The effect of Rbfox2 modulation on retinal transcriptome and visual function

Rbfox proteins regulate alternative splicing, mRNA stability and translation. These proteins are involved in neurogenesis and have been associated with various neurological conditions. Here, we analyzed Rbfox2 expression in adult and developing mouse retinas and the effect of its downregulation on visual function and retinal transcriptome. In adult rodents, Rbfox2 is expressed in all retinal ganglion cell (RGC) subtypes, horizontal cells, as well as GABAergic amacrine cells (ACs). Among GABAergic AC subtypes, Rbfox2 was colocalized with cholinergic starburst ACs, NPY (neuropeptide Y)- and EBF1 (early B-cell factor 1)-positive ACs. In differentiating retinal cells, Rbfox2 expression was observed as early as E12 and, unlike Rbfox1, which changes its subcellular localization from cytoplasmic to predominantly nuclear at around P0, Rbfox2 remains nuclear throughout retinal development. Rbfox2 knockout in adult animals had no detectable effect on retinal gross morphology. However, the visual cliff test revealed a significant abnormality in the depth perception of Rbfox2-deficient animals. Gene set enrichment analysis identified genes regulating the RNA metabolic process as a top enriched class of genes in Rbfox2-deficient retinas. Pathway analysis of the top 100 differentially expressed genes has identified Rbfox2-regulated genes associated with circadian rhythm and entrainment, glutamatergic/cholinergic/dopaminergic synaptic function, calcium and PI3K-AKT signaling.

www.nature.com/scientificreports/ and convey this information to postsynaptic targets in the brain. Transcriptome and functional classification of mouse RGCs identified 40 and at least 32 groups of RGCs, respectively 8,9 . The dendritic arbors of each RGC subtype tile the entire retinal surface so specific features of the image can be detected and communicated to the brain 10 . Dysfunction and degeneration of RGCs and their axons in the optic nerve lead to loss of vision in various optic neuropathies. The most common form of optic neuropathies, glaucoma, affects more than 70 million people worldwide and is the leading cause of irreversible blindness 11 . Investigating Rbfox proteins in the retina is of particular interest to us since they are known to regulate extensive genetic networks in developing and mature neurons. Mutations and deficiencies in Rbfox gene expression have been associated with several neurological conditions including autism spectrum disorder (ASD), mental retardation, epilepsy, ADHD, bipolar disorder, schizoaffective disorder and schizophrenia [12][13][14][15] . Recently, we reported our findings on the expression pattern of Rbfox1 in the retina, evaluated changes in the retinal transcriptome and visual function in response to Rbfox1 downregulation 16 . The current work is a continuation of our study on the role of Rbfox proteins in the retina and presents data on the comprehensive analysis of Rbfox2 expression in differentiating and adult mouse retinas, as well as the effect of Rbfox2 downregulation on the retinal transcriptome and on visual function such as pupillary light reflex and depth perception.

Results
Expression of Rbfox2 in adult and developing retinas. Rbfox2 is expressed in RGCs, ACs and horizontal cells (HCs) of adult mouse retinas. The spatial expression pattern of Rbfox2 in adult mouse retinas was analyzed by immunohistochemistry with anti-Rbfox2 antibodies. Rbfox2 immunostaining was restricted to cells located in the ganglion cell layer (GCL) and inner nuclear layer (INL) of the retina (Fig. 1). Double immunostaining with Rbfox2 and RGC marker Rbpms showed that virtually all Rbpms-labeled RGCs are also Rbfox2 + ( Fig. 1A; pointed with yellow arrows). In addition to Rbfox2 + RGCs, there were Rbfox2-stained cells in the GCL that were not labeled with anti-Rbpms antibodies suggesting that Rbfox2 is also expressed in displaced ACs (dACs; Fig. 1A; pointed with green arrows). The GCL in rodent retinas contains two types of neurons, RGCs and displaced ACs (dACs), in a ratio of approximately 1:1: ~ 45% RGCs and ~ 55% dACs are present in mouse retinas; in rat retinas, these cells are in ~ 50:50 ratio; and in hamster retinas, there are 56% RGCs and more than 40% dACs [17][18][19][20] . In the INL, which contains cell bodies of horizontal cells (HCs), bipolar cells and ACs, as well as Muller glial cells, Rbfox2 + cells were mostly localized to two rows of cells proximal to the inner plexiform layer (IPL). ACs of the mouse retina form two to four rows of cells at the inner margin of the INL 17,21 , suggesting that Rbfox2 + cells in the INL are ACs. Indeed, these cells were colocalized with AP-2α, an AC marker 22 ; virtually all AP-2α + cells were also stained for Rbfox2 ( Fig. 1B; pointed with yellow arrows). However, there were very few Rbfox2 + cells that were AP-2α − ( Fig. 1B; pointed with green arrows). Furthermore, Rbfox2 was also expressed in sparsely distributed cells in the INL that were adjacent to the outer plexiform layer (OPL). These cells were colocalized with calbindin + HCs ( Fig. 1C; pointed with yellow arrows). Calbindin is commonly used as a marker for both ACs and HCs in the mouse retina. The anti-calbindin antibodies generated against calbindin C-terminal peptide that we used in this immunostaining appears to label HCs much stronger than ACs 23,24 . Double staining with antibodies against Rbfox2 and Rbfox1 was performed to evaluate the extent of potential overlap in expression of Rbfox2 and Rbfox1 in the retina ( Fig. 1D; pointed with green arrows). Many cells in the INL and GCL express both isoforms, however significant differences are evident: Rbfox2 but not Rbfox1 is expressed in HCs and there is a wider distribution of Rbfox2 + cells particularly in the INL. There are also cells in the GCL that express only Rbfox2 and, since all RGCs express Rbfox1 16 , these Rbfox2 + cells must be subtypes of dACs ( Fig. 1D; pointed with green arrows). It should be noted that typically all GCL cells are Rbfox2 + , however, in some retinal sections we occasionally see cells in the GCL that are Rbfox2 − ( Fig. 1D; pointed with blue arrows). Since retinal sections represent a small portion of the tissue and may not accurately represent the entire retina, whole mounted retinas were used for quantitative analysis of Rbfox2 + cells. Rbfox2 + /Rbpms + and Rbfox1 + /Rb-fox2 + cells were counted in the superior, inferior, nasal and temporal quadrants of the retinas 1 mm from the center of the optic nerve (Fig. 2). Almost half of Rbfox2 + cells (868.67 ± 25.08 cells/mm 2 ) were colocalized with Rbpms (4200 ± 24.99 cells/mm 2 ). Virtually 100% of RGCs (Rbpms + cells) were Rbfox2 + ( Fig. 2A). Colocalization of Rbfox1 + and Rbfox2 + cells showed extensive overlap in their expression in the GCL but approximately 6-7% of cells express only Rbfox2 (Fig. 2B).
Rbfox2 is expressed in GABAergic ACs. ACs are inhibitory interneurons, which represented in the mammalian retina by morphologically and functionally diverse cell types. Based on the type of neurotransmitter that ACs synthesize and use for their function, they can be classified into two groups: wide-field ACs that express gammaaminobutyric acid (GABA) and narrow-field ACs that express glycine 25,26 . GABA and glycine considered as the primary neurotransmitters of ACs. GABAergic and glycinergic ACs can be further subdivided into subgroups that synthesize other neuromodulators including acetylcholine, dopamine, somatostatin, substance P, neuropeptide Y (NPY), serotonin vasoactive intestinal polypeptide, and endocannabinoids. We used several markers of ACs, including GABA, choline acetyltransferase (ChAT), NPY, early B-cell factor 1 (EBF1), glycine transporter (GlyT1) and vesicular glutamate transporter 3 (vGlut3) to define the subtype identity of Rbfox2 + ACs. The vast majority of GABAergic ACs in the INL were Rbfox2 + ( Fig. 3A; yellow arrows point at GABAergic ACs that express Rbfox2; green arrows point at Rbfox2-negative ACs). It appears that almost all Rbfox2 + cells in the INL were GABAergic, but some GABAergic ACs were Rbfox2 − ( Fig. 3A; GABA + /Rbfox2 − cells are pointed by green arrows). In the GCL, all GABAergic ACs were also Rbfox2 + . A subtype of GABAergic ACs, cholinergic (ChAT + ) starburst ACs (SACs), were all Rbfox2 + in both INL and GCL (Fig. 3B). Extensive overlap of Rbfox2 with NPY and EBF1 expression was also observed, although there were very few Rbfox2 + cells that had very faint or no staining for NPY or EBF1 (Fig. 3C,D) www.nature.com/scientificreports/ (Fig. 3E). vGlut3 + /Rbfox2 + cells were present in abundance in the GCL. However, these cells were not ACs: they were Rbpms + RGCs (Fig. 3F).
Rbfox2 expression during retina development. Proteins of Rbfox family have been associated with neurogenesis and synaptogenesis 27,28 and, therefore, we evaluated expression of these proteins during retinal development. In mouse retinas, the first cell types to differentiate from multipotent progenitor cells are RGCs and HCs, followed in overlapping phases by cone photoreceptors, ACs, rod photoreceptors, bipolar cells and, finally, Müller glia cells 29 . Expression of Rbfox2 was present starting on E12, the earliest stage that we used in this study (Fig. 4A). No significant difference in Rbfox2 expression pattern between E12 and E15 was observed, except a noticeable increase in its staining intensity level at the later stage ( Fig. 4B). At P0, the IPL, which separates GCL from the rest of the retina can be seen (Fig. 4C). At this stage, Rbfox2 staining was present in the GCL as well as in several rows of cells above the IPL. Furthermore, faintly stained Rbfox2 + cells that were most likely HCs were present. At  www.nature.com/scientificreports/ P5, a narrow OPL that separates ONL from INL was apparent (Fig. 4E). Rbfox2 staining patterns at P5, P10, P15, P21 and in the mature retina were very similar: its expression was restricted to dACs and RGCs within the GCL and to HCs, and to first three rows of ACs adjacent to the IPL within the INL (Fig. 4E). Colocalization of Rbfox2 with Rbpms-labeled RGCs and ChAT + SACs are shown in Fig. 4E-H. The earliest stage that Rbpms staining can be reliably observed is at P0 (Fig. 4C). Rbpms staining in P5 and older retinas becomes more prominent and defined. With respect to SACs, which were reported to be the first AC subtype to differentiate from retinal progenitor cells (50% of ChAT + ACs are born between E8 and E14) 21 , we were able to identify the first ChAT + cells with confidence at P5 (Fig. 4D,F). Both Rbpms + and ChAT + cells were also Rbfox2 + at all stages of retinal differentiation.
Histological, behavioral and functional assessment of Rbfox2 −/− animals. Rbfox2 downregulation has no recognizable effect on retinal architecture. Rbfox2 −/− animals were generated by crossing Rbfox2 fl/fl homozygous transgenic mice with Tg(UBC-cre/ERT2)1Ejb mice. The Tg(UBC-cre/ERT2)1Ejb mouse was used in these experiments since it shows robust Cre-mediated reporter expression in RGCs, in dACs, in ACs adjacent to the IPL and in some sparsely distributed cells in the INL adjacent to the OPL, most probably HCs (http://www. infor matic s.jax.org/recom binas e/speci ficit y?id=MGI:37073 33&syste m=senso ry+organ s). Tamoxifen-induced Rbfox2 downregulation and its effect on retinal morphology and architecture was evaluated in adult animals four and nine months after the last administration of the drug. A dramatic reduction in Rbfox2 expression in the GCL and INL was observed (Fig. 5A,B). Very few Rbfox2 + cells in the GCL were detected, indicating that the downregulation of Rbfox2 effectively takes place in both RGCs and dACs. On the representative image shown in Fig. 5A, only six Rbfox2 + cells were present in the GCL, three of which were colocalized with Rbpms. Robust, but relatively modest compared to GCL, downregulation of Rbfox2 was observed in ACs located in the INL. Furthermore, expression of Rbfox2 was completely abolished in HCs (Fig. 5B). Despite this significant downregulation of Rbfox2, no aberrations in overall retinal gross morphology or cellular architecture were detected.
Pupillary light reflex (PLR). PLR test which measures the change in pupil diameter in response to change in light intensity was performed to evaluate the integrity of the non-image forming visual pathway mediated by M1 subtype of intrinsically photosensitive RGCs 30 . Light-dependent pupillary constriction appeared to be normal in both the control and Rbfox2 −/− mice ( Depth perception. A visual cliff test that relies on the innate tendency of animals to avoid the deep side of the field 31 was used to evaluate visual depth perception in Rbfox2-deficient animals (Fig. 7A). Control mice, as expected, had a clear preference for the shallow side of the box and tried to avoid the deep side of the field. Animals with Rbfox2 downregulation were less selective in preferring the shallow side and spent more time on   (Table 1). Highly enriched gene sets in Rbfox2-deficient retinas also included genes associated with microtubule organization, DNA repair and mitochondrial gene expression ( Table 1). The major gene sets/processes that were underrepresented in Rbfox2 −/− retinas include the extracellular matrix, ion transport, neuron differentiation and synapse organization ( Table 2). DEGs associated with RNA metabolism are listed in Tables S3 and S4. Expression levels of several DEGs, including bestrophin 2 (Best2), glutathione S-transferase alpha3 (Gsta3), solute carrier family 13 member 4 (Slc13a4), thioredoxin interacting protein (Txnip), calmodulin like 3 (Calml3), cytochrome P450 monooxygenase (Cyp4a12b), olfactory receptor 12 (Olfr12) and tumor-associated calcium signal transducer 2 (Tacstd2) were arbitrarily chosen for further analysis with real-time PCR to validate RNA-seq data (Fig. 8A). Quantitative PCR data were in agreement with the RNA-seq results. Heatmap of top 100 DEGs (29 downregulated and 71 upregulated genes) identified with EdgeR 32 (p < 0.01) is shown in Fig. 8B. Major pathways identified with the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analyses 33,34 of the top 100 DEGs include circadian rhythm, circadian entrainment, dopaminergic synapse, glutamatergic/cholinergic synapse, calcium signaling, PI3K-AKT signaling and protein processing in ER (Table 3).

Discussion
Here we present data on the expression of Rbfox2 in adult and differentiating mouse retinal cells and the effects of Rbfox2 deficiency on visual function and the retinal transcriptome. This is a continuation of our efforts to characterize the expression and function of Rbfox proteins in the retina; results of a similar work on Rbfox1 were published earlier 16 . We first identified Rbfox1 and Rbfox2 in the retina after analyzing gene expression profiles of RGCs 7 . RGCs collect, process and send both image-forming and non-image forming visual information from the retina to target regions in the brain via their axons in the optic nerve. Here we show that in both developing and mature retinas, Rbfox2 expression is restricted to RGCs, HCs and ACs. ACs and HCs are retinal interneurons involved in forming visual signals by feedback and feedforward inhibition. HCs make synaptic connections with photoreceptors and bipolar cells in the OPL and their activity is essential for establishing center-surround receptive-field properties in the visual pathway. ACs form synaptic connections with bipolar cells and RGCs within IPL and by modulating their activity, they shape their receptive fields' spatial and temporal characteristics 35  Interestingly, a large number of vGlut3 + /Rbfox2 + cells were observed in the GCL. vGlut proteins, which promote the uptake of glutamate into synaptic vesicles to support synaptic transmission, are represented in the retina by three members: vGLUT1 is localized to photoreceptor and bipolar cell terminals, vGLUT2 is restricted to HCs and RGCs and vGlut3 is expressed in a sparse population of glycinergic ACs. In mouse and rat retinas vGlut3 immunolabeled cells were present in the INL 36,37 . However, in human retinas vGlut3 was detected in the GCL  The expression pattern of Rbfox2 in the retina shares similarities with that of Rbfox1: both proteins are present in all RGCs and almost all Rbfox1 + and Rbfox2 + ACs are GABAergic ACs. However, there are notable differences: (a) Rbfox2 but not Rbfox1 is expressed in HCs; (b) Rbfox1 expression in the INL appears to be restricted to ACs that are located in the innermost layer of the INL, whereas Rbfox2-expressing cells are present in several rows occupied by ACs; and (c) Rbfox2 is expressed in a wider population of dACs in the GCL than Rbfox1. Quantitative analysis of Rbfox2 + cells in whole-mounted retinas supports the data observed in cross section: virtually 100% of RGCs are positive for Rbfox1 and, although most dACs are positive for both Rbfox2 and Rbfox1, approximately 6-7% of dACs express Rbfox2 only.
We evaluated Rbfox2 expression during retinal development at E12, E15, P0, P5, P10, P12, P14, P15 and P21. These stages were chosen based on the timeline of retinal cell differentiation and establishment of synaptic connections between them 38 . Expression of Rbfox2 can be seen in differentiating retinal cells at E12. It has been reported that GCL cells born before E11.8 and E12.8 were 98% and 99% RGCs, respectively, and those born after E15.8 were 97% dACs 39 . This suggests that the Rbfox2 + cells that we observed at E12 are almost exclusively RGCs. www.nature.com/scientificreports/ By P0, although the retina was not fully developed yet, the expression pattern of Rbfox2 resembles that of the mature retina. Subcellularly, Rbfox2 appears to be localized predominately to the nucleus irrespective of phases of retinal development. This is different from Rbfox1 subcellular localization, which shifts from cytoplasmic at E12 to predominantly nuclear at P0 and thereafter 16 . The transition in Rbfox1 subcellular distribution coincides with the stage II spontaneous retinal waves of excitation 40 and can be explained by increased expression of the nuclear Rbfox1 isoform necessary for regulation of the alternative splicing of target genes at this stage of retinal development.
The role of Rbfox2 on retinal morphology and visual function was evaluated in Rbfox2 −/− animals. Downregulation of Rbfox2 in adult animals had no effect on gross retinal architecture or retinal cell morphology. The PLR and depth perception were analyzed in Rbfox2 −/− animals for potential abnormalities in visual function. The PLR helps adapt vision to different levels of light intensity by regulating the pupil diameter. Retinal cells that mediate PLR are intrinsically photosensitive retinal ganglion cells (ipRGCs), the melanopsin-expressing cells that comprise ~ 0.2% of the total number of RGCs 41 . The PLR test of Rbfox2-deficent mice showed no detectable abnormalities, suggesting that Rbfox2 downregulation has no significant effect on the function of ipRGCs. The functional integrity of the retino-geniculo-cortical pathway in Rbfox2 −/− animals was evaluated by analyzing their depth perception. Unlike control animals, Rbfox2 −/− mice exhibit no clear preference for the shallow side and as a group spent almost 3 times more time on the deep side compared to controls. It is noteworthy that  www.nature.com/scientificreports/ www.nature.com/scientificreports/ depth perception impairment was also observed in Rbfox1 −/− animals, suggesting that the molecular mechanisms underlying this phenomenon caused by downregulation of the Rbfox1 and Rbfox2 genes are similar. A significant overlap in Rbfox1 and Rbfox2 expression in the retina and the fact that these RNA binding proteins recognize the same (U)GCAUG element within their target genes, suggest that they may be involved in regulation of same set of genes. This appeared redundancy may explain why a deletion of a single gene, Rbfox1 or Rbfox2, has no dramatic effect on overall retinal morphology. The notion that Rbfox1 and Rbfox2 regulate a similar set of genes is also supported by our observation that animals deficient in either Rbfox1 or Rbfox2 have depth perception impairment. Interestingly, the deletion of the third member of Rbox family, Rbfox3, which is normally expressed in most types of RGCs, some types of ACs and HCs, resulted in the reduction of IPL thickness, but the number of RGCs and ACs was normal. Rbfox3 −/− mice had normal PLR and optomotor response, leading authors to conclude that RBFOX3 is dispensable for visual function 42 . Although, these observations imply that Rbfox proteins are redundant and may substitute each other, the differences that we observed in their expression pattern suggest that each of these proteins may have distinct functions during retinal development, as well as in mature retinas.
To identify Rbfox2-regulated genes, we analyzed Rbfox2 −/− and control retinal transcriptome. KEGG pathway enrichment analysis identified Rbfox2-regulated genes involved in circadian rhythm, circadian entrainment, dopaminergic synapse, glutamatergic/cholinergic synapse, calcium signaling, PI3K-AKT signaling and protein processing in ER. Regulation of genes involved in synaptic function was expected since Rbfox proteins have been implicated in neuronal differentiation and synaptic transmission in particular 6,27,43 . However, to our knowledge, the coordinated regulation of several core circadian rhythm proteins, including aryl hydrocarbon receptor nuclear translocator-like (Arntl/Bmal1), basic helix-loop-helix family, member e41 (Bhlhe41/Dec2), cryptochrome 2 www.nature.com/scientificreports/ (Cry2), period circadian clock proteins 1 and 2 (Per1 and Per2), by Rbfox2 is observed for the first time. Expression of these core clock components together with CLOCK play a critical role in cellular rhythm generation. The mechanisms for circadian clock regulation involve transcription-translation feedback loops that self-regenerate with a ~ 24-h rhythm. In mammalian cells, the CLOCK and BMAL1 transcription factors control the expression www.nature.com/scientificreports/ of Per1/2 and Cry1/2, which in turn inhibit CLOCK and BMAL1 44,45 . These four factors regulate the expression of thousands of clock-controlled genes that coordinate the oscillation of cellular metabolism and physiology. Clock genes are linked directly to metabolic syndromes; polymorphisms circadian clock genes CLOCK, BMAL1, Per1/2 and Cry1/2 have been associated with metabolic disorders, including neuropsychiatric or neurodegenerative diseases 46 . In Rbfox2 −/− retinas, Arntl/Bmal1 is downregulated, whereas Cry2 and Per1/2 are upregulated. The expression of Clock, Bmal1, Cry2 and Per1 in mouse retinas have been shown to be predominantly localized to RGCs, ACs and HCs 47 . These cells, as we described in this study, also express Rbfox2, further collaborating the results of RNA-seq data analysis on possible regulation of core circadian clock genes by Rbfox2.
In summary, Rbfox2 expression in mouse retinas is restricted to cells in the GCL and INL. Virtually all types of RGCs and dACs in the GCL were positive for Rbfox2. In the INL, HCs and 2-3 rows of cells proximal to IPL were also Rbfox2 + . Among AC subtypes, Rbfox2 is expressed in GABAergic cells, including NPY + , EBF1 + , Vglut3 + and ChAT + ACs. There is a significant overlap between Rbfox1 and Rbfox2 expression among RGCs and ACs. However, Rbfox2 is more widely distributed among AC subtypes compared to Rbfox1. In the developing retina, Rbfox2 is present as early as E12 and, unlike Rbfox1, it was localized in the nucleus throughout the retinal differentiation. Downregulation of Rbfox2 in adult animals had no effect on retinal architecture, however depth perception in these animals had significant deficiency. Transcriptome analysis identified Rbfox2-regulated genes associated with circadian rhythm/circadian entrainment pathways and dopaminergic/glutamatergic/cholinergic synapse function. for five consecutive days. Tamoxifen or corn oil (vehicle) was administered by oral gavage. To prevent tamoxifen cross-contamination, tamoxifen-treated mice were housed individually 49 . Experimental animals were viable, had a normal growth rate and behaviorally had no apparent anomalies compared to control animals. Age-matched heterozygous Rbfox2 fl/+ mice were used as controls.

Methods
Immunohistochemistry. Immunohistochemistry was performed according to a standard procedure 50 .
Quantification of Rbfox2 + cells in whole-mounted retinas. Cell quantification was performed as described earlier 16,50 . The retinas were dissected from enucleated eyeballs, fixed in 4% paraformaldehyde in 0.1 M phosphate buffer, incubated with 10% serum for 1 h to block nonspecific staining, and then with primary antibodies overnight at 4 °C. After incubation with secondary antibodies overnight at 4 °C, the retinas were mounted flat with several radial cuts on a glass slides with the GCL facing upward. The retinas were divided PLR video recordings and analysis. PLR was performed during the light phase of the light-dark cycle.
Animals were dark-adapted for 1 h before the experiment. For PLR recording mice were restrained with a head post similar to that described by Cahill and Nathans 51 . Animals were exposed to a 30-s stimulus of a NIR-100 light source. Video recordings were obtained by six infrared digital cameras (three per eye). Screen shots of the video were generated every 5 s with VLC Media Player for 1 min video duration. The pupil area was measured with ImageJ software (NIH) and normalized to baseline. PLR for each animal was recorded and analyzed in three independent experiments.

Visual cliff test.
A visual cliff test was used to evaluate depth perception in Rbfox2 −/− animals. The test was performed as described earlier 16,31 . Briefly, a box with a transparent bottom was placed on the edge of the table ("shallow" side) so half of it was suspended above the floor ("deep side"). The illusion of the cliff was created by covering the table and the floor with 1-in. squared black checker linoleum. Both sides of the setup were equally illuminated. The time the animal spent on the shallow and deep sides during 5 min of testing was recorded. Each mouse was tested five times. To eliminate tactile placing responses that may interfere with testing of visual function, the vibrissae of the animals were removed. The glass and central platform were thoroughly cleaned after each test. Six control and ten Rbfox2 −/− animals were used in this experiment. Data are presented as the mean ± SD.
RNA sequencing (RNA-seq) analysis of retinal transcriptome. Retinas from Rbfox2 −/− (n = 4) and control (n = 4) animals were dissected at approximately 10 am. Total retinal RNA was isolated with an RNeasy mini kit (Qiagen, Germantown, MD). Libraries for RNA-seq were prepared with Nugen Universal Plus mRNAseq library preparation kit (Tecan Genomics, Redwood City, CA). The workflow consisted of mRNA enrichment with poly(A) selection, RNA fragmentation, cDNA generation and end repair, and followed by adaptor ligation and PCR amplification. Sequencing was performed on a NovaSeq 6000 (Illumina, San Diego, CA) for 2 × 150 run. Data quality checks were done on an Illumina SAV. Demultiplexing was performed with the Illumina Bcl2fastq2 v 2.17 program. Reads were aligned to the latest mouse mm10 reference genome with the Spliced Transcripts Alignment to a Reference (STAR) software. Total counts of read-fragments aligned to known gene regions within the mouse mm10 refSeq reference annotation were used as the basis for the quantification of gene expression. Fragment counts were derived using the HTS-seq program with the mm10 Ensembl transcripts as a model. Various QC analyses were conducted to assess the quality of the data and to identify potential outliers. Quantitative real-time PCR. Real time PCR was performed according to a standard protocol 16 . Mouse retinal total RNA was extracted with RNeasy mini kit (Qiagen) according to the manufacturer's protocol. Firststrand cDNA synthesized with SuperScript III First-Strand Synthesis System (Thermo Fisher Scientific), primers (Table 4) and the SYBR Green PCR Master Mix (Applied Biosystems/Life Technologies) were used for real-time PCR. The LightCycler 480 II (Roche Applied Science, Mannheim, Germany) was used for cycling and real-time quantitative detection of PCR products. Expression levels of target genes were normalized to the threshold cycle Table 4. Primers used for quantitative real-time PCR.

Gene Forward Reverse
Best2 5′-ACC CCT ACG GAC ACC TAC CTA AT  5′-TGA GGG AAG CAG CAA CAG CTG   Calml3  5′-CAT GAG GCT TCT GGG CTA CAA  5′-GCC AGA ATA AGA TGC AGA AAG TGA   Cyp4a12b  5′-TCA CAG AAC AAG ACA GCT CAA ATG 5′-GGG TGG AGG TGA ACA AGG TGA   Gsta3  5′-TCA GGC AAC TAT AAG TAC ATA GC  5′-GAA TTG ACA CAG ACG CCT TAG A   Olfr12  5′-GGA ATG CTT CTG TGT AGA TCA CTG 5′-GGT AAC TGA GTC CCA CAC TTT CG   Slc13a4  5′-AGC GGA CAG TTA CCA AGC ACA G  5′-GTA GGC TTT CTT GAG AGT TGG  Statistical analysis. Statistical analysis was performed as described earlier 16 . For cell quantification, an unpaired Student's t-test was used. P < 0.05 was considered statistically significant. For the behavior test, a repeated measures ANOVA model was used to analyze the mean difference between the two groups in time spent on the deep side, after controlling for repeated measurements within each animal.