Bidirectional neuronal migration coordinates retinal morphogenesis by preventing spatial competition

While the design of industrial products is often optimized for the sequential assembly of single components, organismal development is hallmarked by the concomitant occurrence of tissue growth and organization. Often this means that proliferating and differentiating cells occur at the same time in a shared tissue environment that continuously changes. How cells adapt to architectural changes in order to prevent spatial interference remains unclear. To understand how cell movements important for growth and organization are orchestrated, we here study the emergence of photoreceptor neurons that occur during the peak of retinal growth using zebrafish, human tissue and human organoids. Quantitative imaging reveals that successful retinal morphogenesis depends on active bidirectional photoreceptor translocation. This leads to a transient transfer of the entire cell population away from the apical proliferative zone. This migration pattern is driven by distinct cytoskeletal machineries, depending on direction: microtubules are required for basal translocation, while actomyosin drives apical movement. Blocking photoreceptor translocation leads to apical overcrowding that hampers progenitor movements. Thus, photoreceptor migration is crucial to prevent competition for space and thereby allows concurrent tissue growth and lamination. This shows that neuronal migration, in addition to its canonical role in cell positioning, is involved in coordinating morphogenesis.


Main
During organ development, the correct number of cells needs to be generated and, upon differentiation, these cells need to be spatially organized to assure tissue function. One way to achieve organogenesis would be the 'bang-bang' approach, a term often used in engineering to describe a system that switches abruptly between two states to minimize costs while at the same time enhancing performance [1][2][3] . In development, this would entail that tissues first grow to their correct size and generate the accurate number of cells and then switch to a phase of differentiation and remodeling. However, such design strategy seem rather an exception in biological systems, for example, the intestinal crypts 4 . In most other instances, including the formation of the pancreas, heart or brain, organogenesis involves a complex choreography during which the expansion of cell number coincides with the onset of differentiation and the establishment of functional tissue architecture [5][6][7] . Thus, it is important to understand how concomitant growth and differentiation are coordinated in a shared environment. During brain development, newborn neurons often leave their initial position and migrate to the locations at which they later function when the tissue is still proliferative and growing. Thus, migrating neurons and remaining proliferating cells need to adjust to changing tissue environment in order to prevent spatial interference. How this is achieved at the cellular and tissue level is not well understood. One part of the CNS that allows to study how tissue growth and differentiation occur synchronously is the vertebrate retina, the part of the brain responsible for visual perception. Here, the organization of retinal neurons into distinct layers ensures proper circuit formation and thereby organ function 8,9 . Further, retinal architecture is conserved across species, allowing the comparison of findings across model systems. Recent studies have shown that in the zebrafish retina, complex migration patterns occur simultaneously at stages when the tissue undergoes significant growth [10][11][12][13] . However, how progenitor and neuronal cell movements are coordinated in the same space has not yet been explored. Thus, we here investigated how photoreceptors (PRs) that emerge at peak proliferation stages 10 form the apical photoreceptor cell layer despite ongoing apical proliferation. We show that all PRs emerging apically exhibit transient basal translocation across the neuroepithelium but return apically for final positioning. This phenomenon is seen in zebrafish, human tissue and human organoids. We further demonstrate that, in zebrafish, basal and apical migration are active and driven by distinct cytoskeletal machineries: microtubules in basal direction and actomyosin in apical direction. While this bidirectional movement is dispensable for proper photoreceptor positioning, it is crucial to prevent competition for space with the apically dividing progenitors. Thereby PR translocation ensures concurrent tissue growth and lamination in zebrafish and humans, indicating that it is a fundamental aspect of retinal morphogenesis.

Photoreceptors undergo bidirectional migration with direction-dependent kinetics
PRs are retinal neurons that are born at the same place at which they later function.
However, recent studies reported a transient population of cells expressing PR markers away from the PR layer in developing zebrafish and human retinal organoids [14][15][16] . To test whether this ectopic population was due to photoreceptor translocation, we imaged zebrafish retinas in vivo at near-physiological conditions using light-sheet microscopy 17 and reporter genes for early neurogenic progenitors (Tg(ath5:gap-RFP)) and differentiating PRs (Tg(crx:gap-CFP)) 18,19 . All time-lapse data was restored using the deep learning-based algorithm CARE to improve signal-to-noise ratio and allow tracing of emerging PRs (Extended Data Fig. 1A; for details, see Methods) 20 . Imaging of whole retinas revealed that, upon apical birth, both daughter cells of Ath5+ neurogenic divisions moved basalward across the neuroepithelium ( Separating apical and basal movement phases showed that basal movement was more efficient than apical movement ( Fig. 1D; Supplementary Video 1) and that apical movement was more variable in duration (Fig. 1E). Further, basally moving PRs displayed more continuous and monotonic movements while apically moving PRs translocated in a more saltatory manner (Fig. 1D). Directionality ratio and mean square displacement (MSD) analysis confirmed these notions ( Fig. 1F and Extended Data Fig. 1D). Nevertheless, both movements occurred in a directed manner, as shown by the positive curvature of the MSD analysis ( Fig.   1F).
Thus, emerging PRs undergo bidirectional migration prior to final apical positioning and basal and apical movements display different kinetics.

Photoreceptor migration is conserved in humans
As retinal architecture is highly conserved in many vertebrates ranging from fish to mammals, we asked whether the bidirectional migration of emerging PRs seen in zebrafish is a fundamental aspect of retinal development by studying PR behavior in a distantly related species, humans. To this end, we stained human fetal retinas (provided by the human tissue bank (HDBR)) against Otx2 and Crx, two transcription factors observed during PR fate determination and differentiation, respectively 23,24 ( Fig. 2A).
At 9 weeks post-conception (PCW), only few Crx+/Otx2+ cells were observed at the apical surface of the central retina while other cells occupied positions along the apicobasal axis (Fig. 2B and E). At 10 PCW and 11 PCW, however, an increasing number of Crx+/Otx2+ PRs was detected along the entire thickness of the neuroepithelium (Fig. 2C-G). As the human retina differentiates in a central-to-peripheral gradient, different areas of the 11 PCW retina were analysed as a proxy for developmental progression. Otx2+ cells at the periphery occupy both apical and basal positions, while in central, more developed regions, cells became more restricted to the apical surface (Extended Data Fig. 2A). Thus, the spatial distribution of emerging PRs during human retinogenesis was similar to what was seen in zebrafish.
Nevertheless, the overall trajectories were strikingly similar to zebrafish, as in both systems basal movement was continuous, whereas the apical translocation was saltatory ( Fig. 2N and Extended Data Fig. 2D).
These results show that PR translocation is comparable in retinas of systems as dissimilar as zebrafish and humans indicating that it is an important feature for retinal development.

Stable microtubules are required for active basal photoreceptor migration but dispensable for apical migration
We noted kinetic differences between basal and apical movements of emerging PRs in both zebrafish embryos and human organoids. Thus, we turned back to the zebrafish system to better understand the cellular mechanisms that drive the PR translocation in the different directions. The kinetics of initial basal translocation of the emerging PRs were strikingly similar to what has been previously reported for their sister cells, RGCs 11 ( Fig. 3A and Extended Data  Fig. 3B). Furthermore, some emerging PRs in control embryos were able to move to basal positions even when a considerable distance between sister cells was observed (Fig. 3C). These findings indicated that PRs moved autonomously with very similar kinetics to sister RGCs. As basal RGC translocation depends on a stabilized microtubule cytoskeleton 11 , we tested whether a similar mechanism was at play for emerging PRs. To this end, the distribution of the microtubuleassociated protein Doublecortin (bactin:GFP-DCX) was analysed during basalward PR movements. Microtubules were enriched in the apical process shortly after their birth and during basal movement (Fig. 3D). Acetylated tubulin staining showed that these apical microtubules were stabilized as previously seen in RGCs (Fig. 3E) 11 .
To test the active involvement of these stabilized microtubules in basal PR translocation, microtubules were depolymerized by colcemid. This resulted in a premature accumulation of emerging PRs at the apical surface (Extended Data Fig. 3D). This finding was substantiated by overexpression of the microtubule-destabilizing protein Stathmin1 11 which lead to erratic movements of emerging PRs and their premature return to the apical surface (Extended Data  Fig. 3C). Together, these results argue that basal translocation of emerging PRs is microtubule-dependent, while their return to the apical surface is not.

Actomyosin is the main driver of photoreceptor apical migration
In addition to microtubules, the actin cytoskeleton has been shown to be responsible for cell and organelle translocation in neuroepithelial and neuronal cells [28][29][30] . Thus, we explored whether actomyosin could be involved in apical translocation of emerging PRs. To this end, crx:gap-CFP expressing retinas were stained for F-actin and active myosin at 42 hpf, when most emerging PRs undergo apical translocation. In some CRX+ cells, enrichment of F-actin and active myosin was observed basal to the nucleus ( Fig. 4A and B). Live imaging of Utrophin

Photoreceptor migration prevents overcrowding of the apical mitotic zone
The fact that bidirectional PR movement is conserved in zebrafish and humans and is highly regulated at the molecular level argued that, despite being counterintuitive, it played an important role for tissue formation.
Reassessment of the trajectories gathered in Fig. 1 revealed that emerging PRs stay away from the apical surface for an average of 300.6 minutes (coefficient of variation (CoV), 25.97%) while the depth of migration varied between 8.7 µm and 44.2 µm (CoV, 31.14%) (Extended Data Fig. 5A). However, depth of migration and duration away from the apical surface did not strongly correlate (Extended Data Fig. 5A) arguing that moving PRs out of the way for a certain amount of time was more relevant than moving them as far basally as possible.
Furthermore, PR migration did not seem to be essential for the determination or maintenance of cellular fate as when migration away from the apical surface was perturbed by Stathmin overexpression, emerging PRs nevertheless positioned correctly and showed typical PR morphology (Fig. 3F). Thus, other non-cell autonomous prerequisites were more likely responsible for the migration patterns observed.
As PRs are one of the first neurons to be born in the developing retina 31 , their generation coincides with a last peak of proliferative growth 10 . Indeed, a substantial number of actively cycling progenitors is still present when PRs start to emerge as confirmed by EdU labelling in zebrafish (Extended Data Fig. 5B) as well as in human organoids ( Fig. 2I-K).
During this proliferative peak, all progenitor divisions occur apically and non-apical divisions disturb tissue integrity 32 . Taking these notions into consideration, we asked whether basal translocation of emerging PRs could be linked to the need for apical surface availability during the proliferative peak. We analysed the fraction of the apical surface that was occupied by apical photoreceptors at this stage ( Fig. 5A). To this end, we quantified tissue surface area  Fig. 5C). This showed that only 11,2% of PRs were located at apical positions at 42 hpf occupying 11,9% ± 2,2% (mean ± SD) of the apical surface area (Fig.   5D). In contrast, in a theoretical scenario in which all PRs born at 42 hpf would not undergo translocation, the apical occupancy would rise to 108,7% ± 13,8% (mean ± SD) (Fig. 5D). We thus speculated that overcrowding of the apical surface could potentially affect incoming progenitor divisions if emerging PRs were not moving out of the way. We tested this notion by inducing a tissue-wide arrest of basal translocation of emerging PRs using colcemid (Extended Data Fig. 3D). This led to PRs occupying the complete apical surface of the tissue (

Discussion
This study shows that active bidirectional migration of emerging PRs in zebrafish and humans is important to coordinate growth and lamination during retinogenesis. We reveal that movements are driven by different molecular motors depending on direction: microtubules in basal direction and actomyosin in apical direction. More precisely, we propose that microtubule polymerisation propels the soma of emerging PRs in the basal direction while concurrent stabilization of microtubules prevents backward movements (Extended Data Fig. 6). In the apical direction, actomyosin contractions regulated by the Rho-ROCK pathway drive the saltatory movement of emerging PRs (Extended Data Fig. 6). While the interplay between actomyosin and microtubules to integrate the movement of different parts of a cell and the different steps of migration has already been suggested in other contexts 33,34 , the use of two distinct cytoskeletal elements by the same emerging neuron to move in different directions has to our knowledge not yet been revealed. Next, we need to unveil the molecular and tissue wide triggers that induce the reversal of migration, an exciting topic for future investigations.
The fact that we discovered a new migratory mode for one the most widely studied neuronal cell types, the PRs 23,35,36 , demonstrates that neuronal migration is far from being understood. Thus, studying different areas of the central nervous system will be essential to refine our understanding of the diversity of neuronal migration phenomena and how they contribute to the generation of the unique cytoarchitectures of the brain.
Generally, neuronal migration has so far mainly been recognized as an important and widespread phenomenon that ensures correct neuronal positioning. This in turn safeguards the formation of functional neural circuits. Not surprisingly, neuronal migration defects can lead to a variety of neurological disorders 37,38 . However, our study shows that even neurons that are born at the same position at which they later reside and function, like the retinal PRs, can undergo translocation before final positioning. While at first sight a translocation resulting in a zero-net displacement seems redundant, it turns out to be essential for overall successful retinal development. Unlike other neuronal translocation phenomena 11,13,39 , this migration is not a prerequisite for correct positioning of the cells. Instead, it is important to prevent local crowding which can induce mechanical stress that in turn can lead to progenitor delamination as seen also at earlier developmental stages 40,41 . Thus, we propose that by preventing the competition for space, migration of the emerging PRs is crucial for the spatiotemporal coordination of tissue growth and lamination. We speculate that when neuronal tissues start to differentiate and remodel while still under active proliferation, strategies need to be deployed that nevertheless ensure apical surface availability for progenitors. In the retina, this takes the form of basal translocation of apically emerging neurons. Thus, neuronal migration in addition to its canonical role for correct cell positioning can be an important mechanism to coordinate cell and tissue wide phenomena during brain morphogenesis. As simultaneous tissue growth and differentiation are a hallmark of many developing organs, it is tempting to speculate that similar relocation strategies of entire cell populations could also be at play elsewhere.
Our finding that this phenomenon is conserved in zebrafish, human tissue and humanderived organoids suggests that it is fundamental to retinal development across evolution. It also argues that retinal morphogenesis, like tissue architecture and gene regulatory networks, is highly conserved. This area of conservation is however much less understood and more comparative studies are needed to probe how tissue growth and differentiation are orchestrated in the developing embryo and how defects of such coordination can lead to a diseased state.

Human iPSC culture and generation of retinal organoids
Human iPSCs (IMR90 clone 4 from WiCell) 42  µM Taurine) and maintained in the cell culture incubator until organoid dissection. The optic vesicles were dissected under a widefield microscope using 5x/0.12 or 10x/0.25 NPLAN objectives (LEICA), and 27-gauge needles (BD). Organoids were transferred to the agarose bed with fresh ROM medium and placed in individual wells to prevent fusion. From this point on, the ROM medium was exchanged every third day. In all stages of the protocol, iPSCs and developing organoids were cultured at 37ºC and 5% CO2 in a humidified incubator (Thermo Fisher Scientific).

Zebrafish husbandry
Wild-type zebrafish and transgenic lines were maintained at 26°C. Embryos were raised at 28.5 or 32°C in E3 medium. Medium was changed daily and supplemented with 0.2 mM 1phenyl-2-thiourea (Sigma-Aldrich) from 8 hours post fertilization (hpf) to prevent pigmentation. Staging was performed in hours post fertilization 44 . All animal work was performed in accordance with European Union directive 2010/63/EU, as well as the German Animal Welfare Act.

DNA cloning and constructs used
The ath5:MRLC-mKate2 and ath5:GFP-UtrophinCH used to label myosin and actin, respectively, in PRs were assembled using Gateway cloning (Thermo Fisher Scientific) based on the Tol2 kit 46 . The human MRLC2 gene from pCS2 + -MRLC2-GFP 47 was used for creation of hsp70:MRLC-mKate2 construct which was then used as template for construction of MRLC2-mKate2 middle entry clone (pME) by PCR using Phusion polymerase (New England Biolabs) and primers with ATT recombination site (shown in lower case): Forward 5'-ggggacaagtttgtacaaaaaagcaggctggCTTCGCTGTCGTTTGTGGTCTCG-3' and reverse 5'-ggggaccactttgtacaagaaagctgggtcTCATCTGTGCCCCAGTTTGC-3'. The pCS2 + vector containing GFP tagged calponin homology domain of human utrophin (GFP-UtrophinCH) 48 was used as template for generation of pME GFP-UtrophinCH by PCR using Phusion polymerase (New England Biolabs) and primers with ATT recombination site (shown in lower case): Forward 5'-ggggacaagtttgtacaaaaaagcaggctggATGGTGAGCAAGGGCGAGG and reverse ggggaccactttgtacaagaaagctgggtcTTAGTCTATGGTGACTTGCTGAG. The pME MRLC-mKate2 and pME GFP-UtrophinCH were combined with the 5′ entry clone containing the ath5 promoter 46 into the destination vector pTol2 + pA R4-R2 backbone 49 to create the ath5:MRLC2-mKate2 and ath5:GFP-UtrophinCH.

Electroporation of retinal organoids
Labeling of human iPSCs-derived retinal organoids for live-imaging was done via electroporation with a transfection reporter (caggs:Lyn-tdTomato) and a photoreceptor reporter construct (hcrx:GFP). DNA preparation and electroporation procedure were adapted from previously published protocol for electroporation of mouse retina explants 51

In vivo labelling of S-phase cells by EdU
To label S-phase cells, zebrafish embryos were incubated at 4ºC for 1 h in E3 supplemented with 500 µm of EdU (ClickiT-Alexa 488 fluorophore kit, Invitrogen) while human retinal organoids were incubated at 37ºC for 1h in organoid medium supplemented with 10 µm of EdU. Embryos and organoids were immediately fixed overnight in 4% PFA and, after antibody staining, incorporated EdU was detected according to manufacturer's protocol.

Whole-mount staining of zebrafish embryos
Zebrafish were dechorionated manually and fixed overnight in 4% PFA in PBS at 4ºC. VectaShield (Vector Labs) were stored at -20ºC.

Light-sheet imaging of retinal organoids
Electroporated retinal organoids were imaged using Lightsheet Z. imaging was performed at room temperature using the 40x/1.2 C-Apochromat water immersion objective (Zeiss).

Image processing and analysis
Image data were processed in ZEN Black and/or Fiji 53 . Spatial drift of live imaging data was corrected using Manual drift correction plugin from Fiji (by Benoit Lombardot, Scientific computing facility, MPI-CBG).

Cell tracking
For analysis of PR translocation in the zebrafish retina, wild-type and Stathmin1 overexpressing PRs labelled by mosaic injection of plasmids containing ath5-driven reporters reporter plasmid were identified based on the expression of the reporter gene ( Fig. 2N and Extended Data Fig. 2D). Full trajectories from birth to final positioning and partial trajectories were obtained by tracing the center of the cell body in 2D maximum projected substacks of the raw data using ImageJ plugin MTrackJ 54 .

Analysis of kinetics of migration
Mean square displacements (MSDs) and directionality ratios were calculated from the first 80 min of basal and apical migration of PRs in Excel (Microsoft) using the open-source computer program DiPer 55 . To estimate directional persistence, the a-value was obtained from the slope of the log-log plots of MSDs and Time interval.

Analysis of the spatial distribution of PRs
The position of PRs cells with respect to the apical surface of developing human retinal organoids was measured manually using Fiji (Fig. 2L). The distance between the center of the cell body and the apical attachment was calculated in Excel.

Quantitative analysis of myosin enrichments
Myosin enrichments basal and apical to the nucleus of apically migrating PRs were measured in 2D sum projected substacks of the raw data. The 2D sum projections were registered to the center of the cell of interest using Fiji plugin for manual registration developed by the MPI-CBG Scientific Computing Facility. Value cutoff for thresholding was determined after median filtering (Median filter size: 1 pixel) by measuring the mean pixel intensity at the center of the cell (2x2 µm square) during the apical migration. The mean of the mean intensity of all timepoints plus two times the mean standard deviation was used as threshold limit. Values below the limit were set to NaN (Not a Number). The fraction of the positive area within the regions of interest (4x2 µm, Width x Height) basal and apical to the nucleus was measured for every timepoint of apical migration. The position of the square ROIs was adjusted manually according to the position of the cell, whereas the size of the ROIs remained constant.

Analysis of apical surface occupancy
To estimate the fraction of the apical surface area occupied by PRs at 42 hpf, the number of apical PRs and the total number of PRs were counted manually in subvolumes of the 42 hpf retina (N = 3 embryos). The tissue surface area in each subvolume was measured using a custom Fiji plugin (Volume Manager, by Robert Haase, Scientific computing facility, MPI-CBG) and visualized in 3D using Fiji plugin ClearVolume (Fig. 5B) 56 . The cross-sectional area of apical PRs was measured at the center of the cell in the XZ orthogonal view (n= 11 cells, N = 4 embryos; Fig. 5C). To calculate the fraction of occupancy by apical PRs, the average crosssectional area was multiplied by the number of apical PRs and subsequently divided by the tissue surface area. To calculate a theoretical occupancy in the case of no migration, the total number of PRs was used.

Statistical analysis
Statistical tests were performed using GraphPad Prism 6, except Dunn's Kruskal-Wallis Multiple Comparisons with Holm-Bonferroni correction, which was conducted in python using scipy.stats v1.3.1 and v0.6.2 scikit-posthocs packages. Two-tailed tests were used and 95% confidence intervals were considered. See figure legends for more information on the P values and sample sizes.

Data availability
All data is available in the manuscript. Fig. 6 Schematic summary of mechanisms driving photoreceptor migration and its relevance for the coordination of tissue growth and organization.

Extended Data
(Upper panel) Photoreceptors (PR) emerge at the apical surface of the retinal neuroepithelium and undergo bidirectional somal translocation that entails fast and directed movement towards the basal lamina followed by a basal pause and saltatory movements towards the apical surface.
Basal and apical migration display different kinetics due to the involvement of distinct cytoskeleton machineries. While basal migration depends on stable microtubules, they are dispensable for apical translocation, which depends on actomyosin contractions at the cell rear.
(Lower left panel) PR bidirectional migration is concurrent with cell proliferation and tissue growth.
(Lower right panel) Bidirectional migration of PRs delays lamination helping to coordinate tissue growth and organization by securing space at the apical surface for incoming divisions of progenitor cells. This bidirectional mode of somal translocation is crucial to maintain tissue integrity and function, as blockage of PR movements leads to overcrowding of the mitotic zone and subsequent progenitor delamination.