Top2b is involved in the formation of outer segment and synapse during late‐stage photoreceptor differentiation by controlling key genes of photoreceptor transcriptional regulatory network

Topoisomerase II beta (Top2b) is an enzyme that alters the topologic states of DNA during transcription. Top2b deletion in early retinal progenitor cells causes severe defects in neural differentiation and affects cell survival in all retinal cell types. However, it is unclear whether the observed severe phenotypes are the result of cell‐autonomous/primary defects or non–cell‐autonomous/secondary defects caused by alterations of other retinal cells. Using photoreceptor cells as a model, we first characterized the phenotypes in Top2b conditional knockout. Top2b deletion leads to malformation of photoreceptor outer segments (OSs) and synapses accompanied by dramatic cell loss at late‐stage photoreceptor differentiation. Then, we performed mosaic analysis with shRNA‐mediated Top2b knockdown in neonatal retina using in vivo electroportation to target rod photoreceptors in neonatal retina. Top2b knockdown causes defective OS without causing a dramatic cell loss, suggesting a Top2b cell‐autonomous function. Furthermore, RNA‐seq analysis reveals that Top2b controls the expression of key genes in the photoreceptor gene–regulatory network (e.g., Crx, Nr2e3, Opn1sw, Vsx2) and retinopathy‐related genes (e.g., Abca4, Bbs7, Pde6b). Together, our data establish a combinatorial cell‐autonomous and non–cell‐autonomous role for Top2b in the late stage of photoreceptor differentiation and maturation. © 2017 The Authors Journal of Neuroscience Research Published by Wiley Periodicals, Inc.


INTRODUCTION
Rod and cone photoreceptors constitute the largest cell population in the mammalian retina (Carter-Dawson and LaVail, 1979;Brzezinski and Reh, 2015;Cepko, 2015). A recent study using single-cell RNA-seq profiling has determined that the percentages of rods and cones in the mouse are 65.6% and 4.2%, respectively (Macosko et al., 2015). Lineage studies showed that both cell types are derived from multipotent retinal progenitor cells (reviewed by Brzezinski and Reh, 2015;Cepko, 2015;Wang and Cepko, 2016). Birth-dating studies have established that the genesis of retinal cell types is a highly SIGNIFICANCE Photoreceptors sense light signals with their outer segments and transduce the signals to other cells in the neural retina, enabling animals to see. Defects in photoreceptors cause their own death, eventually leading to severe eye diseases and often blindness. Extensive studies have revealed that many genes are involved in the differentiation of photoreceptors from retinal progenitor cells. In this study, we show that topoisomerase II beta contributes to the late-stage differentiation/maturation of photoreceptors, especially in the formation of outer segments and synapses by affecting the expression of key genes in the photoreceptor transcriptional network and genes linked to retinopathies. conserved process. In the mouse, the production of cone cells starts around embryonic day 12 (E12) and peaks at E14-E15, and the majority of cone cells are differentiated by birth. Meanwhile, the production of rod cells starts around E14 and peaks at birth, and the majority of rods are generated by postnatal day (P) 8 to 12 (Young, 1985b;Rapaport et al., 2004;Wang et al., 2014). Three major stages are well defined in photoreceptor cell development: 1) cell proliferation and cell fate determination, during which the multipotent retinal progenitors proliferate and their competence is restricted as photoreceptor cell precursors; 2) early differentiation stage, during which genes for morphogenesis and phototransduction are expressed; and 3) late differentiation stage, which includes axonal growth, synapse formation, and outer segment (OS) biogenesis (Swaroop et al., 2010). The development of the OS is well documented with electron microscopic studies (Tokuyasu and Yamada, 1959;Nilsson, 1964;Steinberg et al., 1980). However, the molecular mechanism underlying photoreceptor cell differentiation is not completely understood. Studies have shown that photoreceptor development involves a complex gene regulatory network including several key factors (e.g., Otx2, Crx, Nr2e3, Nrl, Rorb, Opn1sw/mw/lw) Swaroop et al., 2010;Brzezinski and Reh, 2015;Wang and Cepko, 2016). Mutations in one or more of these critical genes cause loss of the OS and further complete loss of photoreceptors that leads to several retinopathies (e.g., retinitis pigmentosa, loss of vision) (Coppieters et al., 2007;Freund et al., 1997;Swain et al., 1997;Weitz et al., 1992).
Topoisomerase II beta (Top2b) is an enzyme that controls and alters the topologic states of DNA during transcription (Wang, 2002). The onset of Top2b expression is observed in progenitor cells that just finished the final division and is actively involved in neural development (Tsutsui et al., 1993;Tsutsui et al., 2001;Lyu and Wang, 2003;Tiwari et al., 2012). Many studies have established a multifaceted role of Top2b in 1) resolving the topological constraints in early-stage neuronal gene expression (Madabhushi et al., 2015); 2) neurite guidance during late-stage neuronal development (Yang et al., 2000;Nevin et al., 2011); 3) cerebral stratification (Lyu and Wang, 2003); and 4) control of developmentally regulated genes in the brain and neural retina (Lyu et al., 2006). We have shown severe defects in differentiation and survival of retinal neurons and M€ uller glia in Dkk3-Cre-mediated Top2b conditional knockout animals (cKO) (Li et al., 2014). However, because of the pleiotropic effect of Top2b knockout on gene expression, the mechanism underlying Top2b functions (e.g., cellautonomous (primary) or non-cell-autonomous (secondary) during retinal development is unclear.
In this study, we show that Top2b plays an important role in photoreceptor development in both a cellautonomous and non-cell-autonomous manner. Transcriptome analysis by RNA-seq reveals that Top2b controls key genes in the photoreceptor regulatory network and genes linked to retinopathies.

Sample Collection
All studies performed were conducted under the strict guidelines and consent of the Institutional Animal Care and Use Committee of Rutgers, the State University of New Jersey. Top2b cKO animals were generated by breeding a strain of Dkk3-Cre mice (Sato et al., 2007) with the Top2b-floxed mice on a 129SvEv background. Three to five animals from each genotype were used for each stage. No animal was excluded. Gender of the animals was not reported because no difference between different sexes was observed. Animals were euthanized with CO 2 inhalation. For immunohistochemistry, retinae were dissected from the animals and fixed with 4% (w/v) PFA for 1 hr. Following fixation, samples were washed three times with 1 3 PBS for 10 min each, soaked in 30% (w/v) sucrose for 2 to 3 days, embedded in cryopreserving media (Tissue Tek V R OCT compound, SAKURA FINETEK USA INC, Torrance, CA), and stored at 2808C. Cryosectioning was performed with a cryostat (Thermo Scientific, Waltham, MA). Sections of samples were air dried for 10 min and stored at 2808C. For RNAseq analysis, P0 and P6 retinae (n 4 for each genotype on each stage) were dissected rapidly free of other ocular tissue and frozen with liquid nitrogen. For total RNA isolation, retinal samples were transferred into 500 ml of TRIzol reagent (Invitrogen), and isolation was performed according to the manufacturer's instructions.

Immunohistochemistry
Immunofluorescence staining was performed on mouse eye sections as described previously (Li et al., 2014). The primary antibodies used in this study are listed in Table I. Images were captured using a Zeiss Axio Imager M1 fluorescence microscope and analyzed using AxioVision 4.8 (Zeiss, Germany). For proper comparison between control and cKO samples, all images were captured from the dorsal-medial region of the retina. Cell counting and measurements were performed in the central region of the retina, and the regions were randomly selected. For each data point, samples from at least three animals were calculated for an average and standard deviation. Two-tailed heteroscedastic Student t-test was employed to calculate P values.

Helium Ion Microscopic Imaging
Retinae from control and Top2b cKO mice were collected, fixed, and sectioned as described previously (Li et al., 2014). Each slide contains 8 to 10 sections, which are 10 to 12 mm thick. For chemical drying, the slides were subjected to sequential dehydration in graded ethanol (Fisher Scientific, Waltham, MA) (50%, 70%, 80%, 90%, and 100%) baths at 20-minute intervals. After the final 20 minutes of ethanol dehydration, solutions of different ratios of ethanol and hexamethyldisilazane (HMDS, UltraPure Solutions, Inc., Castroville, CA) (25%, 50%, 75%, 100%) were added sequentially to replace the inner-tissue solution from ethanol to HMDS. The tissue samples were removed from the 100% HMDS bath, placed in a fume hood with cover for 12 hr, and stored in a newly charged desiccation chamber under vacuum. For helium ion microscopy (HIM; ORION Plus, Carl Zeiss, Peabody, MA), the sample was mounted onto an HIM stub with carbon tape, and images were taken with different fields of view using ORION PLUS software.

Cre-mediated Top2b Knockout and shRNA-mediated Top2b Knockdown
For Top2b gene knockout in postnatal mouse retinae, a CAG-Cre-GFP construct (Addgene plasmid #13776) was injected into the subretinal space of P0 pups of Top2b f/f mice (Lyu and Wang 2003). For Top2b knockdown, an effective Top2b shRNA (shTop2b) and a control plasmid construct using LentiLox 3.7 vector (Azarova et al., 2010) were used. Plasmid DNA injection and electroporation were performed following a published protocol (Matsuda and Cepko, 2004). Briefly, the plasmids were mixed with FastGreen (0.025%), and 0.5 ml of DNA was injected into the subretinal space between the retina and choroid of P0 wild-type mouse pups. After injection, tweezer-type electrodes (model 520; 7-mm diameter; BTX, San Diego, CA) briefly soaked in PBS were placed to softly hold the heads of the pups, and five 80V square pulses of 50-msec duration with 950-msec intervals were applied using a pulse generator ECM830 (BTX, San Diego, CA). Usually DNA was transfected only into right eyes. Transfected retinae were harvested at various stages and sectioned for immunostaining as described (Li et al., 2014).

RNA-seq Analysis
Whole-transcriptome RNA libraries were sequenced using the SOLiD System (Applied Biosystems). Data analysis was performed according to published protocols (Trapnell et al., 2012) with minor modifications. Briefly, colorspace data of raw 50-bp reads were aligned to the mouse genome (mm9 or GRCm37) using Bowtie (Langmead et al., 2009). Gene expression levels were analyzed with Cufflinks, with differentially expressed genes determined by Cuffdiff (Trapnell et al., 2010). Full raw and processed data have been deposited to NCBI with GEO accession number GSE86187. Coverage plots were generated using BEDTools (http://code.google. com/p/bedtools/) with BEDGraph and BigWig files, which were then visualized in the UCSC Genome Browser (http:// genome.ucsc.edu).

Top2b Deficiency Causes Developmental Defects in Photoreceptor Outer Segment Formation
In our previous study, we showed that Top2b cKO (Top2b f/f ; DKK3-Cre) in early retinal progenitor cells leads to defects in differentiation and survival of all retinal cell types Top2b Affects Photoreceptor Differentiation (Li et al., 2014). In this study, we attempt to define the molecular mechanism of Top2b function using the photoreceptor cell as a model. First, retinal samples from the cKO (Top2b -/-) and control (Top2b 1/1 and Top2b 1/2 ) mice at various stages were analyzed by immunostaining with rod marker rhodopsin (Fig. 1A), s-cone marker Opn1sw (Fig.  1B), and m/l-cone marker Opn1mw/lw (Fig. 1C). Rhodopsin expression was detected only in a few cells in the outer nuclear layer (ONL) of both the control and cKO samples starting at E19.5, and became extensive at P7 and P14 (Fig.  1A). At P7 and P14, noticeably thinner ONLs were observed in the cKO retina (Fig. 1A, D), and there was no or little rhodopsin accumulation near the outer limiting membrane (purple arrows in Fig. 1A). By P14, rhodopsin signal was not detected or significantly reduced in the OS with a strong signal in the cell bodies (Fig. 1A, E). In addition, Opn1sw-labeled s-cones showed dramatic degeneration at P0, P7, and P14 (Fig. 1B). Opn1sw 1 cells were observed with shorter processes at P0 (arrowheads in Fig. 1B). No neural process was detected at P7 and P14 (Fig. 1B). The OS was only observed in control samples (purple arrows in Fig.  1B) but not in cKO samples. By P14, the vast majority of Opn1sw 1 s-cones were degenerated in the cKO retinae (Fig. 1B, F). Moreover, examination of the m/l-cones by anti-Opn1mw/lw antibody staining showed that in the cKO opsin was accumulated in small regions, but no OS can be identified (purple arrows in Fig. 1C). A strong signal of Opn1mw/lw was found in cytoplasm (yellow arrows in Fig.  1C). It is also noted that the rhodopsin-and Opn1mw/lwlabeled ONL cells intruded into the inner nuclear layer (INL) in P14 cKO samples (Fig. 1A, C).
To further determine the defects in the OS at a micrometer ultrastructural level, we examined retinal sections using HIM (Zeiss, Germany). In both P14 and P21 control retinae, the OSs of both rod and cone cells were found between the retinal pigment epithelial cell (RPE) layer and the ONL (arrows in Fig. 1G). In contrast, there was no distinguishable OS structure in cKO retinae at these two stages (Fig. 1G). In addition, the retina in cKO samples was largely detached from the RPE.

Aberrant Synapse Formation in Top2b cKO Retinae
Although we reported missing outer plexiform layer (OPL) in our previous study (Li et al., 2014), it was not clear whether it was due to the absence of horizontal cells or interruption of synaptic connections between photoreceptor cells and bipolar cells. The observed phenotypes in the OS lead us to speculate that axonal/dendritic growth and synapse formation of photoreceptor cells may be affected in the cKO mouse retinae. Therefore, retinal sections of the control and cKO animals were stained with antibodies against vGlut1 (a vesicular glutamate transporter) ( Fig. 2A) and synaptophysin (a synaptic vesicle protein) (Fig. 2B). vGlut1 is required for synaptic vesicles with glutamate at the synaptic terminals (Johnson et al., 2003) and is usually found in the presynapses (Sherry et al., 2003). Expression of vGlut1 in control retinae suggested the extension of photoreceptor cell synapse into OPL at P7 and formation of bulb-shaped structures at P14 and P21 (arrowheads in Fig. 2A). In cKO retinae, however, vGlut1 was expanded in the cytoplasm, and no bulb-shaped structure can be found ( Fig. 2A). In addition, synaptophysin is expressed in all vesicular synaptic terminals in both the OPL and inner plexiform layer (IPL) of the retina (Brandstatter et al., 1996;Spiwoks-Becker et al., 2001). Synaptophysin signals were prominent in the synapses within the OPL (dotted lines in Fig. 2B) at P7, P14, and P21. In the cKO, synaptophysin signals at P14 and P21 became more prominent in the cytoplasm of the ONL cells, and the OPL structure was disrupted (Fig. 2B).

Increased Photoreceptor Cell Death in the ONL of Top2b cKO
Defects in axonal/dendritic growth and the formation of synapse and OS usually lead to photoreceptor degeneration (Swaroop et al., 2010). Thus, we analyzed photoreceptor cell death by immunostaining with apoptosis marker-activated caspase 3 (Casp3; Fig. 3A). Compared with the control, a significantly increased number of Casp3 1 cells were observed in the ONL of cKO retinae at P7, P14, and P21 (Fig. 3B).

Cell-autonomous Role of Top2b in Rod OS Formation by Mosaic Cell Analysis
Because the cKO animals showed defects in almost all retina cell types (Li et al., 2014), it is unclear whether the observed defects in photoreceptors were due to cellautonomous or non-cell-autonomous effects. To address this issue, we performed a mosaic analysis by targeted Top2b knockout in postnatal rod photoreceptor cells (rods) by injection and electroporation of plasmid construct CAG-Cre-GFP into the eye in P0 Top2b f/f animals (Lyu and Wang, 2003). Animals were harvested and analyzed at P7. We found that almost all GFP 1 cells (Top2b knockout) were rods and located in the INL, while the majority of DsRed 1 rods (control) were in the ONL (Fig. S1), a phenotype of delayed cell differentiation. In addition, no significant cell death was detected in GFP 1 cells (Fig. S1C). These results support a cell-autonomous role of Top2b in rod cell differentiation. To further confirm this cellautonomous function of Top2b, shRNA-mediated gene silencing analysis was performed. Plasmid DNA shTop2b-GFP was injected and electroporated into P0 retina of the wild-type mice. Transfected retinae were examined at P14 and P21 after P0 electroporation (Fig. 4A). Transfection led to strong reporter GFP expression in rods in the ONL (Fig.  4B). The number of GFP 1 cells was counted, and the percentage of GFP 1 cells in the total number of ONL cells was determined (Fig. 4C). The efficiency of Top2b knockdown was evaluated by immunostaining analysis with anti-Top2b antibody, which showed a reduction or elimination of Top2b expression in the transfected cells (arrows in Fig. 4B). Over 97% of GFP 1 cells at P7 and almost 100% at both P14 and P21 were negative for Top2b staining (Fig. 4C), indicating a successful gene knockdown. Comparing shTop2b knockdown samples with the controls (scrambled shRNA), there was a significant reduction of OS length (Fig. 4F, bars and chart) and disorganized OS (Fig. 4F, white arrows) in transfected GFP 1 cells at P21. In contrast to the dramatic phenotypes observed in cKO animals, there was no significant difference of ONL thickness, ONL cell number (Fig. 4E), or synapses (yellow arrows, Fig. 4F), indicating that Top2b knockdown did not cause a dramatic cell loss. Thus, these results further support a cell-autonomous role of Top2b in rod cell differentiation.

Top2b Deficiency Affects the Expression of Key Genes in a Transcriptional Network for Photoreceptor Differentiation and Maintenance
Top2b is known to regulate gene transcription (Wang, 2002). To determine genes controlled by Top2b, we performed transcriptome analysis using RNA-seq data from control and cKO retinae at P0 and P6. The two time points (i.e., P0 and P6) were selected because 1) P0 is at the peak of rod photoreceptor genesis and cone maturation, 2) P6 is at the middle stage of rod differentiation, and 3) rods comprise the vast majority (94%) of the photoreceptor cell population (Young, 1985a;Brzezinski and Reh, 2015). Thus, sequencing results for these two stages largely represent the genes required for photoreceptor cell differentiation and maturation. Four libraries were constructed: P0 control, P0 cKO, P6 control, and P6 cKO (Table S1). Differentially expressed genes contrasting the control and Top2b cKO samples were determined using Cufflinks and Cuffdiff (Trapnell et al., 2010(Trapnell et al., , 2012. We identified 699 genes at P0 and 180 genes at P6 (Fig. 5A, B; Table S2, S3) that were significantly changed in expression level (jfold changej 1.5 (jlog 2 (fold change)j 0.585), P value 0.05), including Top2b itself (Fig. 6A). These differentially expressed genes are thus referred to as the Top2b Dependent Genes (TDGs). No significant change was detected in the expression of 20 known housekeeping genes (Table S4), confirming that the sequencing and calculation of transcript abundances was of good quality.
To explore the role of Top2b in photoreceptor development, TDGs were analyzed by Ingenuity Pathway Analysis (www.ingenuity.com, Redwood City, CA, USA), a web-based tool for statistical analysis and biological interpretation of gene expression, in parallel with analysis via literature review and functional classification from the Gene Ontology Consortium. Top2b-dependent interactions, pathways, and disease-related gene pools were obtained to predict its regulatory role. Interestingly, the analysis reveals six TDGs (i.e., Crx, Nr2e3, Vsx2, Opn1sw, Glo1, and Smim13) ( Table II) that are critical for rod and cone photoreceptor cell fate determination, differentiation, and maturation Hennig et al., 2008;Swaroop et al., 2010;Cepko, 2015). The altered expression level of Crx and Nr2e3 was shown in coverage plots (Fig. 6B, C). It is likely that Top2b modulates photoreceptor cell differentiation by regulating the transcription of these pivotal genes in the regulatory network (Fig. 6D).

DISCUSSION
Photoreceptors constitute the majority of the retinal cell population, and their development expands from embryonic to neonatal stages (Young, 1985a,b;Rapaport et al., 2004;Brzezinski and Reh, 2015;Wang and Cepko, 2016). In this study, we report a combinatorial cellautonomous and non-cell-autonomous role of Top2b during late-stage photoreceptor differentiation and maturation.

Top2b Functions in the Late-stage Differentiation and Maintenance of Photoreceptor Cells
Dkk3-Cre-mediated cKO resulted in Top2b deletion in retinal progenitor cells and all progeny in all retinal cell types (Li et al., 2014), making it a valuable approach to study Top2b function in the development of all retinal cell lineages including photoreceptors. In Top2b cKOs, both cones and rods were generated but failed to fully differentiate. Deficits exist in i) the OS maturation and maintenance (Fig. 1), ii) accumulated opsin in the cell body ( Fig. 1A-C), iii) synapse formation (Fig. 2), and iv) cell degeneration and cell death in both rods and cones (Fig. 3). The defects in the OS with traditional fluorescence microscopy were confirmed by HIM. HIM provides a new approach to investigate nanometer-scale surface features of biological samples Fig. 4. Targeted Top2b knockdown in photoreceptors reduces OS thickness. DNA mixture containing shTop2b-GFP or scrambled shRNA-GFP constructs was injected and electroporated into subretinal space of P0 mouse to transfect photoreceptor progenitors. A: Schematic shows the timeline of electroporation and sample harvest and analysis at P7, P14, and P21. B: The resulting transfected rod photoreceptors were stained with anti-Top2b antibody. Arrows indicate transfected cells that lack Top2b staining. C: Plots of the number of GFP 1 cells and percentage of GFP 1 cells in the ONL per 100lm region. D: Plot of the percentage of shTop2b-GFP1/Top2bcells. E: Plots of the number of cells and thickness of the ONL in the transfected regions. F: Reduced thickness of the OSL/ISL of shTop2b-transfected retinae at P21. White arrows indicate transfected OS/IS (GFP1). The length of OS was measured as indicated by the vertical bars. Synapses of the transfected cells are similar in the control or shTop2b-transfected samples (yellow arrows). Boxed areas are shown in a higher magnification on the side or in the bottom right corner of each image. OS, outer segment; OSL, outer segment layer; IS, inner segment; ISL, inner segment layer; ONL, outer nuclear layer; INL, inner nuclear layer; GCL, ganglion cell layer. Scale bars 5 50 lm. Data in plots are represented as median 6 max/min (n 3). P values were determined by Student t-test. (Joens et al., 2013). With this technique, detailed differences on surface ultrastructure of photoreceptors between the control and cKO mice were revealed clearly (Fig. 1G). Similar defects were also observed in latestage development in the animals lacking Crx (Furukawa et al., 1999), Nrl (Daniele et al., 2005), or Vsx2 (Rutherford et al., 2004), except that in these animals there was an increase in the number of cone cells. In addition, the lack of OS and the degeneration of photoreceptor cells are usually observed in the opsin-deficient animals (Daniele et al., 2011;Lem et al., 1999). It is believed that mislocalized opsin is frequently associated with retinal degeneration (Deretic, 2006;den Hollander et al., 2008) and has been indicated as a major cause of photoreceptor cell death in the absence of heterotrimeric kinesin-2 function (Louie et al., 2010). The enriched rhodopsin and Opn1sw/mw/lw in photoreceptor cell bodies of cKO retinae (Fig. 1A-C) may explain the dramatic cell loss found in the ONL (Fig. 3). Because cone cells represent 3% to 5% of all photoreceptor cells, the dramatic cell loss observed in the ONL is most likely due to the rod cell phenotype.

Cell-autonomous and Non-cell-autonomous Role of Top2b in Photoreceptor Differentiation
Dkk3-Cre-mediated Top2b cKO provides valuable insight into retinal development. However, it has limitations for addressing cell-autonomous vs. noncell-autonomous and primary vs. secondary effects of Top2b on photoreceptor differentiation. Thus, we performed mosaic analysis using shRNA-mediated Top2b knockdown in postnatal developing photoreceptor cells. Shorter OS was observed in the sample transfected with shTop2b ( Fig. 4F) without dramatic cell loss (Fig. 4D), suggesting that a cell-autonomous function of Top2b is primarily affecting OS maturation and maintenance. Moreover, in contrast to a dramatic synapse degeneration (Fig. 2), cell degeneration, and cell death (Fig. 3) observed in the cKO, Top2b knockdown did not cause a noticeable difference in the thickness of the ONL or morphology of synapses (Fig. 4E, F). Cellautonomous effects alone cannot account for the severe phenotypes observed in the retinae of cKO mice. Therefore, the multilevel defects in photoreceptor differentiation in cKO animals are likely due to both cellautonomous (primary) and non-cell-autonomous (secondary) effects of Top2b. As photoreceptors directly form synaptic connections with the horizontal and bipolar cells, alterations in these two cell types usually lead to defects in photoreceptor cells. A study has shown that the loss of horizontal cells leads to partial rod photoreceptor cell degeneration, retraction of axons from the OPL, and partial cell death (Sonntag et al., 2012). Because defects of the horizontal cells but  not the bipolar cells were observed in Top2b cKO animals (Li et al., 2014), it is likely that horizontal cell degeneration exacerbates the photoreceptor phenotype. However, in the same study (Sonntag et al., 2012), the s-cones remained largely unchanged, which is different from our observation of a dramatic s-cone cell loss (Fig.  1B). It will be intriguing to find out what is responsible for the s-cone dystrophy. Surprisingly, synapse formation was not affected in the mosaic retinae with shTop2b knockdown (Fig.  4F), suggesting that Top2b may not play a cellautonomous role in this process. Top2b is known to be important for axon guidance and neurite growth of motor neurons (Yang et al., 2000). We also found reduced optic nerve and disrupted IPL and OPL in cKO retina (Li et al., 2014). These observations suggest that Top2b might play a different role among various types of neurons.

Top2b Controls Key Genes in the Photoreceptor Gene-regulatory Network
Rods constitute the majority of the cells in the mouse retina (Carter-Dawson and LaVail, 1979;Brzezinski and Reh, 2015;Cepko, 2015). TDGs (e.g., Vsx2, Crx, Nr2e3, Opn1sw, Glo1, Smim13) identified by RNA-seq analysis (Figs. (5 and 6)) are key genes in the regulatory network that dictates photoreceptor differentiation and homeostasis. Thus, Top2b functions as a key modulator at multiple levels in this hierarchical transcriptional regulatory network (Fig. 6E). Among the six photoreceptor-related TDGs, Vsx2 mutation leads to delayed Crx expression and rod development (Rutherford et al., 2004). Crx and Nr2e3 are known to be essential genes in controlling the rod/cone cell fate determination  and development of the OS . These studies are consistent with the observed deficits in photoreceptor cells in Top2b cKO animals, except that  (Gerber et al., 1995;Maugeri et al., 2000;Molday et al., 2000) Note: Genes involved in photoreceptor cell-related retinal diseases are shown in bold.
Top2b Affects Photoreceptor Differentiation mutation of Nr2e3 can lead to excess s-cones . It is possible that Top2b deletion affects Crx and Nr2e3 expression and consequently affects Opn1sw and rhodopsin expression at a later stage during photoreceptor differentiation. Alternatively, Top2b may directly regulate Opn1sw expression by binding to a distant regulatory region 200 kb upstream of the Opn1sw transcription starting site (data from GSM1516578 in Madabhushi et al., 2015). As Opn1sw encodes the blue cone pigment (Nathans et al., 1986), the reduction in Opn1sw expression may also be a contributor to the dramatic loss of the OS and the number of s-cones in the cKO. Several TDGs have been known to function in neurite growth and retinal cell maintenance including Igf1, Creb1, Hras, Mark1, Grin1, Cdkn1a, Serpine2, Ptprz1, Gnao1, Drd2, Pdia3, and Dpysl3. Although neurite degeneration and neural cell death can be separate events (Ikegami and Koike, 2003), our observation favors the idea that the impaired neurite outgrowth accelerates retinal cell death.
Finally, mice lacking Top2b present a transcriptome profile that is prone to several photoreceptor-related retinal diseases such as retinitis pigmentosa, s-cone syndrome, and Bardet-Biedl syndrome (Table III). The identification of retinopathy-related TDGs suggests that Top2b function is required for not only normal development but also the prevention of retinal diseases.

ACKNOWLEDGMENT
We thank Dr. Robert S. Molday for a monoclonal antibody to rhosopsin 4D2, Dr. Gabriella D'Arcangelo for an antibody to synaptophysin, and Drs. Torgny Gustafsson and Slava Manichev for HIM imaging. We also thank the members in Cai lab for proofreading the manuscript and providing helpful suggestions.