Mitochondrial Ca2+ influences photoresponse recovery, metabolism, and mitochondrial localization in cone photoreceptors

Mitochondrial Ca2+ regulates key cellular processes including cytosolic Ca2+ signals, energy production, and susceptibility to apoptosis. Photoreceptors are specialized neurons that have extraordinarily high energy demands and rely on cytosolic Ca2+ signals for light adaptation and neurotransmission. Here we show that unlike other neurons zebrafish cone photoreceptors express low levels of the mitochondrial Ca2+ uniporter (MCU), the channel that allows mitochondrial Ca2+ entry. To determine why MCU expression is kept low, we overexpressed MCU specifically in cones. This increases mitochondrial [Ca2+], causes faster cytosolic Ca2+ clearance, and accelerates photoresponse recovery. Moreover, flux through the citric acid cycle increases despite dramatic changes in mitochondrial ultrastructure and localization. Remarkably, cones survive this ongoing stress until late adulthood. Our findings demonstrate the importance of tuning mitochondrial Ca2+ influx to modulate physiological and metabolic processes and reveal a novel directed movement of abnormal mitochondria in photoreceptors.


Introduction 24
Photoreceptors are highly specialized sensory neurons responsible for vision. In 25 addition to having unique and vulnerable structural features, photoreceptors reside in 26 the retina, a hostile cellular environment. They can be exposed to damaging light 27 radiation, are located near blood vessels with high levels of oxygen, and use more ATP 28 2 than most cells in the body. Despite these chronic stressors, most people retain vision 29 throughout their lives, highlighting the extraordinary ability of photoreceptors to regulate 30 cellular homeostasis and maintain viability. 31 Photoreceptor function and survival depends on Ca 2+ homeostasis. Photoreceptors rely 32 on Ca 2+ as a second messenger to recover from light signals and adapt to constant 33 illumination (Nakatani and Yau, 1988). In darkness they continuously release synaptic 34 vesicles, which requires precise regulation of synaptic Ca 2+ by L-type voltage-gated 35 channels ( Barnes and Kelly, 2002). Both chronic elevations and chronic decreases in 36 cytosolic Ca 2+  kinase and arrestin that cause sustained low intracellular Ca 2+ also cause degeneration 43 of photoreceptors (Chen et al., 1999a(Chen et al., , 1999bLaVail et al., 1987). 44 Cytosolic Ca 2+ in photoreceptors is buffered and regulated by the endoplasmic reticulum 45 (ER) and mitochondria (for review, see Križaj, 2012). In the synapse, Ca 2+ flow through 46 the ER to the synaptic terminal supports sustained CICR-driven synaptic transmission 47 Results 111

MCU expression is specifically limited in retinal tissue 112
We developed a custom antibody against the purified N-terminus of zebrafish MCU and 113 verified the specificity of this antibody using a global zebrafish MCU knock-out (MCU 114 KO, Figure 1A). Using this antibody, we found that MCU protein expression (as 115 normalized to total protein) is significantly higher in brain tissue than retina, which is 116 even lower than heart tissue ( Figure 1A). When normalizing MCU protein to other 117 mitochondrial inner membrane proteins cytochrome oxidase (MTCO1) and succinate 118 dehydrogenase (SDH), the retina more closely resembles heart tissue ( Figure 1B). To 119 see if this low retinal abundance of MCU was reflected in other regulators of the 120 complex, we used RT-qPCR to determine the relative mRNA quantity of the MCU 121 regulators MICU1, MICU2, and MICU3 (which has two isoforms in zebrafish) across the 122 same tissues. Expression of all MICU transcripts in retina is more similar to expression 123 in brain rather than heart ( Figure 1C). 124 To quantify MCU expression specifically in zebrafish cone photoreceptors, we took 125 advantage of a zebrafish cone degeneration model caused by a mutation in the cone 126 phosphodiesterase (pde6c -/-); in this model, cones selectively degenerate while rods 127 and other retinal neurons are preserved (Stearns et al., 2007). We analyzed MCU and 128 mitochondrial membrane protein expression in the pde6c -/mutants compared to their 129 7 138

Overexpression of MCU in cones raises basal [Ca 2+ ] in the mitochondrial matrix 139
We hypothesized that low expression of MCU protein in photoreceptors could protect 140 them from overloading their mitochondria matrix with Ca 2+ . To test the consequences of 141 increased MCU, we expressed MCU-T2A-RFP in zebrafish cones under control of the 142 promoter for cone transducin ("gnat2" or "TαCP") and established a stable transgenic 143 line overexpressing zebrafish MCU in cone mitochondria (Figure 2A). Cones 144 overexpressing MCU (MCU OE) also express RFP in the cytosol, as the T2A sequence 145 stalls the ribosome after MCU has been translated to allow MCU protein release before 146 translation of RFP (Kim et al., 2011). Based on immunoblot analysis we found that 147 retinal MCU protein expression is 102 ± 5 -fold higher in the MCU OE model. This also 148 is reflected in the ratio of MCU to other mitochondrial membrane proteins ( Figure  149 2B,C). Immunohistochemistry confirmed that MCU expressed in this way is localized to 150 cone mitochondria ( Figure 2D). 151 152 Figure 1 contd.
C. qRT-PCR quantification of relative mRNA of MICU proteins (relative to reference gene Ef1α and/or b2m, see methods) across retina, heart, and brain tissues. Bars = standard error. *p<0.05 and ns = not significant using ANOVA followed by Dunnett post-hoc test (comparison to retina). D. Retinal lysate, enriched for mitochondrial proteins, of WT and pde6c -/cone deficient retinas. Each lane is from mitochondrial-enriched lysate of two pooled retinas from a single fish, n=4 fish. Each lane contains 30 µg of protein. Ns = not significant using Welch's t-test.

153
To test the effects of MCU protein overexpression on mitochondrial Ca 2+ content, we 154 crossed MCU OE fish with the gnat2:mito-GCaMP3 transgenic line, which expresses 155 the Ca 2+ sensor GCaMP3 in cone mitochondria (Giarmarco et al., 2017). Mito-GCaMP3 156 fluorescence in live zebrafish larvae is 4.4 fold higher (median, Q1: 3.4, Q3: 6.06 fold) in 157 MCU OE models compared to WT siblings ( Figure 3A). We next prepared ex vivo 158 A. Schematic of MCU OE construct. The cone transducin promoter (TαCP, gnat2) drives expression of zebrafish MCU cDNA in all cone subtypes. The MCU cDNA is tagged with a T2A sequence followed by RFP. The T2A sequence causes ribosomes to stall and release the nascent MCU polypeptide with some added peptides from the T2A sequence before translating the RFP separately. Thus, RFP is present in the cytosol of cones with MCU overexpression.
B. Retinal lysate, enriched for mitochondrial proteins, of WT and MCU OE retinas. Blot was probed with antibodies for MCU, MTCO1, and SDH. 8 µg of protein was loaded per well. Each lane is from mitochondrialenriched lysate of two pooled retinas from a single fish, n=4 fish.
C. Quantification of relative MCU signal as a function of protein concentration and relative to other mitochondrial markers from the gel in B. Both exogenous and endogenous MCU were used for total MCU quantification in the MCU OE retina. *p<0.05 using Welch's t-test.

Overexpression of MCU in cones reduces cytosolic Ca 2+ transients and alters the 169 cone photoresponse 170
Increased mitochondrial Ca 2+ uptake could alter the clearance kinetics of cytosolic Ca 2+ 171 transients. To test this, we prepared ex vivo retinal slices of control and MCU OE 172 gnat2:GCaMP3 zebrafish, which express the Ca 2+ sensor GCaMP3 in the cytosol of 173 cones. We pre-incubated the retinal slices in a 0 mM Ca 2+ solution before introducing a 174 bolus of 5 mM CaCl2. We then used timelapse imaging to monitor the clearance of 175 cytosolic Ca 2+ from the cell body ( Figure 4A). Consistent with increased Ca 2+ uptake by 176 mitochondria, MCU OE cones clear Ca 2+ from the cell body cytosol 2.3 ± 0.1 times 177 faster than their WT siblings, as determined by the decay constant of a single 178 exponential fit ( Figure 4B, Supplemental Figure 4A). We also found that the peak fold 179 change in cone cell body GCaMP3 fluorescence in response to the Ca 2+ bolus is lower 180 in MCU OE compared to their WT siblings ( Figure 4C). To determine whether these 181 changes were due to Ca 2+ uptake via MCU and not another secondary effect, we 182 incubated MCU OE retinal slices in Ru360, a drug that blocks Ca 2+ uptake through the 183 MCU; these experiments significantly restored the WT phenotype (Figure 4A-C). 184

Figure 3: MCU overexpression in cones increases basal [Ca 2+ ]mito.
A. Total cone mitochondrial clusters in a larval zebrafish eye expressing gnat2:mtGCaMP3, a mitochondrial Ca 2+ sensor (green). Dotted outlines demarcate the region of the eye used for fluorescence quantification. Reporting the median with bars = interquartile range, n=15 larvae for both WT and MCU OE. *p<0.05 using Mann-Whitney test.
B. Relative mito-GCaMP3 fluorescence of cone mitochondrial clusters in adult retinal slices of gnat2:mtGCaMP3 fish (WT or MCU OE). Baseline fluorescence was first assayed in the presence of KRB buffer containing 2 mM CalCl2, then ionomycin (5 µM) was added to the slice to allow 2 mM Ca 2+ entry into the mitochondria to saturate the probe. Next, EGTA (5 mM) was added to the solution (keeping [ionomycin] constant) to chelate Ca 2+ and establish the minimum GCaMP3 fluorescence signal. n= 45 mitochondrial clusters (3 fish) for WT and n=42 mitochondrial clusters (3 fish) for MCU OE. Fish were between 3-5 months of age and slices were imaged every 30 s. Shaded region = standard error.    Ca 2+ in cone as compared to rod outer segments (Sampath et al., 1998(Sampath et al., , 1999. We It has been reported that pyruvate dehydrogenase (PDH), isocitrate dehydrogenase 207 (IDH), and α-ketoglutarate dehydrogenase (α-KGDH) have higher activity in the 208   (Denton 2009). How Ca 2+ stimulates α-KGDH and IDH is not as 212 well understood, but it is thought that Ca 2+ increases activity by interacting directly with 213 these enzymes to lower the KM for their substrates (Denton 2009). 214 We investigated the effect of MCU overexpression on PDP1c phosphatase activity by 215 immunoblotting WT and MCU OE retinas with antibodies against phosphorylated PDH 216 (P-Ser293 of E1α) and against total E1α. While we hypothesized MCU OE cones would 217 have a lower P-PDH/total PDH ratio, we observed that it was instead very slightly 218  Legend on following page.
While glucose is a physiologically relevant fuel for photoreceptors, it makes it difficult to 238 observe IDH and α-KGDH activity in isolation because the two are intrinsically linked in 239 the TCA cycle. For this reason, we next used U-13 C-glutamine in the absence of glucose 240 to bypass IDH and directly fuel α-KGDH ( Figure 6A). Since glutamine is not a typical 241 fuel for photoreceptors, we tested concentrations of 13 C-glutamine ranging from 0.1 mM 242 to 2 mM (Supplemental Figure 6A). We found that at most concentrations of 13 C-243 Glutamine, MCU OE retinas produced metabolites downstream of α-KGDH at higher 244 levels compared to WT ( Figure 6B). Increased production of these metabolites further 245 confirms that α-KGDH activity is higher in MCU OE cones. A. P-PDH/total PDH ratio from WT and MCU OE retinas, normalized to WT. MCU OE retinas have a 1.12 ± 0.02fold higher P-PDH/total PDH ratio than WT. n=6 WT and 6 MCU OE retinas. *p<0.05 using Welch's t-test.
B. Diagram showing how labelled carbons from U-13 C-glucose are incorporated through glycolysis and the first round of the TCA cycle. Shaded = labeled carbon, empty = unlabeled carbon.
C. Levels of isotopomers in WT and MCU OE retinas. 'm' signifies the number of 13 C-labeled carbons in each metabolite. 'm2' TCA cycle metabolites are made from one round of the TCA cycle. Data points represent averages from n=3 retinas from 3 different fish. *p<0.05 using Welch's t-test.  also contain what appear to be healthy mitochondria at the ellipsoid region ( Figure 7D). 267

MCU
This electron-lucent phenotype emerges soon after expression of MCU is first 268 detectable at 78 hours post-fertilization and is widespread by 120 hours post-fertilization 269 ( Figure 7E). 3D reconstructions of MCU OE mitochondria suggest that movement of 270 electron-lucent mitochondria away from the ellipsoid region is an active process, as 271 these mitochondria deform the nucleus on their way toward the synapse ( Figure 7F, 272 Supplemental Video 2). 273 A. Diagram showing how labelled carbons from U-13 C-glutamine are incorporated into α-ketoglutarate and downstream metabolites. Shaded = labeled carbon, empty = unlabeled carbon.
B. Levels of isotopomers in WT and MCU retinas supplied with 2 mM 13 C-glutamine for 15 minutes. Data points represent averages from n=3 retinas from 3 different fish. * p < 0.05 using Welch's t-test.
Using electron micrographs of whole larval eyes, we quantified mitochondria that were 274 electron-lucent in the ellipsoid region and those that were mislocalized to the nuclear 275 layer and synapse. With the caveat that the lucent mitochondria in MCU OE models are 276 often much larger than their healthy counterparts, we found that relative to WT cones 277 the MCU OE cones have fewer electron-dense mitochondria and substantially more  Despite the early emergence of mitochondrial abnormalities, we noted that larval cones 297 otherwise remain intact and other retinal cell morphology is unperturbed (Supplemental 298 Figure 8A). Fluorescent imaging of Tg(gnat2:GFP) fish showed that cones are 299 conserved as zebrafish reach maturity at 3 months of age and throughout early 300 adulthood despite their severely altered mitochondrial structure. However, some 301 morphological disturbances (shorter cones and disorganized alignment) are apparent at 302 these earlier ages. Severe cone degeneration eventually occurs later in adulthood; the 303 cone number decreases, and many remaining cones round up and lose their outer 304 segment integrity (Figure 8A,B). Rods remain intact and healthy rod outer segments 305 persist even when cone loss is extensive (Figure 8B, black arrowhead). 306 A. Cone mitochondrial clusters in live larvae expressing gnat2:mtGCaMP3. WT mitochondria were imaged with higher laser settings to show localization. In MCU OE models, mitochondrial clusters were found near the synapse and nuclear layer (white arrows), which was not observed in WT siblings. Scale bar = 5 µm.
B. Timelapse of a migrating mitochondrial cluster (green, white arrow) in live MCU OE gnat2:mtGCaMP3 larvae. Cone cell bodies express cytosolic RFP (magenta). MCU OE mitochondrial clusters slowly move from the ellipsoid region down to the synapse. Full video in supplemental data. Scale bar = 5 µm.
C. EM images of cone mitochondria in WT sibling and MCU OE fish at 1 month of age. Mitochondria in MCU OE cones exhibit swelling, loss of electron density, and disorganized cristae. Scale bar = 1 µm. D. A single cone photoreceptor can contain both healthy mitochondria in the ellipsoid region (black arrow) and electron-lucent mitochondria near the synapse (white arrow). Electron micrograph from MCU larvae at 14 days of age. Scale bar = 2 µm. OS = outer segment, N = nucleus.
E. Electron micrograph of MCU cones at 120 hours of age. Scale bar = 5 µm.
F. 3D reconstruction of an MCU OE cone using serial block-face EM (synapse not shown). Electron-lucent mitochondria displace the nucleus of the cone to move toward the synapse region. OS = outer segment. M -= electron-lucent mitochondria. M + = healthy mitochondria. N = nucleus. Outline (yellow) = cell body.
G. Quantification of cone mitochondrial phenotypes from EM images of whole zebrafish larval eyes (single slice at optic nerve) at 6 days of age. MCU OE cones have fewer "healthy" electron-dense mitochondria, and an increase in lucent mitochondria both in the ellipsoid and outside of this region. All mislocalized mitochondria observed had an electron-lucent phenotype. n=3 larvae for both WT and MCU OE fish. *p<0.05 using a t-test with the Holm-Sidak correction for multiple tests. A. WT sibling and MCU OE retinas stained with Hoescht (blue) and exhibiting fluorescence from gnat2:GFP in all cone types (green) at 3 months, 6 months, and 10 months of age. Cone outer segments are labelled with α-PNA (magenta). Scale bar = 25 µm.
B. WT sibling and MCU OE retinas stained with Richardson's stain at 3 months, 6 months, and 10 months of age. Swollen mitochondria can still be visualized in 3 month and 6 months cones. By 10 months, the cones are very few and have severe morphological disturbances (white arrow); however, rod mitochondria and outer segments remain intact (black arrow). Scale bar = 25 µm.
C. Quantification of double-cone nuclei from the optic nerve to the ciliary margin in WT and MCU OE fish. Counts are an average of dorsal and ventral retinal slices in each fish, n = 3 WT and 3 MCU OE fish. *p<0.05 using a t-test with Holm-Sidak correction for multiple tests.
D. Electron micrograph of fragmented mitochondria in a cone from 10-month-old MCU OE fish. We observed that many of the remaining cones at 10 months have severe mitochondrial fragmentation rather than the electron-lucent phenotype. Scale bar = 2 µm.
Quantification of double-cone nuclei, which are most easily distinguished from rod 308 nuclei, revealed that cones in the MCU OE model are preserved at 3 months and 6 309 months, but severe cone loss occurs by 10 months (Figure 8C not re-populate to WT density after 10 months, but rods remain intact and abundant 317 (Supplemental Figure 8C). Notably, in 1 year old WT sibling retinas, there are multiple 318 cones with large, electron-lucent mitochondria in the clusters; these were similar in 319 appearance to MCU OE electron-lucent mitochondria (Supplemental Figure 8D).   Figure 5C) is surprising 362 considering that the inactivation of cone phototransduction was faster in these cones. 363 The gain of cone phototransduction activation reactions, normally thought to be Ca 2+ -364 independent, is somehow increased by altered Ca 2+ homeostasis due to MCU 365 23 overexpression. We also found that the maximal light response amplitude was 366 decreased in MCU OE cones (Supplemental Figure 5B), possibly due to misalignment 367 or loss of cones at the age tested (7 months). Altogether, these functional changes 368 demonstrate that mitochondrial Ca 2+ uptake in photoreceptors can modulate the kinetics 369 of the photoreceptor response to light, independent of neurotransmitter release and 370 downstream responses. It is likely that this phenomenon is not restricted to zebrafish or 371 cones, as human patients with mtDNA diseases have delayed recovery of the rod 372 photoreceptor response (Cooper et al., 2002). 373

Effect of MCU OE on mitochondrial metabolism in the retina 374
Since photoreceptors experience the highest levels of intracellular Ca 2+ in darkness 375 when O2 consumption is highest, Ca 2+ could play an important role in stimulating 376 increased TCA cycle activity (Okawa Fain 2008, Krizaj Copenhagen 2002. We have 377 previously observed that photoreceptors have higher α-KGDH activity in darkness (Du 378 Rountree 2015). Here, we observe that both IDH and α-KGDH are stimulated by 379 increasing Ca 2+ in cone mitochondria, which supports this hypothesis. 380 An increase in the P-PDH/total PDH ratio is a common metabolic phenotype in MCU KO 381 tissues, so we initially hypothesized that increasing mitochondrial Ca 2+ would decrease 382 the P-PDH/total PDH ratio by stimulating PDP1c (for a review of metabolic phenotypes 383 in MCU KO tissues see Mammucari, Raffaello 2018). However, we found that this was 384 not the case. Since the P-PDH/total PDH ratio also does not decrease when MCU is 385 overexpressed in muscle cells, it is possible that changes in mitochondrial bioenergetics 386 resulting from increased mitochondrial Ca 2+ feed into the complex regulation of PDH 387 (Mammucari, Gherardi 2015). For example, stimulation of α-KGDH and IDH activity may 388 result in higher NADH levels in MCU OE cones, which in turn stimulates PDH kinase to 389 balance increased PDP1c activity. 390

Effect of MCU OE on mitochondrial homeostasis and subsequent cell death 391
While some of our results align with predicted consequences of increased mitochondrial 392 Ca 2+ uptake, our work also revealed some surprising observations. The most 393 unexpected finding in MCU OE models was the presence and selective movement of 394 electron-lucent mitochondria away from the ellipsoid region of cones toward the 395 synapse. We also occasionally observed electron-lucent structures that appeared to be 396 mislocalized mitochondria in WT fish, especially in older animals (Supplemental 7A,B; 397 Supplemental 8D). We hypothesize that this mechanism of mitochondrial sorting and 398 movement occurs in normal cones, but it is more active in MCU OE models due to 399 extensive mitochondrial disruption. This hypothesis is strengthened by recent work with  Figure 8D) suggests that this phenotype may even be a symptom of cumulative 423 mitochondrial stress in cones as they age. This phenomenon is unlikely to be specific to 424 zebrafish cones, as large, electron-lucent mitochondria also have been observed in 425 aging cones in human retinas (Nag and Wadhwa, 2016). Increased resistance to 426 mitochondrial stress may be an adaptation necessary for cones, which need to meet 427 excessively high energetic demands throughout a lifetime. A key finding from our study is that cones are highly tolerant of mitochondrial stress. was performed using gRNA with the following sequence 5'-497 CCTCATACCTGGTGCAGCCCCCC-3' using methods as previously described 498 (Brockerhoff, 2017). For generation of the Tg(gnat2:MCU-T2A-RFP) line, zebrafish 499 MCU cDNA was isolated from WT zebrafish larvae (5 dpf) using the forward primer 5'-500 AGAGATGGCTGCGAAAAGTGT-3' and reverse primer 5'-501 TTCTCATCAGTCCTTGCTGGT-3'. Overhang qPCR methods in conjunction with Fast-502 Cloning were used to add the T2A ribosomal stalling sequence and the RFP protein 503 coding sequence; this was cloned into a pCR8/GW vector (Invitrogen) . 504 Plasmids were assembled using the Gateway-Tol2 system (Villefranc et al., 2007). 505 Expression of MCU-T2A-RFP was driven by the cone transducin alpha promoter (TαCP, 506 gnat2), and the RFP coding sequence was flanked by a polyA tail sequence to increase 507 transcript stability (Kennedy et al., 2007). A destination vector with a sBFP2 heart 508 marker for aid in transgenic identification was obtained from Cecilia Moens (Kremers et 509 al., 2007). The fully assembled construct was injected into embryos at the 1-cell stage 510 with Tol2 transposase mRNA. Larvae mosaic for the transgene were raised to 511 adulthood to identify founder carriers. A single F0 founder was used to generate F1 fish 512 that were screened for a single insertion of the transgene; F2 fish from two F1 substrains 513 with a single insertion were used for analysis in this study. 514 Primers for qRT-PCR. All designed primers were empirically tested to confirm primer 515 efficiency was between 90-110%. Only primers passing this benchmark were used for 516 analysis. Primer sequences for the reference genes EF1a, b2m, Rpl13a, and TBF were 517 identical to previous reports testing zebrafish reference gene stability (EF1a, Rpl13a: 518 and reference gene(s) were used to generate a ΔCt value for each biological replicate. 553 Comparing each tissue of interest to the retina generated a ΔΔCt value; these were 554 converted to a normalized expression level using the 2 -ΔΔCt method (Livak assumptions). 555 Standard error of the ΔCt value for each tissue was propagated to the final comparison 556 using standard error propagation rules. Calculations were based off the geNorm method 557 of qPCR normalization (Vandesompele et al., 2002). The excitation/emission wavelengths used for both GCaMP3 and mito-GCaMP3 were 677 488/510 nm. Timelapses were analyzed using ImageJ + Fiji software (SCR_002285).  A. Immunoblot showing P-PDH and total PDH expression in WT and MCU OE retinas. n=6 WT and 6 MCU OE retinas from 3 different fish.
B. Immunoblot of mtCO1, SDH, and Pyruvate Kinase (PK) in WT and MCU OE retinas. We observed that expression of every protein we probed for (PDH, SDH, mtCO1 and pyruvate kinase) was slightly lower in MCU OE retinas, even when the same amount of protein lysate was loaded. We hypothesize that this is due to the fact that in MCU OE retinas, MCU and RFP comprise a much larger fraction of the total protein, so other proteins appear less abundant when normalizing to total protein. n=4 WT and 4 MCU OE retinas from 4 different fish.
D. Glycolytic intermediates from WT and MCU OE retinas supplied with 13 C-glucose. We observed no trends of altered glycolytic flux over time between WT and MCU OE retinas. * p < 0.05 using Welch's t-test.

Supplemental Figure 6: 13 C-Glutamine titration in WT and MCU OE retinas.
A. Titration of WT and MCU OE retinas supplied with 13 Cglutamine (0.1, 0.2, 0.4, 0.6, 1, and 2 mM) for 15 minutes. m5 citrate (produced from reductive carboxylation) is included to show that only metabolites directly downstream of α-KGDH are produced at higher levels in MCU OE cones. Data points represent averages from n=3 retinas from 3 different fish.*p<0.05 using Welch's t-test.  A. Electron micrograph from WT fish at 5 days of age. Structures similar to the electron lucent mitochondria in MCU OE fish can be sparsely observed in the ellipsoid WT cones (white arrowhead). These were markedly smaller than structures seen in MCU OE cones. Scale bar = 2 µm.
B. Electron micrograph from WT fish at 5 days of age. What appear to be electron lucent, mislocalized mitochondria are sometimes observed in WT fish (white arrowhead). Scale bar = 2 µm.
C. Quantification of mitochondrial clusters containing all electron dense mitochondria (healthy) and any electron lucent mitochondria either in the ellipsoid or mislocalized. Data compiled from a single slice at the optic nerve of larval zebrafish at 5 days of age. Clusters were counted and plotted relative to the distance (µm) of the cluster from the optic nerve. Graphs shown are representative plots from dorsal and ventral sides of a single larvae, but this analysis was performed on n=3 larvae for each group for quantification in D.
D. Quantification of clusters containing lucent mitochondria as a percentage of total clusters observed (Healthy + Lucent (ellipsoid) + Lucent (mislocalized)) from data shown in C. N=3 larvae for WT and MCU OE. *p<0.05 and ns = not significant using Welch's t-test.
E. Electron micrograph of MCU OE fish at 3 months of age. The electron lucent mitochondrial phenotype is preserved as fish age and is not specific to early development. Scale bar = 10 µm.  A. Larval zebrafish sections at 6 days of age stained with Richardson's stain. Aside from mitochondrial disturbance in the photoreceptor layer of MCU OE retinas, normal morphology appears to be conserved. Scale bar = 50 µm.
B. Representative stitched and straightened images of TαCP:GFP (cones, magenta) retinal sections stained with a nuclear stain (Hoescht, cyan). The double cone nuclei that sit on top of the nuclear layer that contains both UV/blue cones and rods was used for quantification. Scale bar = 50 µm.
C. Richardson's stain of representative 1 year old zebrafish retina from WT and MCU OE fish. In MCU OE models, cones are rarely observed and instead rods have proliferated (rod OS = rod outer segments). Scale bar = 25 µm. D. Richardson's stain of 1 year old zebrafish retina from WT fish. Occasionally, cone ellipsoid mitochondria clusters in these older fish are observed to have large, lucent mitochondria like younger MCU OE fish (white arrowheads). Scale bar = 25 µm.