Amoeboid-like neuronal migration ensures correct horizontal cell layer formation in the developing vertebrate retina

As neurons are often born at positions different than where they ultimately function, neuronal migration is key to ensure successful nervous system development. Radial migration during which neurons featuring unipolar and bipolar morphology, employ pre-existing processes or underlying cells for directional guidance, is the most well explored neuronal migration mode. However, how neurons that display multipolar morphology, without such processes, move through highly crowded tissue environments towards their final positions remains elusive. To understand this, we here investigated multipolar migration of horizontal cells in the zebrafish retina. We found that horizontal cells tailor their movements to the environmental spatial constraints of the crowded retina, by featuring several characteristics of amoeboid migration. These include cell and nucleus shape changes, and persistent rearward polarization of stable F-actin, which enable horizontal cells to successfully move through the crowded retina. Interference with the organization of the developing retina by changing nuclear properties or overall tissue architecture, hampers efficient horizontal cell migration and layer formation. Thus, cell-tissue interplay is crucial for efficient migration of horizontal cells in the retina. In view of high proportion of multipolar neurons, the here uncovered ameboid-like neuronal migration mode might also be crucial in other areas of the developing brain.


ABSTRACT:
As neurons are often born at positions different than where they ultimately function, neuronal migration is key to ensure successful nervous system development. Radial migration during which neurons featuring unipolar and bipolar morphology, employ pre-existing processes or underlying cells for directional guidance, is the most well explored neuronal migration mode. However, how neurons that display multipolar morphology, without such processes, move through highly crowded tissue environments towards their final positions remains elusive. To understand this, we here investigated multipolar migration of horizontal cells in the zebrafish retina. We found that horizontal cells tailor their movements to the environmental spatial constraints of the crowded retina, by featuring several characteristics of amoeboid migration. These include cell and nucleus shape changes, and persistent rearward polarization of stable F-actin, which enable horizontal cells to successfully move through the crowded retina. Interference with the organization of the developing retina by changing nuclear properties or overall tissue architecture, hampers efficient horizontal cell migration and layer formation. Thus, cell-tissue interplay is crucial for efficient migration of horizontal cells in the retina. In view of high proportion of multipolar neurons, the here uncovered ameboid-like neuronal migration mode might also be crucial in other areas of the developing brain.

INTRODUCTION:
Migration of newly-born neurons to their designated positions is a key step in the establishment of neural circuits and thereby function of the central nervous system (CNS). A variety of migration modes have been uncovered featuring diverse cell morphologies ranging from unipolar or bipolar to multipolar (Marin and Rubenstein 2003, Kawaji et al. 2004, Ayala, Shu and Tsai 2007, Tanaka et al. 2009, Cooper 2013, Rahimi-Balaei et al. 2018, Gressens 2000. As of yet, most research unveiling the cellular and molecular mechanisms of neuronal migration has focused on radial migration which is limited to movements of unipolar and bipolar neurons, perpendicular to the tissue surface (Angevine and Sidman 1961, Berry and Rogers 1965, Morest 1970b, Morest 1970a, Rakic 1971, Walsh and Cepko 1988, Marin and Rubenstein 2001, Nadarajah et al. 2001, Nadarajah and Parnavelas 2002, Cooper 2013.
During radial migration, migrating neurons determine their direction of movement by two different strategies: 1) sending process(es) which anchor the migrating neurons to the tissue lamina(e), and thereby facilitating the faithful arrival of neurons at their destination (somal translocation) (Nadarajah et al. 2003), or 2) moving along radiallyoriented fibers of neural progenitors known as radial glia, that provide physical scaffolding for migrating neurons (glia-guided migration) (Rakic 1971, Rakic 1972, Edmondson and Hatten 1987, Hatten 1990, Gertz and Kriegstein 2015. In both scenarios, the migrating neurons exhibit elongated unipolar or bipolar morphologies in the direction of travel and move unidirectionally via radial paths to their final positions (Nadarajah et al. 2001, Nadarajah et al. 2003. In contrast, neurons that display multipolar morphology are neither attached to the tissue lamina(e) nor move along radial glia fibers and instead extend multiple dynamic processes in various directions (Stensaas 1967, Shoukimas and Hinds 1978, Nowakowski and Rakic 1979, Gadisseux et al. 1990, Honda, Tabata and Nakajima 2003, Noctor et al. 2004, Cooper 2014. Examples include interneurons of the mammalian neocortex (Nadarajah et al. 2003, Tanaka et al. 2006, Tanaka et al. 2009), pyramidal and granule neurons in the mammalian hippocampus (Kitazawa et al. 2014, Namba, Shinohara andSeki 2019). How multipolar neurons, without any predisposed migratory information or scaffolds reach their destination during brain development is much less explored. It is particularly not known whether and how multipolar neurons adjust their migration mode, path and efficiency to their densely-packed surrounding environment (Bondareff andNarotzky 1972, Sekine et al. 2011). This knowledge gap is partly due to the inaccessibility of many brain regions for in vivo imaging.
One CNS region that allows for such imaging approaches in a quantitative manner is the zebrafish retina (Galli-Resta et al. 2008). The retina consists of five major neuron types; photoreceptor (PR), horizontal cell (HC), bipolar cell (BC), amacrine cell (AC), retinal ganglion cell (RGC), and a single glial cell-type; Müller Glia (MG). During development, retinal neurons move from their birth-site to their functional residence ( Fig 1A), and reproducibly assemble into three distinct nuclear layers; outer nuclear layer (ONL), inner nuclear layer (INL), ganglion cell layer (GCL). Synapses between these nuclear layers form at two nuclei-free layers, known as plexiform layers; the outer plexiform layer (OPL) and the inner plexiform layer (IPL) (Fig 1A').
Retinal neurons follow diverse and complex migratory modes and routes to reach their destinations (Chow et al. 2015, Icha et al. 2016a, Amini, Rocha-Martins and Norden 2017, Amini, Labudina and Norden 2019, Rocha-Martins et al. 2021. So far however, parameters that influence successful migration of the diverse neuron types in their complex and highly dynamic environment are only beginning to be understood (Chow et al. 2015, Icha et al. 2016a, Amini et al. 2019, Rocha-Martins et al. 2021. Especially intriguing is the movement of HCs, the retinal interneurons that modulate information flow from PRs to BCs (Chaya et al. 2017). HCs exhibit a bidirectional and bimodal migration pattern that features a switch from bipolar to multipolar morphology (Edqvist and Hallbook 2004, Weber et al. 2014, Chow et al. 2015, Amini et al. 2019 ( Fig 1B, Sup-Fig 1A, Video 1). Newborn HCs display a bipolar morphology and migrate radially from their apical birth-site to the center of the INL while keeping an attachment to the apical surface (Phase 1). Upon detachment of this anchorage (Sup- Fig 1C, Video 2), HCs acquire a multipolar morphology and migrate with frequent direction changes (Phase 2) (Chow et al. 2015, Amini et al. 2019. During this phase, HCs move deeper into the INL before turning apically towards the HC layer at the upper border of the INL, beneath the OPL (Fig 1B, Sup-Fig 1A-B).
The developing zebrafish retina is a densely-packed environment (Matejcic, Salbreux and Norden 2018) which undergoes structural changes in space and time during neuronal lamination. How multipolar HCs adapt their migration behavior and trajectories to the changing and crowded microenvironment of the retina, remains unexplored. The fact that HCs follow unpredictable migration paths despite an overall directionality (Amini et al. 2019) implies that HC path selection is not intrinsically programmed but that the surrounding environment plays a key role. However, if, how, and to what extend cellular and tissue-wide properties influence HC movements towards layer formation remained unexplored.
We here investigated the cellular and tissue-wide parameters that influence HC migration in the developing zebrafish retina. We show that HCs constantly tailor their migration behavior to the limited space within the densely-packed retina by frequent and reversible amoeboid-like shape changes. We further uncover that changing organization of the developing retina at the single-cell or tissue-scale impairs efficient HC migration and perturbs proper HC layer formation.

HORIZONTAL CELLS DO NOT EMPLOY GLIA-GUIDED MIGRATION
It was shown that HCs lose their apical attachment (Sup- Fig 1C, Video 2) ~2-3 hrs after apical birth and subsequently display multipolar morphology when moving towards their final destination (Chow et al. 2015, Amini et al. 2019. Since many multipolar interneurons in the neocortex use radially-oriented cells featuring bipolar morphology as their migratory scaffold (Cooper 2014), we asked if such a phenomenon is also seen for HCs in the retina. To this end, we examined whether migrating HCs move along BCs or MGs, the radially-oriented retinal cells with bipolar morphology (Fig 1A').
We performed light-sheet time-lapse imaging using double-transgenic zebrafish embryos Tg(vsx1:GFP) x Tg(Ptf1-a:dsRed) labeling BCs and HCs, respectively. The fact that neuronal differentiation in the retina occurs in a wave-like manner (Hu and Easter 1999), allowed us to simultaneously visualize retinal regions hosting immature (unlaminated) and mature (laminated) BCs (Sup- Fig 1D). We noted that in regions without laminated BCs, HCs were either already at (Sup- Fig 1D) or en route towards their final position. Further, many HCs did not follow strictly radial migratory trajectories but instead moved in all three dimensions while frequently changing direction (Fig 1C).
The same was seen in regions that hosted laminated BCs (Fig 1D, Video 3), arguing against steady, direct interaction between radially-oriented BCs and migrating HCs.
Similar observations were made when probing a possible association between migrating HCs and developing MGs, the glial cells of the retina. Immunofluorescence stainings of double-transgenic animals Tg(gfap:GFP) x Tg(Ptf1-a:dsRed) marking MGs and HCs, showed that prior to MG emergence at 48 hpf, GFAP (glial fibrillary acidic protein) was expressed in radially-oriented retinal neurogenic progenitors (Bernardos andRaymond 2006, Rapaport et al. 2004) (Fig 1E, 38 hpf). No GFAP + cells featuring mature MG morphology at the onset (48 hpf) or peak (55-65 hpf) of HC migration were observed (Sup- Fig 1D, 50 hpf). When mature MGs emerged (around 65 hpf), some HCs were still en route towards the apical side ( Figure 1E, 65 hpf), while others had already reached the prospective HC layer ( Figure 1E, 65 hpf). From 70-72 hpf, when GFAP was specifically expressed in mature MGs (MacDonald et al. (Fig 1E, Sup-Fig 1E), the majority of HCs were already integrated into the HC lamina ( Figure 1E, 70 hpf). As seen in embryos expressing BC markers, also in embryos labelled for MGs, HCs followed tangential routes perpendicular to the radial orientation of MG fibers, both before and after MG maturation (Fig 1D, Sup-Fig 1E'-E"). Thus, we conclude that HCs are unlikely to move along the radially-oriented BC or MG fibers.

DOMINATED BY ECM
Cells migrating within tissues can be influenced by mechanical cues from their environment. For example, local gradients in extracellular matrix (ECM) stiffness can guide cell migration in a process termed durotaxis (Lo et al. 2000, Isenberg et al. 2009, Roca-Cusachs, Sunyer and Trepat 2013, Bollmann et al. 2015. In the context of neuronal migration, the ECM can act either as an instructive scaffold along which migration occurs or as a barrier for migrating neurons (Franco and Muller 2011). We thus asked whether ECM components could influence HC migration, focusing on Laminin α1, a glycoprotein that forms fibrous structures (Timpl et al. 1979, Chung et al. 1979, and Neurocan (termed here ssNcan in Tg(ubi: ssNcan-EGFP)) as a biosensor for Hyaluronic acid (HA) (De Angelis et al. 2017, Grassini et al. 2018) a glycosaminoglycan forming hydrogel-like structures. Our immunofluorescence imaging showed that anti-Laminin α1 and HA were only detected in the basement membrane of the retina at all developmental stages (Fig 2A-B, arrowheads) and never within the INL wherein HC migration takes place (Fig 2A-B). Thus, it is unlikely that HCs use ECM scaffolds as a main migratory substrate to reach their destination.
We next asked whether the developing zebrafish retina features mechanical gradients during HC migration, and if yes, whether and how these change in space and time. To address this point, we used Brillouin light scattering microscopy, a non-invasive technique which has been recently applied to a broad range of living biological systems including different areas of the CNS (Girard et al. 2015, Schlussler et al. 2018, Scarcelli and Yun 2012, Scarcelli, Kim and Yun 2011, Mattana et al. 2017). This technique provides information about tissue compressibility by measuring the Brillouin shift values of the sample (Zhang et al. 2017, Prevedel et al. 2019, Elsayad, Polakova and Gregan 2019b, Elsayad et al. 2019a).
The Brillouin shift maps revealed that in comparison to the retina neuroepithelium, the lens had higher Brillouin shift at 50 hpf (Fig 2D'-E') which further increased as development progressed (70 hpf) (Fig 2D''-E''). At 70 hpf, the retina displays a layered organization composed of five different layers from apical to basal: ONL, OPL, INL, IPL, and GCL (Fig 2C'). Notably, Brillouin shift maps of 70 hpf retinae revealed a layered, heterogenous pattern that matched the retinal layers observed in confocal images (Fig 2D''-E''). While the two nuclear-free plexiform layers (OPL and IPL) showed lower Brillouin shifts, the three nuclear layers (ONL, INL, GCL) displayed comparably higher Brillouin shifts. This indicated that a correlation between Brillouin shift maps and the characteristic features of each retinal layer exists, and that Brillouin shift values are influenced by nuclear occupation, as shown previously for fibroblasts (Zhang et al. 2017). However, Brillouin shift values in the INL showed no obvious differences along the apico-basal axis, during (Fig 2D', E') and after HC migration ( Fig   2D'', E''). Thus, HC migration seems to occur in a homogenously compressible environment without notable local compressibility gradients.
As Brillouin shift values were higher in regions with high cell-body density compared to the nuclei-free plexiform layers, we tested INL packing at stages of HC peak of migration (50-65 hpf). To this end, we performed live-imaging of double-transgenic zebrafish embryos Tg(lhx-1:eGFP) x Tg(bactin:mKate2-ras) labeling HCs and membrane of all retinal cells, respectively. No extracellular space between neighboring cells within the INL was detected during periods of HC migration (Sup- Fig 2A-A'). This notion was further supported by confocal imaging of membrane (Tg(bactin:mKate2ras)) and nuclei envelope (Tg(bactin:eGFP-lap2b) of all retinal cells, showing that no space was detected in between the neighboring nuclei in INL (Sup- Fig 2B-B'). Thus similar to the neuroepithelial stage (Matejcic et al. 2018), also during neurogenesis, the developing INL features a dense packing of cells and nuclei, implying that HCs migrate in an environment wherein space is limited.

DEFORMATIONS
We showed that migrating HCs navigate within a crowded environment with neither cellular nor ECM scaffolding structures (Fig 1C, E, Fig 2A-B), or compressibility gradients (Fig 2D-E''). Consequently, HCs are left with two major strategies employed by cells migrating in crowded environments with physical constraints: 1) active generation of migratory tracks by protease-dependent local ECM degradation or remodeling (e.g. used by mesenchymal cells such as fibroblasts during wound healing, and aggressively migrating tumor cells) (Wolf et al. 2007, Krause and Wolf 2015, or 2) amoeboid adaptations of their shape, migration path and direction to the limited available space (e.g. seen for leukocytes like neutrophils and dendritic cells) Sixt 2009, Friedl andWolf 2010).
Given that no major ECM components were detected in the INL (Fig 2A-B), proteasedependent strategies were unlikely to be of major importance in this context. To underline this notion, tissue integrity was probed using double-transgenic zebrafish retinae Tg(lhx-1:eGFP) x Tg(bactin:mKate2-ras) labeling HCs and PRs, and membrane of all retinal cells, respectively. Consistent with lack of ECM in the INL, we did not detect any tissue ruptures or holes, neither at the peak of HC migration nor after HC migration using time-lapse imaging and immunostainings (Sup- Fig 2A-A').
Thus, we considered it unlikely that HCs employ a protease-dependent path generation strategy to move through the crowded retina.
Typically, amoeboid migrating cells undergo cytoplasmic and nuclear deformations which allow them to navigate through crowded environments either in vivo (Friedl, Wolf and Lammerding 2011, Wolf et al. 2003, Salvermoser et al. 2018, Manley et al. 2020 or in narrow channels in vitro. We thus asked whether such deformations accompanied and/or influenced HC migration behavior. To achieve mosaic labeling of cells, we either used blastomere transplantation of Tg(lhx-1:eGFP) to mark cell bodies, or injected trbeta2:tdTomato and LAP2b:eGFP DNA constructs to visualize cell bodies and nuclear envelopes of HCs, respectively. We observed that migrating HCs exhibited a wide variety of cellular and nuclear shape changes, ranging from elongated to bended (Fig 3A-A'').
To quantify these morphological alterations, we used the open-source Icy platform Video 4). We measured perimeter (µm), elongation ratio and sphericity (%) of cell body and nucleus of HCs during migration, as well as during their final mitosis ( Fig   3B), on their basal-to-apical journey. At mitosis, HCs displayed an elongation ratio of ~1 ( Fig 3D) and a sphericity of ~100% (Fig 3E). This implied that cell bodies and nuclei of mitotic HCs adopt a spherical shape (Fig 3A'- To test whether a correlation between encountered tissue obstacles and shape changes of HCs existed, we monitored HC migration (Tg(lhx-1:eGFP)) in relation to the surrounding local environment (Tg(hsp70-H2B:RFP)). We noted that migrating HCs featured cellular deformations when encountering a neighboring cell that entered mitosis (Fig 3F-I, Video 6). In other cases, when the tissue seemed impassable, HCs changed both their direction of migration and cellular shape, often taking less direct routes to their final position. Notably, the adaptive cellular and nuclear deformations shown by HCs were reversible after the physical barrier was circumvented (Fig 3F-I).
Together, these data support the notion that migrating HCs tailor their cellular and nuclear shapes, paths and directionalities to their surrounding tissue environment.
This behavior does not generate migration trails but could rather serve as a spaceadaptation strategy enabling HCs to successfully move within the densely-packed and dynamically complex retina. Many types of amoeboid migration from Dictyostelium to leukocytes (Friedl and Wolf 2010, Trepat, Chen and Jacobson 2012, Arts et al. 2021 show similar morphological changes. Therefore, our results suggest that HCs undergo amoeboid-like migration in the zebrafish retina.
In contrast, we found that HCs feature multiple dynamic protrusions with different directionality as they moved sideways (tangentially), up (towards the apical) or down (towards the basal) within the INL. These protrusions showed different thicknesses, lengths, morphologies (branched vs. unbranched) and orientations, and were dynamically extended and retracted from the cell soma in multiple directions (Fig 4A,   Video 7). At times, two or multiple protrusions were seen to extend from the cell body at the cell front (Fig 4A), another feature also reported in amoeboid migrating cells (Weber et al. 2013, Renkawitz et al. 2019, Kameritsch and Renkawitz 2020. One possible component that could be responsible for front-rear polarity in HCs during migration is the asymmetric distribution of stable F-actin as reported for amoeboid migrating cells including leukocytes (Cassimeris, McNeill and Zigmond 1990), dendritic cells in 3D (Insall andMachesky 2009, Lammermann et al. 2008), neutrophils (Yoo et al. 2010, Manley et al. 2020, Barros-Becker et al. 2017) and neutrophil-like cells in vitro (Cooper, Bennin and Huttenlocher 2008). We thus investigated F-actin distribution in HCs using two distinct actin bioprobes: Lifeact (17 amino acids of yeast Abp140), which labels all filamentous F-actin structures (Yoo et al. 2010, Fritz-Laylin et al. 2017, and utrophin (calponin homology domain) (Utr-CH) that has been shown to preferentially bind to a more stable cortical population of F-actin (Burkel, von Dassow and Bement 2007, Riedl et al. 2008, Yoo et al. 2010, Belin, Goins and Mullins 2014, Barros-Becker et al. 2017. We found that while Lifeact was observed in the cell soma and protrusions ( Fig 4D, Video 9), Utr-CH was absent from the cell soma, membrane-proximal regions and protrusions (Fig 4E, Video 10). Instead, measurements of Utr-CH fluorescent intensity profiles along the cell axis showed its enrichment at the back of the migrating HC ( Fig 4D, Fig 4F', Sup- Fig 4E). This is akin to the structure seen at the cell-rear of amoeboid migrating cells also known as uropod Front-back polarity manifestation of migrating cells is typically accompanied by asymmetric positioning of organelles, including centrosomes (Kupfer, Louvard andSinger 1982, Luxton andGundersen 2011). We thus monitored distribution and dynamics of centrosomes during HC migration, using Centrin:GFP (centrosome marker) and trb2:tdTomato (HC marker) DNA injections (Fig 4G, Sup-Fig 5A-D). Our analysis revealed that HC centrosomes displayed a highly variable and dynamic localization and continuously shifted their positions from the cell-front to the cell-back while occasionally staying in the cell's middle (Fig 4G-H, Sup-Fig 5A). This oscillating configuration suggested that centrosome position did not directly influence the direction of HC movement. Overall, we concluded that similar to amoeboid moving cells, HCs also acquire a polarized morphology with persistent rearward polarization of stable F-actin, most likely without contribution of centrosome position. This further supports the idea that HCs employ amoeboid-like migration strategies to move forward in the zebrafish retina.

CELL MIGRATION AND LAMINATION
We showed that HCs undergo frequent cellular and nuclear deformations (Fig 3A-D) while migrating through the densely-packed INL (Sup- Fig 2A-B''). Consequently, we asked whether changing the properties of nuclei as the biggest and bulkiest cell organelle (Martins et al. 2012, Lammerding 2011, could impact HC migration. Some nuclear properties are determined by the differential expression of type V intermediate filament proteins of A-and B-type lamins which are part of the nuclear lamina (Broers et al. 1997, Gruenbaum et al. 2005, Lammerding et al. 2006, Gerace and Huber 2012, Burke and Stewart 2013. Particularly A-type Lamins (A, C and C2) are inversely correlated to the deformability of the nucleus as increasing their expression levels has been linked to decreasing nuclear deformability (Lammerding et al. 2006, Harada et al. 2014, Rowat et al. 2013, Swift et al. 2013, McGregor, Hsia and Lammerding 2016. It was shown that nuclei in the developing zebrafish retina only express negligible levels of A-type lamins (Yanakieva et al. 2019). Thus, we probed whether and how changing nuclear properties by increasing the expression of Lamin A (LMNA) at the tissue-scale could influence HC migration.
To test this, we generated a zebrafish transgenic line Tg(hsp70:LMNA-mKate2) in which LMNA overexpression in all cells is induced upon heat-shock. We then quantitatively analyzed HC migration and layer formation in this condition. First, we studied HC layer formation 40 hrs after heat-shock, in fixed samples of Tg(lhx-1:eGFP) (as controls) ( Fig 5A) and Tg  x Tg(hsp70:LMNA-mKate2) double transgene retinae (Fig 5B), at 90 hpf, a developmental stage at which HC layer formation is complete in wild-type embryos. In control heat-shocked retinae, all HCs were positioned within the HC layer at this stage (Fig 5A), showing that heat-shock treatment of the embryos does not impair proper HC layer formation. In contrast, in the LMNA overexpressing retinae, many HCs were found at ectopic basal positions, mostly within INL and at times even in the GCL (Fig 5B).
To assess whether the apical migration of the ectopically-positioned HCs is delayed or abrogated as a result of LMNA overexpression, we performed long-term in vivo  Fig 5E). This implies that changing the nuclear laminar composition at the tissue-scale hampers the migration efficiency of HCs and thereby perturbs HC layer formation.

MIGRATION
Since HCs are unlikely to remodel their environment to generate their path, we wondered whether their migration path could be impacted by physical barriers in the tissue. In the developing zebrafish retina, five layers with discrete properties (Fig 2C-E'', Sup Fig 2B-B'') and topographical features emerge during neuronal lamination.
We previously showed that IPL, once it is formed, negatively influences the depth of HC migration, as HCs did not basally pass it (Amini et al. 2019). This suggested that IPL, despite being devoid of cell-bodies may act as a non-passable obstacle for migrating HCs. This is most likely due to the intermingled axonal terminals of BCs, dendritic trees of ACs and RGCs, and MG processes which form a dense neuropil enriched with membrane (Sup Fig 2B-B'', Sup Fig 5F).
To test this hypothesis, we set out to drive HCs to positions below IPL before its formation and explore how HCs deal with IPL once it was formed. To this end, we genetically eliminated RGCs, using a validated atoh7 morpholino (Kay et al. 2004, Kay et al. 2001, Pittman, Law and Chien 2008 that results in RGC depletion, and thereby delays IPL formation. In this condition, ACs are found at the most basal layer (Kay et al. 2001) intermixed with occasional HCs (Weber et al. 2014). Our immunofluorescence stainings of Tg(lhx-1:eGFP) retinae revealed that in atoh7 morphants at 50-55 hpf, the maximal depth of HC migration increased significantly and that many HCs were located adjacent to the basement membrane at the most basal side of the retina, a depth never seen for HCs in control embryos (Fig 6A, C).
Thus, upon interference with RGC emergence, HCs reached deeper positions in the tissue by moving beyond their typical basal stopping-point. This suggests that the RGC layer acts as a basal brake for HCs during their apical-to-basal journey.
Because these basal locations are completely different from INL wherein HC migration occurs, we asked whether ectopically located HCs are able to reach their apical layer beneath OPL. To determine this, we monitored HC migration in atoh7 morphant retinae until 82 hpf. We found that before IPL formation, HCs successfully turned back apically and reached the HC layer in atoh7 morphants, implying that the ability of HCs to move back apically is not abrogated in the absence of RGCs. In contrast, during or after IPL formation, HCs failed to reach the HC layer (Fig 6D-E). Consequently, while all HCs reached the HC layer at 82 hpf in control retinae, many remained ectopically constrained at positions below IPL, in RGC depleted conditions (Fig 6B,C). While, these ectopically located HCs were still able to move in all dimensions (up, down, lateral) below IPL, they failed to migrate apically through it. As a result, HCs remained trapped below IPL until they underwent apoptosis, evidenced by their progressive fragmentations, immobility and ultimately disappearance. Overall, we conclude that IPL despite being more compressible than nuclear layers (Fig 2D-E''), represents a barrier through which migrating HCs are not able to pass when trapped beneath. Thus, IPL likely poses a limit to the morphological adaptability of HCs and their nuclei.

DISCUSSION:
We here showed that HCs employ space adaptation strategies to navigate through the complex and crowded environment of the developing retina. In particular, we revealed that HCs repeatedly and reversibly adjust their cellular and nuclear morphology, and direction of movement to the constraints they encounter within their surrounding densely-packed tissue. Because we found that the migratory behavior and morphology of HCs share many hallmarks of amoeboid migration, we refer to HC migration as "amoeboid-like neuronal migration". To the best of our knowledge, this is the first study describing a neuronal cell type that undergoes amoeboid-like migration in a part of the developing CNS. We further uncovered that changing tissue properties can feedback on the efficiency of ameboid-like neuronal migration and layer formation most likely by influencing the space-negotiation capability of HCs.
As opposed to neocortical neurons including projection neurons which en route to their destination switch from multipolar to bipolar morphology and resume unidirectional radial migration along radial glia fibers (Nadarajah et al. 2003, Nadarajah et al. 2001, Cooper 2014), HCs do not seem to travel along radially oriented progenitors, MGs or BCs. During long stretches of their migration, HCs follow unconventional tortuous migratory tracks while frequently alternating between radial (up and down) and tangential (lateral) routes. Such migration trajectories also set HCs apart from other emerging retinal neurons including PRs (Rocha- Martins et al. 2021) and RGCs (Icha et al. 2016a), which display bipolar morphologies and remain constrained to radial routes due to their anchored processes.
That HCs move without anchorage during most of their journey allows them to move freely in all dimensions, a feature which in combination with HCs' flexible morphodynamic properties assists them to overcome obstacles of moving in the crowded environment by taking stochastic migration tracks (Amini et al. 2019). Thus, while HCs reproducibly and robustly reach the HC layer, their path selection is not intrinsically programed but rather influenced by the cellular surroundings and the tissue-scale parameters encountered in their local environment.
The amoeboid-like migration mode exhibited by HCs is not based on bleb formation but correlates with multiple highly dynamic actin-filled protrusions. The exact nature of these protrusions, and whether they directly drive HC migration, allow HCs to explore potential environmental cues or both, remain to be further explored. Our observation that protrusions are simultaneously extended towards multiple directions, especially during periods in which HCs stay stationary, suggests that they are rather involved in probing the tissue environment and pathfinding than directly propelling the movement.

Such exploratory roles have been proposed in amoeboid moving cells including
Dictyostelium during chemotaxis, leukocytes and neutrophils (Gupton et al. 2005, Wu et al. 2012, Leithner et al. 2016, Vargas et al. 2016, Fritz-Laylin et al. 2017, Gerisch and Hess 1974. Future experiments that specifically interfere with protrusion formation or maintenance will shed light on their exact role in HC migration and layer formation. We currently do not understand the mechanism(s) and forces that move HCs forward.
Many amoeboid migrating cells display a front-rear polarity wherein stable F-actin filaments are asymmetrically enriched at the highly contractile uropod (Hind et al. 2016, Bergert et al. 2012, Lammermann and Sixt 2009. Our finding that migrating HCs display strikingly similar polarized morphology implies that they may also use uropod contraction as a pushing force to move forward. Unraveling the spatiotemporal molecular machineries of cell polarity, force-generation, the signaling and the cytoskeletal elements that drive them will be exciting areas for future studies.
Using in vivo Brillouin microscopy, we showed that the Brillouin shift maps of INL remain relatively homogenous throughout HC migration. That no obvious compressibility gradient was observed along the apico-basal axis of INL during HC migration implies that tissue compressibility could have a permissive rather than an instructive role. The fact that interfering with tissue-wide components such as properties of the nuclear lamina of HCs and their surrounding cells, impedes HC migration efficiency and successful layer formation further argues in this direction.
An additional tissue-wide feature that influences HC migration is the emergence of IPL. We previously reported that the depth of HC migration correlates with IPL emergence and that once it is formed, HCs do not pass beyond it on their apical-tobasal journey (Amini et al. 2019). This together with our finding that HCs get trapped beneath IPL in RGC-depleted retinae, strongly suggests that IPL acts as a steric hindrance through which HCs cannot penetrate in either direction. While this interpretation may seem at odds with the finding that IPL is more compressible than INL according to the Brillouin shift profile, it is possible that the fibrillar arrangement of axonal and dendritic processes within IPL poses a net-like obstacle with low porosity that is below the deformation capability of the HC nuclei. This idea is in line with studies that showed that migration efficiency is optimal at pore diameters that match or range slightly below the diameter of cell's nucleus.
It remains unknown what external cues guide HC migration to ensure that HCs always find their accurate functional position while avoiding entrapment within the crowded retina. Such cues could either come in the form of mechanical gradients or chemical signaling or a combination of both. In the future, it will be important to explore the guidance cues that trigger reorientation of HCs toward the apical side where they later reside and function. It will be important to find their source, to understand how they change in space and time, and how HCs sense, integrate and prioritize these cues within the local structural features of their surroundings to successfully find and reach their destination.
Taken together, this study reveals that in addition to the numerous neuronal migration modes characterized so far, neurons can also undergo amoeboid-like migration in an important part of the developing CNS, the retina. The ability to undergo direction, celland nuclear-shape changes allows HCs to evade rather than degrade encountered barriers in the crowded tissue environment. It will be interesting to explore whether this mode of migration is specific to HCs in the zebrafish retina or conserved in retinae of other organisms and/or other regions of the brain. As multipolar migration modes are observed in many parts of the developing CNS, it is likely that amoeboid-like neuronal migration is widespread in diverse systems. Similarities and differences of what influences amoeboid-like migration in different systems will teach us more about the intricate development of the nervous systems in vertebrates of all kinds.

MATERIALS AND METHODS:
1. ZEBRAFISH WORK

Zebrafish husbandry
Wild-type TL zebrafish (Danio rerio) and transgenic lines were maintained and bred at 26°C as previously described. Embryos were raised at 28.5°C or 32°C and staged in hours post fertilization (hpf) according to (Kimmel et al. 1995). Embryos were kept in E3 medium, which was renewed daily and supplemented with 0.2 mM 1-phenyl-2thiourea (PTU) (Sigma-Aldrich) from 8±1 hpf onwards to prevent pigmentation. All animal work was performed in accordance with the European Union (EU) directive 2010/63/EU, as well as the German Animal Welfare act.

Zebrafish transgenesis
To generate Tg(hsp70:LMNA-mKate2), a stable transgenic line containing heat-shock inducible LMNA 1 nl of hsp70:LMNA-mKate2 (Yanakieva et al, 2019) was injected at 36 ng/ul, together with Tol2 transposase RNA at 80 ng/ul in double-distilled (dd)H2O supplemented with 0.05% phenol red (to visualize the injection material) (Sigma-Aldrich) into the cytoplasm of one-cell stage wild-type embryos. F 0 embryos were raised until adulthood. Germline carriers displaying mKate signal were identified in F0 progeny, after heat-shock treatment at 37°C, for 20 min, at 24 hpf. Carriers were then outcrossed with wild-type fish.

Transgenic lines
Refer to Table S1 for a list of transgene lines.

DNA injections
To mosaically label HCs or express proteins of interest in the zebrafish retina, DNA constructs were injected into the cytoplasm of one-cell stage embryos. Constructs were diluted in ddH2O supplemented with 0.05% Phenol Red (Sigma-Aldrich). Injected volumes ranged from 1 to 1.5 nl. DNA concentrations were 20-30 ng/µl and did not exceed 45 ng/µl when multiple constructs were injected. See Table S2 for a list of injected constructs.

Heat-shock
To induce expression of heat-shock promoter (hsp70) Table S3 for antibodies used.

Blastomere transplantations
Transplantation dishes were prepared by floating a plastic template in a Petri dish that was half-filled with 1% low-melting-point agarose in E3. Once the agarose solidified, plastic templates were gently removed, leaving an agar mould that contained rows of wells to hold embryos. Embryos at stages high to sphere were dechorionated in pronase (Roche) and dissolved in Danieu's buffer. Dechorionated embryos were transferred to wells in agarose molds using a wide-bore fire-polished glass pipet.
Approximately at the 1000-cell stage, cells from the donor embryos were transplanted into the animal pole of the acceptor embryos using a Hamilton syringe. Transplanted embryos were kept on agarose for about 3-5 h and then transferred onto glass dishes that contained E3 medium supplemented with 0.003% PTU and antibiotics (100 U of penicillin and streptomycin, Thermo Fisher Scientific). Transplanted embryos were identified via fluorescence and imaged from 42 hpf for 24-30 hrs.

in vivo light sheet fluorescent imaging
Imaging was performed on a Zeiss Light sheet Z.1 microscope as previously described

Confocal scans
Fixed samples were imaged in a laser-scanning microscope (LSM 700 inverted, LSM 880 Airy upright; ZEISS) or point scanning microscope (2photon inverted; ZEISS) using the 40×/1.2 C-Apochromat water immersion objective (ZEISS). The samples were mounted in 1% agarose in glass-bottom dishes (MatTek Corporation) filled with E3 medium and imaged at room temperature. The microscopes were operated with the ZEN 2011 (black edition) software (ZEISS).

Sample drift correction
First, maximum projected sub-stacks (five z slices) of the raw live images were generated in Fiji. XY-drift of 2D stacks was then corrected using a manual drift correction Fiji plug-in created by Benoit Lombardot (Scientific Computing facility, Max Planck Institute of Molecular Cell Biology and Genetics, Dresden, Germany). The script can be found on (imagej.net/Manual_drift_correction_plugin). The drift corrected movies were then used for tracking migrating HCs.

Tracking migrating HCs
The migrating HCs were manually tracked by following the centre of the cell body in 2D drift-corrected images using MTrackJ plug-in in Fiji (Meijering, Dzyubachyk and Smal 2012).

Deconvolution
The raw LSFM data was deconvolved in ZEN 2014 software (black edition, release version 9.0) using the Nearest Neighbor algorithm. Minimal image pre-processing was implemented prior to image analysis, using open-source ImageJ/Fiji software (fiji.sc).
Processing consisted of extracting image subsets or maximum intensity projections of a few slices. Processed files were analyzed in Fiji.

Morphodynamic analysis of HC migration
To quantify HC cell and nuclear morphodynamics , the free and open-source platform for bioimage analysis Icy (de Chaumont et al. 2012, Manich et al. 202) (http://icy.bioimageanalysis.org) was used to automatically digitize cell and nuclear contours. The "Active Contours" plugin was used to segment the contours of HC cell and nuclear outlines during migration and in mitosis (Materials and Methods).
Because the retina is densely-packed, the segmentation only worked when the cell of interest was singled out from the background and had enough distance from neighboring cells. To meet this goal, we used two different approaches: 1-cell transplantation (see section 1.7. Blastomere transplantations) and 2-DNA injection (see section 1.3.). A step-to-step manual of the protocols and plugins to measure cell and nucleus morphodynamics is available in (http://icy.bioimageanalysis.org) (Manich et al, 2020).
In this study, we extracted the following shape descriptors from Icy analysis: perimeter (µm), sphericity (%), elongation ratio (a.u.). "Perimeter" measures the perimeter of the region of interest (ROI) in micrometers. "Sphericity" is a measure of how similar to a sphere the ROI is. "Elongation ratio" is a scale factor given by the ratio between the first and second ellipse diameters of an ROI. The minimum value is 1 (for a round object).

Protrusion tracking and analysis
Protrusions were analyzed using light-sheet time lapse video recording of Tg(ptf1a:Gal4-VP16, UAS:gap-YFP) at 1 min intervals. The protrusion tips were manually tracked simultaneously with HC centroids using the MTrackJ plug-in in Fiji (section 3.2.). The angle between the protrusion and the direction of HC movement was defined as the angle between the unit vector defined by the direction of HC movement and the unit vector pointing to the protrusion tip from the HC centroid. All plots for this part of the analysis were created in Matplotlib (Hutner, 2007).

Utrophin fluorescence intensity distribution profiles
Utrophin fluorescence intensity distribution profiles of migrating HCs were measured in Fiji by drawing a line (width=3) along the cell axis at each time point. The utrophin signal intensity was measured using the max projection of 3 consecutive central z planes of the cell.

Mounting of zebrafish larvae for in vivo Brillouin
Embryos were anesthetized in MS-222 (0.02% in E3; Sigma-Aldrich) for approximately 20 minutes and placed in a lateral position on a glass-bottom dish suitable for optical imaging. Some specimens were placed on a polyacrylamide gel that acted as a spacer between the glass bottom and the embryo (for gel preparation see (Schlussler et al. 2018). A drop (200 μl) of low-gelling-point agarose (1% in E3, 30°C; Sigma-Aldrich) was used to immobilize the embryo. Immobilized larvae were then immersed in MS-222 (0.02%) and 1-phenyl-2-thiourea (PTU, 0.003%, Sigma-Aldrich) containing E3 during imaging. All embryos were released from the agarose embedding between Brillouin measurements and kept under standard conditions as described in Section 1.1.

Brillouin microscopy set up and data analysis
The Brillouin shift measurements were performed using a custom-built Brillouin

Competing Interest
The authors declare no competing financial interests.