Circuit mechanisms underlying embryonic retinal waves

  1. Christiane Voufo
  2. Andy Quaen Chen
  3. Benjamin E Smith
  4. Rongshan Yan
  5. Marla B Feller  Is a corresponding author
  6. Alexandre Tiriac  Is a corresponding author
  1. Helen Wills Neuroscience Institute, University of California, Berkeley, United States
  2. Department of Molecular and Cell Biology, University of California, Berkeley, United States
  3. School of Optometry, University of California, Berkeley, United States

Abstract

Spontaneous activity is a hallmark of developing neural systems. In the retina, spontaneous activity comes in the form of retinal waves, comprised of three stages persisting from embryonic day 16 (E16) to eye opening at postnatal day 14 (P14). Though postnatal retinal waves have been well characterized, little is known about the spatiotemporal properties or the mechanisms mediating embryonic retinal waves, designated stage 1 waves. Using a custom-built macroscope to record spontaneous calcium transients from whole embryonic retinas, we show that stage 1 waves are initiated at several locations across the retina and propagate across a broad range of areas. Blocking gap junctions reduced the frequency and size of stage 1 waves, nearly abolishing them. Global blockade of nAChRs similarly nearly abolished stage 1 waves. Thus, stage 1 waves are mediated by a complex circuitry involving subtypes of nAChRs and gap junctions. Stage 1 waves in mice lacking the β2 subunit of the nAChRs (β2-nAChR-KO) persisted with altered propagation properties and were abolished by a gap junction blocker. To assay the impact of stage 1 waves on retinal development, we compared the spatial distribution of a subtype of retinal ganglion cells, intrinsically photosensitive retinal ganglion cells (ipRGCs), which undergo a significant amount of cell death, in WT and β2-nAChR-KO mice. We found that the developmental decrease in ipRGC density is preserved between WT and β2-nAChR-KO mice, indicating that processes regulating ipRGC numbers and distributions are not influenced by spontaneous activity.

Editor's evaluation

This paper investigates waves in embryonic mouse retinas. These stage 1 waves have been studied less than the post-natal (stage 2) waves. The paper uses elegant imaging and analysis approaches to monitor calcium signals across the entire retina and to determine the properties of the waves and their dependence on cholinergic and electrical synapses. This contributes substantially to the understanding of how these waves are generated.

https://doi.org/10.7554/eLife.81983.sa0

Introduction

Throughout the developing nervous system, spontaneous activity is observed before neural circuits are fully formed and sensory transduction begins (Akin and Zipursky, 2020; Blankenship and Feller, 2010; Luhmann and Khazipov, 2018; Martini et al., 2021). This activity is implicated in several development events, including cell death, maturation of functional circuits, and refinement of axonal projections (Blanquie et al., 2017a; Fujimoto et al., 2019; Kirkby et al., 2013). This is well studied in the developing visual system, where prior to the maturation of vision, laterally propagating spontaneous depolarizations sweep across retinal ganglion cells (RGCs), a pattern referred to as retinal waves (Wong, 1999). Retinal waves drive eye-specific segregation and retinotopic refinement of retinal projections to the dorsal lateral geniculate nucleus and superior colliculus (Ackman and Crair, 2014; Arroyo and Feller, 2016). Retinal waves also play a role in the maturation of direction-selectivity within the retina (Tiriac et al., 2022) and superior colliculus (Ge et al., 2021; Wang et al., 2009), as well as the development of retinal vasculature (Biswas et al., 2020; Weiner et al., 2019).

Retinal waves are present throughout mouse retinal development, starting as early as embryonic day 16 and persisting until postnatal day 14, which is around the time of eye opening (Blankenship and Feller, 2010; Choi et al., 2021; Feller and Kerschensteiner, 2013). As the retina develops, the circuits that mediate waves change. Stage 2 retinal waves, observed between postnatal day 1 and 10 (P1–10), are mediated via activation of nicotinic acetylcholine receptors (nAChRs) by acetylcholine (ACh) released from starburst amacrine cells (SACs). Stage 3 retinal waves, observed between P10 and 14, are mediated via activation of ionotropic glutamate receptors by glutamate released from bipolar cells.

Stage 1 waves are perhaps the least well understood, yet they are concurrent with many important events in retinal development as well as with retinal projections reaching their targets in the brain (Martini et al., 2021). Stage 1 waves are present in mice between embryonic day 16 and 18 (E16–18) (Bansal et al., 2000) and in rabbit starting at E22 (Syed et al., 2004). In rabbit, stage 1 waves persist in the presence of pharmacological antagonists of fast neurotransmitters and are blocked by gap junction antagonists (Syed et al., 2004). In mice, stage 1 waves consist of large propagating waves and small non-propagating events (Bansal et al., 2000). The application of nAChR antagonist inhibits larger propagating waves (Bansal et al., 2000). Though there is no anatomical evidence of synapses as early as E16–18, recent work has shown that SACs are present embryonically, begin to migrate to the inner nuclear layer (INL), and send projections to the inner plexiform layer (IPL) guided by homotypic contacts (Ray et al., 2018). Hence, cholinergic signaling is likely occurring via the volumetric release of ACh (Ford et al., 2012). The relative contribution of gap junction coupling and cholinergic signaling to the spatiotemporal properties of stage 1 waves remains to be understood.

Here we describe the spatiotemporal properties of stage 1 waves across the whole retina using a novel macroscope. We then identify the role of gap junction and cholinergic circuits on the generation and propagation of stage 1 waves. Next, we explore stage 1 waves in the β2-nAChR-KO mouse, which is the canonical mouse model for studying the role of stage 2 waves in developmental processes, and report that this mouse also exhibits altered stage 1 activity. Finally, we use β2-nAChR-KO mice to demonstrate that the regulation of ipRGC density is a wave-independent process.

Results

Macroscope imaging reveals the spatiotemporal properties of stage 1 retinal waves

The mouse retina at E16–18 exhibits spontaneous correlated transients (Bansal et al., 2000; Syed et al., 2004), termed stage 1 retinal waves, despite the immature state of retinal circuits. At E16–18, several postmitotic cell types are present in the retina, including broad classes of retinal ganglion cells (RGCs) and amacrine cells (ACs), as well as proliferating progenitors which will go on to produce cells such as rods, bipolar cells, and Muller glia postnatally (Figure 1A; Cepko, 2014). By E17, there are no chemical synaptic structures (Hoon et al., 2014), though migrating SACs release ACh (Wong, 1995), and they, along with RGCs, express nicotinic acetylcholine receptors. One potential mode of cell–cell communication is via gap junction coupling between RGCs as well as between progenitor cells (Cook and Becker, 2009), which have been proposed to be the primary substrate mediating stage 1 waves (Catsicas et al., 1998; Syed et al., 2004; Wong et al., 1998).

Figure 1 with 3 supplements see all
Spatiotemporal characteristics of embryonic waves.

(A) Schematized cross section of an E16 retina when stage 1 waves begin. Green cells represent post-mitotic retinal ganglion cells (pRGCs). Magenta cells represent post-mitotic amacrine cells (pACs). Gray cells represent progenitor cells (PCs). GCL, ganglion cell layer; IPL, inner plexiform layer; INL, inner nuclear layer; NBCL, neuroblastic cell layer. (B) Top left: macroscope image of the baseline fluorescence of an E17 GCaMP6s retina. For subsequent analysis, the retina was divided into 11,750 10 µm × 10 µm (not drawn to scale) squares. Number of ROIs changed depending on retina size. Top: ∆F/F traces of calcium transients from four example 10 µm × 10 µm ROIs. Middle: rasterized calcium transients for all ROIs raster plot of calcium transients >50% ∆F/F. Bottom: percentage of ROIs active throughout the time course of the recording (1 hr). (C) Histogram showing the distribution of the percent of ROIs with inter-transient-intervals (ITIs) ranging from 0 to 200 s. (D) Left: heatmap showing the temporal progression and spread of four example waves. Scale depicts timescale of propagation; dark gray = start of propagation and white = end of propagation. Right: histogram showing the distribution of waves with an area ranging from 0 to 100% of the retina. (E) Summary plot of wave speeds. Red dot and line = mean and standard deviation, respectively. n = 30 waves, three retinas, three mice.

Figure 1—source data 1

Inter-transient-interval, wave area, and wave speed.

https://cdn.elifesciences.org/articles/81983/elife-81983-fig1-data1-v3.xlsx

To begin to understand how cholinergic signaling and gap junction coupling govern stage 1 waves, we first used calcium imaging to characterize their spatiotemporal properties. Retinas were isolated from E16–18 mice that were either bath loaded with the organic calcium dye Cal 520 or from mice expressing the genetically encoded calcium indicator GCaMP6s under the Vglut2 promoter (Vglut2;GCaMP6s). The earliest age at which we could detect reliable waves was E16 (Video 1). Stage 1 retinal waves were recorded on retinal whole mounts using a custom-built epifluorescent macroscope.

Video 1
Stage 1 waves.

Calcium imaging of stage 1 waves using the macroscope. Total field of view is 4.7 mm × 4.7 mm. Frame rate 10 Hz. Total length of movie represents 1 min of recording.

To assess the spatiotemporal properties of stage 1 waves, we divided the retinal surface into small square ROIs (roughly 10 µm × 10 µm), about 7 µm apart. ΔF/F traces for each ROI were rasterized based on a detection algorithm that identified the timing of peak changes in fluorescence, which we call transients (Figure 1B). These rasterized transients were used for subsequent analysis. We first computed the time between spontaneous transients by measuring the inter-transient interval (ITI) for each ROI and found that the distribution peaked at around 80 s (Figure 1C). We also described the size of individual waves by computing the sum of transients that occurred simultaneously (i.e., percent of ROIs that participated in each wave). The distribution of wave sizes was broad, ranging from 6 to 90% of the retina, with a mean and standard deviation of 46 ± 21% (Figure 1D).

In addition to propagating waves, we also observed small non-propagating calcium events (defined by transients present in a few neighboring ROIs that never propagated beyond 160 × 160 μm2; Video 2 and Figure 1—figure supplement 1). These events were like those described previously in Bansal et al., 2000. Note that these small non-propagating transients were included in the ITI but not in the analysis of propagating wave sizes.

Video 2
Example of small non-propagating events.

Calcium imaging of stage 1 waves using the macroscope. Total field of view is 4.7 mm × 4.7 mm. Frame rate 10 Hz. Total length of movie represents 15 s of recording. Yellow triangles depict small non-propagating events.

Finally, we calculated the average propagation speed of stage 1 waves to range from 145 μm/s to 237 μm/s (average ± SD: 181 ± 24 μm/s; Figure 1E, Figure 1—figure supplement 2), similar to stage 2 waves we observed with an average speed of 177 μm/s ± 62 μm/s (see Table 1 for wave speed summary data). We also measured the distribution of stage 1 wave initiation sites in WT retinas and saw no evidence of an initiation site bias (Figure 1—figure supplement 3).

Table 1
Stage 1 wave speeds in WT and β2-nAChR-KO, plus stage 2 wave speeds.

Speeds recorded from individual waves. Summary data (µm/s) reported in Figure 3 with additional data to compare with speeds of stage 2 waves. Each row is an individual wave.

WT stage 1 wavesB2-nAChR-KO stage 1 wavesStage 2 waves
174.1777.91205.68
183.9492.40124.44
192.1998.19153.25
162.26117.34120.47
189.84128.52158.59
237.3487.97176.49
179.2492.40263.86
165.6892.06156.09
144.52103.25108.62
182.07103.90299.29
Avg181.1299.39176.68
Stdev24.3914.7262.62

nAChRs and gap junctions are important for setting the frequency and area of stage 1 waves

Previous work based on epifluorescent calcium imaging experiments performed in embryonic mice has shown that retinal waves, as defined by correlated changes in fluorescence, are reduced in frequency and size by curare, a competitive antagonist for nAChRs (Bansal et al., 2000). However, in roughly the equivalent developmental period in rabbit, blockade of all fast neurotransmitter receptor, including nAChRs, had no impact on wave frequency (Syed et al., 2004). Rather, waves in rabbit are blocked after the application of 18β-glycyrrhetinic acid, a gap junction antagonist (Syed et al., 2004).

To determine the relative role of gap junctions and nAChRs on the frequency and area of stage 1 waves in mice, we used two-photon calcium imaging and pharmacology in retinas isolated from E16–18 mice bath loaded with Cal 520. To block gap junctions, we bath applied the gap junction antagonist meclofenamic acid (MFA, 50 μM), which reversibly blocks electrical coupling between retinal interneurons (Veruki and Hartveit, 2009) and developing ipRGCs (Caval-Holme et al., 2019). Application of the gap junction blocker MFA (50 μM) nearly abolished stage 1 waves, causing a significant reduction in frequency of waves and cell participation during waves (Figure 2A-C). We also found a small but significant reduction in wave amplitude, quantified as the average maximum response amplitude of all cells participating in individual waves (Figure 2D).

Figure 2 with 1 supplement see all
Embryonic waves are mediated by gap junction and cholinergic circuits.

(A) ∆F/F time course of spontaneous activity observed in the field of view (FOV) and raster plot of neuronal calcium transients in ACSF (control condition, top) and in the presence of meclofenamic acid (MFA) (bottom). (B) Summary plot showing frequency of waves in control and MFA. Red dots and lines = mean and standard deviation, respectively. Asterisks represents significant effects. n = 8 retinas (six mice); p=4.81e–4. (C) Percent cells that participate in retinal waves in control and MFA (50 µM) (p=3e–3). (D). Mean calcium response of the neurons that participate in waves in control and MFA (p=0.02). (E–H) Same as (A–D) for dihydro-β-erythroidine (DHβE, 8 μM). (F–H) (n = 6 retinas; six mice); (F) p=0.05; (G) p=4.58e–4; (H) p=0.09. (I, J) Same as (A, B) but with hexamethonium (Hex, 100 µM) following a baseline recording. n = 14 retinas (nine mice); p=1.48–2. (K–L) Same as (A, B) but with epibatidine (EPB, 10 nM) following a baseline recording. n = 5 retinas (five mice); p=4.8e–3. All statistical tests here are paired t-tests.

MFA has notable off-target effects including overall cell health after long exposures (Kuo et al., 2016). Previous studies from our lab show that at P6 MFA does not reduce light-induced calcium transients in M1-iRGCs, the ability of M4-ipRGCs to fire action potentials, nor the amplitude of depolarization-induced calcium transients (Caval-Holme et al., 2019). Note this is in contrast to another gap junction antagonist, carbenoxolone, which inhibits light responses in cultured ipRGCs (Bramley et al., 2011). Using whole-cell voltage-clamp recordings, we observed that MFA induced a significant decrease in voltage-activated potassium conductance (Figure 2—figure supplement 1). Note that a reduction in K+ conductance would increase excitability of cells and is therefore unlikely to be the reason for the observed decrease in wave activity in the presence of MFA. Hence, we conclude that MFA’s block on retinal waves is via their impact on gap junctions.

We next assayed the impact on nAChR-antagonists on stage 1 waves. For stage 2 waves, both spontaneous calcium transients and compound excitatory synaptic events are completely blocked by bath application of dihydro-ß-erythroidine hydrobromide (DHβE, 8 μM) (Ford et al., 2012), which preferentially targets nAChRs containing α4 and β2-subunits (Harvey and Luetje, 1996; Harvey et al., 1996). We found that DHβE also dramatically reduced activity associated with stage 1 retinal waves (Figure 2E-H). However, in contrast to stage 2, some waves persisted in the presence of DHβE, but recruited fewer neurons and therefore had smaller areas (Figure 2G). Stage 1 waves were blocked by general nAChR antagonists: both hexamethonium (Figure 2I-J; Hex, 100 μM), a non-selective nAChR antagonist, and epibatidine (Figure 2K-L; EPB, 10 nM), an nAChR agonist that potently desensitizes all nAChRs (Corrie et al., 2020; Spang et al., 2000), blocked stage 1 wave events. Hence, both gap junctions and multiple subunit combinations of nAChRs mediate the initiation and propagation of stage 1 waves .

β2-nAChR knock-out mice exhibit perturbed stage 1 waves

Our results thus far indicate that both the frequency and area of stage 1 retinal waves are modulated by the activation of different subtypes of nAChRs as well as gap junction coupling. To further differentiate the role of different nAChRs, we characterized mice where the β2 subunit of the nicotinic acetylcholine receptor is genetically ablated (β2-nAChR-KO). β2-nAChR-KO mice have severely disrupted stage 2 retinal waves (Bansal et al., 2000; Rossi et al., 2001) and have served as a model system for assessing the role of stage 2 waves in driving different developmental events (Ackman et al., 2012; Burbridge et al., 2014).

We observed that β2-nAChR-KO retinas exhibited stage 1 waves with different spatiotemporal properties than those of WT retinas (Figure 3A and B, see also Video 3). Specifically, β2-nAChR-KO retinas exhibited longer IEIs (Figure 3C and D), and individual waves propagated over smaller areas β2-nAChR-KO (Figure 3E and Figure 1—figure supplement 1). In sharp contrast to WT retina, waves in the β2-nAChR-KO retina were unaffected by the addition of Hex but showed a significant reduction in both area and frequency with the application of MFA (Figure 3D and E). These results suggest that in the absence of β2-nAChRs, gap junctions are the primary remaining component of the stage 1 wave generation mechanism. Furthermore, we found that stage 1 waves in the β2-nAChR-KO propagated at a significantly slower speed than those observed in WT mice (average ± SD: 99 ± 15 µm/s compared to 181 ± 24 µm/s for WT, n = 10 waves per condition, p=3.9073e–8) (Figure 3F), consistent with a distinct mechanism of propagation.

Figure 3 with 1 supplement see all
β2-nAChR-KO mice have reduced stage 1 wave activity.

(A) Top: area plot summarizing the percentage of ROIs active throughout the time course of the recording; bottom: raster plots of ROI calcium transients across conditions. Black: WT control; magenta: β2-nAChR-KO control; teal: β2-nAChR-KO in Hex; yellow: β2-nAChR-KO in meclofenamic acid (MFA). (B) Heatmap showing the temporal progression of a propagating event observed using epifluorescent calcium imaging on the macroscope across experimental conditions. Scale depicts timescale of propagation; dark colors = start of propagation; white = end of propagation. (C) Histogram showing the distribution of the percent of ROIs with inter-transient-intervals (ITIs) ranging from 0 to 200 s. Bars represent summary data across all retinas. Black bars are same data as Figure 1C. Dots represent distributions for individual retinas. mean IEI/ROI in WT and β2-nAChR-KO retinas in control conditions. (D-E) Violin plots summarizing the distribution of mean IEI/ROI (D) and event area (E) across experimental conditions. n = 4 retinas, four mice (WT), n = 5 retinas, five mice (β2-nAChR-KO). Black bar = mean; red bar = median. **p<0.01, *p<0.05. One-way ANOVA, followed by Tukey–Kramer post hoc test . (F) Summary plot of wave speeds. WT are the same data as Figure 1E. Red dot and line = mean and standard deviation, respectively. n = 30 waves, three retinas, three mice per condition p=1.19e–6. Unpaired t-test.

Figure 3—source data 1

Inter-transient-interval, wave area, and wave speed: WT vs. β2-nAChRKO.

https://cdn.elifesciences.org/articles/81983/elife-81983-fig3-data1-v3.xlsx
Video 3
Stage 1 waves in β2-nAChR-Kos.

Calcium imaging of stage 1 waves in retinas isolated from β2-nAChR-KO mice using macroscope. Total field of view is 4.7 mm × 4.7 mm. Frame rate 10 Hz. Total length of movie represents 1 min of recording.

Finally, we wanted to test whether the properties of stage 1 waves in WT or β2-nAChR-KO were influenced by light activation of ipRGCs via the 474 nm imaging light used to excite the calcium dye on the macroscope. ipRGCs have been shown to be responsive to 476 nm light during both embryonic and postnatal development (Emanuel and Do, 2015; Verweij et al., 2019). Previous studies have shown a light-dependent increase in the wave frequency of stage 2 waves in the β2-nAChR-KO during the first postnatal week (Kirkby and Feller, 2013). This light modulation of the spatiotemporal properties of β2-nAChR-KO during the first postnatal week depends on ipRGC melanopsin expression, as well as an increase in gap junction conductance between ipRGCs and other RGCs (Arroyo et al., 2016; Kirkby and Feller, 2013). However, we found no difference in the frequency of waves recorded using a two-photon microscope (based on 920 nm illumination) to those recorded on the macroscope and in either the WT or β2 nAChR-KO embryonic retinas (Figure 3 and Figure 3—figure supplement 1). Hence, we conclude that light activation of ipRGCs does not significantly influence the spatiotemporal properties of stage 1 retinal waves in either WT or the β2 nAChR-KO retinas. These results are consistent with the fact that light stimulation of the retina does not modulate the frequency of stage 2 waves and only begins to do so when conventional photoreceptors come online during stage 3 waves (Tiriac et al., 2018, Renna et al., 2011).

ipRGCs participate in stage 1 waves but their density and distribution are not altered in the β2-nAChRs-KO

RGCs undergo a period of dramatic cell death during the first two postnatal weeks of development, the majority occurring during the first postnatal week (Abed et al., 2022; Braunger et al., 2014). Whether this cell death process is regulated by retinal waves is unknown. We looked specifically at intrinsically photosensitive ganglion cells (ipRGCs) for several reasons. First, ipRGCs have completed proliferation (Lucas and Schmidt, 2019; McNeill et al., 2011) and appear to be fully differentiated by E16 (Shekhar et al., 2022; Whitney et al., 2022), the onset of stage 1 waves. ipRGCs undergo a period of dramatic cell death during the first two postnatal weeks of development, the majority occurring during the first postnatal week. Prevention of cell death profoundly disrupts several important developmental processes in the retina – including spacing of ipRGC somas as well as rod- and cone-mediated circadian entrainment through the activation of ipRGCs (Chen et al., 2013). However, the exact mechanism regulating ipRGC cell death is unknown. Here we assessed the impact of disrupting stage 1 and 2 waves on the density and distribution of ipRGCs.

We first set out to determine whether ipRGCs participate in stage 1 waves. To do this, we conducted two-photon calcium imaging of RGCs in the ganglion cell layer (GCL) of retinas isolated from Opn4Cre/+;Ai9 E16–18 mice (Ecker et al., 2010), which express tdTomato in ipRGCs (Figure 4A) enabling us to assess the differential participation of RGCs and ipRGCs during stage 1 waves (Figure 4B). On average, both RGCs and ipRGCs participated in most waves with no significant differences between the two groups (Figure 4C; average ± SD: RGCs 78.78 ± 21.48%; ipRGCs 84.64 ± 16.86%). We also found no significant differences in the amplitude of the calcium response that RGCs and ipRGCs exhibit in response to stage 1 waves (Figure 4D). Hence, ipRGCs are depolarized by stage 1 waves similarly to stage 2 waves (Caval-Holme et al., 2019; Caval-Holme et al., 2022; Chew et al., 2017).

Stage 1 waves robustly recruit intrinsically photosensitive ganglion cells (ipRGCs) and the general retinal ganglion cell (RGC) population.

(A) Top: example field of view (FOV) where pixel intensity was averaged across all frames (712) to get representative image of Cal 520 bath-loaded RGCs. Pink dots correspond to four example RGCs and blue dots correspond to four example ipRGCs Bottom: same example FOV from the top image, with tdTom signal averaged across all frames. Blue dots the same as those in the top image. (B) Top: traces of the four example RGCs marked by the pink circles in (A). Bottom: traces of four example ipRGCs marked by the blue circles in (A). (C) Violin plot of percentage of waves each cell participated in. n = 20 retinas (13 mice); p=0.09. Unpaired t-test. (D) Violin plot of mean event amplitude/cell/FOV. n = 20 retinas (13 mice); p=0.06. Unpaired t-test. Black bar = mean; red bar = median.

Figure 4—source data 1

Intrinsically photosensitive ganglion cell (ipRGC) participation and mean response amplitude.

https://cdn.elifesciences.org/articles/81983/elife-81983-fig4-data1-v3.xlsx

Since β2-nAChR-KO retinas exhibited reduced retinal activity during both embryonic and early postnatal development, we used this mouse as a model to determine whether normal stage 1 and 2 wave activity is important for regulating the density and distribution of ipRGCs. To this end, we isolated retinas from Opn4Cre/+;Ai9 and Opn4Cre/+;Ai9;(β2-nAChR-/-) at P1 and P7. These retinas express tdTomato in all melanopsin-expressing cells regardless of subtype. Retinas were imaged using the macroscope (Figure 5A and B).

Stage 1 and 2 waves do not contribute to the developmental cell death of intrinsically photosensitive ganglion cells (ipRGCs).

(A) Example epifluorescence images of Opn4Cre/+;Ai9 (WT) retinas at P1 and P7. Scale bars are 500 μm; each field of view is 4.7 × 4.7 mm. (B) Representative 200 × 200 μm2 fields of view for P1 and P7 WT and ß2-nAChR-KO retina. (C) ipRGC densities across different ages and genotypes. P1 WT n = 9 retinas from six mice, P1 ß2-nAChR-KO, 9 retinas, seven mice; P7 WT n = 8 retinas from five mice; P7 ß2-nAChR-KO n = 8 retinas from eight mice. Means represented by blue bars. p=0.012. (D) Cumulative distribution function of nearest-neighbor distances (NND) and regularity index (mean NND/SD) from individual retinas, separated by genotype and age. Means represented by blue bars. p=0.020.

We observed a dramatic decrease in the density of ipRGCs in WT retinas from P1 to P7 (Figure 5B and C; p=1.17 × 10–7), consistent with previous studies of ipRGCs (Chen et al., 2013) and which coincides with peak levels of RGC apoptosis, the primary cause of RGC death during development (Braunger et al., 2014; Young, 1984). We found that β2-nAChR-KO retinas exhibited the same density of ipRGCs at P1 as WT retinas, suggesting that the decrease in activity in stage 1 waves does not regulate ipRGC cell density. Like WTs, β2-nAChR-KO retinas also exhibited a decrease in the density of ipRGCs from P1 to P7 (p=4.99 × 10–11). At P7, we observed a small but significant increase in ipRGC densities at P7 in β2-nAChR-KO mice than in WT mice (568 ± 33 ipRGCs/µm2 in WT vs. 654 ± 78 ipRGCs/µm2 in β2-nAChR-KO, n = 8 retinas in each genotype). We cannot determine whether this small difference is due to the smaller size of retinas in β2-nAChR-KO retinas (Xu et al., 1999) or reflects a true increase in cell number. Overall, these data indicate that the cell death processes that regulate ipRGC density during the first postnatal week persist despite a significant reduction in wave activity.

To determine the impact of this developmental decrease in cell density on the mosaic organization of ipRGCs, we computed the regularity index, which is equal to the average nearest-neighbor distance divided by the standard deviation. A large regularity index is associated with a non-random distribution of somas. Despite the expected increase in nearest-neighbor distance in WT and β2-nAChR-KOs retina between P1 and P7 (Figure 5D, see Table 2), there was only a small decrease in the regularity index. Interestingly, the measured mean regularity indices of 2.8 ± 0.12 (WT P1), 2.9 ± 0.16 (β2-nAChR-KO P1), 2.3 ± 0.13 (WT P7) and 2.6 ± 0.2 (β2-nAChR-KO P1) fall within the range of what would be predicted by a random distribution of cells with soma diameters between 7–10 μm (Keeley et al., 2020). Hence, the decrease in cell density does not appear make the soma organization more ordered. This might be expected since our analysis does not differentiate between ipRGC subtypes, each of which likely forms an independent retinal mosaic. Together these data indicate that although retinal waves provide a robust source of depolarization for embryonic and early postnatal ipRGCs, reducing wave activity does not significantly influence the processes that regulate their density or organization. Whether elimination of all spontaneous activity affects these processes remains to be determined.

Table 2
Nearest-neighbor distances (NND) and regularity indices for intrinsically photosensitive ganglion cells (ipRGCs) labeled in Opn4Cre/+;Ai9 mice.

Data represented as averages ± SD. This data complements quantification in Figure 5.

Median nearest-neighbor distance (µm)Regularity index (NND/SD)Number of retinas
WT P111.59 ± 0.932.83 ± 0.129
β2 P111.51 ± 0.552.89 ± 0.169
WT P722.32 ± 0.582.33 ± 0.138
β2 P720.96 ± 1.402.55 ± 0.208

Discussion

We show that stage 1 retinal waves are a robust source of spontaneous activity in the embryonic retina. Stage 1 waves initiate throughout the retina, propagate over finite regions of varying size, and drive periodic depolarizations of neurons in the immature retinal GCL. In WT embryonic mice, stage 1 waves are abolished in the presence of general nAChR antagonists but persist, albeit with greatly reduced cell participation, in an nAChR antagonist that targets α4β2 containing nAChRs. In the presence of gap junction antagonists, the frequency and participation of cells within waves is also greatly diminished, though some waves still occur. We found that in retinas isolated from embryonic β2-nAChR-KO mice, which exhibit strongly reduced stage 2 wave frequency, waves persisted both in control conditions and in the presence of a general nAChR antagonist, though the area of waves is greatly diminished compared to WT. However, similar to what we observed in WT mice, the frequency of stage 1 waves in β2-nAChR-KO mice was greatly reduced in the presence of a gap junction antagonist. This more striking effect of the gap junction antagonist compared to the general nAChR antagonist indicates that the electrical synapses are sufficient to generate waves in β2-nAChR-KO mice. Finally, we showed that ipRGCs are depolarized by stage 1 waves, but that the decrease in density of ipRGCs across early postnatal development was unaffected in the β2-nAChR-KO mouse, indicating that the proliferation and cell death processes that influence ipRGC density are not dependent on normal patterns of wave activity.

Distinctions and similarities between stage 1 and stage 2 waves

Our findings support a model in which there are both key differences and some similarities between stage 1 and 2 waves. The distribution of ITI is peaked at roughly 80 s for stage 1 waves that is a slightly longer interval than the peak interval reported for stage 2 (compare Figure 1C here to Figure 7D in Ford et al., 2012). Stage 1 and 2 waves also propagate at similar speeds. Key differences are that spontaneous activity between E16 and E18 included small non-propagating transients and propagating waves of a broad range of sizes. In contrast, after P1, there are no small non-propagating transients, and stage 2 waves mostly propagate over large areas of the retina, as observed both in vitro (Feller et al., 1997; Hilgen et al., 2017) and in vivo (Ackman et al., 2012). Note the spatiotemporal properties of stage 2 waves also vary dramatically across the developmental period spanning P1–P10 (Ge et al., 2021; Hilgen et al., 2017; Maccione et al., 2014) and therefore we are making these comparisons to the earlier stage 2 waves.

The similarity in frequency of waves and propagation speed indicates that the mechanisms responsible for initiating and propagating large waves are similar for stage 1 and 2 waves. Indeed, stage 1 and 2 waves share a dependence on nAChR signaling. Starburst amacrine cells (SACs) are the sole source of ACh in the retina. During the period of development concurrent with stage 1 waves, SACs begin to migrate toward the INL and send projections to the IPL via homotypic contacts (Ray et al., 2018). Also during this period, SACs start to express choline acetyltransferase at E17 in the rat retina (Kim et al., 2000), equivalent to E15.5 in mice (Schneider and Norton, 1979), and show a response to nicotine in the fetal rabbit retina between E20 and 27 (Wong, 1995), corresponding to stage 1 and 2 waves in rabbit (Syed et al., 2004). Together these studies support the idea that during embryonic development, SACs are not only releasing ACh but also forming a cholinergic network, similar to that of stage 2 wave propagation (Ford and Feller, 2012). Here, we have shown that stage 1 waves fail to initiate in the presence of the general nAChR antagonists, hexamethonium (Hex) and epibatidine (EPB). This suggests that spontaneous depolarization of SACs is important for wave initiation and that the cholinergic circuits that mediates waves start earlier than expected.

One key difference in ACh signaling between stage 1 and 2 waves is their sensitivity to the specific nAChR antagonist, DHβE. Though DHβE reduced the number of stage 1 waves, many smaller calcium transients persisted. In contrast, DHβE is a potent blocker of all activity during stage 2 waves (Ford et al., 2012). DHβE is an nAChR antagonist with a greater affinity for nAChRs containing α4 and β2 subunits in heterologous systems (Ho et al., 2020; Papke et al., 2010). The fact that all stage 1 wave activity is blocked by Hex and EPB suggests that different subunit combinations of nAChRs, on both SACs and RGCs, are independently contributing to either the initiation or propagation of stage 1 waves.

The different sensitivity to nAChR antagonists between stage 1 and 2 waves is highlighted in the patterns of retinal waves in β2-nAChR-KO mice. β2-nAChR-KO mice have significantly reduced stage 2 cholinergic waves (Bansal et al., 2000; Burbridge et al., 2014; Xu et al., 2015; Xu et al., 2016) and as such have served as the canonical model for studying the role of stage 2 cholinergic waves in eye-specific segregation, retinotopic maps, retinal and collicular direction selectivity, and in the optokinetic reflex (Arroyo and Feller, 2016; Grubb et al., 2003; Thompson et al., 2017; Tiriac et al., 2022; Wang et al., 2009). In contrast, stage 1 waves in the β2-nAChR-KO mice persist, albeit they propagate with slower speed and cover smaller areas of the retina (Figure 3). Despite this difference in spatiotemporal properties, both stage 1 and 2 waves that persist in the β2-nAChR-KO mice are blocked by gap junction receptor antagonist rather than blockers of fast neurotransmitter receptors (Kirkby and Feller, 2013). Recent evidence suggests that embryonic activity influences many aspects of visual system development including eye specific segregation and retinotopic mapping of retinal projections to the brain (Guillamón-Vivancos et al., 2022), as well as cortical innervation of the visual thalamic nucleus (Moreno-Juan et al., 2023). Whether the various visual system phenotypes observed in β2-nAChR-KO mice can be attributed in part to reduced stage 1 waves remains to be determined.

In our hands, global blockade or desensitization of nAChRs completely abolished stage 1 waves. This result appears to conflict with previous studies in mice showing that application of curare, a competitive antagonist for nAChRs, preserves small non-propagating transients (Bansal et al., 2000). One possibility is that in contrast to hexamethonium, curare has mixed affinity for neuronal nAChRs. A second is that curare was acting via other neurotransmitter receptors where it has some cross-reactivity not shared by the receptor antagonists that we used (Spirova et al., 2019; Wotring and Yoon, 1995).

In addition to the ACh signaling, gap junctions also play a role in mediating stage 1 retinal waves. Here, we use the gap junction antagonist meclofenamic acid (MFA), which was previously shown to reversibly block junctional conductance (Veruki and Hartveit, 2009), dye coupling (Pan et al., 2007), and spikelets between developing RGCs (Caval-Holme et al., 2019). Application of MFA led to a significant reduction in both the frequency and size of stage 1 waves. The results we observed in MFA are consistent with the pharmacological studies of stage 1 waves in other species: stage 1 waves in rabbit are insensitive to nAChR antagonists (Syed et al., 2004), as are early waves in developing chick retina (Catsicas et al., 1998). Note that waves in these species were also sensitive to antagonists of various metabotropic receptors, indicating that neurotransmitters are still important for propagating waves in these systems. However, there is the important caveat that MFA can also have some off-target effects that might impact wave propagation (Kuo et al., 2016). Previously we showed that in P6-7 retina MFA did not alter depolarization-induced calcium transients and did not reduce the excitability of RGCs. Here, we show that short applications (10 min) of MFA increased input resistance of RGCs and did not appear to block compound EPSCs associated with waves or voltage-gated sodium and potassium channels. Future experiments with independent measures of gap junction coupling – such as tracer coupling and potentially knockout of connexin proteins – are necessary to have a more complete understanding of how electrical synapses contribute to wave propagation properties.

We also observed a difference between stage 1 and 2 waves in β2-nAChR-KO retinas. Embryonic β2-nAChR-KO retinas still exhibited stage 1 waves but at reduced frequency and size when compared to WT retinas. Though the spatiotemporal properties of stage 2 waves recorded in vitro for P1-P8 β2-nAChR-KO retinas are highly dependent on recording conditions with results ranging from sparse activity to high frequency (Bansal et al., 2000; Stafford et al., 2009; Xu et al., 2016), in vivo waves in β2-nAChR-KO are infrequent and weakly depolarizing (Burbridge et al., 2014). The activity that persists in β2-nAChR-KO retinas both embryonically and postnatally is resistant to nAChR antagonists, as was observed previously (Bansal et al., 2000). Rather the remaining activity in β2-nAChR-KO retinas is completely blocked by MFA, thereby suggesting that stage 1 and 2 waves in β2-nAChR-KO retinas rely solely on gap junctions (Kirkby and Feller, 2013).

We propose a model for stage 1 waves consistent with these observations (Figure 6). Stage 1 wave initiation is dependent on the spontaneous depolarization of SACs, but wave propagation is dependent on both gap junction coupling and nAChR activation. Hence, stage 1 wave initiation is similar to stage 2 waves where the spontaneous depolarization of a SAC activates neighboring SACs, leading to the volumetric release of ACh responsible for wave propagation both in the INL and GCL (Ford et al., 2012). The model for stage 1 wave propagation is also similar to stage 2 wave propagation, though a complete description of how gap junction coupling is mediating stage 2 wave propagation is needed. The pharmacological and genetic block/ablation of gap junctions have yet to reveal a phenotype (Blankenship et al., 2011; Caval-Holme et al., 2019; Kirkby and Feller, 2013; Singer et al., 2001; Torborg et al., 2005). Interestingly, pure gap junction-mediated waves, such as those we observed in the β2-nAChR KO mouse, are considerably slower than waves in WT retina. A recent computational model of gap junction-mediated stage 1 waves, where waves are initiated by RGCs undergoing rare and random depolarizations that propagate entirely via electrical synapses, argues that the speed of propagation is limited by the slow rate at which the junctional currents charge up the membrane capacitance of neighboring RGCs (Kähne et al., 2019). Hence, the faster speed of waves mediated by a combination of nAChRs and gap junctions indicates that diffuse release of ACh leads to faster propagation than electrical synapses alone.

Working model of stage 1 wave initiation and propagation.

Wave initiation set by spontaneously depolarizing cholinergic amacrine cell/SAC, outlined in black, which release ACh and depolarizes neighboring cells, leading to the volumetric release of ACh. Retinal ganglion cells (RGCs) depolarized by nAChR activation (orange) via SAC-induced ACh release. Wave propagation set by RGC depolarization via gap junctional currents (orange) and volumetric ACh release.

Interactions between stage 1 waves and ipRGCs

Early intrinsic light responses of ipRGCs have been implicated in several developmental events (Aranda and Schmidt, 2021), including retinal vascularization (Rao et al., 2013), maturation of circadian circuits (McNeill et al., 2011), and the maturation of the lens to prevent myopia (Chakraborty et al., 2022). Here, we report that ipRGCs are robustly activated by stage 1 waves, similar to our observations that ipRGCs participate in stage 2 waves (Arroyo et al., 2016; Caval-Holme et al., 2022; Kirkby and Feller, 2013). Thus, it is possible that depolarization via stage 1 waves may contribute to some of these ipRGC-dependent developmental processes.

To begin to explore to role of stage 1 and 2 waves in ipRGC development, we monitored the impact of chronically altered waves on the distribution and density of ipRGCs across the retina. Notably, ipRGCs undergo extensive apoptotic cell death, with the peak of apoptosis occurring between P2–P4 (Chen et al., 2013). Prevention of apoptosis during this developmental period doubles the density of ipRGCs and dramatically increases the clumping of M1-ipRGC somas. In some systems, correlated network activity has been implicated in cell proliferation and cell death. For example, retinal wave activity promotes neurite outgrowth and potentially survival among RGCs (Goldberg et al., 2002). Additionally, in the developing primary somatosensory and motor cortices, higher levels of spontaneous electrical activity were shown to have a neuroprotective effect (Blanquie et al., 2017b). However, here we show in the β2-nAChR-KO mouse, which has significantly diminished stage 1 and 2 retinal waves, the normal developmental loss of ipRGCs between P1 and P7 is maintained. Thus, stage 1 and 2 retinal waves are not required for ipRGC apoptosis; however, it is possible that the residual wave activity in β2-nAChR-KO is sufficient to activate pro-survival pathways. A deeper understanding of how spontaneous activity modulates RGC survival pathways (e.g., Ahmed et al., 2022) is warranted.

Methods

Animals

All animal procedures were approved by the UC Berkeley Institutional Animal Care and Use Committee and conformed to the NIH Guide for the Care and Use of Laboratory Animals (AUP-2015-10-8080-2), the Public Health Service Policy, and the SFN Policy on the Use of Animals in Neuroscience Research. For our calcium dye-based calcium imaging, Opn4Cre/+;Ai9 mice were generated by crossing mice Opn4Cre/+ B6.Cg-Gt(ROSA)26Sortm9(CAG-tdTomato) Hze/J mice (Ai9) mice (stock # 007909, Jackson Laboratory, Bar Harbor, ME) to the Opn4Cre/+ reporter mouse (T. Schmidt, Northwestern University, Evanston, IL). For our GCaMP6s-based calcium imaging, we generated Vglut2;GCaMP6s mice by crossing B6J.129S6(FVB)-Slc17a6tm2(cre)Lowl/MwarJ mice (stock # 028866) to B6J.Cg-Gt (ROSA)26Sortm96(CAG-GCaMP6s)Hze/MwarJ mice (stock # 028863). ipRGC density measurements were conducted on P1–P7 mice of either sex using Opn4Cre/+;Ai9.

To obtain mice that were precisely at the correct embryonic age, we set up timed pregnancies and checked vaginal plugs every morning for 4 days after the animals were paired. This approach led to an uncertainty of age of ± 1 day. Hence, we grouped data across E16–18. Since we observed more variance within litters than across litters, we assume age was not a determining factor in our findings. We used the β2-nAChR-KO mouse line in which the β subunit of the nicotinic acetylcholine receptor is knocked out as a genetic model in which cholinergic retinal spontaneous activity is disrupted. For experiments regarding the influence of spontaneous retinal activity on the distribution of ipRGCs across the retina, we used β2-/-::Opn4Cre/+;Ai9 (β2-nAChR-KO) mice, generated by crossing β2-/- (A. Beaudet, Baylor University, Waco, TX) mice to Opn4Cre/+;Ai9 mice to label all melanopsin-expressing cells. All mouse lines are maintained on a C57BL/6 genetic background. All animals used for two-photon calcium imaging experiments and immunohistochemistry were housed in 12 hr day/night cycle rooms.

Retinal preparation

On the day of the experiment, pregnant dams were deeply anesthetized via isoflurane inhalation and fetuses were harvested via a cesarean section. tdTomato-positive fetuses were identified using miner goggles (Biological Laboratory Equipment Services and Maintenance Ltd., model: GFsP-5). Fetuses were kept alive in 50 ml Falcon tubes filled with oxygenated (95% O2 5% CO2) ACSF (in mM, 119 NaCl, 2.5 KCl, 1.3 MgCl2, 1 K2HPO4, 26.2 NaHCO3, 11 D-glucose, and 2.5 CaCl2). Fetuses were then euthanized sequentially by decapitation. Eyes were immediately enucleated and retinas were dissected at room temperature in oxygenated ACSF under a dissecting microscope. Isolated retinas were mounted whole over a 1–2 mm2 hole in nitrocellulose filter paper (Millipore) with the photoreceptor layer side down and transferred to the recording chamber of a two-photon microscope for imaging. The whole-mount retinas were continuously perfused (3 ml/min) with oxygenated ACSF warmed to 32–34°C by a regulated inline heater (TC-344B, Warner Instruments) for the duration of the experiment. Additional retina pieces were kept in the dark at room temperature in ACSF bubbled with 95% O2, 5%CO2 until use (maximum 8 hr).

For the calcium imaging experiments, retinas were bath loaded with the calcium indicator Cal 520 AM (AAT Bioquest) for 1–2 hr at 32°C.

Two-photon calcium imaging

Two-photon fluorescence measurements were obtained with a modified movable objective microscope (MOM) (Sutter instruments, Novator, CA) and made using an Olympus ×60, 1.00 NA, LUMPlanFLN objective (Olympus America, Melville, NY) for single-cell resolution imaging (field of view, FOV: 203 × 203 µm) or a Nikon ×16, 0.80 NA, N16XLWD-PF objective (Nikon, Tokyo, Japan) for large FOV (850 × 850 um) imaging. Two-photon excitation was evoked with an ultrafast pulsed laser (Chameleon Ultra II; Coherent) tuned to 920 nm to image Cal520, GCaMP6s, and tdTomato. Laser power was set between 6.5 and 12 mW for imaging of Cal520 and tdTomato expression. The microscope system was controlled by the ScanImage software (https://www.scanimage.org/). Scan parameters were [pixels/line × lines/frame (frame rate in Hz)]: [256 × 256 (1.48 Hz)], at 2 ms/line. This MOM was equipped with a through-the-objective light stimulation and two detection channels for fluorescence imaging.

Epifluorescent macroscope calcium imaging

Epifluorescent calcium imaging were obtained on a custom-built macroscope with an Olympus XLFLUOR4X/340 4× 0.28 NA objective, a Teledyne Kinetix camera. Collectively, this macroscope has 4.7 mm × 4.7 mm FOV, and 1.5 µm/pixel. All movies were taken at a 10 Hz frame rate and pixels were binned 4 × 4 bringing the resolution down to 5.9 µm/pixel, still maintaining single-cell resolution. Cal520 and GCaMP6s excitation was evoked with a 474 nm LED. A full description and building instructions can be found at https://github.com/Llamero/DIY_Epifluorescence_Macroscope, (Smith, 2022).

Initiation site measurements

Macroscope recordings of stage 1 events were used for the manual detection of initiation sites. Small non-propagating events were identified as local regions of correlated calcium activity with fixed areas and no wave fronts. For this analysis only, we separated waves into small and large events to see whether there were differences in initiation sites. Small propagating waves were identified as regions of correlated calcium activity with no fixed areas and with wave fronts covering up to 25% of the retina. Large propagating waves were identified as regions of correlated calcium activity with no fixed areas and with wave fronts covering up to 90% of the retina.

Pharmacology

We blocked gap junctions via application of the gap junction blocker MFA (50 µM, Sigma-Aldrich). We blocked the nicotinic acetylcholine signaling pathway via application of the broad nicotinic receptor antagonists hexamethonium (Hex, 100 µM, Sigma-Aldrich) and epibatidine (EPB, 10 nM, Sigma-Aldrich) as well as the specific antagonist dihydro-ß-erythroidine hydrobromide (DHßE, 8 µM, Sigma-Aldrich).

The following procedure was used for all pharmacology experiments: We recorded baseline activity in ACSF for 8 min before pharmacological agents were applied to the perfusion system. We then waited 15–30 min for the agents to take effect before acquiring another 8 min recording session.

To attempt to assay off-target effects of MFA, we used whole-cell voltage-clamp recordings to compare voltage-gated ion channels on RGCs in E16–18 retina but found inconsistent results (Figure 2—figure supplement 1). We associate this high variance with a rapid changing complement of ion channels during development and the quick washout of these conductances during whole-cell recordings.

Image analysis of population calcium imaging movies

Movies were preprocessed for motion correction using a MATLAB code from the Flat Iron Institute (https://github.com/flatironinstitute/NoRMCorre; Pnevmatikakis, 2019). The baseline movie frame (F0) was computed by taking the temporal median projection of all the movie frames. Each movie frame (F) was normalized by dividing its difference from the baseline frame (F-F0) by the baseline frame ((F-F0)/F0) to produce a ΔF/F0 movie. For movies taken on the two-photon microscope, circular ROIs were drawn on all cells within the FOV. Additional circular ROIs were drawn for tdTomato+ cells. For movies taken on the macroscope a grid of 10 µm × 10 µm squares, which were spaced 1.5 pixels apart, were drawn over the whole surface of the retina using a custom FIJI macro. The ROIs and the ΔF/F0 movie were then imported into MATLAB for further analysis using custom algorithms. Traces for each FOV and ROI were computed as the mean value of the pixels enclosed by the ROI in each frame of the ΔF/F0 movie.

For transient frequency analysis, event detection was done using the findpeaks function in MATLAB, with the minimum threshold set to greater than at least 10 times the standard deviation of the baseline fluorescence, which corresponds to at least 10% ΔF/F0. The ITI for each ROI was calculated by finding the difference between the frame of each detected transient. This difference was then converted to seconds by multiplying it to the movie’s frame rate.

The area of waves was calculated for all movies taken on the macroscope. To do this, we first z-scored the percent active ROI traces and used the findpeaks function to detect individual waves, with the threshold set to a z-score of 1. After determining the time of waves, we removed any waves that occurred within the first and last 6 s of the movie due to them being edge cases. We then summed the number of active ROIs within 12 s around the wave times. This number was then divided by the total number of ROIs to get the percentage of active ROIs.

To determine whether neurons participate in waves in the two-photon calcium imaging data, we employed the following bootstrapping strategy: we randomly sampled the activity of individual neurons outside of wave times a thousand times to build a baseline of neural activity. We then set a threshold of 95th percentile to statistically determine whether neurons exhibited a greater calcium response during a wave than at rest. We then calculated the percentage of waves each cell participated in and averaged this value for every FOV. Similarly, mean response amplitudes were calculated for each cell and then averaged for each FOV.

For a description of statistics used, please refer to the figure captions.

Analysis of ipRGC densities

To image the density of ipRGCs in fixed retinas, dissected retinas from P1 and P7 mice were fixed in 4% PFA for 30 min. The fixed retinas were subsequently mounted on a slide with vectashield and a cover slip, then imaged within an hour of mounting on the macroscope. For P1 retinas, Z-stacks were acquired by manually turning the focus knob.

We first identified the centroid of each ipRGC. For P7 retinas, where there is more space between cells, we employed the following automatic segmentation. Images were bandpass filtered and somata automatically segmented using the Morpholibj (Legland et al., 2016) classic watershed tool to obtain 8-bit binarized masks. The masks were then processed in MATLAB in order to obtain the centroid locations and nearest-neighbor distances for each soma. For P1 retinas, where there is less space between cells and in fact cells seem to form clusters, automatic segmentation was not possible. Therefore, cells were manually marked using the ImageJ multipoint tool and soma locations exported as a CSV file. For all ages, the centroid data was imported to MATLAB for further analyses.

Density was quantified by dividing the microscope field of view up into 200 ×200 μm squares, manually excluding ones that did not cover the retina or covered partial or damaged parts of the retinas. Out of the resulting squares (~150 per P7 retina, ~100 per P1 retinas), 75 squares (for P7 retinas) and 50 squares (for P1 retinas) were randomly selected and the average density of TdTom+ cells in those squares calculated.

To quantify the nearest-neighbor distances, we used a custom-written MATLAB code that, for each ipRGCs, identified the closest neighbor using the shortest Euclidean distance.

Statistical tests

Details of statistical tests, number of replicates, and p values are indicated in the figures and figure captions. p values < 0.05 were considered significant.

Data availability

We have uploaded the raw data for Figure 5 on Dryad. All other raw imaging data and images are available upon request as they are too large to upload to Dryad. They are residing on our lab server and can be transferred via ftp. Figures 1-4: These data are based on movies acquired from live imaging of activity using a macroscope or a 2-photon scanning microscope. Figure 1: 50 gigabytes Figures 2/4: 180 gigabytes Figure 3: 233 gigabytes Figure 5: These are high resolution fluorescence images acquired from microscope at various z-plane focus planes. 7 gigabytes total (on Dryad). All code and software are available on GitHub: https://github.com/FellerLabCodeShare/Embryonic-Retinal-Waves (copy archived at swh:1:rev:b205760d05ea3c0dcd6aca8ed3f7bcb273f4c5a5). Instructions on how to build a macroscope available at this GitHub repo: https://github.com/Llamero/DIY_Epifluorescence_Macroscope.

The following data sets were generated

References

  1. Book
    1. Feller MB
    2. Kerschensteiner D
    (2013) Retinal waves and their role in visual system development
    In: Rubenstein JLR, Rakic P, editors. Cellular Migration and Formation of Neuronal Connections. Academic Press. pp. 941–953.
    https://doi.org/10.1016/B978-0-12-397266-8.00153-8
    1. Rothe T
    2. Jüttner R
    3. Bähring R
    4. Grantyn R
    (1999)
    Ion conductances related to development of repetitive firing in mouse retinal ganglion neurons in situ
    Journal of Neurobiology 38:191–206.

Decision letter

  1. Fred Rieke
    Reviewing Editor; University of Washington, United States
  2. Marianne E Bronner
    Senior Editor; California Institute of Technology, United States
  3. Fred Rieke
    Reviewer; University of Washington, United States
  4. Guillermina López-Bendito
    Reviewer; Instituto de Neurociencias · UMH-CSIC Campus San Juan · Alicante, Spain, Spain

Our editorial process produces two outputs: (i) public reviews designed to be posted alongside the preprint for the benefit of readers; (ii) feedback on the manuscript for the authors, including requests for revisions, shown below. We also include an acceptance summary that explains what the editors found interesting or important about the work.

Decision letter after peer review:

Thank you for submitting your article "Cellular Mechanisms Underlying Embryonic Retinal Waves" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, including Fred Rieke as Reviewing Editor and Reviewer #1, and the evaluation has been overseen by Marianne Bronner as the Senior Editor. The following individual involved in the review of your submission has agreed to reveal their identity: Guillermina López-Bendito (Reviewer #3).

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission. The reviewers agreed that the topic of the paper was important and appreciated the technical advances that made the work possible. Several important issues came up in review, however, that need to be strengthened before considering the paper further.

(1) Possible off-target effects of gap junction inhibitors should be discussed earlier and in more detail.

(2) The characterization of ipRGC development needs to be better motivated.

(3) Different subtypes of activity during the time period when stage 1 waves are observed need to be defined and distinguished more clearly.

(4) Several data analysis issues need attention.

These and a number of other issues are described in detail in the reviews.

Reviewer #1 (Recommendations for the authors):

Page 7: The description of Stage 1 waves in rabbits talks about "wave frequency" being insensitive to block of fast synaptic transmission, and "blocked" by gap-junction antagonists. It is not clear whether this means all waves, or if "frequency" is included to indicate a subset of waves or properties. Please clarify!

Page 7, bottom: "Waves that persisted … were smaller and had comparable amplitudes …" – clarify that this means smaller in spatial extent.

Page 8, first sentence: revise – not clear how signaling is comprised of subunits.

Page 8, last full paragraph: this paragraph starts by saying that there is a light-dependent modulation of stage 1 wave frequency, but ends by saying that there is not. Is this a difference between WT and KO? If so that could be clarified.

Page 9, bottom, and page 10 top: Sentence starting "Interestingly …" is confusing. It seems to say that the regularity index is regular, but matches expectations if it was random. Please revise.

Reviewer #2 (Recommendations for the authors):

The Introduction is too long and reads more like a review. I suggest limiting it to specific premise, tested hypothesis, and eliminating paragraphs 2 and 3.

Few technical notes:

In the final Results paragraph the cell body size of 7-10 μm is used to estimate ipRGC mosaics. This is too small. Please consider 20-30um.

Across the text, some statistical values are missing.

In Discussion, please note that while MFA does block GJs, it can also reduce spiking activity by a GJ-independent mechanism.

Reviewer #3 (Recommendations for the authors):

I provide here a long list of points that the authors should address. Despite the length of the list, the points I raised and requested do not involve any fundamental error or flaw in the manuscript but are aimed to improve the analysis, organization, and presentation of the dataset.

– It has been shown previously that stage 2 waves span from P1-P10. In this manuscript, the authors nicely set the focus on the embryonic phase of stage 1. The properties of both wave patterns are similar in many aspects (propagation speed, IEIs, and to some extent cholinergic). Thus, it is possible to think that the broad features extracted by the authors from stage 1 retinal waves represent the slow maturation of what we know as stage 2 waves rather than a preceding and complete separate class of retinal waves. Could this be the case? Could the authors speculate in the discussion on what happens in the transition between stages 1 and 2, at P0-P1?

– The authors pool together data from E16, E17, and E18 retinas. Are there any differences in the properties of the waves across time? The authors should at least show this data separately in the supplementary materials to justify pooling the data.

– It seems like the authors recognize the gap-junction component to a lesser extent and stress the cholinergic component. As one of the goals of the paper is to characterize the so-far-unknown stage 1 retinal waves, considered to be mainly gap-junction mediated in the literature, the reader expects to see a larger discussion on this issue and a comparison to the published literature.

– The first part of the Results could be better organized. It is not clear to me if the authors try to explain the relevant methods of analysis, illustrate the different types of activity they found (showing individual examples), or describe activity showing population data. In the second paragraph, the authors describe how they acquired the data, this is fine. In the third paragraph, the authors explain how they analyzed the movies and some properties of the transients detected per ROI and how ROIs participate in waves (although there is not, beforehand, a clear definition of what a wave is). Here, some clarity is needed when the authors use the terms "transients", "events", and "waves". Up to this point, the authors refer to figure 1 where they show data from sample experiments. But, in the last paragraph of this section, they move to population data (and jump suddenly to panels of figure 3). Moreover, they focused on large propagating events, though it seems that sometimes they refer to all the propagating events, and present a summary in Table 1, where surprisingly small waves appeared for the first time without being mentioned in the text. In sum, it would be helpful for the reader if the authors reorganize the text of this section, try to set a clear scope for it, and present the data as smoothly and clearly as possible.

– Are waves detected by looking at the "% Active" plots, by visual inspection of the movies or by an algorithm? In the first case, what happens if there are two or more waves running simultaneously? It would help the reader if the authors explain how waves were detected. If wave detection is manual or visual, why the authors did not use automatic detection? In figure S2, it seems that there are a lot of waves per experiment, too many to be quantified manually or by eye.

– I suggest modifying the title to better fit the data. In the first half of the manuscript, the authors investigate the spatiotemporal properties and pharmacological profile of embryonic retinal waves. Using this information, the authors propose a model for the initiation and propagation of stage 1 waves (figure 6), but none of the cell types of the model were directly assessed in the manuscript. Regarding the ipRGCs, the authors nicely show that these cells participate in the network activity at early developmental stages; however, the data does not show the mechanism related to ipRGCs. Strictly, and in my opinion, the cellular mechanisms are not really established here.

– Why is the frequency of waves in the Hex control (Figure 2B) half the frequency of the other controls?

– The authors should include the mean duration of the waves/events in the description of the different types of activity.

– The authors claim that stage 1 retinal waves persist in WT retinas treated with DHβE. It is true that some activity remains upon DhβE, but, to me, the raster plot looks different than the WT (figure 2D). The statistical test yields no significant differences, but the p-value is amazingly borderline p = 0.05. This means that more experiments are needed to raise robust conclusions. Moreover, since DhβE (figure 2E) reduce the area of the events from 90% to 10% … is it still possible to call them stage 1 retinal waves? In sum, when the authors claim that waves persist, it suggests that normal stage 1 waves were observed in the DhβE-treated retinas; however, what remains looks like something different.

– As stage 1 waves are also affected in β2-nAChR KO mice, which is the "standard" model that has been widely used to assess the function of stage 2 waves, this is a good opportunity for the authors to discuss to what extent the conclusions raised using this mouse in the previous literature needs to be re-interpreted.

Additional comments (Introduction is page 1)

– The term "events" is used both to indicate the activation of an ROI and the correlated activation of multiple ROIs (as in waves). It would be easier for the reader if the authors differentiate them. Examples of the first case can be found in the third paragraph of the results: (i) "… traces for each ROI were rasterized based on an event detection algorithm …" or (ii) "We first computed the time between spontaneous events by measuring the inter-event interval (IEI) for each ROI and …". Examples of the second case can be found in the fourth paragraph: (i) "In addition to propagating events …" or (ii) "These events were like those described previously in (Bansal et al., 2000)". In addition, this disambiguation will clarify what they quantify in the "Events/min" plots. In sum, the authors should somehow clearly differentiate the activation of single ROIs from population activity. Perhaps, "transients" is a good option for the activation of single ROIs. It is not clear from the context which class of events is referred to by "These".

– Why are waves divided into small and large events? This subdivision is not used in any of the quantifications that follow figure 1. In the same line, the authors adequately classified the events in propagating and non-propagating but we don`t know what happens with the latter in the different experimental conditions analyzed. It would be interesting to know whether the properties of non-propagating events are changed or remain the same upon pharmacological or genetic manipulations.

I think that the description of how to classify large waves is not accurate. In Methods: "Large propagating waves were identified as regions of correlated calcium activity with no fixed areas and with wavefronts covering up to 90% of the retina". I guess that the authors meant wavefronts covering areas larger than 25% of the retina, which is the upper limit for small waves. In addition, the authors classify the events as fixed or non-fixed. However, it is not clear to the reader what "fixed" means.

– The number of asterisks used to indicate statistical significance is not consistent (see Figures 2 and 3, for example).

– The algorithm that detects calcium transients within ROIs should be described or referenced.

– Why does the threshold for detection vary from 15% to 50% depending on the experiment? In addition, these thresholds seem very high. Why not use n-times as the SD of the baseline? Is it related to the signal-to-noise ratio?

– I have some comments that may help to refine the Abstract. In lines 2-3, the authors should mention the species. In line 5, as far as I know, stage 1 waves are also observed at least at P0-P1, it is thus not completely correct to use "embryonic retinal waves" as a synonym for stage 1 waves. Perhaps "the earliest pattern of retinal waves" will do it. In line 7, it is not clear to me what the authors mean by "finite". And finally, by reading only the Abstract, it is not evident which could be the relationship between stage 1 waves and the distribution of ipRGCs. Perhaps, the authors should briefly mention here what normally happens to this subpopulation during perinatal development.

– Page 1, line 7. Typo: vision instead of "vison". Page 2, line 26. Typo: characterized instead of "characterize".

– Page 2, lines 29-30. The details of FOV and resolution should go to Methods. The authors emphasize here that the field of view could cover the whole retina while preserving cellular (not subcellular) resolution, this is clear even without the details, increasing the readability of the paragraph. Please, check that "FOV" is spelled out before being abbreviated in the main text.

– Do the authors think that "complex" is a good descriptor for the spatiotemporal properties of stage 1 waves (as stated in the first subheading of Results)?

– I think that the initial paragraph of the first part of Results, about the cell types in the developing retina, does not have a clear link to the following paragraphs of the section. Perhaps, this information fits better in the Introduction. Further, I believe that Bansal and colleagues is not a convenient reference for the last sentence where gap junctions are proposed to be a primary substrate of stage 1 waves. There is only one mention of gap junctions in this publication. The authors may add here the Wong and colleagues (1998).

– Page 3, line 5. I think that it is more appropriate to give the average area of these events in square microns rather than in the number of ROIs occupied by them.

– Page 3, line 23. "E16-18 mouse pups". Better replace mouse pups with "mice" or "mouse embryos".

– Page 4, line 11. Figure 3D is not the correct panel for these data.

– Page 5, line 22. "We observed a dramatic decrease in the density of ipRGCs in WT retinas from P1 to P7 (Figures 5B, C), …". Although differences are evident, this statement needs statistical analysis.

– The authors claim that IEIs are longer in beta2-nAChR-KO than in WT retinas (figure 3E). A statistical analysis should be provided here that directly compares these two populations (Black and pink violins in the panel).

– In the caption of figure 1B, I guess that "an E17" means "an E17 retina".

– In the caption of figure 2, p-values do not correspond to the panels.

– It is convenient that the authors add the maximum number of ROIs represented to the X-axis of raster plots (for instance, 1 at the bottom and say 500 at the top of the axis).

– In Figures 2F and H, the authors show time-series data points; however, there are points without lines in the graphs.

– Rows in Table 1 are individual experiments. Please, add this information to the Table.

– In the speed quantification, the authors explain that the first point of the wavefront is selected at random. I understand then that the speed value for a wave will change every time that the quantification is repeated (because wavefronts do not seem to propagate homogeneously). Is this correct? Or, are measurements repeated many times for every wave and a mean calculated? Also, when the second point is selected, it is not clear to me what "parallel to the direction of the first point" means. This section will benefit from a better explanation of the quantification method. Is this method standard or comparable to other methods that have been used in previous publications?

– Could the authors state in each figure the number of independent experimental units (N): how many retinas from how many mice (for example, information is incomplete in figure 3 and missing in figure 5)?

– "Cook and Bekker, Physiology, 2009" is not listed in the reference list. "Emanuel and Do, 2015" is not listed in the reference list.

[Editors’ note: further revisions were suggested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "Circuit Mechanisms Underlying Embryonic Retinal Waves" for further consideration by eLife. Your revised article has been evaluated by Marianne Bronner (Senior Editor) and Fred Rieke (Reviewing Editor).

The paper has improved in revision, and most of the original issues have been nicely handled. A few issues remain or have been introduced in the revisions.

MFA. The additional discussion of MFA is helpful though, as noted in the paper, ultimately not entirely conclusive. Did you try to wash out the effects of MFA in any of the experiments described here? Even a partial washout would be helpful. I would consider moving the whole cell recordings on lines 155-158 to the Methods since the variability limited the conclusions that can be drawn from them. It does look like K channel activity decreased systematically in MFA, which might be noted.

The motivation for the measurements in the knockout animals is clearer now. The conclusion that compensation causes gap junctions to contribute more strongly to waves in the knockout mice could be highlighted more (it currently comes in the middle of the paragraph starting on line 180). I would consider adding that conclusion to the abstract or introduction and making it clearer in the results.

The abstract is quite long and detailed. I would suggest removing some of the sentences about the experimental approach to the final paragraph of the introduction, and focusing the abstract more on conclusions to make it more generally and easily accessible.

ipRGC development and waves. Since stage 1 waves are modified but not eliminated in the knockouts, some of the conclusions about waves and ipRGC development seem too strong (e.g. lines 250-251, line 265, lines 283-284). The last sentence of the Discussion strikes me as a balanced way to handle this issue, so some of that could come into the earlier statements.

https://doi.org/10.7554/eLife.81983.sa1

Author response

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission. The reviewers agreed that the topic of the paper was important and appreciated the technical advances that made the work possible. Several important issues came up in review, however, that need to be strengthened before considering the paper further.

(1) Possible off-target effects of gap junction inhibitors should be discussed earlier and in more detail.

We have added additional experiments based on whole cells recordings to address some off target effects of MFA but we do make note of the limitations of these new controls since we observed significant variability of voltage-gated conductances across RGCs at this age as well as the limited ability to maintain stable recordings for the requisite time to have within cell controls for MFA. In the revised manuscript we are more explicit about the caveats of using MFA both in results and in discussion.

(2) The characterization of ipRGC development needs to be better motivated.

We felt this was a good cell type with which to start since the reduction of ipRGC due to cell death is well characterized (see Chen et al., Neuron, 2013) while most other studies generalize across all RGCs. We have clarified this motivation in the revised manuscript.

(3) Different subtypes of activity during the time period when stage 1 waves are observed need to be defined and distinguished more clearly.

In the revisions, we tried to develop our analysis so that we could statistically differentiate between small and large waves. Upon further analysis, we found that wave sizes cover a broad range that is not bimodal. Thus, in this revision, we only distinguish between propagating waves and small non-propagating events (as described in revised text, bottom of page 4).

(4) Several data analysis issues need attention.

We hope to have addressed all of these in the revised text.

Reviewer #1 (Recommendations for the authors):

Page 7: The description of Stage 1 waves in rabbits talks about "wave frequency" being insensitive to block of fast synaptic transmission, and "blocked" by gap-junction antagonists. It is not clear whether this means all waves, or if "frequency" is included to indicate a subset of waves or properties. Please clarify!

We have clarified this terminology on page 5. We reproduce sentences here for the benefit of reviewer.

“Previous work based on epifluorescent calcium imaging experiments performed in embryonic mice has shown that retinal waves, as defined by correlated changes in fluorescence, are reduced in frequency and size by curare, a competitive antagonist for nAChRs (Bansal, 2000). However, in roughly the equivalent developmental period in rabbit, blockade of all fast neurotransmitter receptors, including nAChRs, had no impact on wave frequency (Syed et al., 2004). Rather, waves in rabbit are blocked after the application of 18β-glycyrrhetinic acid, a gap junction antagonist (Syed et al., 2004).”

Page 7, bottom: "Waves that persisted … were smaller and had comparable amplitudes …" – clarify that this means smaller in spatial extent.

We have revised text to read (page 6)

“However, in contrast to Stage 2 wave, some waves persisted in the presence of DhβE, but they recruited fewer neurons and therefore had smaller areas (Figure 2B).”

Page 8, first sentence: revise – not clear how signaling is comprised of subunits.

We have revised the text to read (page 6):

“Our results thus far indicate that both the frequency and area of Stage 1 retinal waves are modulated by the activation of different subtypes of nAChRs as well as gap junction coupling.”

Page 8, last full paragraph: this paragraph starts by saying that there is a light-dependent modulation of stage 1 wave frequency, but ends by saying that there is not. Is this a difference between WT and KO? If so that could be clarified.

We previously reported that spontaneous correlated activity in the β2-nAChR KO during the Stage 2 waves was light sensitive while spontaneous correlated activity in WT mice is not (Kirkby et al., 2013). We found this not to be the case in β2-nAChR KO mice Stage 1 waves. We hope the revised paragraph on page 6 clarifies this point.

Page 9, bottom, and page 10 top: Sentence starting "Interestingly …" is confusing. It seems to say that the regularity index is regular, but matches expectations if it was random. Please revise.

We thank the reviewer for point out this error. We have revised the text to say that the regularity indices we measured are consistent with a random distribution.

Reviewer #2 (Recommendations for the authors):

The Introduction is too long and reads more like a review. I suggest limiting it to specific premise, tested hypothesis, and eliminating paragraphs 2 and 3.

We have streamlined the intro. We do not completely eliminate paragraphs 2 and 3 since they provide background on circuits that mediate waves which we thought might be important for context. However, we did trim greater than 30% of the words.

Few technical notes:

In the final Results paragraph the cell body size of 7-10 μm is used to estimate ipRGC mosaics. This is too small. Please consider 20-30um.

We have measured the somas of ipRGCs at these ages and 10 µm is the best estimate value. It is worth noting that given the analysis in Keely et al., larger somas would lead to an increase in the regulatory index associated with a random distribution. Hence, our conclusion that the organization of ipRGCs we observed in WT and β2-nAChR retinas do not vary from random remains valid.

Across the text, some statistical values are missing.

We have added p-values to all Figure captions.

In Discussion, please note that while MFA does block GJs, it can also reduce spiking activity by a GJ-independent mechanism.

We have added discussion regarding off-target effects of MFA earlier in manuscript in presentation of the results, as suggested by Reviewer #1. Please see pages 5 and 10 in revised manuscript.

Reviewer #3 (Recommendations for the authors):

I provide here a long list of points that the authors should address. Despite the length of the list, the points I raised and requested do not involve any fundamental error or flaw in the manuscript but are aimed to improve the analysis, organization, and presentation of the dataset.

– It has been shown previously that stage 2 waves span from P1-P10. In this manuscript, the authors nicely set the focus on the embryonic phase of stage 1. The properties of both wave patterns are similar in many aspects (propagation speed, IEIs, and to some extent cholinergic). Thus, it is possible to think that the broad features extracted by the authors from stage 1 retinal waves represent the slow maturation of what we know as stage 2 waves rather than a preceding and complete separate class of retinal waves. Could this be the case? Could the authors speculate in the discussion on what happens in the transition between stages 1 and 2, at P0-P1?

We revised the discussion significantly based on these comments. Please see pages 9 -11.

– The authors pool together data from E16, E17, and E18 retinas. Are there any differences in the properties of the waves across time? The authors should at least show this data separately in the supplementary materials to justify pooling the data.

We do not feel confident that we could rely on plug checks to estimate age of embryos better than +/- one day. Hence we have grouped across ages. We have assessed whether any variance could be attributed to mice that came from different litters (as a proxy for potential different ages). However we found more variance within litters than across litters and therefore we assume age does not explain the variance. We have added this to the Methods-Animals section.

– It seems like the authors recognize the gap-junction component to a lesser extent and stress the cholinergic component. As one of the goals of the paper is to characterize the so-far-unknown stage 1 retinal waves, considered to be mainly gap-junction mediated in the literature, the reader expects to see a larger discussion on this issue and a comparison to the published literature.

The dependence on nAChR signaling was more surprising to us given the lack of synaptic structures. In the revised manuscript, we think we have made clearer that both gap junction coupling and nAChR signaling are required for Stage 1 waves. Throughout the paper, we have added more discussion about the differential roles of gap junctions and nAChR on Stage 1 waves.

– The first part of the Results could be better organized. It is not clear to me if the authors try to explain the relevant methods of analysis, illustrate the different types of activity they found (showing individual examples), or describe activity showing population data. In the second paragraph, the authors describe how they acquired the data, this is fine. In the third paragraph, the authors explain how they analyzed the movies and some properties of the transients detected per ROI and how ROIs participate in waves (although there is not, beforehand, a clear definition of what a wave is). Here, some clarity is needed when the authors use the terms "transients", "events", and "waves". Up to this point, the authors refer to figure 1 where they show data from sample experiments. But, in the last paragraph of this section, they move to population data (and jump suddenly to panels of figure 3). Moreover, they focused on large propagating events, though it seems that sometimes they refer to all the propagating events, and present a summary in Table 1, where surprisingly small waves appeared for the first time without being mentioned in the text. In sum, it would be helpful for the reader if the authors reorganize the text of this section, try to set a clear scope for it, and present the data as smoothly and clearly as possible.

We have modified significantly the first part of the results to address these concerns. Specifically, we:

1) Rearranged the presentation of the results so that the flow is better. Figure 1 has been rearranged so that captions are called in order and there is no jump to Figure 3.

2) Increased clarity by using only the two terms “transients” and “waves”, and removing the word “event”. Transients and waves are both clearly defined in revised text page 4.

– Are waves detected by looking at the "% Active" plots, by visual inspection of the movies or by an algorithm? In the first case, what happens if there are two or more waves running simultaneously? It would help the reader if the authors explain how waves were detected. If wave detection is manual or visual, why the authors did not use automatic detection? In figure S2, it seems that there are a lot of waves per experiment, too many to be quantified manually or by eye.

We have added information in the methods that describe how waves are detected. Specifically, the waves are automatically detected using a peakfind algorithm in MATLAB on the % Active traces.

– I suggest modifying the title to better fit the data. In the first half of the manuscript, the authors investigate the spatiotemporal properties and pharmacological profile of embryonic retinal waves. Using this information, the authors propose a model for the initiation and propagation of stage 1 waves (figure 6), but none of the cell types of the model were directly assessed in the manuscript. Regarding the ipRGCs, the authors nicely show that these cells participate in the network activity at early developmental stages; however, the data does not show the mechanism related to ipRGCs. Strictly, and in my opinion, the cellular mechanisms are not really established here.

We modified the title to: Circuit mechanisms underlying embryonic retinal waves

– Why is the frequency of waves in the Hex control (Figure 2B) half the frequency of the other controls?

There was variance across preparations. However, we have conducted additional experiments and found that the frequency of waves within the same range as other experiments. The new data is incorporated into part of Revised Figure 2.

– The authors should include the mean duration of the waves/events in the description of the different types of activity.

Duration is convolved with velocity, area and kinetics of the calcium dye. Therefore we restrict our description in the paper to area and velocity. It is unclear to us why this parameter is needed and so we have not added it.

– The authors claim that stage 1 retinal waves persist in WT retinas treated with DHβE. It is true that some activity remains upon DhβE, but, to me, the raster plot looks different than the WT (figure 2D). The statistical test yields no significant differences, but the p-value is amazingly borderline p = 0.05. This means that more experiments are needed to raise robust conclusions. Moreover, since DhβE (figure 2E) reduce the area of the events from 90% to 10% … is it still possible to call them stage 1 retinal waves? In sum, when the authors claim that waves persist, it suggests that normal stage 1 waves were observed in the DhβE-treated retinas; however, what remains looks like something different.

We clarified how we describe this result (bottom of page 5).

“However, in contrast to Stage 2 wave, some waves persisted in the presence of DhβE, but they recruited fewer neurons and therefore had smaller areas (Figure 2B)”

– As stage 1 waves are also affected in β2-nAChR KO mice, which is the "standard" model that has been widely used to assess the function of stage 2 waves, this is a good opportunity for the authors to discuss to what extent the conclusions raised using this mouse in the previous literature needs to be re-interpreted.

We revised the discussion significantly. We highlight page 9 – the text is reproduced here for reviewers convenience:

“The different sensitivity to nAChR antagonists between Stage 1 and 2 waves is highlighted in the patterns of retinal waves in β2-nAChR-KO mice. β2-nAChR-KO mice have significantly reduced Stage 2 cholinergic waves (Bansal et al., 2000; Burbridge et al., 2014; Xu et al., 2015, 2016) and as such have served as the canonical model for studying the role of Stage 2 cholinergic waves in eye-specific segregation, retinotopic maps, retinal and collicular direction selectivity, and in the optokinetic reflex (Arroyo and Feller, 2016; Grubb et al., 2003; Thompson et al., 2017; Tiriac et al., 2022; Wang et al., 2009). In contrast, Stage 1 waves in the β2-nAChR-KO mice persist, albeit they propagate with slower speed and cover smaller areas of the retina (Figure 3). Despite this difference in spatiotemporal properties, both Stage 1 and Stage 2 waves that persist in the β2-nAChR-KO mice are blocked by gap junction receptor antagonist rather than blockers of fast neurotransmitter receptors (Kirkby and Feller, 2013). Recent evidence suggests that embryonic activity throughout the developing sensory system may influence many aspects of visual systems development (Martini et al., 2021; Moreno-Juan et al., 2022). Whether the various visual system phenotypes observed in β2-nAChR-KO mice can be attributed in part to reduced Stage 1 waves remains to be determined.”

Additional comments (Introduction is page 1)

– The term "events" is used both to indicate the activation of an ROI and the correlated activation of multiple ROIs (as in waves). It would be easier for the reader if the authors differentiate them. Examples of the first case can be found in the third paragraph of the results: (i) "… traces for each ROI were rasterized based on an event detection algorithm …" or (ii) "We first computed the time between spontaneous events by measuring the inter-event interval (IEI) for each ROI and …". Examples of the second case can be found in the fourth paragraph: (i) "In addition to propagating events …" or (ii) "These events were like those described previously in (Bansal et al., 2000)". In addition, this disambiguation will clarify what they quantify in the "Events/min" plots. In sum, the authors should somehow clearly differentiate the activation of single ROIs from population activity. Perhaps, "transients" is a good option for the activation of single ROIs. It is not clear from the context which class of events is referred to by "These".

We clarified the wording of transients, events, and waves and defined these terms on page 4.

– Why are waves divided into small and large events? This subdivision is not used in any of the quantifications that follow figure 1. In the same line, the authors adequately classified the events in propagating and non-propagating but we don`t know what happens with the latter in the different experimental conditions analyzed. It would be interesting to know whether the properties of non-propagating events are changed or remain the same upon pharmacological or genetic manipulations.

I think that the description of how to classify large waves is not accurate. In Methods: "Large propagating waves were identified as regions of correlated calcium activity with no fixed areas and with wavefronts covering up to 90% of the retina". I guess that the authors meant wavefronts covering areas larger than 25% of the retina, which is the upper limit for small waves. In addition, the authors classify the events as fixed or non-fixed. However, it is not clear to the reader what "fixed" means.

We no longer make the distinction between small and large waves because there is a continuous distribution of wave areas. See Figure 1 supplement 1.

– The number of asterisks used to indicate statistical significance is not consistent (see Figures 2 and 3, for example).

We fixed this to make it consistent across figures.

– The algorithm that detects calcium transients within ROIs should be described or referenced.

We added this information to the revised methods.

– Why does the threshold for detection vary from 15% to 50% depending on the experiment? In addition, these thresholds seem very high. Why not use n-times as the SD of the baseline? Is it related to the signal-to-noise ratio?

See new figure 1B. Our signal-to-noise is very good for these recordings, S Thresholds varied with dye that is used. In all cases, thresholds > 10 SD were used.

– I have some comments that may help to refine the Abstract. In lines 2-3, the authors should mention the species. In line 5, as far as I know, stage 1 waves are also observed at least at P0-P1, it is thus not completely correct to use "embryonic retinal waves" as a synonym for stage 1 waves. Perhaps "the earliest pattern of retinal waves" will do it. In line 7, it is not clear to me what the authors mean by "finite". And finally, by reading only the Abstract, it is not evident which could be the relationship between stage 1 waves and the distribution of ipRGCs. Perhaps, the authors should briefly mention here what normally happens to this subpopulation during perinatal development.

We have really avoided P0 recordings here and therefore cannot make much of a comment about the transition between Stage 1 and Stage. But I do think it is fair to call embryonic waves “Stage 1”.

We have eliminated the word “finite”

We have added.. “which undergo a significant amount of cell death” too

– Page 1, line 7. Typo: vision instead of "vison". Page 2, line 26. Typo: characterized instead of "characterize".

These have been corrected.

– Page 2, lines 29-30. The details of FOV and resolution should go to Methods. The authors emphasize here that the field of view could cover the whole retina while preserving cellular (not subcellular) resolution, this is clear even without the details, increasing the readability of the paragraph. Please, check that "FOV" is spelled out before being abbreviated in the main text.

We clarified the text but kept some of this methodology wording in the Results sections so that the distinction between transients and waves is clear.

– Do the authors think that "complex" is a good descriptor for the spatiotemporal properties of stage 1 waves (as stated in the first subheading of Results)?

We have removed the word complex and changed the first subheading to

Macroscope imaging reveals the spatiotemporal properties of Stage 1retinal waves.

– I think that the initial paragraph of the first part of Results, about the cell types in the developing retina, does not have a clear link to the following paragraphs of the section. Perhaps, this information fits better in the Introduction. Further, I believe that Bansal and colleagues is not a convenient reference for the last sentence where gap junctions are proposed to be a primary substrate of stage 1 waves. There is only one mention of gap junctions in this publication. The authors may add here the Wong and colleagues (1998).

Our hope here was to motivate the specific pharmacological manipulations we conducted and to better describe what we mean by “immature state of retinal circuits”. We have modified text to hopefully make that clearer. We added the Wong and colleagues, 2018 reference and removed the Bansal reference.

– Page 3, line 5. I think that it is more appropriate to give the average area of these events in square microns rather than in the number of ROIs occupied by them.

We have made this change.

– Page 3, line 23. "E16-18 mouse pups". Better replace mouse pups with "mice" or "mouse embryos".

We have made this change.

– Page 4, line 11. Figure 3D is not the correct panel for these data.

We have made this correction.

– Page 5, line 22. "We observed a dramatic decrease in the density of ipRGCs in WT retinas from P1 to P7 (Figures 5B, C), …". Although differences are evident, this statement needs statistical analysis.

We have added p values for the change in ipRGC density from P1 to P7 in both WT and β2Kos.

– The authors claim that IEIs are longer in beta2-nAChR-KO than in WT retinas (figure 3E). A statistical analysis should be provided here that directly compares these two populations (Black and pink violins in the panel).

We have added this analysis in revised Figures 3D and E.

– In the caption of figure 1B, I guess that "an E17" means "an E17 retina".

We have modified text accordingly.

– In the caption of figure 2, p-values do not correspond to the panels.

We have added p-values to all figures and made this correction in Figure 2.

– It is convenient that the authors add the maximum number of ROIs represented to the X-axis of raster plots (for instance, 1 at the bottom and say 500 at the top of the axis).

We have modified the figure accordingly.

– In Figures 2F and H, the authors show time-series data points; however, there are points without lines in the graphs.

We have removed these points.

– Rows in Table 1 are individual experiments. Please, add this information to the Table.

We added this information to the caption.

– In the speed quantification, the authors explain that the first point of the wavefront is selected at random. I understand then that the speed value for a wave will change every time that the quantification is repeated (because wavefronts do not seem to propagate homogeneously). Is this correct? Or, are measurements repeated many times for every wave and a mean calculated? Also, when the second point is selected, it is not clear to me what "parallel to the direction of the first point" means. This section will benefit from a better explanation of the quantification method. Is this method standard or comparable to other methods that have been used in previous publications?

We clarified the description of the wave speed measurements. This method was used to give an average measurement to wavefront propagation. We implemented several more complex algorithms based on location motion along all edges of wavefronts but given the complexity of Stage 1 waves, this was not productive. Rather we wanted to capture a more macroscopic assessment of wave propagation to compare between WT and β2-nAChR-KO. This approach led to similar Stage 2 velocities reported in previous publications. To describe the method we have added a new Figure 1 Supplemental Figure 2.

– Could the authors state in each figure the number of independent experimental units (N): how many retinas from how many mice (for example, information is incomplete in figure 3 and missing in figure 5)?

We have now added this to the Figure captions for Figures 3 and 5.

– "Cook and Bekker, Physiology, 2009" is not listed in the reference list. "Emanuel and Do, 2015" is not listed in the reference list.

We have added both of these references.

[Editors’ note: further revisions were suggested prior to acceptance, as described below.]

The paper has improved in revision, and most of the original issues have been nicely handled. A few issues remain or have been introduced in the revisions.

MFA. The additional discussion of MFA is helpful though, as noted in the paper, ultimately not entirely conclusive. Did you try to wash out the effects of MFA in any of the experiments described here? Even a partial washout would be helpful. I would consider moving the whole cell recordings on lines 155-158 to the Methods since the variability limited the conclusions that can be drawn from them. It does look like K channel activity decreased systematically in MFA, which might be noted.

We have made the changes you suggested – moving the results based on whole cell to the methods but leaving the summary in the main text.

For this study, we have not systemically checked for a washout of MFA. We do not see any signs of general health decline with the calcium imaging which manifests itself by a steady increase in baseline of the fluorescence (which in the past we have seen with higher concentrations of MFA). For whole cell recording, in two cells, we saw partial recovery of the voltage-gated potassium conductance after 15 minutes of rinse (after 45 minutes in 50 µM MFA) – given washout issues this seems not useful to add to paper. We also do not see big changes in resting potentials of cells or voltage-gated Na conductances. It is important to note that in the past we have found 50 µM MFA was sufficient to block gap junction coupling as assayed with spikelets (Arroyo et al., 2016, Figure 7; Caval-Holme et al. 2019; Figure 4).

The motivation for the measurements in the knockout animals is clearer now. The conclusion that compensation causes gap junctions to contribute more strongly to waves in the knockout mice could be highlighted more (it currently comes in the middle of the paragraph starting on line 180). I would consider adding that conclusion to the abstract or introduction and making it clearer in the results.

We cannot conclude there was compensation – rather gap junctions are the remaining circuit mechanisms for propagating waves in the knockout mice. We have made modifications to Lines 180 and 280 to clarify this point.

The abstract is quite long and detailed. I would suggest removing some of the sentences about the experimental approach to the final paragraph of the introduction, and focusing the abstract more on conclusions to make it more generally and easily accessible.

We have streamlined the abstract.

ipRGC development and waves. Since stage 1 waves are modified but not eliminated in the knockouts, some of the conclusions about waves and ipRGC development seem too strong (e.g. lines 250-251, line 265, lines 283-284). The last sentence of the Discussion strikes me as a balanced way to handle this issue, so some of that could come into the earlier statements.

page 250-251 – cell death primarily occurs P1 – P7 (Stage 2 waves) and during this time waves are nearly absent in b2-nAChR KOs. Cell differentiation occurs during Stage 1 waves. That said, we have modified the sentence but not weakened the conclusion too much. Specifically, we added the following sentence to address your point:

“Whether elimination of all spontaneous activity affects these processes remains to be determined.”

https://doi.org/10.7554/eLife.81983.sa2

Article and author information

Author details

  1. Christiane Voufo

    Helen Wills Neuroscience Institute, University of California, Berkeley, Berkeley, United States
    Contribution
    Conceptualization, Software, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing – original draft, Writing – review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-5299-0913
  2. Andy Quaen Chen

    Department of Molecular and Cell Biology, University of California, Berkeley, Berkeley, United States
    Contribution
    Conceptualization, Software, Formal analysis, Investigation, Visualization, Methodology, Writing – original draft, Writing – review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-4970-045X
  3. Benjamin E Smith

    School of Optometry, University of California, Berkeley, Berkeley, United States
    Contribution
    Software, Methodology, Writing – review and editing
    Competing interests
    No competing interests declared
  4. Rongshan Yan

    1. Helen Wills Neuroscience Institute, University of California, Berkeley, Berkeley, United States
    2. Department of Molecular and Cell Biology, University of California, Berkeley, Berkeley, United States
    Contribution
    Formal analysis, Investigation
    Competing interests
    No competing interests declared
  5. Marla B Feller

    1. Helen Wills Neuroscience Institute, University of California, Berkeley, Berkeley, United States
    2. Department of Molecular and Cell Biology, University of California, Berkeley, Berkeley, United States
    Contribution
    Conceptualization, Supervision, Funding acquisition, Writing – original draft, Project administration, Writing – review and editing
    For correspondence
    mfeller@berkeley.edu
    Competing interests
    Reviewing editor, eLife
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-9137-5849
  6. Alexandre Tiriac

    1. Helen Wills Neuroscience Institute, University of California, Berkeley, Berkeley, United States
    2. Department of Molecular and Cell Biology, University of California, Berkeley, Berkeley, United States
    Present address
    Biological Sciences and Vanderbilt Brain Institute, Vanderbilt University, Nashville, United States
    Contribution
    Software, Supervision, Investigation, Writing – original draft, Project administration, Writing – review and editing
    For correspondence
    alexandre.tiriac@vanderbilt.edu
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-7966-981X

Funding

National Eye Institute (RO1EY013528)

  • Marla B Feller

National Eye Institute (RO1EY019498)

  • Marla B Feller

National Eye Institute (P30EY003176)

  • Marla B Feller
  • Benjamin E Smith

National Eye Institute (K99EY030909)

  • Alexandre Tiriac

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

All authors supported by NIH RO1EY013528, RO1EY019498, and P30EY003176. AT was supported by K99EY030909. BES was supported by NIH P30EY003176. We thank members of the Feller Lab for their comments on the manuscript.

Ethics

All animal procedures were approved by the UC Berkeley Institutional Animal Care and Use Committee and conformed to the NIH Guide for the Care and Use of Laboratory Animals, the Public Health Service Policy, and the SFN Policy on the Use of Animals in Neuroscience Research.

Senior Editor

  1. Marianne E Bronner, California Institute of Technology, United States

Reviewing Editor

  1. Fred Rieke, University of Washington, United States

Reviewers

  1. Fred Rieke, University of Washington, United States
  2. Guillermina López-Bendito, Instituto de Neurociencias · UMH-CSIC Campus San Juan · Alicante, Spain, Spain

Version history

  1. Received: July 19, 2022
  2. Preprint posted: August 15, 2022 (view preprint)
  3. Accepted: February 13, 2023
  4. Accepted Manuscript published: February 15, 2023 (version 1)
  5. Version of Record published: March 6, 2023 (version 2)
  6. Version of Record updated: March 15, 2023 (version 3)

Copyright

© 2023, Voufo et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 2,939
    Page views
  • 261
    Downloads
  • 2
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Christiane Voufo
  2. Andy Quaen Chen
  3. Benjamin E Smith
  4. Rongshan Yan
  5. Marla B Feller
  6. Alexandre Tiriac
(2023)
Circuit mechanisms underlying embryonic retinal waves
eLife 12:e81983.
https://doi.org/10.7554/eLife.81983

Share this article

https://doi.org/10.7554/eLife.81983

Further reading

    1. Computational and Systems Biology
    2. Neuroscience
    Tony Zhang, Matthew Rosenberg ... Markus Meister
    Research Article

    An animal entering a new environment typically faces three challenges: explore the space for resources, memorize their locations, and navigate towards those targets as needed. Here we propose a neural algorithm that can solve all these problems and operates reliably in diverse and complex environments. At its core, the mechanism makes use of a behavioral module common to all motile animals, namely the ability to follow an odor to its source. We show how the brain can learn to generate internal “virtual odors” that guide the animal to any location of interest. This endotaxis algorithm can be implemented with a simple 3-layer neural circuit using only biologically realistic structures and learning rules. Several neural components of this scheme are found in brains from insects to humans. Nature may have evolved a general mechanism for search and navigation on the ancient backbone of chemotaxis.

    1. Neuroscience
    Frances Skinner
    Insight

    Automatic leveraging of information in a hippocampal neuron database to generate mathematical models should help foster interactions between experimental and computational neuroscientists.