Changes in fundus reflectivity during myopia development in chickens

: Previous studies have shown that changes in functional activity in the retina can be visualized as changes in fundus reflectivity. When the image projected on the retina is low pass filtered or defocused by covering the eye with a frosted diffuser or a negative lens, it starts growing longer and develops myopia. We have tested the hypothesis that the resulting altered retinal activity may show up as changes in fundus reflectivity. Fundus reflectivity was measured in chickens in vivo , both in visible (400-800 nm, white) and near ultraviolet (UV) light (315-380 nm). Two CCD cameras were used; a RGB camera and a camera sensitive in near UV light (peak sensitivity at 360 nm). White and UV LEDs, respectively, placed in the center of the camera lens aperture, served as light sources. Software was written to flash the LEDs and record the average brightness of the pupil that was illuminated by light reflected from the fundus. The average pixel grey level (px) in the pupil was taken as a measure of the amount of reflected light while refractive errors were corrected by trial lenses after pupil brightness was corrected for pupil size. It was found that myopic eyes had brighter pupils in UV light, compared to eyes with normal vision, no matter whether myopia was induced by diffusers or negative lenses (48 ± 9 vs. 28 ± 3, p<0.001 and 47 ± 7 vs. 27 ± 2, respectively). Using SD-OCT in alert chickens it was found that the retinal nerve fiber layer (RNFL) and the retinal ganglion cell layer (RGCL) in the central retina became thinner already at early stages of myopia development, compared to controls (31.2 ± 5.8 µm vs. 43.9 ± 2.6 µm, p<0.001 and 36.9 ± 1.2 µm vs. 44 ± 0.5 µm, respectively). While the decrease in RNFL thickness occurred concomitantly with the increase in UV reflectivity, it remains unclear whether these changes were causally linked. Thinning of the RNFL could be due to reduced neural activity in retinal ganglion cells but also due to metabolic changes in the retina during myopia development.


Introduction
Myopia (near-sightedness) is the most common ocular disorder in young people, where distant objects are focused in front of the retina. Its prevalence is currently still rising, especially in Southeast Asia, but also in Europe and the United States. It has been extrapolated that about half of the world population will be myopic by 2050 [1]. Given that only a small percentage of variance of refractive errors (<10 percent) can be explained by genetic variants that were detected in Genome-Wide Association Studies [2][3][4], the major stimulus for myopia development must reside in the visual environment [5][6][7][8][9]. A large number of studies have focused on the biological mechanisms of myopia development and the visual cues that might trigger the enhanced growth of the eyeball (i.e [10].). However, there is currently no biomarker available that could be used to follow potential myopia-related changes in the retina in vivo. In this study, we have analyzed changes in fundal reflectivity and retinal layer thickness that might be associated with early stages of myopia development in the chicken model.

Previous attempts to measure reflectivity of the fundus
Already in 1952, Rushton introduced quantitative densitometry to measure photopigment density based on the spectral characteristics of reflected light in the living human eye [11][12][13][14]. Van Norren and van de Kraats used continuous recordings of fundus reflectivity in white light to analyze the time kinetics of visual pigment bleaching and regeneration [15]. Optical density and topography of the macular pigment were also measured by fundus reflectometry [16]. Orientation and directionality of the photoreceptors (the Stiles-Crawford effect) were explored in vivo by measuring the amount of light reflected by the retina in the pupil plane [17][18][19][20].

Previous attempts to measure the state of retinal activity
The amount of light and spectral distribution of light reflected in a tissue depends not only on pigment densities and distributions but also on the distribution of structures with different refractive indices. Refractive indices in the intracellular space are related to the metabolic state of a cell [21]. Furthermore, biochemical changes may induce structural modifications in the cell that can also change reflectivity [22][23][24][25][26]. In retinal glia cells it was shown that enhanced neural activity increases light transmittance by causing cellular swelling, related to higher extracellular potassium levels. In the rat optic nerve, swelling of astrocytes was elicited by stimulation with flicker light [27]. With increasing amplitudes of the flicker electroretinogram (ERG), blood flow also increases to deliver nutrients and oxygen to the cells with higher metabolic demands. Both in monkeys and humans it was found that a spot of flickering green light projected on the retina reduces fundus reflectivity in the near-infrared range selectively in the stimulated area [26,28,29]. The reasons for these changes have not been fully understood [30][31][32], despite extensive studies by optical coherence tomography (OCT). OCT images are based on differences in refractive index between the different retinal layers. Based on OCT B-scans, Hillman et al [29] proposed that visual stimulation may affect optical path lengths in photoreceptors outer segments. Another explanation for reduced reflectivity was that increased neural blood flow induced higher light absorption by hemoglobin [33][34][35]. Buerk et al. stimulated a large area of the retina (30 degrees) and found that blood flow in the optic nerve head increased already after a few seconds, as did potassium ion concentrations [36]. In summary, these findings support the idea that visual experience, shaping neural activity as well as retinal and choroidal blood flow, can show up as changes in fundus reflectivity.

Evidence for changes in "retinal activity" during myopia induction
To induce deprivation myopia in an animal model, it is sufficient to spatially low pass filter the retinal image [37]. In a low pass filtered image, it can be expected that ON/OFF antagonistic ganglion cells are scarcely stimulated when their receptive field sizes are smaller than the wavelength at the highest transmitted spatial frequency. However, in the presence of high frequency Ganzfeld flicker, rapid temporal luminance modulations are provided which may restore some of the activity of these cells even in the presence of spatially low pass filtered images [38]. In chickens, myopia can be suppressed by stroboscopic light (i.e [39].). The inhibition of myopia in chicks by flicker light with various duty cycles was correlated with the amplitude of Ganzfeld flicker electroretinogram [40]. While it is expected that the amplitude in the electroretinogram depends on the length of the eye [41] and is inversely correlated with its axial length [42], it is more interesting to look into functional changes in the retina when myopia just starts to develop. In fact, changes in the induced component of the mfERG were found in children already when they just started to develop myopia and had still less than 0.5D [43]. Chen et al showed that the delayed mfERG response in myopic subjects was not due to longer eyes but due to changes in retinal processing [44]. It was proposed that reduced and delayed ERG responses in myopic eyes may result from changes in synaptic transmission between photoreceptors and bipolar cells or biochemical changes in the inner retina [44,45] [46] to induce len days in all case reflectivity we during the treat analyses. nto three categ yopic" stage, ( ity and retinal t the induced re ed more than 1 user n = 9, lens when chicks h s felt into this excluded from age" Experiment 2. At the age of 13 days, 7 chickens were unilaterally treated with diffusers. Fundus reflectivity in UV light was measured in the beginning and after 3 and 5 hours of treatment. Ocular biometry and retinal thickness were measured before and after 5 hours of imposed vision blur.

Measurements of refractive state and ocular dimensions
Refractions were determined by automated eccentric infrared photorefraction [47]. Ocular dimensions were determined by A-scan ultrasonography [48] under local corneal anesthesia (2% xylocaine solution) on the last day of the experiment.

Fundus reflectivity
A monochrome camera sensitive in the near UV range (300-420nm; SONY XC-EU50-CE, ALRAD Imaging, Camberley, England) with a peak sensitivity at 365 nm combined with a UV LED with a peak emission at 375 nm (LED375L, 5 mm; ThorLabs) centered in the camera lens aperture was used to measure fundus UV reflectivity. In chicks, the ocular media transmit light to more than 90 percent down to at least 350 nm [49]. A special UV transmitting camera optics was used (UV-Lens f/4, 60 mm, Carl Zeiss, Jena). In the visible range, fundus reflectivity was measured using a RGB CCD camera equipped with a 50 mm lens (DFK21 AU04, The Imaging Source, Bremen, Germany) with high sensitivity between 400 and 780 nm, and peak sensitivity at 500 nm. For measurements in visible light, a high power white LED (Nichia NSPW500CS, 5 mm; Conrad Electronics, Germany) with an emission peak at 460 nm served as light source. Software was developed under Visual C + + 8.0 to flash the LEDs and to analyze the brightness of the pupil, after back-illumination by light reflected from the fundus. Because automated detection of the pupil margins by the software was not always reliable, the operator could manually mark 4 positions at the pupil border and the software performed automatically a circle fit ( Fig. 2(A)). Pixel grey levels in the pupil were averaged and displayed on the screen, together with their standard deviation. The software corrected for the effects of pupil sizes, assuming that pupil brightness increases linearly with pupil area when a point source is used as described by Choi [50]. In the case of UV light, pixels were monochrome (Y800). In the case of the RGB camera, a simultaneous analysis was possible for the RGB channels ( Fig. 2(A)). In the case of the RGB camera it was verified that there was a linear relationship between luminance (in cd/m 2 ) and pixel grey level over the range of measurements in the current study (between 40 to 100; R 2 = 0.99, pixel values = 0.34*luminance + 38).
Chickens were cooperative and the operator could tilt their head in any direction until the eyes were aligned with the camera. After an image was grabbed, the alignment of the eye could be verified by the position of the first Purkinje image in the pupil (see example in Fig.  6). Since many measurements could be done in a short time, enough images were available to select those with the eyes properly aligned. A complicating factor was the varying amount of refractive error. The optical configuration used for the measurements of fundus reflectivity was the same as for isotopic photorefraction [51,52], meaning that light returning from a bright point in the fundus is focused in space at different distances, depending on the refractive state of the eye. Accordingly, light distributions in the camera plane vary with refractive state, causing also differently bright pupils. To overcome this problem, refractive errors had to be corrected by trial lenses, placed in front of the eye (Fig. 2(B)).  in the OCT scans, as well as the thickness of the layers between the inner plexiform layer (IPL) and the retinal pigment epithelium (RPE) (Fig. 3). Three B-scans, selected based on their best quality, were analyzed for each eye. The thickness of the RNFL and the RGCL in the area centralis could be measured in OCT B-scans with small standard deviations in repeated measurements (about 2.2µm). Inter-individual variance was also low (about 3.6µm).

Statistics
Statistical analyses were performed with the commercially available software package R (R 3.3.3, R Core Team, R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria). Averages of repeated measurements were taken for statistical analysis. Differences between treated and contralateral control eyes at each time point were analyzed using paired Student's t-tests. Repeated-measures MANOVA was used to analyze the effect of diffuser and lens treatment on fundus reflectivity and RNFL thickness over time. For comparison between diffuser-treated, lens-treated and control eyes at one treatment point an un-paired Student`s t-test with Bonferroni correction was used. The effects of diffuser or lens wear over increasing refractive error were evaluated by Pearson's correlation coefficient.

Development of myopia
Seven of 10 chickens treated with diffusers and two of the 5 chickens treated with negative lenses developed relative myopia between −0.25 to −6D (average −1.6 ± 1.8D, with a relative myopic shift of maximally 7D, Fig. 4(A)). Two chickens treated with diffusers were poor responders with only a minor myopic shift of only about 1D, and a difference in axial length between treated and control eye of less than 50 µm after 7 days of treatment. The avera 8) after 5 day the lens treate wear (Fig. 4 6. Pupil brightness ated control eyes a flectivity in UV y: 30 ± 5 px, la with diffusers with unobstructe ± 2 px; Fig. 7( eflectivity alre -treated eyes (d px, un-paired it took two mo 1 px; p<0.05; F a was high and ab [8]. This mi Looking at the h diffusers (n = g. 7(G)). MAN of treatment, ti p<0.001 for ea sing UV fundu me and treatme during a flash of and after induction V light remain st day: 27 ± 3 or negative len ed vision (diffu (C), 6(A)). Inte eady after 40 h diffuser vs. con t-test with Bon ore days for a s Fig. 7(B)). In d the amount of ight be due to e time course = 8, Fig. 7(G)) NOVA analysis ime and a tim ach of these var us reflectivity (M ent interaction) (A) near-UV and n of myopia. ned unchanged px, Fig. 7(A)-( nses became br users: 48 ± 9 p erestingly, eye hours, compare ntralateral con nferroni correc significant incr the current stu f myopia was g the fact that th of changes in ), a significant s for repeated m me and treatme riables). Also i MANOVA, p< was found.

Changes in RNFL and RGCL thickness during myopia development
Thickness of retinal layers was analyzed using SD-OCT. In eyes treated with negative lenses and diffusers, thinning of RNFL + RGCL was not yet detectable after 40 hours (Fig. 7(D)) but significant thinning of the RNFL + RGCL occurred between day 3 to 5 of treatment in both diffuser-and lens-treated chicks (34.9 ± 5.3 µm vs. 42.9 ± 5.1 µm and 39.3 ± 3.5 µm vs. 44.7 ± 1.5 µm, respectively; both p<0.05; Fig. 7(E)). Thinning was even more pronounced at the end of the treatment in all nine myopic eyes (7 treated with diffusers 31.2 ± 5.8 µm vs. 43.9 ± 2.6 µm, p<0.001 and two eyes treated with negative lenses 36.9 ± 1.2 µm vs. 44 ± 0.5 µm; Fig. 7(F), n.s.). Looking at the time course of changes in all chickens which responded to the treatment with diffusers (n = 8), a decreased RNFL + RGCL thickness was measured ( Fig.  7(H)). MANOVA analysis for repeated measurements showed a highly significant effect of treatment, time and a time and treatment interaction on RNFL + RGCL thickness (MANO-VA, p<0.01 for treatment, p<0.001 for time and time and treatment interaction). Also in lens treated chicks (n = 5) a significantly decreasing RNFL + RGCL thickness (MANOVA, p<0.05 for treatment) was found. It was striking that, except for the RNFL + RGCL, there was no significant thinning of retinal layers, even though the posterior globe expanded (194 ± 18 µm vs. 220 ± 19 µm). The differences in fundus reflectivity between treated and control eyes were significantly correlated with the amount of induced myopia, both in diffuser and negative lens-treated chickens (R = 0.66; p<0.0001 and R = 0.77; p<0.001, respectively; Fig. 8(A)). Fundus reflectivity in UV light was not only correlated with induced myopia but also with ocular axial length (diffuser (n = 8): R 2 = 0.45, lens (n = 5): R 2 = 0.7). RNFL + RGCL thickness was significantly negatively correlated with the amount of induced myopia in chicks treated with diffusers ( Fig. 8(B), R = 0.69; p<0.001). In chicks treated with lenses, such a correlation was not observed (Fig. 8(B); R = 0.19; p = 0.49). Instead, RNFL + RGCL thickness were significantly negatively associated with axial length in lens treated eyes (diffuser (n = 8): R 2 = 0.28, lens (n = 5): R 2 = 0.89). ctivity and RN thickness of th ). In eyes treate ly due to the sm (Fig. 9)

Discussion
In an attempt to identify potential biomarkers of myopia development that could be used in vivo, we found that fundus reflectivity in near UV light increased during induction of myopia in the chicken model. We also found that the increase in UV reflectivity was correlated with the amount of induced deprivation and lens induced myopia. As a possible morphological correlate we found that the retinal nerve fiber layer (RNFL) and the retinal ganglion cell layer (RGCL) were thinning in correlation with the increase of the UV reflectivity. However, a short term experiment showed that fundus reflectivity was already increased after 5 hours of diffuser wear, clearly too early to generate measurable changes in ocular biometry and retinal layer thickness. These findings suggest that the changes in UV reflectivity were not (only) due to the thinning of the RNFL. In eyes in which myopia was induced by negative lenses, such changes were also observed but they were less pronounced and also delayed in time. However the amount of myopia induced by the lenses was also lower and there were less chickens available in the lenstreated group. Myopia in animal models is inherently variable. In general, imposing defocus by lenses represents a closed-loop trigger of the emmetropization feedback loop and generates therefore less variable refractions than retinal image degradation by diffusers which represent an open-loop condition as the eye has no chance to improve retinal image by any kind of growth [48,60]. The amount of induced deprivation myopia is therefore largely determined by genetic factors. In fact, Chen et al. has shown that chickens can be selectively bred to develop either high or low amounts of deprivation myopia in only two generations [61]. In the current study, it was helpful that the induced deprivation myopia was highly variable because the variability permitted correlations with RNFLs and RGCLs, as well as with fundus reflectivity.

Why might there be selective thinning of the RNFL and the RGCL?
A striking finding was that the inner retinal layers (RNFL + RGCL) were thinning during development of myopia but that these changes could not simply be attributed to stretching since the other retinal layers did not change significantly, even at the end of the experiment when the eyes were clearly longer and myopic (thickness between IPL and RPE -myopic eyes: 194 ± 18 µm vs. controls: 220 ± 19 µm). A possible explanation is that, with diffusers in front of the eyes, the retinal image is low pass-filtered and also has low contrast, both factors reducing spatial information. It is likely that retinal ganglion cells, sampling the retinal image with their circular ON or OFF antagonistic field structures, are less stimulated when diffusers are worn. We hypothesize that less information is transmitted by the ganglion cell axons which could induce smaller axon diameters. Recently, it was shown in the auditory system that axon diameters and myelination are experience dependent [62]. Mice that were partially deprived from acoustic stimulation by wearing ear plugs developed reduced myelination and axons diameters in the trapezoid body. The decline in myelin sheet thickness amounted to up to about 50 percent. In our study, the thinning of the RNFL + RGCL reached about 30 percent (Fig. 8(B)). Lazari et al. [63], discussing the study by Sinclair et al. [62], emphasized that "experience-related reductions of myelin and axons diameters persist also in adulthood". It is known that information flux in an axon is dependent on its diameter. In fact, there might be a dynamic adaptation of "cable volume" to its "information capacity", as has been described by Sterling for the axons of rods and cones in the retina [64]. We propose that the thickness of the RNFL and the RGCL might be related to the average amount of captured visual information. Extrapolated to humans, it could be that reduced RNFL + RGCL thickness might represent a biomarker for a previous history of low pass filtering and defocus. However, it is clear that a direct demonstration is necessary that thinning of the RNFL + RGCL is due to thinning of the ganglion cell axons.
Studies in human subjects have shown that changes in neural retina are not uncommon among myopic subjects [41,65]. Most of the studies measuring RNFL abnormalities were carried out on glaucomatous eyes [66][67][68][69]. Nevertheless, thinning of the RNFL has also pre-viously been reported in myopic human subjects [70][71][72][73][74]. While most studies in humans showed that RNFL thickness is reduced only in high myopia, recent work showed that such changes can already be detected in low and intermediate levels of myopia although only in specific subfields. It was concluded that a specific pattern exists for myopia influencing RNFL thickness [70].

What could cause the changes in UV reflectivity when myopia develops?
The increased UV reflectivity after 40 hours of treatment could be related to a change in layer thickness and /or a re-arrangement inside axons in the nerve fiber layer of the retina. However, in the current study, a wide range of baseline thicknesses was found for the RNFL and RGCL (35 -53 µm) already in untreated eyes. Despite the variance in thickness, fundus reflectivity in UV light was always low in untreated eyes. A thin RNFL and RGCL can therefore not be the only reason for the high fundus reflectivity in eyes with diffusers. In line with our findings, previous studies have shown that changes in RNFL reflectivity occur mainly in the short wavelengths range (<560nm) [75][76][77][78]. In chicks, changes in reflectivity were restricted to the near UV range; they can therefore not be measured in humans since the ocular media scarcely transmit UV light. In addition to changes in thickness, other optical properties may also change such as biochemical parameters in the axons which depend on retinal activity. Huang and colleagues proposed a biophysical model of RNFL reflectivity speckles. They considered axon bundles as cylindrical structures, which actively transport molecules between cells, producing a characteristic speckle pattern. Each change in the dynamics in this mechanism could change the speckle texture. In-vitro studies have shown changes of RNFL reflectivity over time when retinal activity was modulated [79,80]. In other studies it was also found that differences in reflectivity can occur e.g. during axonal degeneration in rat model of glaucoma [81]. In our experiment we used incoherent light. An advantage of our method is that changes in RNFL and RGCL in the chicken can be easily detected by measuring UV reflectivity of the fundus in vivo. Therefore, UV reflectivity can be considered a biomarker for myopia development in the chicken (Fig. 11). Another possible explanation for the changes in fundus reflectivity could be changes in biochemical messengers during the development of DM and LIM, for instance decreased dopamine levels or downregulation of the transcription factor Erg-1 which should change retinal metabolism and ultrastructure, also affecting reflectivity [82,83].

Summary
In chickens, f myopia. At th ganglion cell changes were fiber layer ma nerve when th enced poor re such changes chickens by s ingen.