Dose-dependent regulation of horizontal cell fate by Onecut family of transcription factors

Genome duplication leads to an emergence of gene paralogs that are essentially free to undergo the process of neofunctionalization, subfunctionalization or degeneration (gene loss). Onecut1 (Oc1) and Onecut2 (Oc2) transcription factors, encoded by paralogous genes in mammals, are expressed in precursors of horizontal cells (HCs), retinal ganglion cells and cone photoreceptors. Previous studies have shown that ablation of either Oc1 or Oc2 gene in the mouse retina results in a decreased number of HCs, while simultaneous deletion of Oc1 and Oc2 leads to a complete loss of HCs. Here we study the genetic redundancy between Oc1 and Oc2 paralogs and focus on how the dose of Onecut transcription factors influences abundance of individual retinal cell types and overall retina physiology. Our data show that reducing the number of functional Oc alleles in the developing retina leads to a gradual decrease in the number of HCs, progressive thinning of the outer plexiform layer and diminished electrophysiology responses. Taken together, these observations indicate that in the context of HC population, the alleles of Oc1/Oc2 paralogous genes are mutually interchangeable, function additively to support proper retinal function and their molecular evolution does not follow one of the typical routes after gene duplication.


Introduction
The retina represents a neural tissue in the eye and transmits information about the light stimulus to target locations in the brain [1]. Neuroretina is composed of six types of neurons: rod and cone photoreceptors, horizontal cells (HCs), bipolar cells (BCs), amacrine cells (ACs), retinal ganglion cells (RGCs); and of one glial cell type: Müller cells (MGCs) [2][3][4]. Müller glia cells maintain the retinal homeostasis and define the retinal boundaries and polarity. Therefore, loss of MGCs would cause defects in retinal lamination and even result in retinal degeneration [5]. In the mature retina, retinal cell are arranged in three layers: outer nuclear layer (ONL) containing rod and cone photoreceptors, inner nuclear layer (INL) composed of HCs, BCs, MGCs, and ACs and ganglion cell layer (GCL) consisting of RGCs and displaced ACs [2][3][4]. Synaptic connections are organized into two separated layers: thin outer plexiform layer (OPL) and inner plexiform layer (IPL). OPL is composed of HC and BC projections and a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 analysis of all possible Oc1/Oc2 genotypes indicate that the abundance of HCs is highly dependent on the dose of Oc proteins. Other retinal cell types were significantly affected only in mice with simultaneous deletion of both Oc transcription factors. The only exception were the cones, which were also reduced in Oc1 -/and Oc1 -/-; Oc2 +/genotypes. The retinal intensityresponse functions measured by flash electroretinogram recording were also increasingly impaired in mice with the reduced Oc gene dose. Taken together, this study suggests that HCs are exquisitely sensitive to the level of Onecut transcription factors that are required for proper retinal function.

Experimental animals
Previously described mouse strains: BAC transgenic mRx-Cre [36], Oc1 fl/fl [37] and Oc2 +/- [23] were used in this study. Oc1 was specifically deleted in the retina by mRx-Cre, Oc2 was deleted constitutively. Housing of mice and in vivo experiments were performed in compliance with the European Communities Council Directive of 24 November 1986 (86/609/EEC) and national and institutional guidelines. Animal care and experimental procedures were approved by the Animal Care Committee of the Institute of Molecular Genetics (no. 71/2014). Mice were euthanized by cervical dislocation. This work did not include human subjects.

Hematoxylin-Eosin (H&E) staining
Whole eyes from mice at stage P18 were removed, fixed in 4% paraformaldehyde (w/v) at 4˚C overnight, dehydrated and embedded in paraffin. Eyes were sectioned at 5 μm by fully motorized rotary Leica RM2255 microtome. Sections were dewaxed, stained with H&E and mounted in 80% glycerol. Images were obtained by light microscopy imaging using a Nikon Diaphot 300 inverted microscope and processed by Fiji-ImageJ and FastStone Image Viewer.

Cell counting
Fluorescence images were acquired with a Leica SP5 confocal microscope using 40x objective, zoom factor 2 for calbindin, Lim1/2, Ap2α, Sox9, Chx10, Pax6 and Tbr2 and with 63x objective for cone arrestin. Images were then processed by Fiji-ImageJ and Adobe Photoshop. DAPI and marker-positive retinal cells were counted in whole acquired single eye fields of the central retinal sections, obj. 40x 193,75 x 193,75 μm, obj. 63x 246,03 x 246,03 μm. Only horizontal cells were counted in whole retinal sections. For each mouse minimum of 4 sections were counted. Number of P18 mice that were analyzed is specified in graphs separately for each staining. Statistical significance was determined by Welch´s ANOVA test in GraphPad Prism. If differences between the means were statistically significant (p<0.05), Dunnett´s T3 multiple comparison test was performed. In cases where less than 3 mice were counted, statistical significance was not performed for these samples.

Failure of OPL development in Oc-deficient retinas
To assess the phenotypic consequences of Oc1 and Oc2 deficiency, we investigated retinas at the histological level ( Fig 1A). Retinal sections were acquired from mice at stage P18, when retinogenesis is complete. All retina-specific cell types are developed at this stage and mice have open eyes as eye opening occurs at P13-15 [40,41]. Although all retinas displayed proper lamination, OPL localized between ONL and INL was reduced ( Fig 1A). In Oc1 +/-; Oc2 +/and Oc2 -/mice, OPL was slightly reduced and in Oc1 -/mice was absent in most parts. In Oc1 -/-; Oc2 -/-, Oc1 -/-; Oc2 +/and Oc1 +/-; Oc2 -/animals, OPL was not even distinguishable, ONL fused with INL and formed one layer. Moreover, Oc1 -/-; Oc2 -/retinas had thinned INL, lower GCL density and slightly thinned ONL compared to the wild-type retina. IPL in Oc1 -/-; Oc2 +/was also partially reduced. Previous studies have shown that an Oc1 -/-; Oc2 -/retina has a decreased number of retinal cell types that are localized in the INL [11,31] and thinned optic nerve compared to wild-type mouse [31].
These findings indicate that Oc proteins deficiency results in reduced OPL in the mouse retina. The level of reduction depends on the Oc1/Oc2 genotype and thus on the level of the Oc proteins. The reduced OPL in Oc1 -/and Oc2 -/retinas and collapse of OPL in the Oc1 -/-; Oc2 -/retina are consistent with previous findings [7,11,31]. OPL is formed by HC and BC projections and photoreceptor terminals [7]. The reduction or even absence of this layer may indicate abolished HC and/or cone population, so we further examined these cell types by immunohistochemistry.

HCs are sensitive to the dose of Oc proteins
To provide further insight into the phenotype of Oc-deficient retinas, we used immunohistochemistry to quantify the number of HCs that are localized between the ONL and INL. P18 retinal sections were stained with antibodies against calbindin and Lim1/2 ( Fig 1B). The number of HCs was dramatically reduced in all Oc mutants ( Fig 1C). HCs positive for calbindin staining decreased to 26% in double heterozygotes and were lost in mutants with one functional Oc allele, Oc1 -/-; Oc2 +/and Oc1 +/-; Oc2 -/-. The number of Lim1/2 + cells, which is another marker of HCs, decreased to 31% in double heterozygotes Oc1 +/-; Oc2 +/and Lim1/2 + cell were not detectable in mutants with only one functional Oc1/2 allele. Dramatically reduced number of HCs in Oc1 -/and Oc2 -/mutants and loss of HCs in Oc1 -/-; Oc2 -/mouse retinas are consistent with current studies [7,11,31,32].
Our data indicate that HCs are sensitive to the dose of Oc1 and Oc2 proteins. Half of the functional Oc1/Oc2 alleles led to more than 50% decrease of HCs. The Oc1/Oc2 double heterozygote was not able to compensate for the missing alleles and HCs were rapidly reduced according to the overall dose of Oc proteins. These findings implicate that a markedly reduced dose, or even absence, of Oc proteins leads to the loss of HCs in the mouse retina and that HCs are dependent on the dose of Oc proteins.

Examination of other retinal cell types localized in the INL in Oc-deficient retinas
Furthermore, we wanted to examine the consequences of Oc protein deficiency in other retinal cell types. We used immunohistochemistry to analyze individual retinal cell types localized in the INL (Fig 2A), in addition to HCs that were analyzed in Fig 1. The following markers were used: Ap2α for amacrine cells, Sox9 for Müller glia cells, Chx10 for bipolar cells and Pax6 for amacrine and Müller glial cells. Marker-positive cells were then quantified ( Fig 2B). We did not notice statistically significant reduction of marker-positive cells in double heterozygote Oc1 +/-; Oc2 +/-. Mice with one functional Oc2 allele (Oc1 -/-; Oc2 +/-) showed significantly reduced number of Sox9 + cells that decreased to 81%. Other marker-positive cells (Ap2α + , Chx10 + , Pax6 + ) were not significantly decreased in Oc1 -/-; Oc2 +/-. Statistical significance was not assessed for mice with one functional Oc1 allele (Oc1 +/-; Oc2 -/-) where less than three mice for each staining (Ap2α, Sox9, Chx10, Pax6) were analyzed. In agreement with previous studies [7,31], significant reduction of marker-positive cells localized in INL, with exception of horizontal cells, was not observed in Oc1 -/or Oc2 -/mice. Statistically significant reduction of marker positive cells was observed in compound Oc1/2 double mutant mice (Oc1 -/-; Oc2 -/-). Ap2α + cells decreased to 82%, Sox9 + cells to 66%, Chx10 + cells to 71%, and Pax6 + cells to 78%. H&E staining of compound Oc1/2 double mutant showed slightly thinner INL compared to the wt (Fig 1). Based on this observation, we also counted number of DAPI-positive cells localized in the INL. We counted DAPI-positive cells in wt and all generated Oc-deficient genotypes ( Fig 2C). DAPI + cells localized in the INL were counted and analyzed separately for each marker (Ap2α, Sox9, Chx10 and Pax6) and each genotype. DAPI + cells were counted in the same eye sections as particular marker positive cells. DAPI-positive cells in Oc-deficient mice were normalized to DAPI-positive wt control and marker-positive cells were normalized to marker-positive wt control. We found that DAPI-positive cells were significantly reduced in Oc1 -/-; Oc2 -/mice and proportionally decreased to the similar level as marker-positive cells. DAPI-positive cells in the Oc1 -/-; Oc2 -/retina were decreased to 78% in Ap2α staining, to 70% in Sox9 staining, to 72% in Chx10 staining and to 77% in Pax6 staining. In Oc1 -/-; Oc2 +/mice, DAPI-positive cells were significantly reduced to 82% only in Sox9 staining. In this case, DAPI + cells also decreased to similar level as marker positive cells. DAPI + cell were not significantly decreased in double heterozygote Oc1 +/-; Oc2 +/-, in Oc1 -/and Oc2 -/mice. Oc1 +/-; Oc2 -/mice were not statistically evaluated as less than three mice per staining were analyzed.  marker and genotype. Percentage of Ap2α + , Sox9 + , Chx10 + and Pax6 + cells in the INL was not significantly affected in all studied mutant genotypes. Statistical analysis was not performed for Oc1 +/-; Oc2 -/mice.
Taken together, these results suggests that amacrine cells, Müller glial cells and bipolar cells were significantly decreased in the Oc1 -/-; Oc2 -/retina, but overall percentage representation of individual retinal cell types remained similar compared to wt mice.

Cone photoreceptors and Tbr2 + retinal ganglion cells are influenced by deletion of Oc1 and Oc2
Oc1 and Oc2 are expressed during the retinal development in precursors of RGCs and cone photoreceptors [11,29], and we therefore analyzed these cell types in P18 retinal sections by immunohistochemistry (Fig 3A). Cone photoreceptors staining using an antibody against cone arrestin and retinal ganglion cells were stained by anti-Tbr2 antibody. Tbr2 is expressed in a subset of RGCs and plays an important role in formation and maintenance of Opn4/melanopsin-expressing ipRGCs-intrinsically photosensitive retinal ganglion cells [42]. These cells constitute an important component of the non-image-forming visual system. The number of marker-positive cells was then quantified (Fig 3B). Cone photoreceptors detected by staining against cone arrestin were significantly reduced: in Oc1 -/retina to 80%, in Oc1 -/-; Oc2 -/retina to 62% and in Oc1 -/-; Oc2 +/retina to 68%. Tbr2 + cells were markedly decreased only in case of Oc1 -/-; Oc2 -/to 58% (ns, p = 0,125). Quantification of DAPI-positive cells in the GCL (Fig 3C) revealed that DAPI-positive cells were reduced in Oc1 -/-; Oc2 -/to 67% (ns, p = 0,051). Tbr2 + cells normalized to DAPI cells in the GCL (S1 Fig) showed slight decrease without significance.

Electroretinogram (ERG) recording
For assessment of in vivo function, ERG-waveforms were recorded under scotopic ( Fig 4A) and photopic (Fig 4D) conditions. Scotopic intensity-response functions show no effect in case of Oc2 -/and a slight amplitude reduction of a-and b-waves in case of Oc1 +/-; Oc2 +/- (Fig  4F), but a similarly distinct loss of b-wave amplitudes under photopic conditions (Fig 4E). However, a-wave amplitudes, corresponding to the cone photoreceptor function, show only a mild amplitude reduction. For both, rod and cone function, a marked decrease in the ERG signal was observed in Oc1 -/-, Oc1 -/-; Oc2 -/-, Oc1 -/-; Oc2 +/and Oc1 +/-; Oc2 -/mice. Additionally, in all Oc-deficient mice the a-wave was better preserved than the b-wave, consistent with a predominant dysfunction of the inner retina (primarily BCs). Interestingly, although the amplitudes were markedly reduced, there was little variation in the waveform component implicit times under scotopic as well as photopic conditions. This finding may be explained by a

Discussion
HCs represent a small part of the retinal cell population, only about 0.2% [43], but they have a very important function. Horizontal interneurons provide a negative feedback to photoreceptors and connection to bipolar cells [44,45] and are necessary to maintain the working range at different illumination levels. We investigated Oc-deficient mice in which a decreased number of HCs corresponded to the Oc1/Oc2 genotype and thus to the dose of Oc1/Oc2 proteins. The level of reduction of HCs corresponds to the extent of thickness reduction of the INL containing the cell bodies and OPL where the synapses of HCs are located. Mice with a markedly reduced number, or even absence of HCs had markedly reduced thickness of, or even absent OPL. HCs synapsing with photoreceptors in OPL are required for their survival. It was shown that the loss of the majority of HCs leads to an age-dependent degeneration of photoreceptors in Oc1 -/mice [7]. OPL disappeared at the age of 5 months and ONL was markedly thinned, being only one-third thickness of the wild-type at 8 months, indicating disrupted survival of photoreceptors. Scotopic ERG recording of 5-month-old mice confirmed the reduced photoreceptor function in Oc1 -/mice [7]. Ablation of HCs using the diphtheria toxin receptor system in adult mouse retinas leads to a significantly decreased number of photoreceptor nuclei, disappearing OPL, thinning of the retina and a marked reduction of the scotopic ERG bipolar cell component and an almost absent photopic ERG [46]. These findings point out the important role of HCs in photoreceptor survival. The number of horizontal interneurons is dependent on the dose of Oc proteins, which are also expressed in the cone precursors during development [11,29]. A previous study discovered that Oc1 -/-; Oc2 -/mice failed in cone genesis at E14.5; nevertheless, cones were developed at E17.5, but their total number was decreased by 30%. This indicated that cones are developed in two ways, one that is dependent on Oc transcription factors and one that is either compensatory or completely independent of them [31]. In accordance with the decreased number of cone photoreceptors at E17.5 in double mutant mice, we observed a significant reduction of cone numbers, by 39%, in Oc1 -/-; Oc2 -/mice at P18. In Oc1 -/-; Oc2 +/mice at P18, the cone numbers were significantly reduced, by 32%. Other studies showed reduction of cone photoreceptors in double mutants by 23% in P18 mice [11], by 34% in P5 mice, with a reduction of 30% persisting in adulthood [31]. Cones in other Oc-deficient genotypes were not significantly influenced. This could be explained by the fact that there exists a threshold dose of Oc proteins required for the proper cone genesis during development.
Gene duplication represents a major mechanism through which new genetic material is generated during molecular evolution. Ancestral vertebrate genome has undergone two rounds of whole genome duplication [47] leading to the emergence of paralogs that were essentially free to undergo the process of neofunctionalization, subfunctionalization, or degeneration (gene loss) [48]. Genetic redundancy created by duplication enables neofunctionalization of one of the gene duplicates that can undergo a series of coding or regulatory mutations without harming the vitality of the organism since the other copy preserves the original function. In contrast, during the subfunctionalization process both gene copies can accumulate degenerative mutations provided that any defects in one gene copy are complemented by the other copy. Duplication-degeneration-complementation model proposes that the functionality of the original gene is distributed among the two copies, neither of which can be lost, as they perform non-redundant functions [49]. Last but not least, some duplications may be lost as they are either detrimental or not beneficial to the host organism. Our study of the functional role of two paralogous genes, Oc1 and Oc2, for retina anatomy, cell composition and function revealed an unexpected gene interaction. We found that in the context of HCs development, the four Oc1/Oc2 alleles appear equivalent and redundant, as similar defects are obtained in Oc1 -/-, Oc2 -/and Oc1 +/-; Oc2 +/double heterozygote mice. This is in contrast to Oc1/Oc2 alleles non-redundant function in other tissues like pancreas and endocrine system [50]. In addition, a single functional allele of either Oc1 or Oc2 (in the presence of a complete knockout of the other paralogue) is not able to support HCs development indicating strong dosage requirement. Although we cannot exclude that Oc1/Oc2 paralogs could have other, more or less divergent functions not adressed in our study, we conclude that Oc1/Oc2 paralogous genes function additively to support HCs fate and so their molecular evolution does not follow one of the typical routes after gene duplication.

S1 Fig. Presence of individual retinal cell types in the INL and GCL in Oc-deficient mice.
Counts of Ap2α, Sox9, Chx10, and Pax6 positive cells were normalized to counts of DAPI-positive cells in the INL and counts of Tbr2 positive cells were normalized to counts of DAPI-positive cells in the GCL. Marker positive and DAPI + cells were counted in the same sections, separately for each genotype. One point in graph means value for one mouse. Mean frequency ± SD. Statistical significance was not determined for Oc1 +/-; Oc2 -/where less than 3 mice were analyzed for each staining.