures in Simple vs. Cristae simulations.
C) Light concentration profiles for various refractive index configurations using the example cone shown in this figure. Simple (n mito = 1.40) is the configuration for simulations used elsewhere in this study. For Simple simulations (blue), increased n mito shortens the focal length but decreases peak light concentration; conversely, lower n mito increases light concentration, but at longer focal lengths. For Cristae simulations, the addition of the inner membrane without otherwise modifying the average mitochondrial refractive index (i.e., n m = n i = 1.40) decreased light concentration by a small amount. From there, balanced changes to the refractive indices of the two compartments (i.e., increases in n m with matching decreases in n i; indicated by gray lines) produced only modest effects, even with n changes of ± 0.05 (profiles in grayscale colors). In contrast, unbalanced refractive index adjustments (green, magenta) had effects equivalent to increasing or decreasing the average refractive index of mitochondria; e.g., n m = 1.43 with n i = 1.35 produced focusing changes similar to a that of a slight decrease in overall refractive index.

D) Bar graphs indicating the average changes in light-gathering measures for Simple and
Cristae simulations across all cone reconstructions (n = 9 models) in comparison to the canonical Simple (n mito = 1.40) configuration. Interestingly, note that OSR light intensity in Simple simulations was decreased both by an increase or a decrease in n mito due to the cumulative trade-off between peak concentration and focal length.

D)
Bar graphs depicting the change in light-gathering measures for "round" or "mega" simulations compared to the "normal" configuration across all cone models (n = 9). Note that light concentration by "round" mitochondria was inferior to the normal configuration, whereas the "megamitochondrion" configuration was optically superior. OSR concentration factor statistics are the result of one-sample t-tests computed for all cone models in each condition (Round or Mega mitochondria) to a mean of zero (i.e., an average zero change in OSR concentration factor).  Repeat 10,000x Repeat 10,000x

Fig. S8. Data structure and bootstrap statistics for experimental light concentration by cones.
A) Structure of sampled imaging data for cone photoreceptor light concentration in active and hibernating GS. n a1 represents the number of photoreceptors analyzed for active GS slice 1, etc.

B)
Bootstrap resampling scheme for statistical analysis of OSR light concentration differences between active and hibernating GS cone photoreceptors. The 99% confidence interval for ΔC OSR is reported as the 0.5% and 99.5% percentile of the 10,000 values thus generated.