Pax2a regulates angiogenesis to facilitate fusion of the optic fissure

ABSTRACT Formation of the vertebrate retina involves timely and precise fusion of the optic fissure, failure of which results in the congenital blinding disorder coloboma. The optic fissure forms in consequence of optic cup morphogenesis and has a functional role as a conduit for migration of endothelial vasculature during hyaloid vasculogenesis. Recent studies have linked hyaloid vasculature precursor cells to be potential triggers, initiating optic fissure fusion. In order to examine this link, we analyzed the molecular events associated with optic fissure fusion and hyaloid vasculogenesis in the zebrafish pax2a mutant coloboma model. By means of wholemount immunohistochemistry, we determined that optic fissure fusion initiation, as indicated by laminin remodeling, is preceded by a spike in F-actin signaling. In contrast, the spike in F-actin activity and the disassembly of the basement membrane (BM) were misregulated in pax2a-/- embryos. To examine how the fusion process is disrupted in the absence of pax2a function, we compared transcriptomic profiles between pax2a-/- and wildtype (WT) embryos. This analysis uncovered a novel connection between regulation of angiogenesis and fusion. Loss of pax2a function resulted in significant reduction of talin1 expression, a known regulator of endothelial vasculature migration associated with fissure fusion. In addition, loss of pax2a also resulted in increased expression of an anti-angiogenic protease, ADAMTS1. 3D confocal and live imaging of retinal hyaloid vascularization in Tg[kdrl:mCherry] embryos indicated a significant deficit in pax2a-/- embryos while pharmacological inhibition of VEGF using DMH4 or ADAMTS1 mRNA overexpression phenocopied the pax2a mutant phenotype. Taken together, we propose that Pax2a positively regulates tln1 while negatively regulating ADAMTS1 expression. This enables timely hyaloid vascularization of the retina which in turn directly signals to initiate fissure fusion via cytoskeletal rearrangements and subsequent BM remodelling. AUTHOR SUMMARY Formation of a fully functioning human eye requires precise and coordinated cellular mechanisms during very early development. One such mechanism is responsible for the formation of the eyeball structure from a flat sheet of cells to a hemispherical organ. In order to facilitate the biophysical properties of this shape change, an opening on the ventral side of the eye forms and must be closed in order to properly form the eyeball. Failure of this event leads to pediatric blindness in the form of a disease termed Coloboma. While it has been studied by researchers for almost a century, we still lack a significant understanding of how and why some children are born with this disease. In this study, we aimed to better understand how the closure of this opening is triggered by the developing tissues. What we found is, that non-eye cells which are found within this opening and are making vasculature networks to nourish the eye, are likely triggers for the fusion event. We found that regulating their migration and proliferation, also known as angiogenesis, is required for closure of this opening and therefore proper development of the eye.

confirm that in the absence of hyaloid vasculature, ectopic retinal OFs fail to initiate 121 propose that accumulation of F-actin is necessary for the initiation of OF BM 214 remodeling. We therefore next sought to investigate the mechanism of OF F-actin 215 accumulation. 216 217 Pax2a retinal transcriptomic profile during optic fissure fusion. 218 Although pax2 has been studied in many systems, there are currently no obvious 219 transcriptional targets of pax2 that would directly point to mechanistic regulation of OF 220 fusion. In an effort to understand the absence of OF BM remodeling and mis-regulation 221 of F-actin observed in pax2a -/embryos we compared whole eye transcriptomic profiles 222 between WT and pax2a -/embryos. Eyes from 48hpf WT and pax2a -/embryos were 223 isolated and total RNA was subsequently purified and sequenced using Illumina 224 sequencing. 48hpf was chosen for our experimental time point as pax2a -/embryos are 225 easily phenotyped at this age thanks to observable heart malformations. When 226 comparing three replicates for WT and pax2a -/embryos using RSEM software, we 227 detected 1215 transcripts significantly upregulated (>95% confidence interval) and 1202 228 transcripts significantly downregulated (Fig 3A and 3D). Gene ontology analysis 229 indicated a wide spread of biological function being affected, including cytoskeletal 230 signaling, adhesion and developmental processes (Fig 3B, E). In fact, the distribution of 231 biological function between up and downregulated genes was highly similar (Fig 3B, D). 232 A list of the top 20 up and down regulated genes is outlined in Figure 3C and F. Results 233 of all the statistically significant up and downregulated genes identified in our assay are 234 presented as supplementary data (Table 1 and Table 2). 235 When examining our list of downregulated targets, one, talin1 (tln1), stood out in 236 particular. Not only was tln1 one of the most downregulated targets (~200 fold 237 decrease) it was also recently shown to be necessary for the migration of hyaloid 238 vasculature cells into the OF and subsequent fusion (22). As such, we next sought to 239 investigate the relationship between pax2a, tln1, vasculogenesis and optic fissure fusion The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/678185 doi: bioRxiv preprint To validate our RNAseq data, we performed wholemount in situ hybridization 245 (WISH) for tln1 comparing WT siblings to pax2a -/mutant embryos. WT expression of 246 tln1 was observed in the optic fissure between 28 and 48hpf coinciding with pax2a 247 expression ( Fig 4A, Fig S3). In pax2a -/embryos, OF tln1 expression appears 248 significantly reduced compared to WT while retaining similar expression in periocular 249 regions ( Fig 4A). Tln1 expression is also reduced in the mid brain-hind brain boundary, 250 another region of strong pax2a expression (Fig 4A, S3A). Our WISH data was further 251 supported by qPCR results showing a significant decrase in tln1 expression at 32hpf 252 within the fissure. However, over the next six hours of imaging it is apparent that pax2a -258 /embryos have significantly fewer mCherry expressing cells pass through the OF (Fig  259   4C, Movie 1). To quantify this effect, we fixed WT and pax2a -/-Tg[kdrl:mCherry] 260 embryos at 24, 32, 36 and 48hpf, collected 3D confocal stacks and counted the number 261 of mCherry positive cells found within the OF (Fig 4D). The data indicate that in pax2a -/-262 embryos there is a significant reduction in the number of mCherry expressing cells at 263 both 32 and 36hpf ( Fig 4E). Furthermore, using 3D rendering, we noted that in 48hpf 264 pax2a -/embryos the hyaloid vasculature established in the back of the lens is reduced 265 in size and lacks proper connections to the newly forming choroidal and superficial 266 vasculature systems (Movie 2-3). Overall, our data indicate that pax2a -/embryos exhibit 267 decreased expression of tln1 and impaired hyaloid vascularization of the optic fissure. 268 269

Inhibition of VEGF signaling impairs optic fissure fusion mechanics. 270
Based on our discovery of impaired OF hyaloid vasculature in pax2a -/embryos, 271 we next examined whether this phenomenon is associated with failure of optic fissure 272 fusion. Hence, we turned our attention to vascular endothelial growth factor (VEGF) 273 signaling. VEGF, the ligand for vascular endothelial growth factor receptor (VEGFR), is 274 a prime candidate for regulating the migration and proliferation of hyaloid vasculature 275 . CC-BY 4.0 International license is made available under a The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/678185 doi: bioRxiv preprint cells. In support of this notion, when examining expression patterns for VEGF ligands, 276 vegfaa and ab, ba, c and d, we found vegfaa, ab and c to be expressed in the retina and 277 periocular regions (Fig S4A). To inhibit VEGF signaling, we took advantage of the 278 dorsomorphin derivative DMH4, which has been shown in zebrafish to selectively inhibit 279 VEGF signaling independent from BMP (38). Based on published working 280 concentrations, we conducted a dose response to examine DHM4 effects on hyaloid 281 vasculature using Tg[kdrl:mCherry] as a readout (38). Treatment of embryos from 12-282 24hpf, ranging from 1-100µM, resulted in a dose dependent reduction of mCherry signal 283 in the developing retina ( Fig S4B). We decided to use the 100µM concentration for 284 subsequent experiments as this concentration was able to completely inhibit 285 vascularization of the retina up to 72hpf without any significant impact on overall embryo 286 health (Fig 5A, S4C). Embryos were treated starting at 12hpf and examined for fissure 287 fusion status via whole mount laminin IHC at 24, 32, 36, 48, 56 and 72hpf ( Fig 5B). 3D 288 confocal imaging revealed a persistence of laminin signal within the fissure up and 289 including 72hpf in all embryos examined ( Fig 5B). Measurement of the laminin signal 290 generated very similar results to what we observed in pax2a -/embryos ( Fig 5C). This is 291 highlighted by retention of laminin signal at 48hpf and persistence of laminin in 100% 292 (n=49) of embryos treated up to 72hpf (Fig 5B-C). In addition to laminin, we also imaged 293 and quantified OF F-actin levels during the treatment (Fig 5B). Similar to pax2a -/-294 embryos DMH4 treatment prevented the accumulation of OF F-actin between 24-32hpf 295 ( Fig 5D). This was in contrast to DMSO treated embryos which exhibited the 296 accumulation of F-actin in the fissure between 24-32hpf like wildtype embryos. DMH4 297 treatment did not affect the size of the retina nor the apposition of retinal lobes (Fig S2). 298 Lastly, we also observed a significant decrease of tln1, but not pax2a, expression in 299 DMH4 treated embryos (Fig S4E,F). This suggests that tln1 is likely expressed by the 300 vasculature cells within the OF. Based on these findings, we conclude that inhibition of 301 VEGF signaling results in failure of F-actin accumulation and basement membrane 302 breakdown, presumably due to the absence of hayloid vasculature cells in the OF. 303 304 Anti-angiogenic protease adamts1 is upregulated in OF of pax2a mutant embryos. 305 . CC-BY 4.0 International license is made available under a The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/678185 doi: bioRxiv preprint Having established a connection between pax2a function and recruitment of 306 hyaloid vasculature cells into the optic fissure, we next sought to examine the 307 mechanism behind it. We therefore re-examined our transcriptomic comparison of WT 308 and pax2a -/eyes, with a focus on angiogenic regulation. This new examination 309 revealed misexpression of a disintegrin and metalloproteinase with thrombospondin 310 motifs 1 (adamts1), a secreted protease known to be involved in regulating VEGF 311 signaling (39). Being a member of the ADAMTS family, adamts1 contains three 312 thrombospondin motifs enabling it to directly bind and sequester VEGF (31). 313 Furthermore, the protease activity of adamts1 has bene shown to target 314 thrombospondin 1 and 2 for cleavage, ultimately liberating their active forms which bind 315 to and block VEGF from binding the VEGF receptor (30). Our RNAseq data indicated 316 that adamts1 expression was upregulated almost two-fold in pax2a -/eyes (table 1). This 317 would suggest that the expression of adamts1, and therefore its anti-angiogenic activity, 318 is normally kept in check by pax2a. Pax2a has not been reported to harbor 319 transcriptionally repressive function, so this may involve an indirect mechanism. To 320 confirm our RNAseq results we performed WISH, examining adamts1 expression at 28 321 and 36hpf. In agreement with our transcriptomic analysis, we observed an upregulation 322 in adamts1 in the ventral region of pax2a -/eyes, specifically surrounding the optic 323 fissure ( Fig 6A). A statistically significant increase in adamts1 expression was also 324 confirmed using qPCR ( Fig 6B). To test whether upregulation of adamts1 expression 325 would have direct effects on hyaloid vascularization and optic fissure fusion we injected 326 Tg[kdrl:mCherry] embryos with adamts1 mRNA. We subsequently examined hyaloid 327 vasculature using 3D confocal imaging between 24-48hpf. Injection of adamts1 mRNA 328 resulted in a significant decrease in the number of kdrl:mCherry cells found within the 329 optic fissure ( The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/678185 doi: bioRxiv preprint Having mimicked the vasculature phenotype of pax2a mutants by over 337 expressing adamts1, we next determined whether it had any effects on OF fusion. To 338 do so, we performed whole mount laminin and F-actin staining in adamts1 mRNA 339 injected embryos at 24, 32, 36, and 48hpf ( Fig 7A). Confocal imaging of the fissure 340 indicated that adamts1 mRNA injection also phenocopied pax2a -/associated 341 persistence of laminin and therefore failure of optic fissure fusion. Quantified levels of 342 laminin persisting at 48hpf was similar to pax2a -/embryos ( Fig 7B). Similar to pax2a -/-343 and DMH4 treated embryos, we again noted a lack F-actin accumulation between 24-344 32hpf in adamts1 overexpressing embryos ( Fig 7C). As was observed with pax2a -/-, 345 timing for apposition of the lobes was not affected in adamts1 overexpressing embryos 346 ( Fig S2). Taken together, our data suggest that maintaining proper levels of adamts1 347 expression is necessary to enable VEGF signaling and subsequent recruitment of 348 hyaloid vasculature required for initiation of OF BM remodeling. 349 350 Hyaloid vasculature is the source of mmp2 necessary for OF BM remodeling. 351 The above data confirm and support a model where hyaloid vasculature drives or 352 initiates the OF fusion process. However, a missing key to this model is the mechanism 353 by which vasculature cells induce fissure fusion. James et al, along with others, have 354 suggested vasculature cells may be a source of BM remodeling enzymes, such as 355 matrix metaloproteases (mmp). To investigate this further, we used WISH to examine 356 mmp expression within the OF between 24-48hpf. Our data indicated that mmp2, 14a 357 and 14b were expressed within the fissure between 28-36hpf (Fig S5 and Famulski lab 358 unpublished data). Mmp2 had recently been associated with OF fusion in the mouse 359 while evidence from an mmp2 metalloproteinase activity probe in zebrafish indicated 360 mmp2 activity is present in the developing eye and likely OF (40, 41). mmp14 is an 361 activator of mmp2 and its co-expression with mmp2 within the fissure suggests mmp2 is 362 in fact active (42,43). To test whether mmp2 fits within our model we assayed mmp2, 363 mmp14a and mmp14b expression in pax2a -/and DMH4 treated embryos. In both 364 cases, OF expression of mmp2, mmp14a and mmp14b was reduced as observed by 365 WISH and confirmed by qPCR (Fig 8A, B). This suggested that all three are expressed 366 by the hyaloid vasculature cells. To confirm this, we performed 2 color WISH for mmp2 367 . CC-BY 4.0 International license is made available under a The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/678185 doi: bioRxiv preprint and kdrl. Confocal imaging verified that mmp2 expression is co-localized with that of 368 kdrl within the OF (Fig 8C). Conversely, when performing 2 color WISH for mmp2 and 369 rorB, a retina specific probe, we did not detect co-localization of the signals. We 370 therefore conclude that the source of mmp2, 14a and 14b enzymes in the OF are 371 hyaloid vasculature cells. Finally, to determine whether mmp2 activity is necessary for 372 OF fusion we treated embryos with ARP101, an mmp2 specific inhibitor previously 373 shown to be effective in zebrafish (44). Treating embryos from 24-48hpf inhibited OF 374 BM remodeling in a dose dependent manner (Fig 8D,E, S6). When compared to DMSO, 375 embryos treated with 20µM ARP101 either completely or partially retain their OF BM up 376 to 48hpf. Mmp2 activity is therefore necessary for BM remodeling in the OF. To 377 determine when mmp2 activity is required we began the ARP101 treatment between 378 24-32hpf and assayed BM status at 48hpf ( Fig 8F). Treatments started later than 30hpf 379 had little to no effect on OF BM remodeling, indicating that mmp2 activity is required 380 between ~24-30hpf ( Fig 8F). This finding also correlates with the observed timing of 381 mmp2 expression in the OF ( Fig S5). However, because we cannot determine exactly 382 how quickly ARP101 inhibits enzymatic activity in our assay, active mmp2 may persist 383 in the OF longer than 30hpf. Additionally, work from other labs has identified several molecular components 414 associated with cell-cell adhesion and epithelial sheet fusion to function within the 415 fissure, including a-catenin, n-cadherin, and netrin (22,24,(48)(49)(50)(51). However, the timing 416 and the molecular mechanism organizing and regulating these components remains 417 uncharacterized. Finally, the elephant in the room has always been the identity of the 418 BM remodeling mechanism. To date, only adamts16 has been functionally examined in 419 context of fissure fusion, mmp2 activity was detected during mouse OF fusion and 420 mmp23bb was implicated in the fusion process from transcriptomic data (41,50,52). 421 While attempting to address the mechanistic aspects of steps 3 and 4 using the 422 pax2a noi model we first uncovered a relationship between F-actin and BM remodeling. we discovered that BM remodeling is in fact preceded by an increase in F-actin signal 426 within the OF (Fig 1). When assayed in pax2a -/embryos the F-actin accumulation is 427 absent while the hyaloid vasculature in the OF is also diminished. The timing of F-actin 428 . CC-BY 4.0 International license is made available under a The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/678185 doi: bioRxiv preprint accumulation coincides with the active migration of hyaloid vasculature cells through the 429 fissure (Fig 4). Furthermore, we show pharmacologically, via VEGF inhibition, that 430 hyaloid vasculature is necessary for the accumulation of F-actin and OF BM 431 remodeling. Transcriptionally, we show that the decrease in hayloid vasculature in 432 pax2a -/embryos coincides with a significant decrease in tln1 expression. Tln1 is a key 433 regulator of endothelial cell migration, recently shown to be directly involved in OF 434 fusion (22). In fact, loss of tln1 function has been previously shown to decrease in 435 fissure associated vasculature and inhibit OF BM remodeling (22). We also show a 436 significant decrease in tln1 expression upon VEGF inhibition (DMH4 treatment), 437 suggesting that tln1 expression within the OF is associated with hayloid vasculature. 438 Talin1 is known to be a direct link between the actin cytoskeleton and the BM via 439 integrin (53, 54), and has been associated with the formation of adhesion junctions (55, 440 56). One could therefore hypothesize several different models for its role in optic fissure 441 fusion. We predict that tln1 is regulating the ability of hayloid vasculature cells to migrate 442 properly through the OF, however we cannot rule out the possibility that tln1 directly 443 participates in the OF fusion process. The absence of tln1 expression is most likely a 444 correlation to the decrease of hayloid vasculature cells within the fissure of pax2a -/-445 embryos. Our transcriptomic analysis was performed using whole eye tissue, which 446 would have included the hayloid vasculature. As such, the decrease in tln1 expression 447 is likely indicative of decreased numbers of hayloid vasculature cells which require tln1 448 expression for proper migration into the fissure and or proliferation. 449

450
Our study strongly supports the notion that vasculature plays an integral part in OF 451 fusion. As such, modulation of angiogenesis represents a potential mechanism to 452 ensure proper vascularization of the OF. In our transcriptomic comparison of WT and 453 pax2a mutants we did in fact discover a connection to angiogenesis. In the absence of 454 pax2a function we observed an upregulation of adamts1. Examining adamts1 455 expression in the eye further confirmed an upregulation of expression in the ventral 456 regions pax2a -/retinas. Encoding 3 thrombospondin (thsb) motifs able to directly bind 457 and sequester VEGF, in addition to targeting thsb1 and 2 for cleavage into their active 458 VEGF inhibiting form, adamts1 is well known for it's anti-angiogenic function in cancer, 459 . CC-BY 4.0 International license is made available under a The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/678185 doi: bioRxiv preprint aortic and renal biology (57). In our study, we show that upregulating adamts1 460 expression reduces the number of hyaloid vasculature cells in the OF. Up regulation of 461 adamts1 also prevents the accumulation of F-actin and subsequent BM remodeling. 462 Taken together we propose that the reduction in hayloid vasculature observed in 463 pax2a -/embryos likely stems from an upregulation of adamts1 leading to suppressed 464 VEGF signaling and therefore limited vascularization of the OF. 465

466
The last aspect of our proposed model pertains to the mechanism of OF BM 467 remodeling. It had been suggested over the years that vasculature may be the source of 468 BM remodeling activity during OF fusion. To that end, we have discovered that mmp2, 469 mmp14a and mmp14b are all expressed within the hayloid vasculature during OF 470 fusion. In fact, timing of their expression matches precisely with the initiation of BM 471 remodeling. When we examine pax2a mutants or inhibit hayloid vasculature completely 472 we no longer detect expression of either of these mmps. Furthermore, using ARP101, a 473 specific mmp2 inhibitor, we show that mmp2 activity is necessary for OF BM remodeling 474 and this activity is specifically required between 26-32hpf. Mmp2 has been implicated 475 in OF fusion previously. Mouse studies have shown mmp2 expression within 476 macrophages residing in the OF. Furthermore, recent examination of mmp2 activity, 477 using a reporter construct, indicates that mmp2 is active in the eye and likely the OF 478 (40). Mmp14 is a known to activate mmp2 activity by cleaving the inhibitory pro-peptide 479 of mmp2. As such, co-expression of mmp2 and 14 is further indicative of mmp2 playing 480 a functional role in OF BM remodeling. 481 To date, the only BM remodeling enzyme to be associated with OF fusion is 482 adamts16. Loss of adamts16 function in zebrafish, via morpholino, led to a coloboma 483 like phenotype which the authors credited to the inability to degrade laminin in the 484 fissure (52). We have not been able to reproduce adamts16 expression within the 485 fissure. Recent studies in mice, zebrafish and chick have compared expression in OF 486 cells pre, during and post fusion (24,50,58,59). Surprisingly, none of these studies The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/678185 doi: bioRxiv preprint 14a and 14b signal within the optic fissure (yellow arrowhead) appears decreased in 604 pax2a -/and DMH4 treated embryos. B) qPCR analysis confirms a decrease in 605 expression of mmp2, 14a and 14b in both pax2a -/and DMH4 treated embryos. C) Two 606 color wholemount in situ hybridization simultaneously examining mmp2 and kdrl, or 607 mmp2 and rorB expression at 32hpf. DNA was stained with DAPI. Scale bar = 50µm. 608 Clear overlap of signal is observed for mmp2 and kdrl, but not mmp2 and rorB. D) 609 Whole mount Immunohistochemistry was used to visualize laminin (red) in ARP101 610 The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/678185 doi: bioRxiv preprint  The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/678185 doi: bioRxiv preprint (Alexa Fluor® 488 -Abcam -1:1000), DAPI 1:1000, and phalloidin (Alexa Fluor® 555 -720 1:50) were incubated overnight at 4 o C in the dark. Tg[kdrl:mCherry] embryos were 721 treated with proteinase K as described above and stained with DAPI 1:1000 overnight. 722 The embryos were washed 2 times in PBST for 15 minutes and visualized using a 723 Nikon C2+ confocal microscope equipped with a 40X (1.15NA) water immersion 724 objective. Embryos were embedded in 1.2% low melting point agarose on glass bottom 725 35mm dishes (Fluorodish, World Precision Instruments). Images were captured in steps 726 of 3.5 microns for a total of 31.5 microns using Nikon Elements software. Image 727 adjustment, such as cropping and brightness/contrast was performed using Adobe 728 Photoshop. 729 730

Analysis of Fluorescence signal 731
Fiji software (https://fiji.sc) was used to measure the fluorescence intensity of laminin 732 and phalloidin signal from raw image data. In order to account for variability in staining, 733 normalization values were measured for laminin and actin pixel intensity where an area 734 directly outside of the choroid fissure was measured and a ratio was generated between 735 the two values ( Fig S1B). In cases where the fissure edges were farther apart then the 736 size of the box used for analysis, the box was divided into two and used to measure The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/678185 doi: bioRxiv preprint distinct heart defects only observed in pax2a -/embryos, as verified by previous 751 genotyping experiments. Dissected eyes were suspended in 1 mL of trizol and sheared 752 with a 22-gauge needle. Samples were incubated at room temperature for 5 minutes, 753 then 200 L of chloroform was added and vortexed for 1 minute. Samples were then 754 centrifuged at 12,000 g for 15 minutes at 4℃. Aqueous phase was removed and put into 755 fresh RNAse free tube. 3 L of Glycoblue was added to samples and vortexed for 5 756 seconds. 500 L of 100% Isopropanol was then added and vortexed for an additional 10 757 seconds. Samples then were incubated at room temperature for 10 minutes. 758 Subsequently, samples were centrifuged at 12,000 g for 10 minutes at 4℃. Supernatant 759 was removed and 1 mL of 75% EtOH was added and vortexed for 10 seconds. Samples 760 were then centrifuged at 7,500 g for 5 minutes at 4℃. Supernatant was removed and 761 samples then underwent a pulse centrifugation and any remaining supernatant was 762 removed. Samples were air-dried for 7 minutes at room temperature under a fume 763 hood. 20 L of RNAse free ddH2O was added to the samples and mixed until pellet 764 dissolved. Finally, samples were incubated at 60℃ for 12 minutes and stored at -80℃. 765 The RNA then underwent a DNase treatment using the DNA-free Kit. 2 L of 10X 766 DNaseI Buffer, and 1 L of rDNaseI was added to the entire 20 L sample. Samples 767 were incubated at 37℃ for 20 minutes, and then 2 L of DNase Inactivation Reagent 768 was added. Samples were then incubated at room temperature for 2 minutes, mixing 769 samples 3 times. Samples were centrifuged at 10,000 g for 1.5 minutes. Supernatant 770 was then removed and placed into fresh RNase free tube and stored at -80℃. Whole-mount in situ hybridization was performed as previously described (62). RNA 779 probes were generated using PCR with T7 promoter sequence linkers and 780 subsequently transcribed [DIG or FITC labeled] using T7 polymerase (Roche). Primer 781 . CC-BY 4.0 International license is made available under a The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/678185 doi: bioRxiv preprint sequences are all found in Table 3. Images were captured using a Nikon Digital sight 782 DS-Fi2 camera mounted on a Nikon SZM800 stereo scope using Elements software. 783 Dissected eyes from 24, to 72hpf embryos were mounted in 70% glycerol and imaged 784 under DIC using a Nikon TiE compound microscope equipped with a 20X (0.7NA) 785 objective and Elements software. Image adjustment, such as cropping and 786 brightness/contrast was performed using Adobe Photoshop. 787 788 qPCR analysis 789 32hpf embryos were anesthetized with tricane, tail tips were collected for genotyping 790 and the heads, dissected just posterior of the eyes, were fixed in RNAlater. After 791 genotyping, embryos corresponding to WT and pax2a-/-were pooled, 5-10 embryos, 792 and total RNA isolated using a RNAaqueous kit (Ambion). DMSO and DMH4 treated 793 embryos were harvested in the same fashion absent any genotyping. qPCR was 794 performed as previously described (63). were imbedded in 1.1% low gelling agarose in 1-inch glass bottomed Flourodish cell 806 culture dishes (World Precision Instruments) and covered in embryo media, 3-amino 807 benzoic acidethylester (tricaine) to anaesthetize the embryos and 1-phenyl 2-thiourea 808 (PTU) to inhibit pigmentation. Z-stacks 75µm thick with a step size of 2.5µm were 809 captured over the course of 6 hours at 10 minute intervals. The time lapse data were 810 reconstructed in 3D using Elements software. Image adjustment, such as cropping and 811  The copyright holder for this preprint (which was not peer-reviewed) is the author/funder.  The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/678185 doi: bioRxiv preprint