The evolution of the vertebrate eye remains enigmatic (Lamb et al., 2007; Lamb, 2013). This is in particular due to the limited number of extant species at the invertebrate/vertebrate lineage transition (at the base of the chordate lineage). One such notable example is represented by cephalochordates (also known as amphioxus or lancelets), the most basally branching chordate clade. This clade includes three genera – Branchiostoma, Asymmetron and Epigonichthys and in total about 30 different species (Poss and Boschung, 1996). To date, the genomes of the three species of Branchiostoma have been sequenced (Holland et al., 2008; Putnam et al., 2008; Huang et al., 2014; Marletaz et al., 2018). Their analysis confirmed that cephalochordates did not undergo the two rounds of whole-genome duplication that occurred in vertebrates (Holland et al., 2008; Putnam et al., 2008; Huang et al., 2014; Marletaz et al., 2018). The phylogenetic relationship within the extant amphioxus lineage was investigated providing divergence time estimates. Molecular dating analysis based on whole-genome nuclear transcriptome revealed a divergence time of ∼120 Ma for Asymmetron lucayanum and Branchiostoma floridae (Yue et al., 2014), being comparable with that of the marsupial/placental split (Benton et al., 2009). This is somewhat less than ∼162 Ma estimated from mitochondrial gene sequences (Nohara et al., 2005; Kon et al., 2007). In either case, informative evolutionary divergence applies to Branchiostoma and Asymmetron pair (Yue et al., 2016). Much more recent diversification was found within the Branchiostoma genus (Igawa et al., 2017).

Comparison of differences and similarities between cephalochordates and vertebrates provides valuable information about ancestral chordate traits as well as vertebrate-specific innovations. Cephalochordate body plan possesses typical chordate characters like segmented muscles, dorsal hollow neural tube, notochord, perforated pharynx with gill slits and through gut - traits that can be found also in vertebrates. On the other hand, cephalochordates lack several vertebrate-specific characters, e.g., neural crest cells, paired appendages (fins or limbs), or well-developed paired sensory organs – eyes and ears (reviewed in Bertrand and Escriva, 2011). Amphioxus possesses four distinct photoreceptive organs: frontal eye, lamellar body, Joseph cells and dorsal ocelli (reviewed in Pergner and Kozmik, 2017). The frontal eye is considered as homolog of vertebrate eye based on its topology, ultrastructural morphology, gene expression pattern and circuitry (Lacalli et al., 1994; Lacalli, 1996; Vopalensky et al., 2012; Suzuki et al., 2015; reviewed in Pergner and Kozmik, 2017). The lamellar body is putative homolog of vertebrate pineal gland, mainly on the basis of its ultrastructural morphology, circuitry and topology (Ruiz and Anadon, 1991; Lacalli et al., 1994; Zieger et al., 2017). Possible vertebrate counterparts of Joseph cells and dorsal ocelli (both formed by rhabdomeric photoreceptors) are still matter of debate (reviewed in Pergner and Kozmik, 2017). The frontal eye is a single simple organ located at the tip of amphioxus cerebral vesicle – homolog of the vertebrate brain. Due to its simple organization, the frontal eye does not possess image-forming capacity. It is formed by only about six photoreceptor cells located in one horseshoe-shaped transversal row (Lacalli et al., 1994; Lacalli, 1996; Vopalensky et al., 2012). These cells were also designated as Row1 cells. Ultrastructural characterization of frontal eye demonstrated that Row1 photoreceptors are of ciliary morphology as are rods and cones (photoreceptors) in vertebrate retina (Lacalli et al., 1994). Row1 photoreceptors ultrastructure is, however, less elaborate than that of vertebrate photoreceptors (Lacalli et al., 1994). Anteriorly from Row1 photoreceptors lie nine pigment cells, arranged in three rows each consisting of three pigment cells (Lacalli et al., 1994). Posteriorly from photoreceptors lie three rows of putative homologs of vertebrate retinal interneurons and/or possibly also retinal ganglion cells (RGCs). These are the so-called Row2, Row3 and Row4 cells, respectively (Lacalli et al., 1994). The arrangement of these rows of cells is less organized than that of Row1 putative photoreceptors (Lacalli et al., 1994). The proposed homology between the frontal eye and vertebrate eyes was strongly supported by molecular studies in one cephalochordate species, B. floridae, showing that orthologous genes are expressed in both the frontal eye and vertebrate retina and pigmented epithelium (Vopalensky et al., 2012). Transcription factors Pax2/5/8, Mitf and Otx were detected in the developing row of frontal eye pigment cells (Vopalensky et al., 2012). Otx, Pax6, c-opsins and Gi subunit of trimeric G proteins were found to be expressed in frontal eye photoreceptors (Vopalensky et al., 2012). Taken together, these results were in concordance with gene expression profile of vertebrate photoreceptors and retinal pigmented epithelium (RPE) but provided limited information about possible homology between Row2, Row3, and Row4 cells of frontal eye and vertebrate retinal interneurons and/or RGCs. In vertebrates, Pax6 is expressed in most of the interneurons and RGCs and is required for their proper differentiation (Marquardt et al., 2001; Klimova and Kozmik, 2014). B. floridae Pax4/6 expression was not detected in Row2 cells but was instead found to be expressed in Row3 and Row4 cells (Vopalensky et al., 2012). Vopalensky et al. (2012) demonstrated that the Row2 neurons in B. floridae were positive for serotonin (alternatively called 5-hydroxytryptamine- further in the text marked as 5HT). More recently it was shown that some of the Row3 and Row4 neurons utilize glutamate as neurotransmitter (Pergner and Kozmik, 2017). Another criterion that can help to address homology between particular neuronal subtypes in the vertebrate retina and amphioxus frontal eye, is directionality of their axonal projections (used for example by Lacalli, 1996). Amacrine and horizontal cells form in vertebrate retina branches connecting neuronal layers laterally, while bipolar and ganglion cells form relay connections between retinal layers and visual processing center (Erskine and Herrera, 2014). Additionally, majority of RGCs in most vertebrates form contralateral projections (Erskine and Herrera, 2014). Based on electron microscopical studies and immunofluorecent imaging in B. floridae it was shown that Row2 cells form ipsilateral projections resembling bipolar cells (Lacalli, 1996; Vopalensky et al., 2012). Row3 cells form lateral contacts with each other resembling amacrine and horizontal cells, but also short irregular projections distantly resembling bipolar cells or RGCs (Lacalli, 1996). Some Row4 cells form contralateral projections resembling typical contralateral RGCs projections forming chiasma complex (Lacalli, 1996). Moreover, Lacalli (1996) observed that Row4 cells are in contact with Row2 cells. EM study also indicated that Row2 cell processes do not form any kind of terminals which questioned their ability to serve as interneurons (Lacalli, 1996). This could mean that they are involved in some kind of visual circuitry beginning with Row1 cells as photoreceptors, Row2 as interneurons, and Row4 as homologs of RGCs. Vopalensky et al. (2012) nevertheless observed that Row2 cell projections lead to an area of amphioxus cerebral vesicle that might be homologous to vertebrate visual processing center (tectum). The same was observed by Suzuki et al. (2015). A recent study focusing on homologies between amphioxus cerebral vesicle and particular regions of the vertebrate brain demonstrated that Row2 cell processes might indeed terminate in putative amphioxus proto-tectum of mesencephalic origin (Albuixech-Crespo et al., 2017) (reviewed in Pergner and Kozmik, 2017). It is, however, impossible to make any definitive conclusion about possible homology between Row3 and Row4 cells and vertebrate retinal interneurons and/or RGCs.

Limited data exist about the development of amphioxus frontal eye region and differentiation of neural progenitors into individual “retina” cell types. Expression of Otx and Pax4/6 genes whose orthologs are known for their crucial role in vertebrate eye development was detected in the early neurula stage in developing cerebral vesicle of B. floridae (Williams and Holland, 1996; Glardon et al., 1998). Both of these genes are nevertheless likely involved not only in frontal eye development but also in the development of other parts of amphioxus nervous system. It is of note, that differentiated c-opsins-positive photoreceptors were observed in B. floridae frontal eye not earlier than by the larval stage at 48 h post-fertilization (hpf) (Vopalensky et al., 2012). No information is currently available about the role of signaling cascades in the development of amphioxus frontal eye. Yet, conservation of gene regulatory networks orchestrated by specific signaling pathway may aid the general effort to assign homology between cell types across divergent species. In the present study we analyzed involvement of Notch signaling in the development of amphioxus photoreceptors and putative interneurons of the frontal eye. It is well established that Notch signaling regulates development of vertebrate photoreceptors. Inhibition of Notch signaling was shown to result in an increased numbers of photoreceptors in the mouse retina (Jadhav et al., 2006; Nelson et al., 2007). The previous molecular study of Vopalensky et al. (2012) was limited to a single amphioxus species, B. floridae, leaving uncertainty about general conservation of the observed gene expression fingerprint across all cephalochordates. In the present study we therefore provide comparative molecular data using three amphioxus species, B. floridae, B. lanceolatum, and the divergent A. lucayanum. Our results extend previous observations (Vopalensky et al., 2012) and provide novel insight into neural circuitry and development of the frontal eye. Combined, our study strengthens the previously proposed homology between amphioxus frontal eye and vertebrate eyes.

The first aim of our study was to determine, whether previously reported frontal eye molecular fingerprint in B. floridae (Vopalensky et al., 2012) is conserved among cephalochordate species. The two species we selected for analysis are represented by a closely related B. lanceolatum, and the most distantly related A. lucayanum. We found that the overall molecular fingerprint of frontal eye pigment cells, photoreceptors and putative interneurons (Row2, Row3, and Row4 cells) was conserved among the three cephalochordate species (summarized in Figure 7A). However, we observed several conspicuous differences. Heterochrony in the appearance of the frontal eye between B. lanceolatum and B. floridae corresponds reasonably well with the overall differences in developmental timing. These two species develop at different temperatures which leads to a generally faster development of B. floridae embryos growing in warmer water (summarized in Fuentes et al., 2007). Heterochrony in frontal eye development observed between B. floridae and A. lucayanum is unexpected since both species inhabit relatively close geographical localities and develop at similar temperatures. No melanin-based pigmentation of the frontal eye in A. lucayanum was detected even by 8 dpf while the same type of pigmentation in the 1st dorsal ocellus was readily observed (Supplementary Figure 7). Lack of cross-reactivity of antibodies directed against B. floridae ciliary opsins toward A. lucayanum c-Ops1 and c-Ops3 precluded detection of their expression in the frontal eye as possible evidence of its functionality. Nevertheless, our results showed that 5-HT-positive Row2 cells are present in 8 dpf-old A. lucayanum larva, and Otx-positive photoreceptors produce glutamate (Supplementary Figure 4) indicating that both photoreceptors and Row2 cells are terminally differentiated and physiologically active by this developmental stage. The absence of frontal eye pigmentation, however, challenged the potential role of frontal eye in light-guided behavior of A. lucayanum, when compared with previously published work on Florida amphioxus. The study of Stokes and Holland (1995) demonstrated that directional light source triggers B. floridae larvae to adopt a specific vertical orientation in the water column which allows photoreceptors to be shielded by the pigment cells of the frontal eye (Stokes and Holland, 1995). Although not shown experimentally, we speculate that due to lack of pigmentation 8 dpf-old A. lucayanum larva might be unable to orient suggesting that the frontal eye can only function as a non-directional photoreceptive organ at this stage of development.

Previous studies of amphioxus frontal eye did not yield much information about putative interneurons Row3 and Row4 cells. So far, only Pax4/6 was found to be expressed in Row3 and Row4 cells. We chose to analyze expression of amphioxus Lhx1, Lhx3, and Brn3 due to their well-described function in vertebrate retina. Lhx1 was shown to be necessary for proper development of horizontal cells (Poche et al., 2007), Lhx3 is expressed in bipolar cells (Elshatory et al., 2007), and Brn3 is crucial for development of RGCs (Liu et al., 2000; Quina et al., 2005). We found that cells in Row3 and Row4 express both Lhx1 and Lhx3 (summarized in Figure 7B). In contrast, Brn3 expression was not detected in any of either Row3 or Row4 cells. Lack of Brn3 staining does not exclude the presence of putative homologs of RGCs in this area of the frontal eye but it makes it currently less likely. One cannot rule out that amphioxus homologs of RGCs are represented by dorsal ocelli, a scenario based primarily on dorsal ocelli morphology and the fact that they express rhabdomeric type opsin melanopsin (Koyanagi et al., 2005) (see also discussion in Pergner and Kozmik, 2017). Combined, our data suggest that cellular populations in Rows3/4 may be homologous to bipolar or horizontal cells (summarized in Figure 8).

To provide more insight into neuronal circuitry, we aimed to determine the nature of neurotransmitters utilized by individual frontal eye neurons. In vertebrates, most of the cells forming visual circuitry – photoreceptors, some populations of bipolar cells and most RGCs – utilize excitatory neurotransmitter glutamate (reviewed in Kolb, 2011). The same can be observed in amphioxus, where photoreceptors, Row3 and Row4 cells (putative interneurons) were found to be positive for glutamate (Figure 4). Some populations of amacrine cells in the vertebrate retina were found to be positive for inhibitory neurotransmitters glycine and GABA. Their proposed role is to modulate the transmission of the visual signal. We were unable to detect GABA or glycine in the region of putative frontal eye interneurons. Glycinergic neurons were instead found in the region where processes of Row2, Row3, and Row4 cells terminate. Here, glycinergic neurons might act as a neuromodulator of frontal eye circuitry. The presence of GABA was restricted to PMC motoneurons and might thus act in the modulation of motoneurons response to light stimuli. Our previous study revealed direct innervation and indicated the paracrine release of serotonin from Row2 cells in the tegmental neuropil in the posterior cerebral vesicle of B. floridae, reminiscent of retinohypothalamic projections in the vertebrates (Vopalensky et al., 2012). In the present study, we confirmed the restricted expression of 5-HT in Row2 cells using two other cephalochordate species, B. lanceolatum, and A. lucayanum, indicating a conserved nature of this molecular signature. The usefulness of serotonin-positivity in Row2 cells for drawing homologies between cell-types in the frontal eye and vertebrate retina nevertheless remains limited. 5-HT is produced by vertebrate photoreceptors, as a precursor for melatonin synthesis, and by amacrine cells, but the role of 5-HT as a neuromodulator in the vertebrate retina is not well understood (reviewed in Masson, 2019). Serotogenic neurons in the vertebrate brain are known to act as modulators of GABAergic and glutamatergic neurons (Ciranna, 2006). One can speculate that Row2 cells might act similarly as modulators of signal from photoreceptors to Row3 and Row4 cells. Row3 and Row4 cell axons terminate in the region of cerebral vesicle that likely serves as a visual processing center homologous to vertebrate optic tectum. In vertebrates, the expression of several POU transcription factors was observed in the developing tectum (Brombin et al., 2011). These included Brn2 (Pou3f2) and Brn4 (Pou3f4) genes, and a gene from Brn3 (Pou4f) family. Using specific antibodies we detected expression of Brn1/2/4, a single amphioxus ortholog of the vertebrate Brn2 and Brn4, in the putative tectum. The expression of amphioxus orthologs of Brn3 and Lhx3 was found in PMC motoneurons correlating well with the known role of Brn3 and Lhx3 in the development of vertebrate motoneurons (McEvilly et al., 1996; Thaler et al., 2002). Combined, the available data suggest that the frontal eye circuitry might be built upon similar logics as in vertebrates. We propose that amphioxus frontal eye circuitry starts with Otx positive photoreceptors utilizing excitatory neurotransmitter glutamate. Then it continues with Pax4/6 positive Row3 and/or Row4 cells (putative interneurons) utilizing also glutamate. Serotonin positive Row2 interneurons as well as glycine positive neurons posterior to Row4 cells might serve as inhibitory neurons modulating frontal eye response. Visual stimulus is then processed by neurons, which develop in Brn1/2/4 positive putative tectum. This would mimic the situation in vertebrates where photoreceptors and neurons transmitting the signal - bipolar and ganglion cells – utilize glutamate and the glycinergic amacrine cells serve as inhibitory modulators.

To provide an insight into the gene regulatory mechanisms underlying frontal eye development and specification of its unique cell types we performed pharmacological treatments during amphioxus embryonic and larval development. We specifically focused on Notch signaling cascade, since in vertebrates the inhibition of Notch signaling leads to an increase in the number of photoreceptors (Jadhav et al., 2006). Our analysis has shown that Notch inhibition similarly results in an increased number of photoreceptors in the amphioxus frontal eye. In addition, following Notch inhibition we observed a highly significant increase in the number of the putative interneurons present within Row3 and Row4 (summarized in Figure 9A). In this context it is interesting to note that overexpression of Delta (ligand of Notch signaling) in vertebrates leads to a decrease of specific later-born retina interneurons (bipolar and amacrine cells) (Dorsky et al., 1997). In contrast, the number of early born neurons (ganglion or horizontal cells) was increased after Delta overexpression indicating that the phenotype was caused by changing the competence of retinal progenitor cells. Precise timing of pharmacological interventions allowed us to uncover windows of competence for differentiation of frontal eye cell types (summarized in Figure 9B). Notch inhibition at the earliest developmental stage (neurula stage N2) led to increased number of photoreceptors, Row3, and Row4 cells, and a complete absence of Row2 cells. The responsiveness of photoreceptor lineage to Notch signaling was lost by the neurula stage N3. In contrast, the fate of Row2, Row3, and Row4 could still be affected by DAPT treatment performed at N3 but not at larval stage L1. Finally, differentiation of melanin-producing pigment cells was sensitive to Notch inhibition applied in any of the examined stages. Differential time window sensitivity is likely related to the onset of differentiation of a particular cell type. Combined, Notch signaling is involved in the regulation of correct ratios of neuronal cell types within amphioxus frontal eye, i.e., the function that appears to be analogous to that described for the vertebrate retina.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

JP and ZK designed the study, conceived the experiments, and wrote the manuscript. JP, ZK, and AV performed the experiments. IK and ZK provided new reagents. All authors have read and approved the manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.