Pax2a regulates angiogenesis to facilitate mmp2-dependent basement membrane remodeling of the optic fissure

Vertebrate retinal development requires timely and precise fusion of the optic fissure (OF). Recent studies have suggested hyaloid vasculature to be involved in optic fissure fusion. In order to examine this link, we analyzed OF fusion and hyaloid vasculogenesis in the zebrafish pax2anoi mutant line. We determined that OF basement membrane (BM) remodeling, normally preceded by F-actin accumulation is mis-regulated in pax2a−/− embryos. Comparing transcriptomic profiles of pax2a−/− and wildtype eyes we discovered a novel connection between regulation of angiogenesis and fusion. Pax2a−/− eyes exhibited a significant reduction of talin1 expression, a regulator of hyaloid vasculature, in addition to increased expression of an anti-angiogenic protease, adamts1. Using 3D and live imaging we observed reduced OF hyaloid vascularization in pax2a−/− embryos. Additionally, pharmacological inhibition of VEGF signaling or adamts1 mRNA overexpression phenocopied the pax2a−/− vasculature, F-actin and BM remodeling phenotypes. Finally, we show that hyaloid vasculature expresses mmp2 which is necessary for remodeling the fissure BM. Taken together we propose a pax2a driven mechanism that restricts anti-angiogenic activity of adamts1 enabling hyaloid vasculature invasion of the OF and delivery of the BM remodeler mmp2.


10
1. Department of Biology, University of Kentucky  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36 Key Words: Optic fissure, coloboma, pax2, hyaloid vasculature, anti-angiogenesis, 37 mmp2 38 39 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 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.
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 ( 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 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 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   Dechorionated embryos were fixed with 4% PFA in PBS at room temperature for 3h 712 and washed with PBST 4 times for 5 minutes. Embryos were then permeabilized with 713 Proteinase K, 30µg/mL 10 minutes for 24-28hpf, 50µg/mL 15-20 minutes for 32-48hpf 714 and 75µg/mL 20 minutes for 56-72hpf, washed 2 times in PBST for 5 minutes and 715 blocked overnight at 4 o C with 10% sheep serum, 0.8% Triton X-100 and 1% BSA in 716 PBS. Primary mouse anti-laminin antibody (ThermoFisher -1:100) in blocking buffer 717 (1% sheep serum, 1% BSA and 0.8% Triton X-100 in PBS) were incubated overnight at 718 (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    B C E F biological adhesion biological regulation cellular component of organization or biogenesis cellular process developmental process immune system process localization locomotion metabolic process multicellular organismal process reproduction reponse to stimulus biological adhesion biological regulation cellular component of organization or biogenesis cellular process developmental process immune system process localization locomotion metabolic process multicellular organismal process reproduction reponse to stimulus rhythmic process