β-SNAP activity in the outer segment growth period is critical for preventing BNip1-dependent apoptosis in zebrafish photoreceptors

BNip1, which functions as a t-SNARE component of the syntaxin18 complex, is localized on the ER membrane and regulates retrograde transport from Golgi to the ER. BNip1 also has a BH3 domain, which generally releases pro-apoptotic proteins from Bcl2-mediated inhibition. Previously we reported that retinal photoreceptors undergo BNip1-dependent apoptosis in zebrafish β-snap1 mutants. Here, we investigated physiological roles of BNip1-dependent photoreceptor apoptosis. First, we examined the spatio-temporal profile of photoreceptor apoptosis in β-snap1 mutants, and found that apoptosis occurs only during a small developmental window, 2–4 days-post-fertilization (dpf), in which an apical photoreceptive membrane structure, called the outer segment (OS), grows rapidly. Transient expression of β-SNAP1 during this OS growing period prevents photoreceptor apoptosis in β-snap1 mutants, enabling cone to survive until at least 21 dpf. These observations suggest that BNip1-mediated apoptosis is linked to excessive activation of vesicular transport associated with rapid growth of the OS. Consistently, knockdown of Ift88 and Kif3b, which inhibits protein transport to the OS, rescued photoreceptor apoptosis in β-snap1 mutants. Treatment with rapamycin, which inhibits protein synthesis via the mTOR pathway, also rescued photoreceptor apoptosis in β-snap1 mutants. These data suggest that BNip1 performs risk assessment to detect excessive vesicular transport in photoreceptors.


Results
ER-targeted Bcl2 effectively prevents photoreceptor apoptosis in coa mutants. Our previous study on zebrafish coa mutants suggested a model in which disassembly failure of syntaxin-18 cis-SNARE complexes facilitates BNip1′s interaction with Bcl2 through its BH3 domain on the ER membrane, sequestering Bcl2 from Bax (Fig. 1A) 2 . To confirm this model, we examined whether BNip1′s pro-apoptotic activity is initiated on the ER membrane. Since zebrafish BNip1 interacts with Bcl2 in vitro 2 , we designed ER-targeted Bcl2 (Bcl2-ER), in which Bcl2 transmembrane (TM) is replaced with BNip1 TM. Furthermore, N-termini of Bcl2-ER were tagged with a fluorescent protein, either EGFP or mCherry (Fig. 1A). We confirmed that EGFP-Bcl2 was located in cytoplasm, ER, and mitochondria, whereas EGFP-Bcl2-ER was predominantly found in ER (Fig. 1B, C).
Next, we generated a zebrafish transgenic line, Tg[hs:mCherry-Bcl2-ER] that expresses mCherry-Bcl2-ER under control of the heat shock promoter, and combined it with coa mutants. We conducted TUNEL and zpr1 antibody labeling to evaluate photoreceptor apoptosis and survival. As we showed previously 2 , overexpression of Bcl2 by heat shock treatment at 36/48/60/72/84 h-post fertilization (hpf) significantly rescues photoreceptor apoptosis in coa mutants at 96 hpf (Fig. 1D, S1A). Overexpression of Bcl2-ER by heat shock treatment at 36/48/60/72/84 hpf also significantly rescues photoreceptor apoptosis in coa mutants at a similar level of Bcl2 overexpression (Fig. 1D, E, S1A, S1B). Thus, apoptotic activity of BNip1 is activated on the ER membrane and overexpression of Bcl2 on the ER membrane effectively suppresses its apoptotic activity.
To exclude the possibility that overexpression of Bcl2-ER by the heat shock promoter might cause Bcl2-ER to leach out of the ER, we used another zebrafish mutant, pinball eyes (piy), which carries a missense mutation in the DNA primase subunit 1 (prim1) gene. In this mutant, DNA replication stress is abnormally activated, leading to activation of p53-dependent apoptosis in retinal neurons 15 . In piy mutants, p53 activates downstream BH3 proteins such as PUMA, which directly suppress Bcl2 in cytoplasm or on mitochondrial membranes, leading to activation of Bax on mitochondrial membranes 15 . We combined piy mutants with transgenic lines Tg[hs:  or Tg[hs:mCherry-Bcl2-ER], conducted heat shock treatment at 36/48 hpf and examined retinal apoptosis at 60 hpf. In the piy mutant without the transgene, almost all retinal neurons underwent apoptosis by 60 hpf (Fig. S1C). On the other hand, retinal apoptosis in piy mutants was suppressed by expression of mCherry-Bcl2, but less effectively by mCherry-Bcl2-ER (Fig. S1C, S1D), suggesting that Bcl2-ER is specifically located on the ER membrane. These observations support the current model that lack of β-SNAP activity causes accumulation of the syntaxin-18 cis-SNARE complex, which facilitates the interaction between the BNip1 BH3 domain and Bcl2 on the ER membrane in coa mutant photoreceptors.
Photoreceptor apoptosis occurs during the OS growth period in coa mutants. Next, we examined photoreceptor apoptosis in coa mutants at developmental stages from 48 to 96 hpf. OSs of rods and cones were labeled with a transgene, Tg[XlaRho:XP-GFP], which expresses GFP-tagged Xenopus Peripherin 2 under Figure 1. Bcl2-ER significantly rescues photoreceptor apoptosis in coa mutants. (A) DNA construction of EGFP-tagged Bcl2, EGFP-tagged Bcl2-ER, and EGFP-tagged BNip1. EGFP-tagged Bcl2-ER was generated by replacement of the Bcl2 TM domain with a BNip1 TM domain. (B) Retinas injected with mRNA encoding ER-mKO (magenta) and either mRNA encoding EGFP-tagged Bcl2, EGFP-tagged Bcl2-ER, or EGFP-tagged BNip1 (green). Fluorescent images shown in squares of upper panels, and their green and magenta channels (middle panels). Bottom histograms indicate spatial profiles of EGFP and mKO signals along the line shown in the middle panels. Like EGFP-tagged BNip1, the spatial profile of EGFP-tagged Bcl2-ER signals correlates with that of ER-mKO signals, whereas EGFP-tagged Bcl2 signals are not correlated. Scale: 20 μm. (C) Retinas injected with mRNA encoding MT-mKO (magenta) and either mRNA encoding EGFP-tagged Bcl2, EGFP-tagged Bcl2-ER, or EGFP-tagged BNip1 (green). Fluorescent images shown in squares of upper panels, and their green and magenta channels (middle panels). Bottom histograms indicate spatial profiles of EGFP and mKO signals along the line shown in the middle panels. Like EGFP-tagged BNip1, peaks of EGFP-tagged Bcl2-ER signals do not match those of MT-mKO signals, whereas EGFP-tagged Bcl2 signals are broader than MT-mKO signals. Scale: 20 μm. (D) Retinas of wild-type and coa mutant embryos combined with transgenic lines Tg[hs:mCherrytagged Bcl2] or Tg[hs:mCherry-tagged Bcl2-ER]. Tg+ and Tg− indicate transgenic and non-transgenic embryos, respectively. Green color indicates zpr1 antibody signals, which label double-cone photoreceptors. The retinal ganglion cell layer (RGCL), inner plexiform layer (IPL), INL and photoreceptor cell layer (PCL) are shown. Right two columns indicate higher magnification of the PCL shown in the left two columns. The interface between pigmented epithelium (PE) and the neural retina is shown as a white, dotted line. Arrowheads indicate rescued photoreceptors. Scale: 50 μm (left two columns) and 10 μm (right two columns). (E) Percentage of zpr1-positive area relative to total retinal area. Both Bcl2 and Bcl2-ER partially, but significantly, increase the zpr1-positive fraction in coa mutants, suggesting that as with Bcl2, Bcl2-ER effectively rescues photoreceptor apoptosis in coa mutants. Means ± SD. Two-way ANOVA with the Tukey multiple comparison test. ***p < 0.005.

Scientific Reports
| (2020) 10:17379 | https://doi.org/10.1038/s41598-020-74360-x www.nature.com/scientificreports/ control of the Xenopus rhodopsin promoter 16 , and anti-red opsin antibody 17 , respectively. These rod and cone OS markers were first detected at 60 hpf in wild-type retinas. The size and density of OSs progressively increased from 60 to 96 hpf, suggesting that the OS actively grows from 48 to 96 hpf in zebrafish ( Fig. 2A). zpr1 is a marker of double cone-type photoreceptors (red and green cones) in zebrafish 18 . TUNEL was applied to wild-type and coa mutant retinas and counterstained with Tg[XlaRho:XP-GFP] and zpr1 antibody. In coa mutants, photoreceptor apoptosis occurs after 60 hpf, and then promptly spreads into the whole retinal region until 84 hpf. Most photoreceptors were eliminated by 96 hpf (Fig. 2B). On the other hand, such concentrated apoptosis was not observed in wild type retinas. Thus, photoreceptor apoptosis occurs from 60 to 96 hpf in coa mutants when the OS grows rapidly. In zebrafish, OS length in cones does not increase drastically after 8 dpf and it reaches adult OS size at 15 dpf, whereas the rod OS rapidly increases in size between 12 and 20 dpf and is maintained until the adult stage 19 . We evaluated the OS growth rate of rods and cones. Cone OS area was measured in wild-type retinal sections labeled with anti-green opsin antibody 20 , from 3 to 8 dpf (Fig. 2C). Cone OS size increased linearly from 3 to 5 dpf, and plateaued after 5 dpf, although small transient increases were observed at 7 dpf (Fig. 2D), which may result from a balance between synthesis of new OS and daily phagocytosis of old parts of the OS by pigmented epithelium. Next, rod OS area was measured in wild-type retinal sections labeled with fluorescent Tg[XlaRho:XP-GFP], from 3.5 to 28 dpf (Fig. 2E). Since rod OS genesis is markedly enhanced in the ventral retina in zebrafish because of retinoic acid signaling 21,22 , we measured rod OS size in the dorsal retina. Rod OS size initially increased until 4.5 dpf, then plateaued around 10 μm 2 from 4.5 to 8.5 dpf, and again started to increase until 28 dpf (Fig. 2F), indicating two phases for rod OS growth. Thus, rod and cone apoptosis occur in coa mutants during their initial OS growth from 2 to 4 dpf.
Overexpression of β-SNAP1 during the OS growth period prevents photoreceptor apoptosis in coa mutants. Since OS growth depends on synthesis of proteins and lipids and their transport to the OS, it is likely that photoreceptor apoptosis is linked to excessive activation of intracellular vesicular transport. We determined the timing of the critical period of β-SNAP1 activity to prevent photoreceptor apoptosis. A DNA construct that expresses N-terminal mCherry-tagged β-SNAP1 under control of the heat shock promoter was injected into zebrafish eggs at the one-cell stage. We confirmed that a one-hour pulse of heat shock treatment at 48 hpf induced mCherry expression at 60 hpf, which declined to 50% at 72 hpf and disappeared at 84 hpf (Fig. S2). Heat-shock treatment at 60 hpf induced mCherry expression at 72 and 84 hpf (Fig. S2C). Heat-shock treatment at 72 hpf induced mCherry expression at 84 hpf (Fig. S2C). Thus, a one-hour pulse of heat shock treatment introduces ectopic β-SNAP1 expression for at least 24 h. Next, we injected the DNA construct into coa mutants, and repeated the heat shock treatment at 12-h intervals from 36 to 72 hpf at four different starting points (36,48, 60, 72 hpf). Cone and rod survival was evaluated at 84 hpf by labeling with zpr1 antibody and fluorescent signals of Tg[XlaRho:XP-GFP], respectively (Fig. 3A). Neither heat shock treatment at 72 nor 60/72 hpf recovered cone and rod survival in coa mutants at 84 hpf ( Fig. 3B-D). However, either 4 heat-shock treatments at 36/48/60/72 hpf or 3 heat-shock treatments at 48/60/72 hpf effectively recovered cone and rod survival in coa mutants ( Fig. 3B-D). These results suggest that β-SNAP1 activity after 48 hpf is required to prevent photoreceptor apoptosis in coa mutants at 84 hpf.
To evaluate rod survival, we introduced a zebrafish transgene Tg[XlaRho:XP-GFP] into wild-type and coa mutant lines carrying Tg[hs:mCherry-β-SNAP1]. Since rod OS genesis are markedly enhanced in the ventral retina in zebrafish 21,22 , we evaluated rod OS size by measuring XP-GFP positive area in the dorsal retina (Fig. 4B). In wild-type retinas with heat-shock treatment from 36 until 132 hpf, rod OS area progressively increased from 6 to 28 dpf (Fig. 4D). In coa mutants with heat-shock treatment from 36 until 132 hpf, rod OS area was similar to that of wild type at 6 dpf, but decreased after 10 dpf, and disappeared at 28 dpf (Fig. 4D, E). Thus, rods survived in coa mutants at 6 dpf when β-SNAP1 was overexpressed from 36 to 132 hpf, but degenerated after 6 dpf. Since the rod OS actively grows from 10 to 28 dpf (Fig. 2F), rod degeneration after 6 dpf is consistent with our model that BNip1-dependent photoreceptor apoptosis is linked to the OS growing period.
There are four snap genes, α-snap1, α-snap2, β-snap1, and β-snap2, in zebrafish, but only β-snap1 mRNA is expressed in retinal photoreceptors during embryonic stages 2 . We confirmed that only β-snap1 mRNA is expressed in retinal photoreceptors at 10 dpf and adult stages (Fig. S3A, S3B) and that α-snap1, α-snap2, and β-snap2 are not upregulated in surviving photoreceptors of 19 dpf coa mutants with overexpression of β-SNAP1 from 36 to 108 hpf (Fig. S3C), excluding the possibility that α-SNAP1, α-SNAP2, or β-SNAP2 is ectopically expressed in photoreceptors at later stages in coa mutants in compensation for loss of β-SNAP1 and prevents cone photoreceptor apoptosis. Thus, cone photoreceptors survive after 6 dpf in the absence of SNAP activity in zebrafish. Taken together, these data suggest that BNip1-mediated photoreceptor apoptosis is specifically activated during the OS growth period.

Müller cells are not reprogrammed for neuronal regeneration, but rod progenitors increase in 21 dpf-coa mutants with overexpression of β-SNAP1 during the OS growth period.
Overexpression of β-SNAP1 from 36 to 108 hpf rescued cone survival in coa mutants until 21 dpf; however, rods continued to degenerate (Fig. 4C, E). In coa mutants with overexpression of β-SNAP1 from 36 to 108 hpf, apoptosis was observed in the surviving photoreceptor cell layer of the central retina as well as in the CMZ at 21 dpf, whereas apoptosis was rare in wild-type sibling retinas (Fig. 5A). However, the density of apoptotic cells in the surviving photoreceptor cell layer of coa mutants was slightly higher than that of wild type; however the difference was not significant (Fig. 5B). This suggests that apoptosis is effectively inhibited in the surviving photoreceptor cell layer in coa mutants with overexpression of β-SNAP1 during the initial OS growth period, although slightly increased apoptosis may be caused by ongoing degeneration of rods. In contrast, the density of apoptotic cells in the CMZ of coa mutants was significantly higher than that of wild-type siblings (Fig. 5B), suggesting that these CMZ photoreceptors are in the OS growing stage and undergo apoptosis.
In zebrafish, after photoreceptor damage, Müller cells are reprogrammed to assume a retinal progenitor cell state and to generate neuronal progenitor cells, which subsequently differentiate into all retinal cell types 23 . Rod progenitors normally produce rods after the embryonic stage throughout life 24,25 , and their proliferation is also activated in response to photoreceptor damage. Although apoptosis in the surviving photoreceptor cell layer was not significantly higher in coa mutants with overexpression of β-SNAP1 during the initial OS growth period (Fig. 5B), we examined whether a retinal regeneration program is activated. Labeling with anti-PCNA antibody can visualize both rod progenitor cells and proliferative reprogrammed Müller cells. In the central region of wild-type retinas at 21 dpf, the average number of rod progenitor cells and proliferative Müller cells were 12 and 2.33 per section, respectively (Fig. 5C, D). In coa mutants, the number of proliferative Müller cells increased in the central retina as well as the CMZ, although the difference was not significant. On the other hand, the number of CMZ rod progenitor cells was significantly higher than that of wild type; however, the number of central rod progenitor cells was similar to that of wild type (Fig. 5C, D). Interestingly, in coa mutants with overexpression of β-SNAP1 during the initial OS growth period, the number of rod progenitors increased twofold in the central retina, but was similar to that of wild type in the CMZ. Furthermore, the number of proliferative Müller cells was also similar to that of wild type in both the CMZ and the central retina (Fig. 5C, D). In summary, Müller cells are not reprogrammed for neuronal regeneration, but rod progenitors markedly increase in central retinas of coa mutants with overexpression of β-SNAP1 during the initial OS growth period. This is consistent with a subtle, but non-significant increase in apoptotic cell density in the surviving photoreceptor cell layer in coa mutants with overexpression of β-SNAP1 during the initial OS growth period at 21 dpf; however, this subtle increase of apoptotic cell density may trigger proliferation of rod progenitor cells.
Reduction of vesicular transport from the ER to the OS suppresses photoreceptor apoptosis in coa mutants. Next, to determine whether photoreceptor apoptosis in coa mutants is functionally linked to excessive vesicular transport, we examined whether photoreceptor apoptosis is rescued in coa mutants when vesicular transport from the ER to the OS decreases. In zebrafish and mice 26,27 , intracellular transport of photoreceptive proteins to the OS through the connecting cilium is mediated by Intraflagellar transport protein 88 (Ift88) 12 and kinesin-2 family proteins, such as Kif3b 13 . In zebrafish ift88 and kif3b mutants, the OS fails to form, but photoreceptors are still maintained at 96 hpf 12 and 120 hpf 13 , respectively. We designed MO-ift88 and MO-kif3b in accordance with previous reports and used the same concentration. When MO-ift88 was injected into wild type, embryos showed a typical downward curled body shape linked to ciliary defects at 54 hpf (Fig. S4A). Alternative splicing was also inhibited, because MO-ift88 targets a splicing site (Fig. S4B). Third, photoreceptors do not undergo apoptosis at 84 hpf (Fig. S4C). Thus, MO-ift88 effectively inhibits ciliary transport functions, but photoreceptor degeneration phenotypes appeared later than 84 hpf. We inhibited functions of Ift88 and Kif3b by Second, we treated wild-type and coa mutant embryos with rapamycin. Rapamycin inhibits the mTOR pathway, which promotes protein synthesis and transcription in response to nutrients and energy levels 14 . Rapamycin treatment generally inhibits protein synthesis, so zpr1 antibody signals would have been very weak. Thus, we evaluated photoreceptor survival with Rhodamine-conjugated phalloidin labelling. Phalloidin visualizes actin fibers associated with the outer plexiform layer (OPL), the outer limiting membrane (OLM), and connecting cilium, which outlines the photoreceptor cell layer. Rapamycin treatment from 36 to 96 hpf partially, but significantly, recovered photoreceptor survival in coa mutants (Fig. 6D, E). Since reduction of protein synthesis in the ER is likely to decrease vesicular transport to the OS, these data suggest that decreased vesicular transport from the ER to the OS suppresses photoreceptor apoptosis in coa mutants. These data support the possibility that excessive activation of vesicular transport activates BNip1 proapoptotic activity.
The ER stress response is not activated in coa mutants during the OS growth period. Depletion of β-SNAP activity compromises disassembly of cis-SNARE complexes after vesicular fusion, but eventually arrests recycling of SNARE molecules, which may activate the ER stress response. The ER stress response, also known as the unfolded protein response (UPR), restores ER homeostasis by increasing the protein-folding capacity in ER and decreasing protein load on the ER 11 . The ER stress response has three ER stress sensors: inositol-requiring protein-1α (IRE1α), activating transcription factor-6 (ATF6), and protein kinase RNA-like ER kinase (PERK). In response to ER stress, IRE1α promotes excision of a 26-nt intron of xbp1 mRNA to generate a stable transcription factor XBPs 28 . PERK inhibits general protein translation through phosphorylation of eukaryotic translation initiator factor-2 (elF2α), but also activates selective translation of the transcription factor, ATF4, which subsequently contributes to induction of apoptosis through upregulation of CAAAT/enhancerbinding protein homologous protein (CHOP) 29 . Tunicamycin treatment induces the ER stress response, such as splicing of xbp-1 and elevation of chop1 mRNA expression (Fig. 7A). However, neither splicing of xbp-1 nor elevation of chop1 mRNA expression was observed in wild-type or coa mutant heads at 60 hpf ( Fig. 7A-C). Thus, the ER stress response is not activated in coa mutants at the stage when BNip1-dependent apoptosis actively occurs. The BNip1-mediated apoptotic pathway is activated earlier than the ER stress response in zebrafish coa mutants.

Discussion
Photoreceptors undergo BNip1-dependent apoptosis in zebrafish β-snap1 mutants 2 . BNip1 is a t-SNARE component of the syntaxin-18 complex, which regulates retrograde transport from the Golgi to the ER. BNip1 also harbors a BH3 domain, which generally induces Bax-dependent apoptosis through an interaction with antiapoptotic protein Bcl2. BNip1 monomer does not induce apoptosis, but co-expression of other components of the syntaxin-18 complex enhances BNip1 pro-apoptotic activity. β-SNAP1 normally promotes disassembly of the cis-SNARE complex, and the syntaxin-18 cis-SNARE complex accumulates abnormally in β-snap1 mutants. From these observations, we proposed that BNip1 pro-apoptotic activity is activated through formation of the syntaxin-18 cis-SNARE complex (Fig. 8A). However, it is still a mystery what physiological conditions cause the syntaxin-18 cis-SNARE complex to accumulate in photoreceptors.
BNip1 is an ER-resident protein and the syntaxin-18 cis-SNARE complex is formed on the ER after vesicular fusion of retrograde-transported vesicles into the ER membrane. In this study, we examined whether BNip1 are positive and negative controls, respectively. Cone photoreceptors and rod OS were visualized by labeling with zpr1 antibody (green) and GFP signals from Tg[XlaRho:XP:GFP] (magenta), respectively. In coa mutants with Tg[hs:mCherry-β-SNAP1], photoreceptors were maintained until 21 dpf, but degenerated at 28 dpf. Scale: 50 μm. (C) Percentage of zpr1-positive area relative to total retinal area. There was no significant difference in cone survival between wild-type embryos with/without Tg[hs:mCherry-β-SNAP1] and coa mutant embryos with Tg[hs:mCherry-β-SNAP1] at 6, 10, and 15 dpf. At 21 dpf, the fraction of surviving cones in coa mutants with Tg[hs:mCherry-β-SNAP1] started to decrease although it was still significantly higher than that of coa mutants. At 28 hpf, surviving cones markedly decreased in coa mutants with Tg[hs:mCherry-β-SNAP1] in a similar level to coa mutants. Means ± SD. Two-way ANOVA with the Tukey multiple comparison test, and Multiple t-test. ***p < 0.005. (D) Higher magnification of photoreceptor cell layers indicated by squares of panel (B). Magenta channel is shown in the right. The rod OS is increased in wild type during development. However, the rod OS in coa mutants was similar to that of wild type at 6 dpf, decreased at 10-21 dpf, and disappeared at 28 dpf. (E) Percentage of XP-GFP-positive area relative to total area of the photoreceptor cell layer. In wild type controls, the XP-GFP-positive fraction increases progressively during development. However, in coa mutant embryos with Tg[hs:mCherry-β-SNAP1], the fraction is similar to that of wild type at 6 dpf, significantly decreased after 10 dpf and became 0% at 28 dpf. Means ± SD. Two-way ANOVA with Sidak's multiple comparison test, and Multiple t-test. *p < 0.05, ***p < 0.005. www.nature.com/scientificreports/ pro-apoptotic activity is activated on the ER membrane. Overexpression of ER-targeted Bcl2 effectively inhibited photoreceptor apoptosis in β-snap1 mutants, implying that BNip1 pro-apoptotic activity is activated on the ER membrane. Next, to discover what physiological conditions induce BNip1-dependent apoptosis, we investigated the spatiotemporal profile of photoreceptor apoptosis in β-snap1 mutants. We found that almost all photoreceptors, including both cones and rods, undergo apoptosis during a short developmental window from 60 to 96 hpf in β-snap1 mutants. In zebrafish, cone differentiation starts at the ventronasal retina around 50 hpf and spreads to the entire retina by 72 hpf 30 . Recent cell lineage analysis of zebrafish retina revealed that cones are generated from 40 to 80 hpf 31,32 . On the other hand, rods begin to be generated at the ventronasal retina around 50 hpf, but their propagation to the dorsal and temporal retina proceeds until later stages than that of cones 30 . Thus, photoreceptor apoptosis occurs in the early stages of their differentiation. The most prominent feature during the early stages of photoreceptor differentiation is the formation of the OS. We examined the growth rate of cone and rod OSs. The cone OS grows rapidly from 3 to 5 dpf, reaching a plateau after 5 dpf. On the other hand, the rod OS initially grows until 4.5 dpf. Then its size is maintained at a plateau from 4.5 to 8.5 dpf, and it again starts to grow from 8.5 until 28 dpf, suggesting two growth phases. Consistently, it has been reported that the OS length in cones does not drastically increase after 8 dpf, whereas the rod OS rapidly increases in size between 12 and 20 dpf 19 . Thus, both cones and rods undergo apoptosis in β-snap1 mutants during their initial OS growth period from 2 to 4 dpf. Although there is a second expansion phase of the rod OS, it is likely that the initial expansion of the rod OS is enough to trigger rod apoptosis in β-snap1 mutants. Next, to determine the critical period of β-SNAP activity for photoreceptor survival, we overexpressed β-SNAP1 during different time windows in β-snap1 mutants using the heat shock promoter. Transient expression of β-SNAP1 from 36 to 132 hpf is enough to prevent cone and rod apoptosis in coa mutants at 6 dpf. Interestingly, cones continue to survive until 21 dpf. However, rods progressively degenerate after 6dpf. Since overexpressed β-SNAP1 protein can be maintained at least for 24 hpf after heat-shock treatment, it is likely that cones are maintained in the absence of β-SNAP activity after 6 dpf. On the other hand, rods failed to be maintained in the absence of β-SNAP after 6 dpf, probably because rod OS genesis is still active from 8.5 to 28 dpf. These data suggest that BNip1-dependent apoptosis is associated with OS growth of rod and cone photoreceptors.
Since protein and lipid synthesis increase during the OS growth period, their intracellular transport must be elevated during the OS growth period. Increased vesicular transport is likely to trap β-SNAP molecules on vesicular fusion sites, which subsequently decrease the relative contribution of β-SNAP to vesicular fusion events on the ER membrane, leading to accumulation of syntaxin-18 cis-SNARE complexes on the ER membrane (Fig. 8B). This scenario is consistent with our current model of the BNip1-mediated apoptosis mechanism. Indeed, ER-targeted Bcl2 effectively rescues photoreceptor apoptosis in β-snap1 mutants, suggesting that BNip1 BH3 activation occurs on the ER membrane. Second, we did not observe activation of the ER stress response in β-snap1 mutants at 2.5 dpf, suggesting that depletion of β-SNAP activity is detected by BNip1 prior to activation of the ER stress response. Third, we examined whether decreased intracellular transport rescues photoreceptor apoptosis in β-snap1 mutants. Knockdown of ciliary transport regulators, Ift88 and Kif3b, partially, but significantly, rescues photoreceptor apoptosis in β-snap1 mutants. Furthermore, rapamycin treatment, which inhibits protein synthesis through suppression of the mTOR pathway 14 rescued photoreceptor apoptosis in β-snap1 mutants. These observations suggest that BNip1 provides risk assessment to detect excessive activation of vesicular transport in photoreceptors. Since arrest of the anterograde transport pathway primarily causes retention of synthesized proteins in the ER and activates the ER stress response, BNip1 and the ER stress response may cooperatively determine an appropriate level of vesicular transport during photoreceptor development and homeostasis (Fig. 8B).

Figure 5.
Photoreceptor apoptosis and neuronal regeneration in coa mutants with overexpression of β-SNAP1 during the OS growth period. (A) TUNEL of wild-type and coa mutant retinas with overexpression of β-SNAP1 during the initial OS growth period. Cones are visualized with zpr1 (green). Yellow and red circles indicate TUNEL signals in photoreceptors of central retinas and retinal CMZ, respectively. Higher magnification images of all TUNEL signals indicated by white squares are shown in the right side. Scale: 100 μm. (B) The number of TUNEL-positive cells per 10,000 μm 2 in photoreceptors and CMZs of 21-dpf, wild-type siblings and coa mutants with overexpression of β-SNAP1 during the OS growth period. TUNEL density is slightly increased in rescued coa mutant photoreceptors, but does not differ significantly from that of wild-type siblings. On the other hand, TUNEL density in the CMZ is significantly higher in coa mutants than in wild-type siblings. Means ± SD. Two-way ANOVA with Sidak's multiple comparison test. ***p < 0.005. (C) Anti-PCNA antibody labeling of wild-type and coa mutants with overexpression of β-SNAP1 during the initial OS growth period. Arrowheads and arrows indicate rod progenitor cells and reprogrammed proliferative Müller cells, respectively. Compared with wild-type control (upper), in coa mutants (middle), the number of reprogrammed proliferative Müller cells was increased in the central retina. In coa mutants with overexpression of β-SNAP1 during the initial OS growth period (bottom), the number of rod progenitor cells is markedly increased in the central retina; however, reprogrammed proliferative Müller cells were only observed in the peripheral retina and not in the central retina. Scale: 50 μm. (D) The number of PCNA-positive cells per section. In coa mutants with overexpression of β-SNAP1 during the initial OS growth period, numbers of proliferative Müller cells in both the CMZ (red bars) and the central retina (purple bars) were similar to that of wild type. Rod progenitors in the CMZ were not significantly different from those of wild type (blue bars). Interestingly, rod progenitors in the central retina were twice as numerous as in wild type (green bars). Means ± SD. Two-way ANOVA with the Tukey multiple comparison test. *p < 0.05. Fish. Zebrafish (Danio rerio) were maintained according to standard procedures 33    The ratio of the spliced form relative to the non-spliced form of xbp-1 mRNA. We carried out three independent sets of PCR reactions for wild type, coa mutants, DMSO and Tunicamycin treatment. The same set of PCR reactions are connected with the line. The difference of coa mutant heads relative to wild-type heads, and Tunicamycin-relative to DMSO-treated embryos in xbp1 spliced/non-spliced mRNA ratio were evaluated by Ratio paired t-test, two-tailed. p** < 0.01. There was no significant difference between wild type and coa mutants. (C) Ratio of chop1 mRNA relative to ef1α mRNA. The difference of coa mutant heads relative to wild-type heads, and Tunicamycin-relative to DMSO-treated embryos in chop1 mRNA expression were evaluated by Ratio paired t-test, two-tailed. p* < 0.05. There was no significant difference between wild type and coa mutants. Calculation of the zpr1-positive and cell death areas relative to total retinal area. The percentage of zpr1-positive area relative to total retinal area (Figs. 1D, E; 3B, C; 4B, C; 6A, B, C) was calculated using one cryosection image containing the central retina per eye, as previously described 2 . Means and standard deviations were calculated from data obtained in four or six retinal images from more than two embryos. Statistical analysis was done using two-way ANOVA with the Tukey multiple comparison test (Figs. 1E; 3C; 4C; 6B, C). The percentage of TUNEL-positive area relative to total retinal area (Fig. S1A, B) was calculated using cryosection images containing the central retina for each eye, as previously described 2 . Means and standard deviations were calculated from data obtained in four retinal images from more than three embryos. Statistical analysis was done using two-way ANOVA with the Tukey multiple comparison test (Fig. S1B).
The extent of the cell death area relative to the total retinal area (%) in piy mutants (Fig. S1C, D) was determined using one plastic section image containing the central retina for each eye. Using Image-J software (NIH), we demarcated the outline of the cell death area, which was represented by pyknotic nuclei, on each plastic section image. Then we compared its size with total retinal size. Means and standard deviations were calculated from data obtained for six retinal images from three embryos. Statistical analysis was done using one-way ANOVA with the Tukey multiple comparison test (Fig. S1D).
Calculation of the XP-GFP-positive area relative to the total area of the photoreceptor cell layer. The percentage of the XP-GFP-positive area relative to total area of photoreceptor cell layer (Figs. 2E, F; 3B, D; 4D, E) was calculated in the dorsal retina using one cryosection image containing the central retina for each eye, similar to the method previously described 2 . Means and standard deviations were calculated. Statistical analysis was done using one-way ANOVA with the Dunnett's multiple comparison test (Fig. 2F), and two-way ANOVA, the Sidak's multiple comparison test (Fig. 3D, 4E).
Calculation of TUNEL density in the photoreceptor cell layer and the CMZ. The number of TUNEL-positive cells was counted in the surviving photoreceptor cell layer or the CMZ of 19 dpf coa mutant and wild-type sibling embryos carrying the transgene Tg[hs:mCherry-β-SNAP1] with heat-shock treatment at 36/48/60/72/84/96/108 hpf (Fig. 5A). Three independent wild-type and four independent coa mutant retinas were used for TUNEL (red) and anti-zpr1 antibody labeling (green). The average number of TUNEL-positive cells was determined in the surviving photoreceptor cell layer and the CMZ, respectively, using 4 independent sections per individual retina. The zpr1-positive area and the CMZ area were outlined and their areas were determined using Image-J software (NIH). TUNEL signal density was calculated as the number of TUNEL per 10,000 μm 2 . Means and standard deviations were calculated. Statistical analysis was done using two-way ANOVA with Sidak's multiple comparison test (Fig. 5B).
Estimation of OS size during development. Using Image-J software (NIH), the OS size of individual cones was calculated from the outline of the green opsin-positive area on cryosection images of wild-type retinas labelled with anti-green opsin antibody at 3, 4, 5, 6, 7 and 8 dpf. The OS size of indivisual rods was calculated from the outline of the XP-GFP-positive area on cryosection images of wild-type retinas at 3.5, 4.5, 5.5, 6.5, 7.5, 8.5, 10, 15, 21 and 28 dpf. Means and standard deviations were calculated. Statistical analysis was performed using one-way ANOVA with the Tukey multiple comparison test (Fig. 2D) and the Dunnett's multiple comparison test (Fig. 2F).