EML1 is essential for retinal photoreceptor localization and light detection


 Calcium regulates the response sensitivity, kinetics and adaptation in photoreceptors. In striped bass cones, this calcium feedback includes direct modulation of the transduction cyclic nucleotide-gated (CNG) channels by the calcium-binding protein CNG-modulin. However, the possible role of EML1, the mammalian homolog of CNG-modulin, in modulating phototransduction in mammalian photoreceptors has not been examined. Here, we used mice expressing mutant Eml1 to investigate its role in the development and function of mouse photoreceptors using immunostaining, in-vivo and ex-vivo retinal recordings, and single-cell suction recordings. We found that the mutation of Eml1 causes significant changes in the mouse retinal structure characterized by mislocalization of rods and cones in the inner retina. Consistent with the fraction of mislocalized photoreceptors, rod and cone-driven retina responses were reduced in the mutants. However, the Eml1 mutation had no effect on the dark-adapted responses of rods in the outer nuclear layer. Notably, we observed no changes in the cone sensitivity in the Eml1 mutant animals, either in darkness or during light adaptation, ruling out a role for EML1 in modulating cone CNG channels. Together, our results suggest that EML1 plays an important role in retina development but does not modulate phototransduction in mammalian rods and cones.


Introduction
Absorption of a photon by the visual pigment in vertebrate rod and cone photoreceptors triggers the activation of a transduction cascade that ultimately results in hyperpolarization of the cells and reduction in the release of neurotransmitter. The continuous function of photoreceptors over a wide range of light intensities with the constantly changing light conditions requires adaptation of their signaling. This is accomplished by controlling the gain of the phototransduction cascade by inhibitory calcium feedback mechanisms in these cells 1,2 . The dominant components of this feedback in both rods and cones involve regulation of cGMP synthesis via a couple of calcium-binding guanylyl cyclase activating proteins 3,4 and regulation of the lifetime of the activated visual pigment via the calcium-binding protein recoverin 4,5 . A relatively less studied aspect of this calcium feedback is the direct modulation of the CNG channels in the plasma membrane of rod and cone photoreceptors. Previous studies in amphibian rods have shown that calcium can modulate the sensitivity of these channels by activating a cytoplasmic protein 6,7 . However, the protein mediating this effect had remained unidentified. Studies with mouse rods expressing a mutant CNG B1-subunit with the calmodulin binding site deleted, have shown normal rod physiology, ruling out calmodulin as the calcium regulator of rod channels 8 . Recently, in striped bass, a novel protein termed CNG-modulin was shown to directly modulate the transduction CNG channels in cones 9 and later in zebrafish cones its homolog Eml1 (echinoderm microtubule associated protein 1 (EMAP1)-like), was shown to modulate these channels 10 . Based on these findings, we hypothesized that EML1 may play a role in the modulation of CNG channels in mammalian photoreceptors.
In this study, we sought to address this question by examining the effect of a mutation in the Eml1 gene 11 on mouse photoreceptors. We found that loss of functional EML1 protein did not affect the sensitivity or light adaptation of mouse rods and cones. Interestingly, our study revealed that the absence of fully functional EML1 causes abnormal structural development of retina, characterized by dramatic mislocalization of some rods and cones to the inner retina. The mislocalized cells survived beyond the course of retinal development but their number decreased over time. Notably, the photoreceptors that were properly localized in the outer nuclear layer (ONL) retained normal function. Thus, while proper EML1 function appears important for the lamination of the outer retina, it does not seem to regulate directly the function of rod and cone photoreceptors. In the course of the preparation of this paper, another group published a study using an unrelated Eml1 mutant that also demonstrated the role of EML1 for the proper localization of photoreceptors in the developing mouse retina 12 .

Results
Eml1 is expressed throughout the mouse retina As Eml1 has been shown to modulate the sensitivity of cones in zebrafish retina 10 , we first sought to determine if Eml1 is expressed in the cones of mouse retina. For this, we performed an in situ hybridization assay on 8 weeks-old wild type retina using an Eml1 probe. We found that Eml1 was expressed throughout the whole retina. The Eml1 transcript was predominantly present in the photoreceptor layer, with relatively sparse expression in the other retinal layers (Fig. 1A). Rods constitute 97% of the photoreceptors in the mouse retina 13 , making it hard to discern the possible expression of Eml1 in mouse cones. To address this question, we examined the expression of Eml1 in the Nrl knockout (Nrl -/-) retina, which lacks rods and is populated exclusively by cone-like photoreceptors 14 . Similar to the wild type retina, we found robust expression of Eml1 transcripts in the photoreceptor layer of the Nrl knockout retina (Fig. 1B). Thus, our results demonstrate that in the mouse retina Eml1 is expressed in both rods and cones.
An insertion mutation in Eml1 introduces a premature stop codon and a shorter transcript in mouse retina Homozygous HeCo mice carrying a spontaneous Eml1 mutation have been found to exhibit a subcortical heterotopia in the brain with associated hydrocephalus and cognitive impairment 11 . In our colony, we bred these mice with C57BL6 wild-type mice to generate heterozygous and then homozygous and wild type mice to study the role of EML1 in rod and cone response. The homozygous Eml1 mutant mice were fertile and showed no apparent behavior abnormalities. The genetic screening of the mutant mice revealed an insertion of several hundred base pairs in the intron 22 of the Eml1 gene which introduced a premature stop codon in exon 23 (Fig. 1C), potentially leading to a shorter transcript. To test that possibility, we performed real time PCR analysis using primers that amplified all known Eml1 transcripts.
At postnatal day 14 (P14), this RT-PCR analysis identified Eml1 transcripts in both wild type and mutant retinas (Fig. 1D). The Eml1 mutant retinas completely lacked the full-length transcript, and instead, gained a shorter spliced variant as reported 11 , confirming the expression of mutant products.
Eml1 mutation leads to reduction in the scotopic light response Because Eml1 was expressed in both rods and cones in the mouse retina, we set out to investigate the possible role of EML1 in both photoreceptor types, starting with the rods. We used in-vivo electroretinography (ERG) recordings to obtain rod-driven (scotopic) responses from control wild type and Eml1 mutant mice that were 8 weeks-old. We found that both the scotopic a-wave and b-wave responses from the Eml1 mutant mice (Fig. 2B) were reduced compared to controls ( Fig. 2A). The maximum scotopic a-wave response in the mutants was reduced to 35% of the controls (Fig. 2C) and the corresponding b-wave response was also recorded to be 34% of the wild type response (Fig. 2D). Thus, the Eml1 mutation caused a dramatic suppression of the rod-driven photoresponses.
Because of the ubiquitous expression of Eml1 in the mouse retina, we considered the possibility that it could modulate not only the rod photoreceptor responses but also the responses from rod bipolar cells.
To investigate the possible regulation of rod bipolar cell function by EML1, we compared the relative amplitudes of scotopic b-wave responses from mutant and control mice to their corresponding a-waves at all test flash intensities. We found that the ratio of scotopic b-wave amplitudes and a-wave amplitudes was comparable for Eml1 mutants and controls (not shown). Thus, the reduction on b-wave amplitude of the mutant mice was proportional to their a-wave amplitude reduction, indicating that the Eml1 mutation affected selectively the function of rod photoreceptors and produced no detectable change in signaling from a-wave to b-wave.
We next tested the rod response using ex-vivo transretinal recording, which allowed us to pharmacologically isolate the photoreceptor response. Using this more precise and reproducible recording method, we also found a substantial reduction in the rod response of Eml1 mutant rods compared to controls (Fig. 3A, B). Comparison of their intensity-response curves revealed a 49% reduction of the maximal response amplitude of Eml1 mutant rods ( Fig. 3C; Table 1). Interestingly, the normalized family of flash response curve was shifted slightly to the left in Eml1 mutants as compared to the controls ( Fig. 3C, inset) indicating slightly higher fractional sensitivity in the mutant rods. Consistent with this, the test flash intensity required to produce half-maximal response, I1/2, was found to be also slightly lower in the Eml1 mutants than the controls (Table 1). If EML1 in rods modulates the Ca 2+ -sensitivity of the transduction CNG channels, it would be expected that light adaptation in the mutant rods would be compromised so that they will desensitize more steeply than control rods. However, we did not find any notable difference in the sensitivity of the mutants and controls during light adaptation in a series of backgrounds with increasing intensity (Fig. 3D). Thus, EML1 does not appear to modulate rod sensitivity in darkness or in background light.
To determine the reason for the reduced maximal rod-driven responses observed both in-vivo and exvivo, we next performed single-cell suction recordings from control and Eml1 mutant rods. Surprisingly, we did not find any notable difference in the individual rod response between the control and the mutants ( Table 2). The saturated flash responses were nearly identical in amplitude (Fig. 4A, B; Table 2). The I1/2 values were also comparable in the mutant rods and controls ( Fig. 4C; Table 2), indicating normal sensitivity in the Eml1 mutant rods. Similarly, there was no significant difference in the kinetics of the dim flash response of Eml1 mutant rods and controls ( Fig. 4D; Table 2). Thus, our single-cell recordings from rods demonstrate normal function of individual Eml1 mutant rods, comparable to that of control rods.
This finding suggested that the abnormally small responses obtained from eyes or whole retinas of Eml1 mutant mice are not caused by intrinsic differences in the functional properties of mutant and control rods, but rather could be the result of a change in the total number of rods generating the overall retina response. This possibility was evaluated by analysis of the structure of the Eml1 mutant that revealed a surprisingly aberrant lamination in the outer retina. The effect of EML1 on retinal lamination is examined in detail below.
Eml1 mutation leads to reduction in the photopic light response We next characterized the function of cone photoreceptors in the Eml1 mutants. This was of particular interest because the effect of Eml1 on cones had already been demonstrated in zebrafish retina 10 . We performed transretinal recordings from control and Eml1 mutant retinas from 8 weeks-old mice in Gnat1 -/background which allowed us to isolate the cone-driven component of the retina response 17 . We found that, similar to the case of rod-driven responses, the cone-driven transretinal responses were also substantially suppressed in Eml1 mutant eyes (Fig. 5A, B). The maximal cone-driven response in Eml1 mutant mice was 40% of that in control retinas (Table 3). However, to our surprise, the sensitivity of cones was not affected by the Eml1 mutation so that the intensity-response curves for Eml1 mutant and control cones were comparable (Fig. 5C). Consistent with this, both I1/2 and the fractional sensitivity of darkadapted cones were comparable for mutant and control retinas ( Table 3). The kinetics of the cone dim flash responses were also comparable overall (Fig. 5C, inset), with only slightly slower time to peak and integration time in the mutant cones (Table 3).
Finally, we examined the effect of the Eml1 mutation on light adaptation in mouse cones. If EML1 modulates the cone CNG channel conductance, its mutation would be expected to compromise cone light adaptation and cause steeper decline of cone sensitivity with increasing background light intensity 4 .
However, we found that the sensitivity of mutant cones was comparable to that of controls over a wide range of background light conditions (Fig. 5D). Together, these results clearly demonstrate that unlike in the case of fish, EML1 plays no role in modulating cone phototransduction in darkness or during light adaptation in the mouse retina.

Loss of Eml1 function downregulates rod-specific phototransduction proteins
Our results show that the Eml1 mutation leads to reduction in the light response of both rods and cones in the whole retina while responses from individual rods remain normal. One possible explanation for this apparent discrepancy is the presence of two populations of photoreceptors in the Eml1 mutant retinaa group of photoreceptors that preserve normal function, and another distinct group where photoresponses are suppressed or completely absent. As a first step in determining the cause of the reduction of the whole retina responses, we examined the overall expression of several phototransduction proteins in control and Eml1 mutant retinas from 8 weeks-old mice. Western blot analysis of the whole retina lysates for -actin showed that control and Eml1 mutant retinas contain identical amount of this common housekeeping protein (Fig. 6A). Thus, -actin was used as loading control. We attempted to detect the EML1 protein with a commercially available polyclonal antibody (PA5-30016) generated against the N-terminal polypeptide corresponding to the 32-349 amino acid region of EML1. However, only a protein band corresponding to the short EML1 isoform with molecular weight of approximately 85-89kDa was identified in both control and mutant samples (Fig. 6B). Thus, the antibody was not able to recognize the long EML1 isoform in wild type mouse retina. Notably, we found a reduction in the expression of rhodopsin (Fig. 6C), the α-subunit of transducin (Gtα; Fig. 6D), the γ-subunit of transducin (Gtγ; Fig. 6E) and the γ-subunit of phosphodiesterase (PDEγ; Fig. 6F) in the mutants as compared to the controls. The reduction in the expression of these phototransduction proteins is consistent with the reduction in the rod-driven response from whole retina observed in the mutants ( Fig.   2 and 3). However, the normal responses that we obtained by suction electrode recordings from individual rods ( Fig. 4A, B) indicate that these rods are likely to express phototransduction proteins at normal levels.
Together, these results appear to be consistent with the notion of two separate populations of photoreceptors, one with normal function that is accessible for suction recordings, and another with suppressed photoresponses and reduced expression of phototransduction proteins that is inaccessible for suction recordings.

Mutation in Eml1 leads to structural impairment of retina
In order to explain the reduction of ERG responses, we next considered the possibility of morphological changes in the retina caused by the Eml1 mutation. To assess that, we did a preliminary screening using optical coherence tomography (OCT) and found some striking differences between the mutant and wild type retinal lamination. Specifically, the ONL appeared thinner and the central section of the retina, corresponding to outer plexiform layer (OPL) and inner nuclear layer (INL), appeared substantially intermixed in the Eml1 mutant compared to control retinas (Fig. 7A vs. 7E). To investigate this apparent difference, we next stained the retinas for hematoxylin and eosin and compared thickness of the ONL of Eml1 mutant and control retinas. We found that the ONL was thinner in the mutants compared to controls  Fig 7K) and contained nuclei that were smaller than usual and similar to the photoreceptor nuclei in the ONL in size and appearance (Fig. 7F). However, unlike the persistent difference in ONL thickness in older animals, the thickness of the mutant INL gradually declined with age ( Fig. 7F-H) and in 5 months-old animals was comparable for that of control retinas (Fig. 7D vs. 7H; Fig. 7L). The ONL thickness peaked at P21 and remained stable thereafter (Fig. 7M). The aberrant disorganization in the INL was already present at P14, shortly after the time of eye opening. Surprisingly, the INL thickness increased until P21, but then by the 8 th week had decreased back to P14 levels, where it remained stable at 5 months of age (Fig. 7N).

Identification of mislocalized cells in Eml1 mutant retina
The mislocalized cells in the INL resembled photoreceptors by their nuclear morphology. Thus, we hypothesized that these mislocalized cells are photoreceptors which could have been trapped in the INL during development. We tested this hypothesis by screening the cells in the INL for the expression of the photoreceptor markers rhodopsin, for rods, and cone arrestin, for cones. Consistent with our hypothesis, we found rhodopsin and cone arrestin expression in the INL where most of the small mislocalized nuclei were located (Fig. 8A). The number of rhodopsin-expressing cells in the INL decreased with age and they were barely noticeable at around 5 months. The outer segment length measured at 8 weeks and 5 months was also significantly shorter in the mutants as compared to controls (Fig. 8B).
To substantiate the surprising possibility that the cone arrestin-positive cells in the INL are mislocalized cones, we also sought to determine whether there is a corresponding reduction in the density of cones in the ONL of mutant mice. We quantified the cone arrestin positive cells at P14, 8 weeks and 5 months and found that their density in the ONL was significantly reduced in the Eml1 mutants at all ages (Fig. 8C, left panel) and consistent with the morphology, the cone density remained stable there from 8 th week onwards. As expected, no cone cells were found in the INL of control retinas. In contrast, in Eml1 mutant retinas, cones were observed in the INL, with density peaking at P14 and then declining significantly by the 8 th week and stabilized onwards (Fig. 8C, right panel). These results confirmed that a substantial number of the rods and cones were misplaced to the INL in the retinas of Eml1 mutant mice. The reduced photoreceptor response from mutant retinas, but normal responses from individual rods, in the ONL suggests that the mislocalized photoreceptors provide little to no contribution to the overall retinal response. Indeed, we attempted to record responses from photoreceptors mislocalized to the INL using a suction electrode, but were not able to observe any responses (data not shown). Together, these results indicate that EML1 is important for the normal structural development of retina that supports photoreceptor function.

Discussion
In this study, we investigated the effect of an insertion mutation in the Eml1 gene on the visual function of mouse photoreceptors. Our results demonstrate that the EML1 protein, which is predominantly expressed by photoreceptors in the retina, does not have a role in regulating the sensitivity of rods and cones. This is surprising because previous studies in zebrafish have shown that Eml1 regulates sensitivity of cones 10  The cells in the developing retina undergo programmed apoptosis triggered by a number of factors including but not limited to the cell type, their interaction with their environment and the maturation stage 15 . Interestingly, we observed that during the course of development and aging of the animals, the number of mislocalized cells in the INL in Eml1 mutants gradually decreased (Fig. 7N) and by 5 months of age, the Eml1 mutant retinas attained normal lamination (Fig. 7A). This could be either due to the migration of mislocalized cells back to the ONL or to the death of these mislocalized cells. In the former case, the migratory cells would be expected to restore at least partially the photoreceptor responses and the thickness of the ONL while in the latter case they would not affect the overall retina photoresponses.
The findings that the thickness of ONL (Fig. 7M, N) and the ONL cone cell density (Fig. 8C) remain unchanged in the Eml1 mutants after reaching adulthood argue against migration of mislocalized photoreceptor from INL to the ONL in mutant with age. To confirm this, we also tested how the physiological responses of Eml1 mutants are affected by age. As previously discussed, both the a-wave and the b-wave (Fig. 2) at 8 weeks of age were significantly suppressed compared to these in control mice.
Notably, we observed no evidence for age-driven increase in either the a-wave or the b-wave amplitudes in the mutants relative to controls (not shown). Thus, the most likely explanation for the gradual loss of mislocalized photoreceptors in the INL is not their recruitment to the ONL but rather degeneration. To confirm this, we performed a TUNEL assay on the Eml1 mutant retinas at 5 months and found TUNELpositive cells there (not shown), suggesting that some of these mislocalized cells were going into apoptosis. Together, these findings suggest that the mislocalized cells fail to migrate to ONL in the Eml1 mutant retinas and instead gradually degenerate. Electrophysiology. For physiology experiments, all animals were dark-adapted overnight prior to the day of experiment. For in-vivo ERG recordings, the animals were anesthetized using a cocktail of Ketamine (100mg/kg) and Xylazine (20mg/kg). Pupils were dilated using 1% atropine sulphate ophthalmic solution (Akorn, Inc., Lake Forest, IL) followed by application of 2.5% Gonak TM hypromellose ophthalmic demulcent solution (Akorn, Inc., Lake Forest, IL) to retain the moisture during the recording. The visual responses to flash stimuli were then recorded using a clinical ERG setup (LKC Technologies; Model UBA-4200c) adapted for mice.
For ex-vivo transretinal recordings, the animals were euthanized by CO2 and then eyes were enucleated under dim red light followed by dissection under infrared illumination. The dissection was performed in a dish containing oxygenated Ames medium (Sigma). The eyeball was cut close to the limbus and then the retina was gently detached from the posterior eye cup by tearing the sclera and RPE using forceps. The retinas were stored in oxygenated Ames medium in a dark chamber at room temperature until recording.
Recordings were conducted using previously described methods 17

Figure 2
Effect of Eml1 mutation on retinal function. In-vivo ERGs responses in scotopic conditions. Representative scotopic ash responses from wild type (A) and Eml1 mutant (B) mice. Flash intensities for both panels were (in Cd • s m-2): 2.5 x 10-5, 2.5 x 10-4, 2.5 x 10-3, 2.5 x 10-2, 0.25, 2.5, 20 and 250. For comparison, the responses to a ash of 2.5 x 10-4 Cd • s m-2 are highlighted in red in the two panels. Averaged intensity-response data for the a-wave responses (C) and b-wave responses (D) from wild type (black squares) and Eml1 mutant (red circles) mice. The continuous lines represent a t to the data with the Naka-Rushton function.   Ensemble-average normalized responses of control and Eml1 mutant cones plotted as a function of ash intensity and tted by the Naka-Rushton function. (C, inset) Averaged normalized dim ash responses of control and Eml1 mutant cones. (D) A plot of average normalized sensitivity as a function of background light intensity from the control and Eml1 mutant cones in light adapted conditions. The lines are ts calculated using a Weber-Fechner function.

Figure 7
Morphology of Eml1 mutant retinas. (A, E) Comparison of 8 weeks-old wild type and age matched Eml1 mutant OCT screening respectively. H&E staining from wild type retina at P14, 8 weeks and 5 months of age (B, C, D respectively) and age-matched Eml1 mutant retina (F, G, H respectively). The quanti cation of ONL thickness as a function of the distance from the optic nerve head shown as spider plots in wild type retinas (triangles) and Eml1 mutant retinas (circles) at P14 (I) and at 5 months (J). The quanti cation of INL thickness as a function of the distance from the optic nerve head shown as spider plots in wild type retinas (triangles) and Eml1 mutant retinas (circles) at P14 (K) and at 5 months (L). The temporal pattern of relative ONL (M) and INL (N) thickness measured at 500, 1000 and 1500 μm from the optic nerve head (ONH) at P14, P21, 8 weeks and 5 months in Eml1 mutants.