Two neuronal peptides encoded from a single transcript regulate mitochondrial complex III in Drosophila

Naturally produced peptides (<100 amino acids) are important regulators of physiology, development, and metabolism. Recent studies have predicted that thousands of peptides may be translated from transcripts containing small open-reading frames (smORFs). Here, we describe two peptides in Drosophila encoded by conserved smORFs, Sloth1 and Sloth2. These peptides are translated from the same bicistronic transcript and share sequence similarities, suggesting that they encode paralogs. Yet, Sloth1 and Sloth2 are not functionally redundant, and loss of either peptide causes animal lethality, reduced neuronal function, impaired mitochondrial function, and neurodegeneration. We provide evidence that Sloth1/2 are highly expressed in neurons, imported to mitochondria, and regulate mitochondrial complex III assembly. These results suggest that phenotypic analysis of smORF genes in Drosophila can provide a wealth of information on the biological functions of this poorly characterized class of genes.

Together, these approaches suggest there may be hundreds, possibly 161 thousands, of unannotated smORF genes. However, these "omics" methods do 162 not tell us which smORFs encode peptides with important biological functions. peptides. Therefore, continued functional characterization is needed to tackle the 176 enormous number of predicted smORF peptides. 177 178 Here, through an effort to systematically characterize human-conserved smORF 179 genes in Drosophila (in preparation), we identified two previously unstudied 180 smORF peptides CG32736-PB and CG42308-PA that we named Sloth1 and 181 Sloth2 based on their mutant phenotypes. Remarkably, both peptides are 182 translated from the same transcript and share amino acid sequence similarity, 183 suggesting that they encode paralogs. Loss of function analysis revealed that 184 each peptide is essential for viability, and mutant animals exhibit defective 185 neuronal function and photoreceptor degeneration. These phenotypes can be 186 explained by our finding that Sloth1 and Sloth2 localize to mitochondria and play 187 an important role in complex III assembly. Finally, we propose that both peptides 188 bind in a shared complex. These studies uncover two new components of the 189 mitochondria and demonstrate how functional characterization of smORFs will 190 lead to novel biological insights. 2013). Furthermore, it is well known that paralogs are often found adjacent to 214 each other in the genome due to tandem duplication (TAYLOR AND RAES 2004). 215 Therefore, we propose that sloth1 and sloth2 are paralogs translated from the 216 same transcript.

218
Sloth1 and Sloth2 closely resemble their human orthologs (SMIM4 and 219 C12orf73), based on sequence similarity, similar size, and presence of a 220 transmembrane domain ( Figure 1B). Like Sloth1 and Sloth2, SMIM4 and 221 C12orf73 also have subtle amino acid sequence similarity to each other ( Figure  222 1B). In addition, sloth1 and sloth2 are conserved in other eukaryotic species 223 ( Figure 1C). Remarkably, sloth1 and sloth2 orthologs in choanoflagelate, sea 224 squirt, and lamprey exhibit a similar bicistronic gene architecture as Drosophila 225 ( Figure 1C, Supplemental File 1). In contrast, sloth1 and sloth2 orthologs in 226 jawed vertebrates (e.g. mammals) are located on different chromosomes (e.g. 227 human Chr.3 and Chr.12, respectively). Interestingly, we only found one ortholog 228 similar to sloth2 in the evolutionarily distant Plasmodium, and two orthologs 229 similar to sloth2 in Arabidopsis, which are located on different chromosomes 230 ( Figure 1C). Therefore, we hypothesize that the sloth1 and sloth2 ORFs 231 duplicated from an ancient single common ancestor ORF and became unlinked 232 in animals along the lineage to jawed vertebrates.

234
We next investigated sloth1 and sloth2 translation parameters and efficiency, 235 since their ORFs are frameshifted relative to each other ( Figure 1A) and they are 236 not separated by an obvious internal ribosome entry site (IRES) ( VAN DER KELEN 237 et al. 2009). Remarkably, only five nucleotides separate the stop codon of the 238 upstream ORF (sloth1) and the start codon of the downstream ORF (sloth2) 239 ( Figure 1A). Therefore, sloth1 should be translated first and inhibit translation of 240 sloth2, similar to the functions of so-called upstream ORFs (uORFs) (THOMPSON 241 2012). However, sloth1 has a non-optimal Kozak sequence 5' to the start codon 242 (ACACATG) and sloth2 has an optimal Kozak (CAAAATG) (CAVENER 1987). 243 Therefore, scanning ribosomes may occasionally fail to initiate translation on 244 sloth1, in which case they would continue scanning and initiate translation on 245 sloth2, known as "leaky scanning" translation (THOMPSON 2012).

247
To test this translation model, we constructed an expression plasmid with the 248 Renilla Luciferase (RLuc) reporter gene downstream of sloth1 (sloth1-RLuc), 249 while retaining non-coding elements of the original transcript (5' UTR, Kozak 250 sequences, 5bp intervening sequence) ( Figure 1D). By transfecting this reporter 251 plasmid into Drosophila S2R+ cells, along with a Firefly Luciferase (FLuc) control 252 plasmid, we could monitor changes in translation of the downstream ORF by the 253 ratio of RLuc/FLuc luminescence. Using derivatives of the reporter plasmid with 254 Kozak or ATG mutations, we found that translation of the downstream ORF 255 increased when translation of sloth1 was impaired ( Figure 1E). Reciprocally, 256 translation of the downstream ORF was decreased when sloth1 translation was 257 enhanced with an optimal Kozak. These results suggest that sloth1 inhibits 258 translation of sloth2, and that balanced translation of both smORFs from the 259 same transcript might be achieved by suboptimal translation of sloth1. 260 261 sloth1 and sloth2 are essential in Drosophila with non-redundant function 262 263 To determine if sloth1 and sloth2 have important functions in Drosophila, we 264 used in vivo loss of function genetic tools. We used RNA interference (RNAi) to 265 knock down the sloth1-sloth2 bicistronic transcript. Ubiquitous expression of an 266 shRNA targeting the sloth1 coding sequence (Figure 2A) lead to significant 267 knockdown of the sloth1-sloth2 transcript in 3 rd instar larvae ( Figure 2B), as 268 determined by two different primer pairs that bind to either the sloth1 or sloth2 269 coding sequence. Ubiquitous RNAi knockdown of sloth1-sloth2 throughout 270 development lead to reduced number of adult flies compared to a control ( Figure  271 2C). This reduced viability was largely due to adult flies sticking in the food after 272 they eclosed from their pupal cases ( Figure 2D). Escaper knockdown flies were 273 slow-moving and had 30% climbing ability compared to control flies ( Figure 2E).

276
We confirmed our RNAi results using CRISPR/Cas9 to generate somatic 277 knockout (KO) flies. By crossing flies ubiquitously expressing Cas9 (Act-Cas9) 278 with flies expressing an sgRNA that targets the coding sequence of either sloth1 279 or sloth2 ( had significantly reduced viability compared to controls ( Figure 2G). Furthermore, 283 escaper adults had short scutellar bristles ( Figure 2H) and frequently appeared 284 sluggish. Importantly, similar phenotypes were observed when targeting either 285 sloth1 or sloth2. 286 287 Next, we further confirmed our loss of function results using CRISPR/Cas9 in the 288 germ line to generate KO lines for sloth1 and sloth2. These reagents are 289 particularly important to test if sloth1 and sloth2 have redundant function by 290 comparing the phenotypes of single and double null mutants. We generated four 291 KO lines ( gene Gal4 that removes sloth1 and sloth2 coding sequences (Gal4-KI). Since 295 sloth1 and sloth2 are on the X-chromosome, we analyzed mutant hemizygous 296 male flies. All four mutant lines were hemizygous lethal, which were rescued by a 297 genomic transgene ( Figure 2I,), ruling out off-target lethal mutations on the X-298 chromosome. Like RNAi and somatic KO results, rare mutant adult escaper flies 299 had slower motor activity ( Figure 2J) and short scutellar bristles ( Figure 2K). 300 Furthermore, the short scutellar bristle phenotype and slower motor activity could 301 be rescued by a genomic transgene ( Figure 2J, K).

303
The  Figure 2A). We found that 312 sloth1-KO lethality could only be rescued by {sloth1-Δsloth2}, and vice versa, 313 sloth2-KO lethality could only rescued by {Δsloth1-sloth2} ( Figure 2L). 314 Furthermore, single ORF rescue transgenes were unable to rescue the lethality 315 of dKO and Gal4-KI lines ( Figure 2L). Third, we used the Gal4/UAS system 316 (BRAND AND PERRIMON 1993) to rescue mutant lethality with ubiquitously 317 expressed cDNA transgenes. These results showed that single ORF KOs could 318 only be rescued by expression of the same ORF ( Figure 2L). Similar results were 319 found by expressing cDNAs encoding the human orthologs ( Figure 2L). In all, 320 these results show that both sloth1 and sloth2 are essential, have similar loss of 321 function phenotypes, are not functionally redundant with one another, and are 322 likely to retain the same function as their human orthologs.

324
Loss of sloth1 and sloth2 leads to defective neuronal function and 325 degeneration 326 327 Since loss of sloth1 and sloth2 caused reduced adult mobility and climbing 328 defects ( Figure 2E, J), we speculated that the two peptides normally play an 329 important role in the brain or muscle. To determine where sloth1 and sloth2 are 330 expressed, we used the Gal4-KI line as an in vivo transcriptional reporter. Gal4-331 KI mobility defects and lethality could be rescued by expressing the entire 332 bicistronic transcript (UAS-sloth1-sloth2) ( Figure 2J, L), or coexpression of both 333 smORFs as cDNA (UAS-sloth1 and UAS-sloth2) ( Figure 2L). Thus, the Gal4-KI 334 line is likely an accurate reporter of sloth1 and sloth2 expression. By crossing 335 Gal4-KI flies with a UAS-GFP fluorescent reporter, we observed strong GFP 336 expression in larval ( Figure 3A, B) and adult brains ( Figure 3C). In addition, Gal4-337 KI is expressed in motor neurons at the larval neuromuscular junction (NMJ) 338 ( Figure 3D) and in larval brain cells that are positive for the neuronal marker Elav 339 ( Figure 3E).

341
We then tested if sloth1 and sloth2 were important for neuronal function by an amplitude similar to that of genomic rescue animals ( Figure 4B). However, 353 upon repetitive light stimulation, ERG amplitudes were significantly reduced 354 ( Figure 4B), suggesting a gradual loss of depolarization. Similar results were 355 observed when young flies were raised in 24hr dark ( Figure 4C). Moreover, ERG 356 traces also showed a progressive loss of "on" and "off" transients ( Figure  In addition to measuring neuronal activity, we analyzed dKO neurons for changes 371 in morphology and molecular markers. Confocal imaging of the NMJ in dKO 3 rd 372 instar larvae did not reveal obvious changes in synapse morphology or markers 373 of synapse function (Supplemental Figure 4). In contrast, using transmission 374 electron microscopy (TEM) of sectioned adult eyes, we observed reduced 375 photoreceptor number and aberrant morphology such as enlarged 376 photoreceptors and thinner glia in dKO animals ( Figure 5A-C), suggestive of 377 degeneration. These phenotypes were rescued by a genomic transgene, but not 378 with single ORF rescue constructs ( Figure 5A-C, Supplemental Figure 5). 379 Furthermore, these phenotypes were similar between young and aged flies, as 380 well as aged flies raised in the dark ( Figure  if this mechanism is occurring in dKO photoreceptors, we imaged Rh1 protein 384 levels using confocal microscopy. We observed Rh1 accumulation in 385 degenerating dKO photoreceptors in 4 week aged flies exposed to light ( Figure  386 5D). However, Rh1 accumulation was milder in 4 week aged flies raised in the 387 dark (Supplemental Figure 6). These results point out that light stimulation, and 388 hence activity, enhance degeneration due to Rh1 accumulation in dKO animals.

390
Sloth1 and Sloth2 localize to mitochondria and their loss impairs normal 391 respiration and ATP production 392 393 Mitochondrial dysfunction in Drosophila is known to cause phenotypes that are 394 reminiscent of loss of sloth1 and sloth2, such as pupal lethality, reduced neuronal 395 activity, photoreceptor degeneration, and Rh1 accumulation in photoreceptors 396 ( Sloth1 and Sloth2 proteins colocalized with the mitochondrial marker ATP5α 409 ( Figure 6A). Furthermore, Sloth1-FLAG and Sloth2-FLAG were enriched in 410 mitochondrial fractions relative to cytoplasmic fractions ( Figure 6B). Similar 411 results were observed using stable S2R+ cell lines that express streptavidin 412 binding peptide (SBP) tagged Sloth1 or Sloth2 under a copper inducible promoter 413 (MT-Sloth1-SBP and MT-Sloth2-SBP) ( Figure 6C).

415
Next, we raised antibodies to Sloth1/2 to determine their endogenous 416 localization. Using two independently generated antibodies for each peptide, 417 immunolocalization in larval brains from wild-type or sloth1/2 dKO animals 418 showed no overlapping signal with a mitochondrial marker and no clear signal 419 above background (Supplemental Figure 7). Furthermore, we did not detect 420 Sloth1 or Sloth2 bands of the expected molecular weight on western blots from 421 wild-type S2R+ whole cell lysates or isolated mitochondria using anti-Sloth1, anti-422 Sloth2, anti-SMIM4, or anti-C12orf73 (Supplemental Figure 8A-C). In contrast, 423 anti-Sloth1 western blots of mitochondria isolated from 3 rd instar larvae and adult 424 thoraxes showed a <15kDa band that is absent from sloth1/2 KO or RNAi 425 samples (Supplemental Figure 8D), suggesting this band corresponds to 426 endogenous Sloth1. Unfortunately, anti-Sloth2 failed to detect a similar band 427 under the same conditions (Supplemental Figure 8D).

429
Since our Sloth1/2 antibodies may not be sensitive enough to detect the 430 endogenous peptides, we generated a stable S2R+ cell line expressing sloth1/2 431 transcript under a copper inducible promoter (MT-sloth1/2) and induced 432 expression for 16hrs. Anti-Sloth1 and anti-Sloth2 western blots of mitochondria 433 isolated from MT-sloth1/2 cells detected <15kDa bands that did not appear in 434 wild-type S2R+ cells, and thus are likely Sloth1 and Sloth2 peptides translated 435 from the overexpressed sloth1/2 transcript (Supplemental Figure 8B). 436 Furthermore, Sloth1 and Sloth2 were enriched in MT-sloth1/2 mitochondrial 437 fractions relative to cytoplasmic fractions ( Figure 6D), similar to the results 438 obtained with FLAG and SBP-tagged peptides ( Figures 6B-C). Based on their 439 amino acid sequence, Sloth1 and Sloth2 are predicted to run at 9.3kDa and 440 6.7kDa, respectively. While Sloth1 does appear to run larger than Sloth2, both 441 peptides run ~2kDa larger than expected ( Figure 6D).

443
A method of assaying defects in mitochondrial function is measuring cellular 444 oxygen consumption from live cells with a Seahorse stress test. Since this 445 typically involves assaying a monolayer of cells, we generated KO S2R+ cell 446 lines using CRISPR/Cas9. Compared to control cells, single KO and double KO 447 S2R+ cells (Supplemental Figure 9A, B) had reduced basal respiration ( Figure  448 7A, B), ATP production (Supplemental Figure 9C), and proton leaks 449 (Supplemental Figure 9D). Results were similar for single KO and dKO lines. 450 These results suggest that both sloth1 and sloth2 are required to support normal 451 mitochondrial respiration in S2R+ cells.

453
Next, we assayed sloth1 and sloth2 mutant flies for defects in mitochondrial Indeed, dKO larvae had ~60% ATP compared to control larvae, which was 458 rescued by a genomic transgene ( Figure 7C). Impaired mitochondrial function 459 can also lead to cellular stress responses, such as increased expression of the mitochondria with normal cristae ( Figure 7E). In contrast, mitochondrial number 468 was increased in mutant photoreceptors in aged animals ( Figure 7E, 469 Supplemental Figure 10A) and decreased in mutant photoreceptors in young 470 animals ( Figure 7F, Supplemental Figure 10B). In all, these data suggest that 471 Sloth1 and Sloth2 localize to mitochondria and are important to support 472 respiration and ATP production.  Figure 8A). Similarly, the complex III band was diminished in 492 mitochondria isolated from sloth1/2 knockout 3 rd instar larvae ( Figure 8B). This 493 change was rescued by a wild-type genomic transgene, but not single paralog 494 transgenes ( Figure 8B). Next, we detected individual respiratory subunits by 495 SDS-PAGE and western blotting of isolated mitochondria. Using antibodies that 496 recognize UQCR-C2, the fly homolog of human complex III subunit UQCRC2, we 497 found that the ~40kDa band corresponding to UQCR-C2 was diminished in 498 mitochondria isolated from sloth1/2 RNAi adult thoraxes ( Figure 8C), as well as 499 sloth1/2 knockout 3 rd instar larvae ( Figure 8D). 500 501 To CG10075-HA ( Figure 8E) and RFeSP-HA ( Figure 8F) binding to anti-FLAG 508 beads. In contrast, Sloth2-FLAG pulled-down CG10075-HA and RFeSP-HA 509 weakly or was at background levels ( Figure 8E,F). Together, these results 510 suggest that Sloth1/2 are required for proper complex III assembly, mediated 511 through physical interaction with complex III subunits. 512 513 Sloth1 and Sloth2 act in a stoichiometric complex 514 515 We speculated that Sloth1 and Sloth2 could physically interact, based on the 516 observation that both share the same loss of function phenotypes and subcellular 517 localization. Indeed, some paralogs bind to the same protein complex 518 (SZKLARCZYK et al. 2008) and there is a tendency for proteins in the same 519 complex to be co-expressed (PAPP et al. 2003). To confirm this putative 520 interaction between Sloth1 and Sloth2, we used co-immunoprecipitation and 521 western blotting. This revealed that Sloth1-FLAG could immunoprecipitate 522 Sloth2-HA ( Figure 9A), and reciprocally Sloth2-FLAG ( Figure 9B) could 523 immunoprecipitate Sloth1-HA. Interestingly, the levels of tagged peptide in cell 524 lysates were higher when the opposite peptide was overexpressed ( Figure 9A,B). 525 Proteins To test this possibility for Sloth1/2, we overexpressed either sloth1 or sloth2 in 533 vivo. Low-level ubiquitous overexpression (using da-Gal4) of either UAS-sloth1 534 or UAS-sloth2 cDNA had no effect on adult fly viability ( Figure 2L). To increase 535 expression levels, we used the strong ubiquitous driver tub-Gal4. Whereas 536 tub>sloth1 flies were viable as adults, tub>sloth2 animals were 100% pupal lethal 537 ( Figure 9C). However, tub>sloth2 animals could be rescued to adulthood by co-538 expression of sloth1. Importantly, this rescue was not due to dilution of the Gal4 539 transcription factor on two UAS transgenes, since co-expression of UAS-540 tdtomato did not rescue tub>sloth2 lethality. Finally, tub-Gal4 overexpression of 541 the entire sloth1-sloth2 bicistronic transcript resulted in viable adult flies. In all, 542 these results suggest that Sloth1 and Sloth2 interact in a complex where their 543 stoichiometric ratio is important for normal function. 544 545 Discussion 546 547 Here, we have assigned new functions to two previously uncharacterized smORF 548 peptides. Sloth1 and Sloth2 appear to be distantly-related paralogs, yet each is 549 important to support mitochondrial and neuronal function in Drosophila. We 550 propose a model where Sloth1 and Sloth2 peptides are translated from the same 551 transcript, imported into mitochondria where they interact with each other and 552 complex III to promote its assembly ( Figure 10) observations support leaky scanning, and we propose a model whereby both 573 peptides are translated because sloth1 contains a non-optimal Kozak sequence.

575
The presence of sloth1 and sloth2 orthologs in many eukaryotic species suggest 576 that their function is likely broadly conserved. Indeed, we could rescue the 577 lethality of sloth1 and sloth2 mutant flies by expressing their human counterparts. 578 Interestingly, Plasmodium and Arabidopsis only have homologs with similarity to 579 sloth2. Perhaps sloth2 maintained functions more similar to its common ancestor 580 with sloth1. We were unable to identify homologs in some eukaryotes such as 581 yeast, though their amino acid sequence may simply be too diverged for 582 detection using bioinformatic programs such as BLAST.

584
The physical interactions of Sloth1-Sloth2, Sloth1-RFeSP, and Sloth1-CG10075, 585 and complex III assembly defects in sloth1/2 loss of function animals, suggest 586 that Sloth1/2 together regulate complex III assembly. Indeed, Sloth1 is 587 bioinformatically predicted to localize to the mitochondrial inner membrane 588 in neurons and could rescue sloth1/2 lethality, it is likely these peptides play 627 important roles in other cell types. For example, publicly available RNA-seq data 628 suggest that they are ubiquitously expressed (Flybase). In addition, neuronal 629 expression of sloth1 or sloth2 was unable to rescue mutant lethality ( Figure 2L). 630 Furthermore, we observed sloth1/2 loss of function phenotypes in dissected adult 631 thoraxes, which are composed of mostly muscle. At present, there are no 632 reported human disease-associated mutations in SMIM4 and C12orf73. 633 Mutations in these genes might not cause disease, or they might cause lethality. 634 It is also possible that the lack of functional information on these genes has 635 hampered identification of disease-associated mutations.

637
There is great interest in identifying the complete mitochondrial proteome (   Images of the NMJ were acquired on a Zeiss Axio Zoom V16 or a Zeiss 780 1259 confocal microscope. Images were taken from muscle 6/7 segment A2. Images 1260 were processed using Fiji software. Quantification of bouton number from NMJ 1261 stained with anti-HRP and anti-Dlg1 was performed by manual counting of 1262 boutons in an entire NMJ for wild-type (N=8) and dKO animals (N=7). A T-test 1263 was used to determine significance. 1264 1265 For imaging whole larvae, wandering 3 rd instar larvae were washed with PBS and 1266 heat-killed for 5min on a hot slide warmer to stop movement. Larvae were 1267 imaged using a Zeiss Axio Zoom V16 fluorescence microscope.

1269
For imaging the adult brain, ~1 week old adult flies were dissected in PBS and 1270 whole brains were fixed in 4% paraformaldehyde for 20min. Fixed brains were 1271 permeabilized in PBT, blocked for 1hr in 5% normal goat serum (S-1000, Vector 1272 Labs) at room temperature, incubated with anti-HRP 647 overnight at 4˚C, 1273 washed with PBT and PBS, and incubated in mounting media (90% glycerol + 1274 10% PBS) overnight at 4˚C. Adult brains were mounted on glass slides under a 1275 coverslip using vectashield (H-1000, Vector Laboratories Inc.). Images of adult 1276 brains were acquired on a Zeiss 780 confocal microscope. Images were 1277 processed using Fiji software.

1279
For confocal microscopy of adult photoreceptors, the proboscis was removed 1280 and the head was pre-fixed with 4% formaldehyde in PBS for 30 min. After pre-1281 fixation, eyes were removed from the head and fixed an additional 15 minutes. 1282 Fixed eyes were washed with PBS 3x for 10 min each and permeabilized in 0.3% 1283 Triton X-100 in PBS for 15 min. Permeabilized, fixed samples were blocked in 1X 1284 PBS containing 5% normal goat serum (NGS) and 0.1% Triton X-100 for 1 h 1285 (PBT). Samples were incubated in primary antibody diluted in PBT overnight at 1286 4°C, washed 3x with PBT, and incubated in secondary antibodies in NGS for 1hr 1287 at room temp the next day. Following secondary antibody incubation, samples 1288 were washed with PBS and were mounted on microscope slides using 1289 vectashield. Samples were imaged with LSM710 confocal with 63X objective and 1290 processed using Fiji software. 1291 1292 S2R+ cells transfected with Sloth1-FLAG or Sloth2-FLAG were plated into wells 1293 of a glass-bottom 384 well plate (6007558, PerkinElmer) and allowed to adhere 1294 for 2 hours. Cells were fixed by incubating with 4% paraformaldehyde for 30min, 1295 washed with PBS with .1% TritonX-100 (PBT) 3x 5min each, blocked in 5% 1296 Normal Goat Serum (NGS) in PBT for 1hr at room temperature, and incubated in 1297 primary antibodies diluted in PBT-NGS overnight at 4˚C on a rocker. Wells were 1298 washed in PBT, incubated with secondary antibodies and DAPI and washed in 1299 PBS. Plates were imaged on an IN Cell Analyzer 6000 (GE) using a 20x or 60x 1300 objective. Images were processed using Fiji software. TEM of Drosophila adult retinae were performed following standard electron 1314 microscopy procedures using a Ted Pella Bio Wave processing microwave with 1315 vacuum attachments. Briefly, whole heads were dissected in accordance to 1316 preserve the brain tissue. The tissue was covered in 2% paraformaldehyde, 2.5% 1317 Glutaraldehyde, in 0.1 M Sodium Cacodylate buffer at pH 7.2. After dissection, 1318 the heads were incubated for 48hrs in the fixative on a rotator at 4˚C. The pre-1319 fixed heads were washed with 3X millipore water followed by secondary fixation 1320 with 1% aqueous osmium tetroxide, and rinsed again 3X with millipore water. To 1321 dehydrate the samples, concentrations from 25%-100% of Ethanol were used, 1322 followed by Propylene Oxide (PO) incubation. Dehydrated samples are infiltrated 1323 with gradual resin:PO concentrations followed by overnight infiltration with pure 1324 resin. The samples were embedded into flat silicone molds and cured in the oven 1325 at 62°C for 3-5 days, depending on the atmospheric humidity. The polymerized 1326 samples were thin-sectioned at 48-50 nm and stained with 1% uranyl acetate for 1327 14 minutes followed by 2.5% lead citrate for two minutes before TEM 1328 examination. Retina were viewed in a JEOL JEM 1010 transmission electron 1329 microscope at 80kV. Images were captured using an AMT XR-16 mid-mount 16 1330 mega-pixel digital camera in Sigma mode. Three animals per genotype per 1331 condition were used for TEM. At least 30 photoreceptors were used for organelle 1332