Retinal slice culture setup
As specifically detailed in the methods section, early postnatal mouse eyes were enucleated, placed in warm culture media (DMEM/F12), and the retina was dissected away from the other ocular tissues. While this process was repeated for additional eyes, the previously isolated retinae were maintained as retinal cups submerged in DMEM/F12 and incubated at 37 °C. Once all retinae were isolated, they were then embedded in 6.5% agarose made with DMEM/F12 and 200 μm retinal cross-sections were sliced on a vibratome. Using a paintbrush, the slices were then mounted directly against the glass of a glass bottom dish and a thin layer of 1.5% agarose/media was poured over the retinal slice to maintain it against the glass. Once the agarose solidified, liquid culture media was added to the dish and the retina was ready to be imaged on an inverted microscope (summarized in Fig. 1).
Analysis of cell death and proliferation
To determine whether our slice culture protocol results in significant retinal cell death that would preclude meaningful analysis and interpretation, we performed Zombie Red™ staining followed by confocal imaging of fluorescence (Fig. 2a-h). Cultures of P0 retinae were stained at two time points: immediately after cultures were prepared (0 h in culture) and after 16 h of live confocal imaging (16 h of laser exposure).
Zombie Red™ (ZR) is an amine reactive fluorescent dye that is non-permeant to live cells, but permeant to cells with compromised membranes (usually dead or dying cells). The dye is optimally excited by 561 nm laser light and has a fluorescence emission peak at 642 nm (Biolegend Ca# 423109). At 0 h in culture, clear ZR+ pixels were observed throughout the ganglion cell layer (GCL) (Fig. 2a, c, e, g). ZR labeling in the GCL was entirely expected because, once the optic nerve is severed during retinal dissection, ganglion cell survival is compromised [25, 38]. Further, these data serve as a positive control for the specificity of ZR labeling of dead/dying cells.
To determine whether slice culture conditions impact RPC survival, ZR+ pixels were counted throughout the entire the neuroblastic layer (NBL) of retinal slices at 0 h and 16 h in culture and we found no significant difference (Fig. 2i). Furthermore, when we compared cell death in laser exposed tissue to an adjacent, non-laser exposed section of the same size, we observed no significant difference (Fig. 2j). These data demonstrate that, after 16 h of live imaging, the NBL of retinal slice cultures does not suffer extensive cell death that would hinder interpretation of the INM time lapse movies. Subsequent live imaging of INM, utilizing the Fucci reporters (see below), confirmed this conclusion as we failed to detect significant RPC nuclear fragmentation in any of our time lapse movies (Additional files 6, 7 and 8).
We next performed an assessment of RPC proliferation in our slice cultures and compared that to RPCs in vivo. Specifically, to measure the incidence of S-phase entry, P0 and P1 pups were injected with 5-ethynyl-2′-deoxyuridine (EdU) 6 h before retinae were processed for cryosections and EdU labeling. After 6 h of imaging in culture, P0 retinal slices were incubated with EdU and this was followed by an additional 6 h in culture before EdU labeling of cryosections. As expected, EdU labeling was observed throughout the NBL of the P0 and P1 retinae and the slice cultures (Fig. 2k-m). Quantification of the #EdU+/#DAPI+ pixels revealed no statistically significant decrease in S-phase entry of the slice culture versus either the P0 or the P1 time points (Fig. 2n). These data suggest that our retinal slice imaging protocol does not significantly impact RPC proliferation.
Analysis of interkinetic nuclear migration within retinal slice cultures
As a first test of our protocol, we performed live imaging of RPC INM. INM describes the periodic movement of the nucleus within a cell that occurs in phase with cell cycle progression and is a common feature of pseudostratified epithelia such as the developing retinal neuroepithelium [5, 11, 35]. INM was first described in the neural tube of chicks and pigs [43] and has since been shown to occur in a variety of tissues in species ranging from mammals to sea anemones [15, 29, 34, 40, 46]. While this evolutionary conservation indicates importance for INM in development, its precise requirement during retinal development remains unclear. However, recent studies of the zebrafish retinae have begun to elucidate the cellular and molecular mechanisms driving INM.
Initially, INM was described by an “elevator model” whereby the nuclei of cells during G1 phase undergo smooth basal migration and, upon G2 entry, nuclei migrate apically until division occurs at the apical end of the cell [43, 46] (Fig. 3a). However, previous live imaging studies performed on zebrafish retinae suggest that nuclei of cells during G1 phase migrate stochastically with a slight basal drift and S phase migration is entirely stochastic (Fig. 3b). G2 INM is indeed directed apically in the zebrafish retina prior to mitosis in the ventricular zone [22, 34]. Based on these findings, we sought to determine whether the same pattern of RPC INM occurs in the mouse.
To track INM, we utilized the Fucci cell cycle reporter mouse, which fluorescently labels nuclei based on cell cycle stage [41]. The Fucci mouse contains two transgenes each expressing a fluorescent protein. One is a variant of monomeric Kusabira Orange (KO2) and labels nuclei of cells in G1 phase. However, it has also been reported that KO2 is expressed in post-mitotic cells [41]. The other fluorescent protein is monomeric Azami Green (AzG), which labels nuclei of cells in late S and G2 phase. Cells in early S phase express both proteins (Fig. 4a). To determine precisely where the transgenes are expressed in the developing retina, we imaged fixed retinal tissue sections from Fucci mice. Since the AzG signal is weak after fixation, we performed immunohistochemistry using an antibody against AzG. The KO2 reporter did not require immunofluorescence for histological detection. Sections of P0 retinae revealed that KO2 is indeed expressed in the developing inner nuclear layer (INL) and GCL, which is comprised of postmitotic retinal neurons (Fig. 4b, arrowheads). As expected, the NBL is densely populated with RPCs in different stages of the cell cycle. We observed KO2+ G1 phase RPCs, AzG+ late S/G2 phase RPCs, and a small number of KO2+/AzG+ early S phase RPCs (Fig. 4b, arrows indicate KO2+/AzG+ cells).
We next performed immunofluorescence for MCM6, which labels RPCs in all stages of the cell cycle [2]. Most MCM6+ cells are co-labeled with one or both Fucci markers, with the exception of apically localized rounded nuclei preparing for division (Fig. 4c, white arrowhead). Cell cycle staging was confirmed by phospho-Histone H3 (PH3) labeling of RPC mitosis, which showed a lack of co-localization with AzG (Fig. 4d). We also noticed KO2+ cells within the presumptive HC layer and ONL that were not MCM6+ (Fig. 4c, blue arrowheads) consistent with the perdurance of KO2 in post-mitotic neurons. HC expression of KO2 was confirmed by Calbindin overlap (Fig. 4e). These cells, indicated by their KO2 brightness relative to the RPCs, were excluded from subsequent experiments that aimed to track INM within the NBL.
For culture and live confocal imaging of P0 Fucci retinal slices undergoing INM, Z-stacks were acquired every 10 min (Additional file 6). As shown in individual images taken from the time lapse, AzG+ cells were clearly observed migrating apically within the NBL (Fig. 5a, cyan and white arrows). With the spot tracking function in Imaris software, we tracked the migration of both AzG+ and KO2+ nuclei. All AzG+ nuclei migrated apically (Fig. 5b), but some apical migration was preceded by a period of stochastic migration, which was likely during late S phase (color-coded black in Fig. 5b). The KO2+ nuclei moved erratically, but most had an overall basal displacement (Fig. 5c). To further describe the INM patterns as random or directional, we next calculated the mean square displacement (MSD) of nuclear trajectories as a function of elapsed imaging time. For particles subject to random diffusion, the MSD is a linear function of elapsed time. When movement is directional, the MSD forms a curved line over time. When we sampled the first 80 min of each trajectory, the MSD of KO2+ and AzG+ nuclei were both nonlinear, which suggests these nuclei migrate in a directional manner (Fig. 5d and e). These findings suggest that the pattern of mouse RPC INM is similar to INM of the zebrafish retina [22, 34]. It also validates our slice culture method as being capable of capturing an essential event during mouse retinogenesis.
As an additional assessment of overall tissue health and tolerance of the live imaging protocol, we quantified the rate of G2 apical INM at different periods throughout the time lapse movies. Apical tracks of G2 nuclei were clustered into three groups (0–200 min, 200–400 min, and 400–600 min in culture) and the average apical velocity (μm/min) of nuclei was quantified for each period. We compared the average velocities among each group using a single factor ANOVA and found no statistically significant difference (Fig. 5f). Therefore, the live imaging protocol does not significantly impact RPC INM in G2 suggesting that cell cycle kinetics are not overtly affected by culture conditions or laser exposure.
Analysis of Cyclin D1
−/− RPC interkinetic nuclear migration
Despite being tightly coupled to the cell cycle, chemical inhibition of INM does not cause cell cycle arrest. Rather, cells continue to proliferate, but mitosis is no longer restricted to the apical end of the ventricular zone [28, 31, 49]. In contrast, chemically blocking cell cycle progression halts INM in the rodent brain and zebrafish neuroepithelia [1, 22, 45, 47]. Additionally, studies of zebrafish mutants with an RPC cell cycle period twice as long as wild type exhibit an INM rate that is also proportionally slowed [4, 50].
Information as to how the core cell cycle machinery interfaces with the cellular mechanisms driving INM is only beginning to emerge. The most established role is for Cyclin dependent kinase 1 (CDK1) which, along with CYCLIN A or CYCLIN B, is well-known to regulate the entry into G2 and progression toward M-phase [24, 37]. Pharmaceutical inhibition of cdk1 in zebrafish resulted in RPCs that stalled in G2 and failed to undergo apical INM suggesting that cdk1 is necessary for INM [45]. Remarkably, inhibition of the cdk1 inhibitor wee1 resulted in precocious cdk1 activity and apical migration of RPCs that are presumably in S-phase. Thus, cdk1 activity is both necessary and sufficient for apical INM [45]. It is not yet clear whether a similar role for CDK1 exists during mouse INM or whether other G2-specific genes directly drive INM.
The best, in vivo-characterized mouse RPC cell cycle regulator is CYCLIN D1 [9, 10, 13, 44]. Although its canonical role is to promote S phase entry, CYCLIN D1 is expressed throughout the cell cycle in RPCs, and its loss extends retinal cell cycle length [2, 9, 10]. Specifically, these studies used thymidine analog incorporation and antibody staining in fixed mouse retinal tissue to demonstrate that the length of G1, G2, and M phases combined was extended in the absence of Cyclin D1. Given the tight coupling of cell cycle timing and INM rate, we hypothesized that apical migration during G2 phase in Cyclin D1−/− (KO) retinae would be slower than in wildtype (WT) retinae.
To test this, we imaged (at least 12 h per retina) WT (n = 3) and KO (n = 3) retinae expressing AzG (Additional files 7 and 8). Both WT and KO AzG+ RPCs exhibited the expected apical displacement characteristic of G2 phase (Fig. 6a-b). Using the spot tracking function in Imaris, we tracked the position of 56 AzG+ nuclei from WT RPCs and 54 nuclei from KO RPCs over time. As mentioned previously, AzG labels nuclei of cells during late S and G2 phase. Thus, to analyze G2 apical migration more specifically, we omitted data points prior to the first indication of directed apical migration and plotted the relative nuclear position over time (Fig. 6c-d). We found that AzG+ nuclei in the KO retinae had slower average and slower maximum apical velocities (Fig. 6e-f). Taken together with evidence from previous studies [9, 10], these data suggest that CYCLIN D1 may indirectly regulate INM rate by maintaining cell cycle timing during G2 phase and support conclusions from zebrafish studies that suggest INM rate and cell cycle length are coupled [4, 50].
Analysis of horizontal cell dynamics
Finally, we sought to determine whether our live imaging protocol is suitable for capturing discrete developmental events among retinal neurons. As a proof-of-concept, we imaged early postnatal horizontal cell (HC) dynamics. We generated Cx57-iCre+/tg; Rosa26R-mTmG+/tg mice [16, 32] in which early postnatal HCs express a membrane-bound green fluorescent protein (eGFP) that clearly highlights HC neurites. The remainder of the retina, without Cre-mediated recombination, is labeled with a membrane-bound tdTomato fluorescent protein (Fig. 7a).
Live imaging of HCs at P2 revealed soma within the presumptive HC layer that showed little movement whereas the neurites were extremely dynamic (Fig. 7a-b and Additional files 9, 10 and 11). Specifically, the more laterally-oriented neurites of adjacent HCs exhibited extension and retraction toward one another and generally maintained significant overlap (Fig. 7c, arrowheads). In contrast, vertically-oriented neurites of the same HCs underwent continuous extension and retraction but appeared to exhibit little or no overlap with the neighboring cell (Fig. 7c, colored arrows and Fig. 7d, colored filaments). To further analyze this cellular behavior, we next measured the territorial overlap between neurites of adjacent HCs (Fig. 7e) [30]. The HC territory over time was determined followed by the overlap of those areas between adjacent cells (Fig. 7f-g). In order to visually highlight territories of overlap, the delineated HC territories over time were shown with a 50% transparency overlaid on the original fluorescence images. Thus, the brighter white regions, covering more horizontally-oriented neurites, indicate overlap whereas we did not detect overlap between vertical neurites (Fig. 7e and Additional file 12). Previous live, multiphoton microscopy of mouse HCs in explant cultures identified these vertical neurites as transient processes that mediate homotypic repulsive interactions driving HC mosaic formation [18].