Monocular enucleation alters retinal waves in the surviving eye
Activity in neurons drives afferent competition that is critical for the refinement of nascent neural circuits. In ferrets, when an eye is lost in early development, surviving retinogeniculate afferents from the spared eye spread across the thalamus in a manner that is dependent on spontaneous retinal activity. However, how this spontaneous activity, also known as retinal waves, might dynamically regulate afferent terminal targeting remains unknown.
We recorded retinal waves from retinae ex vivo using multi-electrode arrays. Retinae came from ferrets who were binocular or who had one eye surgically removed at birth. Linear mixed effects models were used to investigate the effects of early monocular enucleation on retinal wave activity.
When an eye is removed at birth, spontaneous bursts of action potentials by retinal ganglion cells (RGCs) in the surviving eye are shorter in duration. The shortening of RGC burst duration results in decreased pairwise RGC correlations across the retina and is associated with the retinal wave-dependent spread of retinogeniculate afferents previously reported in enucleates.
Our findings show that removal of the competing eye modulates retinal waves and could underlie the dynamic regulation of competition-based refinement during retinogeniculate development.
Dorsal lateral geniculate nucleus
Retinal ganglion cell
Retino-retinal projecting retinal ganglion cell
Starburst amacrine cell
Spike time tiling coefficient
Developing nascent neural circuitry undergoes modifications in an activity-dependent manner [1, 2, 3, 4, 5, 6]. Neural activity that is essential for early stages of visual system development originates from spontaneous processes [1, 4, 5, 6] and appears to facilitate circuit refinement by driving Hebbian-like competition for synaptic partners between innervating neurons [7, 8, 9]. This activity-dependent refinement results in the precise mapping of sensory areas, for example by establishing eye-specific laminae and fine-scale retinotopy across visual areas [1, 4].
Patterned spontaneous retinal activity (i.e. retinal waves) occurs primarily during periods of functional blindness [10, 11, 12, 13] and is characterized by periodically occurring domains of retinal ganglion cell (RGC) activity that slowly propagate across the retina in a wave-like fashion. This spatiotemporal feature of retinal waves leads to a high level of correlated activity between neighboring RGCs and very little correlated activity between RGCs that are distant from each other. Retinal waves have been shown to play a critical role in the establishment of eye-specific laminae in the dorsal lateral geniculate nucleus (dLGN) [7, 8, 14, 15, 16, 17, 18, 19, 20], as well as fine-scale retinotopy [8, 21] and receptive field size [8, 22, 23] in the dLGN and superior colliculus. For example, when an eye is lost early in development, retinogeniculate afferents from the surviving eye spread across the dLGN in a retinal wave-dependent manner . This study demonstrated that retinal waves drive both inter-eye  and intra-eye competition for synaptic space in the dLGN. However, it remains unclear how retinal waves might facilitate retinogeniculate expansion. One possibility is that the loss of the competing eye alters retinal waves to guide this process.
Here we show that in ferrets when we surgically remove a competing eye, retinal waves in the surviving eye were altered. Primarily, retinal wave associated bursts of action potentials by retinal ganglion cells in the surviving eye were shorter as were the number of spikes contained in these bursts. The shortening of bursts also decreased levels of pairwise RGC correlation. Thus, a significant reduction in levels of correlated RGC activity during retinal waves is associated with removal of the competing eye.
Based on these data, we propose a model where the presence of the competing eye reduces intra-eye competition for synaptic space in the dLGN by increasing correlated RGC activity, which facilitates the formation of eye-specific laminae during inter-eye competition. Conversely, the absence of the competing eye promotes the expansion of retinogeniculate laminae by reducing pairwise RGC correlations and increasing intra-eye competition. In this way, adjustments to the duration of RGC bursts during retinal waves could dynamically optimise competition-based retinogeniculate refinement during the establishment of eye-specific laminae.
Time-pregnant fitch-coat ferrets were received at mid to late gestation, giving birth 2–3 weeks later (Marshall BioResources, NY, USA; RRID:SCR_015489). Food and water were provided ad libitum. All procedures were authorized by the University of California, Davis (RRID:SCR_012713) Institutional Animal Care and Use Committee (IACUC) and performed in accordance with national and international standards for humane animal research as set forth by the National Institutes of Health (RRID:SCR_011417), Institute of Laboratory Animal Research (RRID:SCR_006872), USDA (RRID:SCR_011486), and Assessment and Accreditation of Laboratory Animal Care, International (RRID:SCR_015496).
Neonatal ferrets of either sex were anesthetized with isoflurane at P1. After topical lidocaine was applied, the eyelids of one eye were separated, and the muscles and connective tissue of the eyeball were blunt dissected. Hemostats were used to clamp the optic nerve after which it was severed and the eyeball removed. Antibiotic ointment was applied to the orbit, and sterile gelfoam was inserted to stem any subsequent bleeding. A liquid suture was applied to seal the eyelids. Before the animal fully awakened a single dose of buprenex was administered intramuscularly (0.02 mg/kg) as a postoperative analgesic. The monocular enucleation procedure typically took under 5 min. Age-matched littermates served as controls.
Multielectrode array recordings
Ferrets were euthanized with a lethal dose of pentobarbital (0.1–0.2 ml) via an inter-peritoneal injection. An eye was enucleated, and the retina was removed and stored in ice-cold buffered and oxygenated media (M7278, Sigma-Aldrich, USA; RRID:SCR_008988). A piece of the retina was placed RGC side down onto a 60-channel MEA (MEA2100 System, Multi-Channel Systems, Germany; RRID:SCR_014809), and held in place with a piece of dialysis membrane (Spectrapore 132,130, Spectrum Labs, USA; RRID:SCR_015488). The tissue was superfused with buffered media at 1–2 ml/min at 34 or 37 °C. The array electrodes were 30 μm in diameter and arranged in an 8 × 8 rectilinear grid with an interelectrode spacing of 200 μm. At this distance, the signal for a given cell appeared on only one electrode, so each isolated cell was assigned the spatial coordinates of the electrode on which it was recorded. Analog data were acquired at 20 kHz per channel simultaneously from each electrode. After the retina had been placed on the MEA, the tissue was allowed to acclimate for at least 45 min. When retinal waves appeared stable, recordings were performed for 20 min.
Raw data were digitally filtered with a 125-Hz high-pass filter (four-pole Butterworth) for sorting spike events. A threshold of six STD was set for each channel and 1 ms of data before, and 4 ms after a threshold-crossing event were stored for each negative-slope event. These candidate spike waveforms were then sorted with Offline Sorter (Plexon, USA; RRID:SCR_000012) using the first three principal components of the spike waveforms. Coincident events within 0.5 ms of each other that were detected on at least 90% of the channels were attributed to perfusion noise and removed. Clusters were first identified using an EM cluster algorithm  then manually edited for clustering errors. Typically, each electrode recorded the activity of one to three cells.
Analysis of RGC burst properties
RGC bursts were identified as previously described . All burst analyses were carried out using custom scripts written in Matlab (Mathworks, USA; RRID:SCR_001622). The beginning of a burst was defined as the point in an RGC spike train when the inter-spike interval (ISI) was less than 0.1 s. Subsequent spikes with ISIs less than 1 s were included in the burst, whereas an ISI of greater than 1 s denoted the end of the burst. If two bursts occurred within 5 s of each other, they were merged.
The properties of bursts identified by this algorithm were then averaged for each cell. Firing rate, burst duration, the number of spikes within a burst, the percentage of spikes within bursts, burst frequency, burst ISI, the percentage of burst time above 10 spikes/s, the percentage of bursts in waves, and bursts per wave were all quantified.
Analysis of wave properties
Retinal waves were identified in a similar way to that previously described . All wave analyses were carried out using custom scripts written in Matlab (Mathworks, USA; RRID:SCR_001622). MEA recordings were divided into 1 s time bins. The beginning of a retinal wave was defined as the time bin when greater than 5% of all cells were bursting and considered over when less than 2.5% were bursting.
The position of a wave over time was the center of mass of the cells participating in the wave in each time bin. Wave speed was defined as the average change in wave position over time. Wave spread was defined as the average percentage of new electrodes that detected bursting cells in each time bin. Waves that lasted for less than 3 s were not included in analyses of wave speed or spread.
The size of a wave was defined as the average percentage of electrodes that detected bursting cells across the duration of a wave.
All statistical analyses were carried out in Matlab (Mathworks, USA; RRID:SCR_001622). The sample sizes required for this study were estimated based on previous studies [22, 26]. For descriptive statistics, we used mean ± STD, or mean ± SEM where indicated. For box plots, the height of the boxes extended between the 25th and the 75th percentiles of the data. The horizontal bar and cross mark signified the median and mean, respectively. For plotting, outliers were defined as data points 1.5 times higher than or 1.5 times lower than the interquartile range and were shown as circles. The box plot whiskers extended to the most extreme data points that were not considered outliers. Outliers were not excluded from analyses. We considered P values less than 0.05 as significant. Significance values for comparisons of burst property means were calculated by fitting hierarchical linear mixed-effects models to cell data where the condition (monocular or binocular) and recording temperature (34 or 37 °C) were the fixed-effects, and recording/retina was the random-effect to correct for the non-independence of recorded cells. Significance values for comparisons of wave properties were also calculated by fitting linear mixed effects models as described above, except when comparing wave frequencies where the model only included terms for condition and temperature. In cases where samples were lognormal, we carried out a log transformation to bring samples to a normal distribution. In other cases, sample distributions had downward skews and were transformed with the exponential function. Comparing STTC values between enucleation conditions was similarly carried out using a hierarchical linear-mixed effects model where condition and temperature were fixed-effects, RGC pair distance was a covariate, and recording/retina was a random effect. STTC values were averaged for all RGC pairs by distance for each retina, resulting in a single value for each unique RGC pair distance. Before fitting the model, distance values were log transformed to improve linearity as shown in Fig. 6. All figures display data in their transformed state.
Monocular enucleation has multiple effects on retinal waves
The waves recorded from the retinae of monocular and binocular ferrets appeared at first glance to be qualitatively similar (Fig. 1b). However, with further analysis, it was found that retinal waves were notably different in several ways following early monocular enucleation. The largest and most significant effects observed were those on RGC burst duration and the number of spikes within a burst (Fig. 1d-e). Compared to retinae from binocular ferrets, those from enucleates had RGCs whose bursts of action potentials were approximately 30% shorter in duration (binocular, 2.39 ± 1.32 s; monocular, 1.67 ± 0.99 s; mean ± STD; binocular, N = 1178 cells, 13 retinae; monocular, N = 1001 cells, 11 retinae; T(2168) = 4.91, P = 9.9609 × 10− 7, linear mixed-effects model with log transformation) (Fig. 1d). The reduction in burst duration was for the most part consistent across retinae recorded at temperatures of either 34 °C (binocular, 2.76 ± 1.42; monocular, 1.82 ± 0.91; mean ± STD; binocular, N = 731, 7 retinae; monocular, N = 592, 6 retinae; T(1317) = 3.90, P = 9.952 × 10− 5, linear mixed-effects model with log transformation) or 37 °C (binocular, 1.79 ± 0.83; monocular, 1.46 ± 1.07; mean ± STD; binocular, N = 447, 6 retinae; monocular, N = 409, 5 retinae; T(850) = 3.37, P = 0.00078, linear mixed-effects model with log transformation) (Fig. g-h). The impact on burst duration due to enucleation lead to the number of spikes within a burst to be reduced by approximately 25% (binocular, 60.16 ± 53.26; monocular, 44.87 ± 39.82; mean ± STD; binocular, N = 1178 cells, 13 retinae; monocular, N = 1001 cells, 11 retinae; T(2168) = 2.92, P = 0.00357, linear mixed-effects model with log transformation) (Fig. 1e). As expected, given that fewer spikes were contained within bursts, the overall firing rate of RGCs was reduced with enucleation (binocular, 1.19 ± 1.19 spikes/s; monocular, 0.86 ± 0.84 spikes/s; mean ± STD; binocular, N = 1178 cells, 13 retinae; monocular, N = 1001 cells, 11 retinae; T(2176) = 2.49, P = 0.0128, linear mixed-effects model with log transformation) (Fig. 1f).
Shortening RGC burst duration reduces pairwise RGC correlation levels
Based on the hypothesis that a shortening of burst duration should reduce pairwise correlation levels between RGCs during waves (Fig. 6a), we carried out additional measurements to determine if levels of correlated activity decreased with enucleation. To quantify pairwise RGC correlations we calculated the “spike time tiling coefficient” (STTC) , which is a pairwise correlation measure that is bounded and insensitive to firing rate. Retinae from enucleates and binocular ferrets showed pairwise RGC correlations that fell as a function of the distance between cell pairs consistent with what has been previously described  (Fig. 6b). However, as predicted, levels of correlated RGC activity during waves were reduced across RGC pair distances for retinae from enucleates (binocular, N = 401 distances, 13 retinae; monocular, N = 341 distances, 11 retinae; T(738) = 2.85, P = 0.00452, linear mixed-effects model) (Fig. 6c).
We have previously shown that retinal waves are critical for the targeting of retinogeniculate afferents following the removal of competing inputs to the dLGN . While this work demonstrated that aspects of afferent terminal targeting during retinogeniculate refinement are retinal wave-dependent, it remained unclear what retinal wave properties might be necessary for this process. This study aimed to elucidate what retinal wave properties could dynamically guide afferent terminal targeting when an eye is lost. We show that the removal of the competing eye alters the duration of retinal wave associated RGC bursts, which has impacts on RGC correlation. Since studies have shown that correlation plays a critical role in retinogeniculate refinement, our finding is consistent with the hypothesis that retinal wave activity can dynamically guide retinogeniculate refinement while taking into consideration the presence of inter-eye competition.
We should note that since monocular enucleation is the complete removal of an organ, there is essentially no sham surgery that can fully replicate its potential side effects. Thus, our study cannot authoritatively rule out effects on retinal waves due to stress the ferrets may have experienced due to the monocular enucleation procedure. However, there is data to suggest that noxious stimuli at this age are unlikely to have large impacts on the brain. Studies have shown that newborn mammals that undergo extended periods of brain development ex utero are hyporesponsive to noxious stimuli in the first two postnatal weeks . Additionally, we found that newborn ferrets that underwent the short enucleation procedure healed quickly and did not display any developmental stunting or signs of distress. Ultimately, future experiments utilizing more targeted interventions are required to elucidate further how monocular enucleation affects retinal wave activity.
The role of burst duration and RGC correlation
RGC bursts are important for refinement of the retinogeniculate pathway [4, 9, 17, 28, 29]. Thus, the changes in RGC burst duration following monocular enucleation (Fig. 1) may indicate a role of the competing eye in influencing the refinement of the retinogeniculate pathway of the other. Additionally, we found that RGC burst duration scales RGC correlation (Fig. 6). Removal of an eye thus results in reductions in correlation level that are associated with the expansive targeting seen in the dLGN following monocular enucleation . Since a large body of experimental and theoretical work has supported a role for pairwise RGC correlations in retinogeniculate refinement [8, 9, 18, 26, 28, 31, 33, 34], we believe this finding is unlikely to be coincidental. Based on the evidence that afferent terminal targeting is guided by competition , we propose a model where RGC burst duration scales RGC correlation to dynamically guide afferent targeting within the dLGN during visual system development:
Consistent with our model, the importance of low intra-eye competition for eye-specific segregation in binocular ferrets was recently demonstrated . In this study, we used an immunotoxin to ablate starburst amacrine cells (SACs) that are responsible for retinal wave generation. This SAC ablation resulted in a reduction of RGC correlation similar to that seen following monocular enucleation, with fewer SACs leading to less RGC correlation. In binocular ferrets where laminae appeared normal, SAC ablation levels were symmetric across eyes. However, in ferrets where one eye’s retinogeniculate projection was larger, SAC ablation was lower in that eye (i.e., SAC ablation was asymmetric). This result demonstrated that when intra-eye competition is increased in one eye relative to the other due to asymmetric SAC ablation, the eye with increased intra-eye competition (less RGC correlation) is hindered in its ability to compete for synaptic space in the dLGN and loses territory to the eye with lower intra-eye competition (more RGC correlation). Similarly, recent studies in mice have used transgenic lines to investigate the role of retinal wave size in retinogeniculate refinement. In two different transgenic mouse lines, neighboring RGC correlation levels were reduced but not eliminated [33, 34]. Consistent with our model, these studies showed that reduced RGC correlation disrupted eye-specific segregation [33, 34] and resulted in an expanded ipsilateral projection in binocular transgenic mice . Additionally, in transgenic enucleates where competing contralateral inputs were absent, fine-scale retinogeniculate refinement appeared normal . We must note, however, that while the above studies demonstrated the importance high RGC correlation levels for the establishment of eye-specific laminae, the nature of the effects for ferrets and mice were different. In ferrets, moderate increases in relative intra-eye competition shrank eye-specific laminae and had minor effects on eye-specific segregation , while in transgenic mice ipsilateral projection size increased for both eyes and eye-specific segregation was disrupted [33, 34]. The difference can be explained by the high levels of RGC correlation found in ferrets relative to mice . The reduction to the mouse’s already relatively low RGC correlation levels in the above transgenic lines may have prevented effective inter-eye competition [33, 34] and resulted in expanded ipsilateral projections  like observed in monocular ferrets where no inter-eye competition is present [8, 37], or in binocular ferrets when EPI treatment completely blocks retinal waves [7, 8].
Surprisingly, bursts that occurred outside of waves were affected differently by monocular enucleation. For non-wave bursts, burst ISIs were shorter and the percentage of burst time above 10 spikes/s was greater (Fig. 4). However, it is important to note that non-wave bursts made up less than 3% of all bursts in either condition (Fig. 2) and do not appear to have any significant impact on overall levels of pairwise RGC correlation (Fig. 7). For these reasons, the observed changes to non-wave burst properties are unlikely to be related to the altered retinogeniculate refinement in enucleates. It is hard to speculate on why these effects are observed for non-wave bursts and not wave bursts, but after additional studies are carried out, it may prove valuable in understanding the neurobiological mechanisms by which monocular enucleation alters burst properties.
A signal for the presence of the competing eye?
Removing an eye alters retinal waves in the one that is spared. However, this study is unable to elucidate the neurobiological mechanism underlying such effects. It is important to note that we observe differences in retinal waves due to monocular enucleation ex vivo, indicating that the competing eye must be inducing relatively long-lasting effects in the opposing retina. One candidate mechanism is inputs from the competing eye onto neuromodulator releasing amacrine cells, which modify synaptic connectivity or other cell membrane properties . Retino-retinal projecting retinal ganglion cells (rrRGCs) have been identified in several vertebrate species [39, 40, 41] and are greatest in number during early visual system development in rodents . While a direct projection between the retinae to signal the presence of a competing eye is perhaps the most parsimonious explanation for the effects on retinal waves reported here, there is no direct evidence that rrRGCs modulate retinal wave activity. Future experiments to target these cells carefully, with either ablation or silencing, will be necessary to understand their role, if any, in visual system development.
Our results demonstrate a novel phenomenon whereby the removal of a competing region of the central nervous system influences the patterned spontaneous neural activity of another. When the competing eye is absent, it shortens the RGC burst duration of the surviving eye. This effect on RGC burst duration scales the levels of RGC correlation in the developing retina, and reduced correlation levels coincide with the retinal wave-dependent spread of the retinogeniculate projection following the loss of an eye. Based on these novel findings and their association with the retinal-wave-dependent anatomical remodeling found in enucleates, we propose the hypothesis that the presence or absence of the competing eye dynamically scales afferent competition to guide retinogeniculate refinement during visual system development.
We would like to thank members of the Cheng lab for providing valuable scientific feedback over the course of the study.
NIH National Eye Institute (RRID:SCR_011411) grants EY011369 (to H.-J.C.) and EY012576 (to J. S. Werner) supported this research.
Availability of data and materials
The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.
SWF and H-J C conceived the project. SWF performed surgical procedures, collected MEA data, and analyzed electrophysiological data. AN performed surgical procedures and collected MEA and anatomical data. SWF and H-JC wrote the manuscript. All authors read and approved the final manuscript.
Ethics approval and consent to participate
All procedures were authorized by the University of California, Davis Institutional Animal Care and Use Committee (IACUC) and performed in accordance with national and international standards for humane animal research as set forth by the National Institutes of Health, Institute of Laboratory Animal Research, USDA, and Assessment and Accreditation of Laboratory Animal Care, International.
Consent for publication
The authors declare that they have no competing interests.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.