Cephalopod Retinal Development Shows Vertebrate-like Mechanisms of Neurogenesis

Summary:

Coleoid cephalopods, including squid, cuttlefish and octopus, have large and complex nervous systems and high acuity, camera-type eyes. These traits are comparable only to features that are independently evolved in the vertebrate lineage. The size of animal nervous systems and the diversity of their constituent cell types is a result of the tight regulation of cellular proliferation and differentiation in development. Changes in the process of development during evolution that result in a diversity of neural cell-types and variable nervous system size are not well understood. Here, we have pioneered live-imaging techniques and performed functional interrogation to show the squid, Doryteuthis pealeii, utilizes mechanisms during retinal neurogenesis that are hallmarks of vertebrate processes. We find that retinal progenitor cells in the squid undergo nuclear migration until they exit the cell cycle. We identify retinal organization corresponding to progenitor, post-mitotic and differentiated cells. Finally, we find that Notch signaling may regulate both retinal cell cycle and cell fate. Given the convergent evolution of elaborate visual systems in cephalopods and vertebrates, these results reveal common mechanisms that underlie the growth of highly proliferative neurogenic primordia. This work highlights mechanisms that may alter ontogenetic allometry and contribute to the evolution of complexity and growth in animal nervous systems.

eTOC Blurb

Napoli, Daly et al. find that progenitor cells in the squid retina undergo interkinetic nuclear migration and that Notch signaling may regulate differentiation, similar to vertebrate processes. These results reveal the convergent evolution of cellular mechanisms that underlie the growth of highly proliferative neurogenic primordia.

Graphical Abstract

Introduction

Coleoid cephalopods (e.g. squid, cuttlefish and octopus) are charismatic invertebrates known for their expansive behavioral repertoire and large nervous systems, which have independently evolved from similar features found in vertebrate species^1,2 (Figure 1A). Although vertebrates have been well studied, the developmental changes that underlie the evolution of large, complex nervous systems are not well understood. Across animals, we know that in some cases neurogenesis is regulated through invariant numbers of precursors with fixed lineages (i.e. C. elegans), and in others, cell lineages are plastic and proliferation is temporally controlled (i.e. the vertebrate retina)^3–5. Live-imaging has revealed that neurodifferentiation and the regulation of growth is not only a genetic process but a cell biological process, dictated by orchestrated cell behaviors and tissue architecture. However, in vivo observations of cell behavior during neurogenesis have been primarily limited to traditional model species. To elucidate how changes in neurogenesis may contribute to the evolution of large and complex nervous systems, we sought to understand neurodifferentiation in the cephalopod. One of the striking features of the cephalopod nervous system is their highly acute camera-type eyes. The visual system has proven to be a powerful context to learn fundamental aspects of neural development and therefore we focused our investigations on retinal differentiation in the squid Doryteuthis pealeii.

Figure 1: The evolution of nervous system complexity in the cephalopod lineage.

(A) Phylogeny of animals with approximate neuron number in the brain and photoreceptor cell number in the retina where available. Tripedalia cystophora, which does not have a central nervous system, include an average of two counts of the ring nerve as well as rhopalia neuron counts, excluding retinal associated neurons. Photoreceptor cell calculations in Tripedalia cystophora include all retina associated cells in a single rhopalia^{10, 64–75}. (B) A hatching stage squid, Doryteuthis pealeii (stage 29)¹⁸. The red arrowhead points at one of the eyes. (C) A schematic of the hatching stage eye of D. pealeii. R: retina; L: lens; OS: outer segment; V: vitreous space; PR: photoreceptor nuclear layer; SC: support cell layer; BM: basal membrane. (D) A simplified schematic photoreceptor cells and support cells organization in the squid retina. See also Figure S1

The cephalopod retina is composed of two layers divided by the basal membrane (Figures 1B–D). Photoreceptor cell bodies are found posterior (further from the lens) to the basal membrane and support cells are found anterior (closer to the lens). The photoreceptor cells extend projections that pierce through the basal membrane to the anterior of the retina to form the outer segment, which is a region of expanded membrane, packed with proteins required for phototransduction (Figure 1D)^6,7. This arrangement is notably inverted relative to the vertebrate retina, which has photoreceptor cells and their outer segment in the posterior (Figures 2A, S1). Cephalopod photoreceptor cells also extend axons out of the retina and synapse directly on the optic lobe^6,7. The function of the support cells is not well understood but thought to act as photoreceptor sustaining glial-like cell as well a potential stem cell population for ongoing growth^8,9. Both photoreceptors and support cells are pigmented¹⁰. Previous work has shown that both support cells and photoreceptor cells are derived from bilateral, early emerging retinal primordia ⁹.

Figure 2: Growth and cell cycle during neurogenesis.

(A) Schematic of the vertebrate and cephalopod eye at adult, juvenile and early stage development. The early neurogenesis stage is prior to obvious organization and differentiation has occurred. Box shows enlargement of the vertebrate pseudostratified epithelium, apical to the left. Grey ovals are the nuclei distributed across the tissue early in development, darker grey nuclei are undergoing mitosis. Mitoses in the vertebrate retina are found in the posterior and in the squid retina they are found in the anterior (the apical side of the epithelium in each instance). Squid eye schematics are depicting adult eye, Arnold stages 29 and 23. R: retina; L: lens; OS: outer segment; V: vitreous space; PR: photoreceptor nuclear layer; SC: support cell layer; BM: basal membrane. (B) 10 minute BrdU pulse and immediate fix (Time 0) BrdU/PH3 antibody stain shows BrdU incorporation on the basal side of the epithelium (S-phase) and PH3 stain on the apical side of the epithelium (M-phase). Scale 50um. (C) Example of a BrdU+/PH3+ nucleus found on the apical side of the retina, 80 minutes after BrdU pulse. (D) Proportion of PH3 positive nuclei that are BrdU− and BrdU+ at 10 minute time intervals after an initial 10 minute BrdU pulse. SEM (standard error of the mean) is shown. (E) Example of a BrdU+/PH3− nucleus found on the apical side of the retina at 110 minutes after BrdU pulse. The incidence of apical BrdU+/PH3− nuclei is low. Scale 10um for both D and F.

Previous observations from fixed tissue suggest that the cephalopod retinal primordium is a pseudostratified epithelium, which is unusual for an invertebrate neurogenic tissue^9,11 (Figure 2A). This type of neuroepithelium is characteristic of central nervous system development in vertebrates, including the retina^12,13. A pseudostratified epithelium is a monolayer composed of elongated cells with nuclei distributed along the apicobasal axis. Cells in pseudostratified epithelia undergo nuclear movements correlated with the cell cycle called interkinetic nuclear migration¹⁴. During nuclear migration, mitosis occurs on the apical surface, which corresponds to the posterior side of the developing retina in vertebrates¹⁵ (Figure 2A). Eventually, cells delaminate and differentiate within the boundaries of the epithelium. These cell behaviors are required for proper nervous system development, although their function is not completely understood¹⁶. Supporting that a similar mechanism may occur in the squid, fixed time point data suggests that mitosis occurs only on the apical side of the developing retina (anterior in the eye), inverted relative to vertebrates9 (Figures 2A, S1). This inversion is a result of early epithelial morphogenesis. The vertebrate retina is the result of an invagination of the evaginated forebrain and the squid retina is a result of the invagination of a superficial placode¹⁷ (Figure S1). Here, we present descriptive and functional evidence that cephalopod neurogenesis shows exceptional similarity to vertebrate processes, suggesting that the deployment of common developmental mechanisms led to the evolution of convergent visual organ expansion.

Results

Cell cycle in the developing squid retina

Previous work in the squid retina showed that at Arnold stage 23, a three hour bromo-deoxyuridine (BrdU) pulse, an analog of thymidine incorporated in S-phase, resulted in incorporation throughout the retina^9,18. If nuclei are undergoing interkinetic nuclear migration, this long pulse would provide sufficient time for movement after incorporation. To more precisely understand cell cycle state in the early retina, we performed a 10 minute BrdU pulse experiment, fixed immediately, and performed immunofluorescence for phosphohistone H3 (PH3), a marker for mitosis (Figure 2B). BrdU incorporation is restricted to the posterior of the retina after 10 minutes. In addition, PH3 positive nuclei, in late G2 and M phase, are on the apical surface of the retina, supporting previous observations9. The substantial number of BrdU negative nuclei in the anterior retina suggests a population of cells that are in G1, G2 or post-mitotic.

If cells are undergoing interkinetic nuclear migration, we expect BrdU positive nuclei to eventually migrate apically (anterior) to divide. To test this hypothesis, we chased our 10 minute BrdU pulse, sampling embryos every ten minutes for 110 minutes, again co-labeling with PH3 (Figure 2C). At 70 minutes after the initial 10 minute pulse, BrdU positive, PH3 positive cells are on the apical side of the retinal epithelium, suggesting that nuclei do migrate apically (Figures 2D & 2E). If proliferative cells found on the basal side of the epithelium join a post-mitotic population on the apical side of the epithelium, we expect to find BrdU positive, PH3 negative cells apical after mitosis. Assuming most mitoses are not terminal, we also expect to find few BrdU positive, PH3 negative nuclei. Indeed, at 110 minutes we find rare BrdU positive, PH3 negative cells apical in the retinal epithelium (Figure 2F). Together, these data suggest that cells in the cell cycle have nuclei residing in the basal retina and these nuclei migrate apically to divide. Some cells exit the cell cycle and their cell bodies remain anterior. These data are consistent with cells undergoing interkinetic nuclear migration. Although high resolution data describing neurogenesis across species is limited, we hypothesize that this is an ancestral developmental mechanism that has been convergently deployed in the context of neurogenesis in both cephalopods and vertebrates.

Live-imaging observations of nuclear migration in the squid retinal epithelium

To fully understand the dynamics of these cell behaviors, it is necessary to observe development in vivo. This required us to generate a cell-resolution live-imaging protocol in cephalopods. We injected fluorescent Dextran to broadly label cell membranes and performed >9 hour time course experiments of the squid retina at stage 23 (Figure 3A). Within these datasets, we were able to track individual nuclei migrate from the basal lamina to the apical side of the retina, mitose, and migrate basally again (Figures 3B, 3C, Video S1, S2, S3). These data confirm the process of interkinetic nuclear migration in the squid retina. From these data we were able to calculate distance traveled and nuclear velocity and assess that migratory behavior is quantitatively similar to movements in the vertebrate retina. The mean nuclear velocity in the zebrafish and mouse retina is approximately 0.3 um/min (Figure 3D)^19–22.

Figure 3: Live-imaging of nuclear migration and mitosis in the retinal epithelium at stage 23.

(A) Schematic of squid embryo live-imaging. Scale is 50um. (B) False-colored time-lapse of a single mitosis in the squid retina at stage 23 showing the apical migration of a nucleus in the epithelium, mitosis, and basal migration. Apical is up, basal is down. Scale is 10um. (C) Graph of nuclear movement as tracked using the Trackmate Fiji plugin⁷⁶. 13 nuclei tracked in 3 embryos. Tracks aligned to the highest point in migration. (D) Graph of nuclear velocity calculated from the cells tracked in C. (E) PH3 immunohistochemistry on actin polymerization inhibited squid retina at stage 23. 5uM Cytochalasin D and DMSO control treatment for 7 hours and immediate fix. We see a significant increase in cells outside the mitotic zone in our CytoD treated embryo, suggesting actin polymerization is require for nuclear migration. Scale is 50um. (F) Quantification of PH3+ nuclei outside the mitotic zone. (G) PH3 immunohistochemistry on microtubule polymerization inhibited squid retina at stage 23. 5uM Nocodazole and DMSO control treatment for 7 hours and immediate fix. Relative to control, we find an accumulation of PH3 positive nuclei in the mitotic zone as well as a population of PH3 positive nuclei away from the mitotic zone, suggesting that microtubules do play some role in nuclear migration. (H) Quantification of PH3+ nuclei outside of the mitotic zone. (I) Example of symmetric and asymmetric cell divisions observed in live-imaging experiments. Measurement of angle of division shown. Scale is 10um. (J) Quantification of symmetric (60°−90°) and asymmetric (30°−60°) cell divisions, n=70 mitoses from videos of 5 embryos. (K) Radial histogram quantification of division angles. Each dot represents a single mitosis. See also Video S1, S2, S3

The cytoskeletal elements contributing to nuclear migration have been interrogated in a number of developmental contexts and their contribution has been shown to vary ^{12,14,20,23,24}. To understand the proteins involved in nuclear migration in the squid retina, we treated stage 23 embryos with either the actin polymerization inhibitor cytochalasin D or the microtubule polymerization inhibitor nocodazole for seven hours and fixed immediately. The impact of these inhibitors on nuclear migration was assessed using immunofluorescence for PH3 (Figures 3E–H). In control embryos, as expected, the majority of PH3 positive nuclei are in the mitotic zone, (MZ), defined as the region of the epithelium containing nuclei abutting the apical membrane. We also find a small percentage outside the mitotic zone in controls, which we hypothesize are at the end of G2, migrating toward the apical surface. Cytochalasin D and nocodazole treated retinas have PH3 positive nuclei displaced from the mitotic zone, suggesting that both actin and microtubule polymerization are required for nuclear migration (Figures 3E–H). We also find an accumulation of arrested mitotic nuclei at the apical surface after nocodazole treatment, which has been previously observed when microtubules are not required for nuclear migration²⁵. These data suggest heterogeneity in the cytoskeletal mechanism required for nuclear movement.

In addition to nuclear movement, we were also able to assess the angle of cell division in the epithelium relative to the apical surface. The plane of cell division that is 60–90 degrees relative to the apical surface are considered symmetrical, while 0–60 degrees is considered asymmetrical^26,27. Symmetrical and asymmetrical cell division is an essential aspect of regulating self-renewal, delamination, cell cycle exit and cell fate commitment during neurogenesis in multiple organisms^28–30. This phenomena is variable across bilaterian species and neurodevelopmental contexts, inside and outside pseudostratified epithelia and have been previously reviewed³. However, when angled cell divisions are observed, symmetrical divisions are commonly self-renewing. Most mitoses observed at stage 23 in the squid retina were symmetrical, defined by having a division plane angle between 60–90 degrees relative to the apical surface (Figure 3I, Video S2). We hypothesize that these divisions are self-renewing. Approximately 10% of cell divisions are asymmetric, angled 30–60 degrees relative to the apical surface (Figures 3I, 3J & 3K, Video S3). We hypothesize that these asymmetric cell divisions may contribute to cells exiting the cell cycle. We did not observe cell divisions perpendicular to the apical surface, (0–30 degrees) as observed in other systems^26,27.

Molecular identity during retinal neurogenesis

To better understand the molecular state correlated with cell cycle organization observed in our BrdU and live-imaging experiments, we sought to identify markers that define differentiation trajectories in the cephalopod retina. We analyzed spatiotemporal expression of thirteen candidate neurogenesis genes in the retina (Figures 4A, S2 & S3). In early retinal development (stage 21), the transcription factor DpSoxB1 is homogeneously expressed across the retinal epithelium. At stage 23 the epithelium is divided across the apical-basal axis with DpSoxB1 segregated to the basal side of the retina and the receptor tyrosine kinase, DpEphR, segregated to the apical side. At stage 25 we see the first evidence of the terminal differentiation markers DpRhodopsin and DpRetinochrome, a cephalopod photopigment and photo-isomerase. At stage 27, the first evidence of the basal membrane is apparent, which is correlated with a loss of strict segregation of DpSoxB1 and DpEphR expression. We find that DpSoxB1 expression is never found posterior to the basal membrane, however, DpEphR is found on both sides of the membrane until hatching (stage 29). At stage 29, DpEphR is isolated to a subset of cells posterior to the basal membrane. Terminal differentiation markers span both sides of the basal membrane at stage 29. We hypothesize that DpRhodopsin is expressed exclusively in the photoreceptor cells and the mRNA is being trafficked to the outer segment and that DpRetinochrome is expressed in both the photoreceptor cells and support cells.

Figure 4: Molecular marker expression suggests retinal differentiation trajectories.

(A) Staged mRNA expression in the developing cephalopod retina: DpSoxB1, DpEphR, DpRhodopsin, DpRetinochrome. Nuclear stain (SYTOX) shown in green. Schematic of stages shown. Arrowheads and black and white dotted lines identify the basal membrane. Red dotted lines identify the posterior retinal boundary. DpSoxB1 is homogeneously expressed across the retinal epithelium at stage 21. At stage 23, DpSoxB1 is basal, DpEphR is apical, correlating with BrdU positive and BrdU negative nuclei respectively. At stage 25, terminal differentiation markers, DpRhodopsin and DpRetinochrome, are evident. At stage 27 the basal membrane is evident and the stage 23 and 25 mRNA segregation dissolves. mRNA segregation is re-established as DpSoxB1 expression is never found posterior to the basal membrane and DpEphR is found on both sides of the membrane at stage 29. Terminal differentiation markers span both sides of the basal membrane at stage 29. Scale is 50um. RP, Retinal Placode; L, Lens; OS, Outer Segment; BM, Basal Membrane; PR, Photoreceptor Cells; SC, Support Cells; V, Vitreous Space; A, Apical; B, Basal. Stage 21, 23, & 25 embryo anterior is down. Stage 27, 28, & 29 embryo dorsal is up. (B) Hypothesis of cell differentiation trajectories. Color key is used in C. (C) High magnification summary of time course gene expression data. At stage 21 the retina is homogenous. At stage 23 the retina is divided into apical and basal regions. At stage 27, the first nuclei are found behind the basal membrane and strict segregation of apical and basal expression is lost. We hypothesize that the cell-bodies migrate or mix, with apical cells moving to the posterior of the retina and basal cells moving to the anterior of the retina. This resorting may move the post-mitotic cells to the posterior where they will become the first photoreceptor cell bodies to migrate behind the basal membrane. At stage 28, the number of photoreceptor cells behind the basal membrane has increased and gene expression shows increased regional segregation. At stage 29, differentiation markers are robustly expressed. See also Figures S2, S3, Table S1

As found in other cephalopods, we report a lack of expression of the canonical neurogenic ELAV in the retina, DpELAV, an RNA binding protein commonly found in differentiating neurons across species^31–38. However, we do find that both cephalopod-specific DpELAVL paralogs are expressed in the developing retina and brain (Figure S3). bHLH factors DpNeuroD and DpNeuroG, both with orthologs that have roles in neural specification and differentiation in many animals, are expressed in the lateral lips and brain in D. pealeii and other cephalopods but are completely absent from the retina^39–41 (Figure S3).

This gene expression survey suggests differentiation trajectories in the retina (Fig. 4B, 4C). DpSoxB1 is a marker for retinal progenitor cells that are undergoing interkinetic nuclear migration in the pseudostratified epithelium. Homologs of SoxB1 are required to maintain neural precursor cells in vertebrates (Sox1, 2 and 3) and neuroblasts in Drosophila (SoxNeuro and Dichaete)^42–45. Cells differentiating into photoreceptor cells transition to DpEphR positive cells and become post-mitotic. We hypothesize that DpSoxB1 is intracellularly segregated to the cell bodies within the epithelium while DpEphR positive cells are the first cells to delaminate from the epithelium. Eventually, DpEphR expressing cells begin to express terminal differentiation markers and migrate from the apical side of the epithelium to behind the basal membrane. This leads to the dissolution of early retinal gene expression segregation and the emergence of terminally differentiated DpRhodopsin- and DpRetinochrome-expressing photoreceptor cells. Cells differentiating into support cells also start as DpSoxB1 positive retinal progenitor cells. This self-renewing population is maintained throughout development. As photoreceptor cells migrate behind the basal membrane, DpSoxB1 positive support cells begin to express DpRetinochrome, and are found anterior in the retina, and remain in the cell cycle⁹. The DpSoxB1/DpRetinochrome positive support cells may be a stem cell population contributing to ongoing growth.

Notch regulates retinal progenitor identity in the cephalopod

Our data show that DpSoxB1 and DpEphR expression are similarly spatially segregated in the epithelium to cells in S-phase and post-mitotic cells, respectively (Figures 2B, 5A, 5B & 5C). Our BrdU chase experiments suggest that DpSoxB1 cells likely become DpEphR cells over time. Notch signaling is a well-known regulator of cell cycle exit and differentiation. Previous work showed that loss of Notch signaling leads squid retinal cells to exit the cell cycle⁹. We were interested in understanding the role of Notch signaling in regulating cell fate trajectories. We first performed in situ hybridization for Notch signaling pathway members (Figures 5D & S4A). We find that DpNotch and DpHes-1 expression is regionally correlated with DpSoxB1 expression and progenitor identity (Figures 5D & S4A). We also find DpNotch expressed in mitotic, PH3 positive, nuclei on the apical side of the retina (Figure 5E). Thus, inheritance of Notch mRNA may be an important part of maintaining progenitor cell identity.

Figure 5: Notch signaling is required to maintain retinal progenitor cell identity.

(A) DpSoxB1 expression on the basal side of the retina at stage 23 in retinal progenitor cells. (B) DpEphR expression on the apical side of the retina at stage 23 in post-mitotic cells. Scale 50um. (C) Schematic of nuclear migration within the pseudostratified retinal epithelium. Red (DpSoxB1) and yellow (DpEphR) color corresponds to gene expression and summary Figure 3B and 3C. (D) DpNotch and DpHes-1 expression correlates with DpSoxB1 expression in the retinal progenitor cell population. Scale 50um. (E) DpNotch expression on the apical side of the retinal epithelium correlates with PH3 expression. Scale 50um. (F-H) DAVID GO-annotations of RNA-seq results of DAPT inhibited eyes. Embryos treated with 20uM DAPT and DMSO control at stage 23 for 24 hours. Eyes dissected for sequencing. All results shown have a p-value < 0.05. Purple is down-regulated genes, yellow is upregulated. (F) Neural and sensory genes. (G) Cell-cycle genes. (H) Signaling pathway genes. Color Key is shared for all RNA-seq results. Names of genes are best BLAST hits in Homo sapiens or genes phylogenetically assessed here (fig. S2) or in previous work. DpSoxB1 was found downregulated and DpEphR was found upregulated in the dataset, but these changes were not deemed statistically significant. (I) In situ hybridization of Control and DAPT inhibited retinas. Embryos treated with 20uM DAPT and DMSO control at stage 23 for 24 hours and fixed immediately. Nuclear stain (SYTOX-green) shown in green in the first column. Selected down-regulated genes in DAPT-treated embryos in the purple box on the left. Selected up-regulated genes in DAPT-treated embryos in the yellow box on the right. DpSoxB1 expression, the retinal progenitor marker, is lost from the retina in DAPT-treated embryos. DpEphR expression, the neuroblast marker, has expanded expression throughout the retina in DAPT-treated embryos. See also Figure S2, S4, S5, Table S1 and S2.

To assess the role of Notch in cell fate we bathed embryos in the gamma-secretase inhibitor, N-[N-(3,5-Difluorophenacetyl)-L-alanyl]-S-phenylglycine t-butyl ester (DAPT). DAPT impacts Notch signaling by inhibiting gamma-secretase from cleaving the Notch intracellular domain and entering the nucleus. DAPT has been shown to impact Notch signaling in the squid⁹. Embryos were treated with DAPT or DMSO starting at stage 23 for 24 hours. Eyes were dissected and pooled. RNA was extracted from three experimental and control samples and sequenced. We found 2242 genes differentially expressed between DAPT and control eyes with a p-value of less than 0.05 (Figures 5F–H & S4B–F, Table S1). Comparing DAPT-treated to control eye samples showed both up- and down- regulation of neural and sensory markers, suggesting a shift in cell type identity (Figure 5F). In addition, cell cycle related genes were downregulated, as expected⁹ (Figure 5G). Finally, DAPT versus control eyes showed significant changes in cell signaling genes (Figure 5H).

To confirm changes in gene expression identified in the RNA-seq analysis, we performed in situ hybridization studies on control and DAPT-treated embryos (Figure 5I & S6). Specifically, we find a loss of DpNotch expression, a downregulation of DpRhodopsin, and a complete loss of DpSoxB1 in DAPT treated retinas (Figure 5I). We also observe a gain of expression of DpEphR and synaptic transmission related gene, DpSec-1, in the posterior retina, as well as ectopic expression of DpVEGFR in a subpopulation of cells in DAPT treated retinas (Figure 5I). The complete loss of DpSoxB1 and the ectopic expression of DpEphR shows that DAPT inhibition not only results in changes in cell cycle state, but also changes molecular fate in the squid retina. Unlike the consequence of Notch inhibition in the vertebrate retina, the cells do not appear to prematurely terminally differentiate but arrest at the DpEphR step in the trajectory⁴⁶. These data suggest that Notch signaling is required to maintain retinal progenitor identity in addition to regulating cell cycle exit in the squid.

Discussion

We have characterized the epithelial organization and oscillatory nuclear behaviors of retinal progenitor cells in squid. We have defined molecular markers and the regionalization of progenitor, post-mitotic, and differentiated cells in the retina, and we have shown that Notch signaling regulates both cell cycle and molecular identity during retinal development (Figures 4B, 4C and 6A). These findings are the first detailed analysis of the developmental cell biology in the largest invertebrate nervous system. We can use this as a powerful, independent example of nervous system elaboration to enable a greater understanding of mechanisms that allow for the evolution of complex sensory organs.

Figure 6: Summary of Notch signaling and early cell cycle state in animal neurogenesis within pseudostratified epithelia.

Organization of neurogenic epithelia shows similarity between the cephalopod retina and early vertebrate neurogenesis. Summarized Notch expression is mRNA unless otherwise noted. These yellow colored cells show the organization of the first post-mitotic cells during development. (A) Cartoon cross-section of the squid retina. Summary of retinal neurogenesis and Notch mRNA expression found in the current study. DpNotch mRNA is enriched basally (or posterior) and the first post-mitotic cells are found apically (or anterior). (B) Cartoon cross-section of the developing vertebrate retina. Notch mRNA expression has been observed expressed in a gradient enriched on the apical side (or posterior)⁵⁷. Notch protein has been found uniformly expressed across the epithelium⁶¹. Retinal ganglion cells are the first cells to exit the cell cycle, differentiate and migrate to the basal/anterior side of the retina⁷⁷. (C) & (D) are examples of conveyor belt neurogenesis⁵⁸. (C) Cartoon cross-section of the outer proliferative zone in the Drosophila optic lobe. The proliferative epithelium is Notch positive^78–80. Neuroblasts that have exited the epithelium continue to divide. (D) Cartoon cross-section of the juvenile zebrafish retina. Ciliary marginal zone contribution to juvenile and adult retinal neurogenesis and Notch expression is summarized . The proliferative zone is segregated from the differentiated cells. See also Figure S2.

The cell behaviors and organization we describe in the squid retina are unusual because of their unique similarity to early neurogenesis in vertebrate species^47,48 (Figures 6A & 6B). Pseudostratified epithelia and interkinetic nuclear migration have been observed in multiple tissue types and across multiple species and are thought to maximize mitoses during proliferative phases in development ^{12,16,23,49–53}. In the nervous system, pseudostratification was historically considered a vertebrate-specific developmental trait, responsible for the large size of the central nervous system ^12,50–53 (Figures 6B). Currently, the only other invertebrate neurogenic tissue where interkinetic nuclear migration has been well described is in the Drosophila optic lobe (54) (Figures 6C). In this case, the apicobasal distance of migration is significantly smaller than vertebrate neurogenic tissues and our observations in squid (Figures 6D) ^16,54. Invertebrate neurogenesis, in the Drosophila optic lobe and elsewhere, is currently characterized by individual cells delaminating or ingressing from a neuroepithelium, sometimes dividing, and then migrating away from the progenitor population to organize and differentiate^3,31,54–57. Neurogenesis in the Drosophila optic lobe has been compared to “conveyor belt” neurogenesis found in the zebrafish ciliary and tectal marginal zones, where, during long term growth, cells are progressively added from the edge of a pseudostratified stem cell population (58) (Figures 6D). This is unlike embryonic vertebrate neurogenesis and the process described in the present study, where organization and differentiation within the neuroepithelium allows for tissue-level morphogenesis, such as neural tube closure or optic cup formation⁴⁹.

In addition to maximizing proliferation, our data suggest an alternate function for pseudostratified epithelia. We show that Notch signaling is maintaining molecular identity of progenitor cells in the squid retina, and that the loss of Notch signaling leads to expression of the post-mitotic marker DpEphR. Notably, we find the mRNA for genes associated with progenitor cell identity (DpNotch, DpHes-1, DpSoxB1, and DpSoxB2) localized basally within the epithelium. This asymmetric localization is similar to mRNA gradients that have been reported during vertebrate neurogenesis^59,60. Currently, the function of these mRNA gradients are not understood⁶¹. Our data reveal that, DpNotch mRNA, localized basally, is found apically during mitosis, indicating that intracellular mRNA localization may enable the regulated segregation of RNA during symmetric and asymmetric cell divisions in densely packed epithelia (Figures 5E).

Although we have now confirmed that pseudostratified neurogenesis occurs in the retina, little is known about neurogenesis in the cephalopod brain. It has been reported that cells migrate from the lateral lip, an embryonic neurogenic tissue, to regions within the brain³⁶. Pseudostratification of the lateral lips is not apparent, indicating that this development may resemble canonical invertebrate neurogenesis^36,62. However, with a significantly larger brain than other spiralians, requiring long distance cell migration, it is likely that cephalopods have evolved lineage specific mechanisms to manage this process.

The evidence we have generated demonstrates that vertebrate-like cell behaviors during neurogenesis are not exclusive to the chordate lineage and identifies shared developmental tools deployed to build large nervous systems. This is a novel example of developmental process homology or the use of an ancestral developmental mechanism that may be independent of character identity⁶³. This sheds light on cell behaviors and transcriptional modifications that may contribute to changes in allometry and cell type variation found across the diversity of animal nervous systems. Ultimately, this work illuminates the necessity for phylogenetic sampling of cell-resolution expression and live-imaging data to better understand fundamental mechanisms in developmental biology.

STAR Methods

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Kristen Koenig (kmkoenig@fas.harvard.edu)

Materials availability

All unique/stable reagents generated in this study are available from the lead contact

Data and code availability

RNA-seq data have been deposited at GenBank and Dryad Digital Repository are publicly available as of the date of publication. Accession numbers are listed in the key resources table. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request. This study did not generate any unique code.

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
BrdU antibody	Abcam	Cat# Ab6326
Phosphohistone H3	Sigma Aldrich	Cat# 06–570
Goat anti rat Alexa Fluor 488	Invitrogen	Cat# A-11006
Goat anti rabbit Alexa Fluor 647	Invitrogen	Cat# A-31573
Bacterial and virus strains
Biological samples
Chemicals, peptides, and recombinant proteins
Nocodazole	Sigma Aldrich	Cat# M1404–2MG
Cytochalasin D	Sigma-Aldrich	Cat# C8273–1MG
TRIzol	Invitrogen	Cat# 15596026
Dextran Alexa Fluor 488 10,000 MW	Invitrogen	Cat# D22910
Sytox Green	Invitrogen	Cat# S7020
Mounting Medium for IHC	Abcam	Cat# ab64230
Critical commercial assays
gDNA eliminator mini-spin column	Qiagen	Cat# 1030958
Kapa mRNA-Hyper Prep kit with Poly-A selection	Roche	Cat# 08098115702
pGEM-T Easy vector	Promega	Cat# A1360
Deposited data
DAPT:Control RNA-seq raw data	This paper	GenBank BioProject accession PRJNA794156
D. pealeii mRNA sequences	This paper	OM481467-OM481487
Multiple sequence alignments and phylogenetic trees	This paper	https://doi.org/10.5061/dryad.rr4xgxdc4
Assembled and annotated embryonic transcriptome for mapping	9	https://orcid.org/0000-0001-6093-2262
Experimental models: Cell lines
Experimental models: Organisms/strains
Doryteuthis pealeii	Marine Resources Center, Marine Biological Labs	N/A
Oligonucleotides
Primers sequences, see Table S2	This paper	N/A
Recombinant DNA
Software and algorithms
FIJI 2.1.10	76	https://imagej.net/software/fiji/downloads
Coordinate Shift	Housei Wada	https://signaling.riken.jp/en/en-tools/imagej/
Trackmate 7.7.2	92	https://imagej.net/plugins/trackmate/
TrimGalore! v3	93	https://www.bioinformatics.babraham.ac.uk/projects/trim_galore/
Bowtie2	94	http://bowtie-bio.sourceforge.net/bowtie2/index.shtml
Kallisto	95	https://pachterlab.github.io/kallisto/
DESeq2	97	https://bioconductor.org/packages/release/bioc/html/DESeq2.html
DAVID	98	https://david.ncifcrf.gov/
GOPlot	99	https://sourceforge.net/projects/gplot/
gplots	100	https://cran.r-project.org/web/packages/gplots/index.html
VolcaNoseR	101	https://huygens.science.uva.nl/VolcaNoseR/
BLAST	86	https://blast.ncbi.nlm.nih.gov/Blast.cgi
MAFFT v.7.450	87	https://mafft.cbrc.jp/alignment/software/
FastTree2 v.2.1.11	89	http://www.microbesonline.org/fasttree/
IQ-TREE 2 v.2.1.0	90	http://www.iqtree.org/
Photoshop 2020	Adobe	https://www.adobe.com/products/photoshop.html
Geneious	Dotmatics	https://www.geneious.com/
Other

Experimental model and subject details

Doryteuthis pealeii egg sacks were obtained from the Marine Biological Labs. Egg sacks were kept at 20 degrees Celsius in 20 gallon aquaria in artificial seawater under a day/night cycle. Although not required, European guidelines for cephalopod research were followed.

Method details

Cloning and in situ Hybridization

Primers were designed using Primer3 in the Geneious software package version 2020.04 (https://www.geneious.com) and primer sequences are reported in Table S2. Genes were cloned into pGEM-T Easy vector and confirmed with Sanger sequencing and DIG-labeled RNA probes were synthesized as previously reported⁹. Embryos were fixed overnight at 4 degrees Celsius, washed and dehydrated stepwise into 100% ethanol. Embryos were embedded, paraffin sectioned and in situ hybridization was performed as previously reported⁸³. All in situs were replicated in at least three embryos, across multiple separate in situ experiments. Slides were stained overnight with SYTOX-Green 1:1000 overnight, mounted using ImmunoHistoMount (Abcam) and imaged on a Zeiss Axioskop 2.

BrdU Experiments

Embryos were bathed in BrdU (10 mM) in pen-strep seawater for 10 minutes as previously described⁹. Embryos were fixed immediately and the remaining embryos were moved into pen-strep seawater. Groups of 10–15 embryos were fixed every ten minutes for two hours. Embryos were fixed overnight at 4 degrees and washed out into PBS-Tween. Embryos were stepped into 30% sucrose and embedded in tissue freezing medium and sectioned as previously described⁹. Antigen-retrieval and immunofluorescence were performed as previously described⁹. BrdU antibody (Abcam ab6326) at concentration 1:250. Phosphohistone H3 antibody (Sigma-Aldrich 06–570) was used at concentration 1:300. Secondary antibodies goat anti-rat Alexa Fluor 488 and goat anti rabbit Alexa Fluor 647 (Invitrogen). Sections were counterstained with SYTOX Green at 1:1000. Sections were imaged on a Zeiss LSM 880 or Zeiss LSM 980.

Live-imaging experiments

Embryos were dissected from egg cases. Previous studies have shown non-specific labeling of cell membranes or cell walls with exposure to fluorescently labeled Dextran^84,85. This is also the case in squid. The vitreous space was injected using a pico-liter microinjector with Dextran Alexa Fluor 488 10,000 MW (D22910) at stage 23. Ubiquitous membrane labeling was apparent within a half hour of injection. Embryos were embedded in 1% low melt agarose in seawater and mounted in cover glass bottom dishes (100503–366). Embryos were immersed in pen strep seawater. Embryos were imaged on a Zeiss 880. Embryos were imaged every ten minutes for at least nine hours.

Ex ovo and Drug Treatments

Embryos were treated in 20uM DAPT solution or DMSO control in filter-sterilized, Pen-Strep seawater starting at stage 23 for 24 hours and fixed immediately as previously described⁹. Embryos were bathed in 5uM Nocodazole (M1404–2MG, Sigma-Aldrich) and 5uM Cytochalasin D (C8273–1MG, Sigma-Aldrich) for 7 hours and fixed immediately.

Quantification and statistical analysis

Phylogenetic Analysis

Genes were first identified by using annotated sequences from model organisms from major lineages for BLAST⁸⁶ into a custom local database of the D. pealeii transcriptome in Geneious. For top hits the entire sequence in the D. pealeii transcriptome was retrieved, the longest ORF was extracted and translated, then the amino acid sequence was trimmed for coding sequence. To find related sequences, BLASTp was used, searching both the Uniprot database in NCBI and retrieving only select vertebrate and D. melanogaster hits. BLASTp was performed again using the non-redundant protein database, and searching specifically for cephalopods, select mollusks, and Limulus. Trees that were not well resolved after these steps required an additional round of BLASTp, this time including more spiralian and ecdysozoan hits. Full sequences (or as long as is available) were aligned with our D. pealeii sequences for each tree using MAFFT v.7.450 in Geneious⁸⁷. The only exception was our Sox tree where we used the alignment from⁸⁸, which only included the HMG box of Sox proteins. This alignment focused on early metazoan species, so we added select vertebrates, mollusks, and ecdysozoans as described above, but trimmed sequences to include the HMG box for all. For all alignments we checked sequence redundancy and proper outgroups Fast Trees were made using FastTree2 v.2.1.11⁸⁹. We constructed maximum-likelihood trees on the FASRC Cannon cluster supported by the FAS Division of Science Research Computing Groat Harvard University. We exported relaxed Phylip formatted alignment files and used IQ-TREE 2 v.2.1.0 with the following settings: iqtree2 -s ALIGNMENT.phy -st AA -nt AUTO -v -m TEST -bb 1000 -alrt 1000 ⁹⁰. Unrooted trees were visualized as rooted by known outgroups and labeled by known annotated orthologues.

Imaging analysis

Image analysis was performed in Fiji⁹¹. Intensity range was adjusted in Fiji to better identify cell membranes. Drift correction was performed using Fiji plugin CoordinateShift (written by Housei Wada, https://signaling.riken.jp/en/en-tools/imagej/). Nuclear tracking was performed both manually and using Fiji plugin Trackmate^91,92. Tracks were visualized and distance and velocity measurements were obtained from Trackmate and plotted graphically, normalizing to the highest point in migration. Image analysis for BrdU and PH3 was performed manually on central retinal sections of at least 3 embryos per time point and between 3–6 retinas. +/−SEM (standard error of the mean) was calculated and shown. Image analysis for PH3 in CytoD and Nocodazole treated embryos was performed manually on at least 3 embryos per treatment, with 4–6 central sections counted. +/− SEM was calculated and shown.

RNA-seq and Bioinformatics

Stage 23 embryos were treated with the gamma-secretase inhibitor DAPT at 20um in filter sterilized sea water for 24 hours and Control embryos were treated with the equivalent amount of DMSO, as previously described⁹. DAPT and Control eyes were dissected, pooled and macerated in TRIzol and stored at −80 degrees Celsius. RNA was extracted using a standard TRIzol (Invitrogen #15596026) chloroform extraction and passed through a gDNA eliminator mini-spin column (Qiagen #1030958). RNA was precipitated with isopropanol and then again precipitated with ethanol and checked for quality. Library prep and sequencing was performed at the Bauer Core at Harvard University. RNA-seq libraries were generated using the Kapa mRNA-Hyper Prep kit with Poly-A Selection (Roche, Basal) and were sequenced on the Illumina NovaSeq (>70 million 2×150 bp sequences) (Illumina, San Diego, CA).

Sequence quality control was performed according to the best practice recommendation on the Harvard FAS Informatics pipeline (https://informatics.fas.harvard.edu/best-practices -for-denovo-transcriptome-assembly-with-trinity.html). Erroneous kmers were removed from the paired end Illumina dataset using rCorrector. Reads with Ns or other low complexity pairs were removed using a custom Python script provided by the Harvard Informatics GitHub (FilterUncorrectabledPEfastq.py). Adapters and low quality bases were removed using TrimGalore!⁹³. Reads that mapped using Bowtie2 to the rRNA databases SILVA 132 SSURef Nr99 tax and SILVA 132 LSUParc tax were removed⁹⁴. Pseudomapping was performed by Kallisto with 100 bootstraps to a previously published whole embryo transcriptome ^95,9. Transcript abundances were imported using tximport (https://bioconductor.org/packages/3.7/bioc/vignettes/tximport/inst/doc/tximport.html#use-with-downstream-bioconductor-dge-packages ) into DESeq2 using the Kallisto abundance.h5 files. Differential gene expression was determined by importing transcript level abundances and gene level offset using DESeqDataSetFromTximport⁹⁶. The DESeq2 pipeline was run and differentially expressed genes were considered with a p value of .05 and log2 fold change of above 1 and below −1 ⁹⁷ (Table S1). All genes shown in the chord plots in Figure 4 meet these criteria with the exception of Notch which was included with a log2 fold change of −0.83. GO annotations were identified using the DAVID functional annotation tool⁹⁸. Chord plots were generated using the R package GOPlot ⁹⁹ and gplots ¹⁰⁰. The volcano plot was generated using VolcaNoseR ¹⁰¹.

Supplementary Material

3

Video S1. Whole retina time course experiment, Related to Figure 3 An optical section of a developing stage 23 squid retina, imaged every 10 minutes for 10 hours. Each frame is a single plane from a z-stack of a representative long-term live-imaging experiment. Cell membranes are labeled with dextran. The apical side of the epithelium is up, and the basal side of the epithelium is down. Anterior of the eye is up, posterior of the eye is down. The mitoses that are visible in the plane are highlighted with blue arrowheads. The time is shown in the left corner in hours:minutes. Scale is 25um.

4

Video S2.Example of a symmetrical cell division, Related to Figure 3 An example of a symmetrical cell division within the retinal epithelium isolated from a long-term live imaging experiment. The first evidence of two symmetrical daughters is highlighted with two blue asterisks. The embryo was imaged every 10 minutes. Apical is up of the epithelium is up. Anterior of the eye is up. Scale is 10um.

5

Video S3. Example of an asymmetrical cell division, Related to Figure 3 An example of an asymmetrical cell division within the retinal epithelium isolated from a long-term live imaging experiment. The first evidence of asymmetrical daughter cells is highlighted with two blue asterisks. The embryo was imaged every 10 minutes. Apical is up in the epithelium. Anterior of the eye is up. Scale is 10um.

6

Table S1. Notch versus control Deseq2 results table, Related to Figure 5 and STAR Methods

Highlights.

• Retinal progenitor cells in the squid undergo interkinetic nuclear migration.

• Progenitor, post-mitotic and differentiated cells are transcriptionally defined.

• Notch signaling may regulate both retinal cell cycle and cell fate in the squid.