Multilineage Differentiation Potential of CNS Cell Progenitors in a Recent Developed Gilthead Seabream (Sparus aurata L.) Nervous Model

Neural Progenitor Cells (NPCs) have gathered more and more attention in the field of Neural Stem Cells (NSCs). However, the multilineage differentiating behavior of these cells and their contribution to tissue regeneration, almost in lower vertebrate taxa, remain unknown. Since the early 1970s, many comparative studies have been performed using immunocytochemical screening on the brains of several vertebrate taxa, including teleosts, in order to identify these cells, even if the data are sometimes contrasting. This study aims: (1) to investigate in vitro the potential proliferative role of NPCs and Radial Glia Progenitors (RGP) in seabream neurogenesis; (2) to reveal the strict ability of fish NSCs to undertake the multilineage development and differentiation in neurons, astrocytes and oligodendrocytes.


Introduction
The generation of new neurons in the adult Central Nervous System (CNS), so called adult neurogenesis, is a fundamental feature of the vertebrate brain during the post-embryonic neural growth [1]. It is well known that in mammals and birds, constitutive turnover of neurons occurs not only during embryogenesis but also during adult stages [2][3][4][5]. Adult neurogenesis has been observed in rodents and songbirds, where newborn neurons are added throughout life to highly restricted spatial regions in the telencephalon [5][6][7][8][9]. In vivo in mammals, endogenous Neural Progenitor Cells (NPCs), almost generated along the Sub Ventricular Zone (SVZ) in the adult brain, have been proposed as a potential source of newborn neurons for the neural tissue repair after various brain insults, such as ischemic stroke or traumatic brain injury [10,11].
Dissimilarly to birds and mammals, low non-mammalian vertebrates, as fish, amphibians and reptiles, display a considerable amount of adult neurogenesis in more numerous brain regions, even though such differences are poorly understood and investigated [11]. The adult CNS of teleost fish exhibits a high capacity for neuronal regeneration after injury [4]. Comparative studies in zebra fish and mammals show that the telencephalic Ventricular Zone (VZ) in the adult zebra fish brain generates NPCs that share characteristics with the NPCs in the mammalian SVZ: they migrate tangentially and differentiate into mature neurons into the Olfactory Bulb (OB) [5,[11][12][13][14][15]. Recently, Kishimoto et al. [11] developed an in vivo fish model of telencephalic injury to demonstrate the strong ability of adult zebra fish to undergo neuronal regeneration. Prodigious regenerative processes were found in the injured zebra fish telencephalon, with the proliferation of endogenous NPCs, their lateral migration and neuronal differentiation at the injury site [11].
In zebra fish, as in rodents, neuroblasts become quiescent after larval development; while in the adult brain, often after brain injury, a Notch signaling pathway prompted by growth factors regulate the transition of NSCs cells to a proliferative state, into neuroblasts toward neurogenesis. However, the extrapolation on the NSCs self-renewing capabilities arose mainly from in vivo studies of rodents and mammals, and we are still lacking evidences on the truly proliferative ability of multipotent NSCs and whether these cells can undergo multilineage differentiation to develop mature and self-renewing neurons or glial cells in brain.
Nevertheless, none of these studies on fish has addressed in a definite manner the involvement of radial glia (RDG) in the generation of new neurons, by interplaying with the full development of the fish nervous tissue in an in vitro microenvironment. Due to several discrepancies, arose the need for improving differentiation procedures able to generate highly pure populations of mature neurons and glial cells in in vitro culture systems of lower vertebrates.
For this reason, we recently developed a seabream nervous model to study neuronal differentiation and regenerative processes in vitro. Using this model, we aimed: (1) to investigate in vitro the potential proliferative role of NPCs and Radial Glia Progenitors (RGP) in fish neurogenesis; (2) to reveal underlying the strict ability of fish NSCs to undertake the multilineage development and differentiation in neurons, astrocytes and oligodendrocytes.
This study provides a novel seabream nervous model for studying a series of regenerative processes involved in fish brain neuronal regeneration, giving new insights for the treatment of injury-impaired neurogenesis, even in mammals.
Primary neuroglial cultures were rapidly isolated from freshlyenucleated brains [35]. After three washings with three volumes of HBSS (WS), brains were crushed by pressing a glass pestle into a stainless-steel sieve with 380 µm mesh for mechanical disruption. Cell suspension was filtered and collected with Digestion Medium (DM) lacking enzymes (7 mM CaCl 2 , 200 IU/ml penicillin, 200 µg/ ml streptomycin, 200 µg/ml amphotericin B, 100 µg/ml gentamicin, 10 mM HEPES, 25 mM NaHCO 3 in Leibovitz's L-15). The homogenate was then extracted by adding a cocktail of four enzymes including 0.05% Collagenase Type IV, 0.03% Hyaluronidase Type IV-S, 0.3% Dispase Type II, 0.03% DNase Type I for 20 min dissociations at 20°C to the homogenate in DM. The enzyme digestion was blocked adding Leibovitz's L-15 medium supplemented with 10% Foetal Bovine Serum (FBS; Bio Whittaker, Lonza Walkersville Inc., Italy). The digestion mixture was filtered through 104 µm and 60 µm stainless-steel filters, supernatant was discarded, and the pellet was resuspended in cold 1X Phosphate-Buffered Saline (PBS) 5% Fetal Calf Serum (FCS) and centrifuged twice at 150 g for 10 min at 4°C. The white pellet was recovered, washed and dissociated by adding 0.45 µg ml -1 DNase I in cold PBS to dissociate cell clumps by Pasteur pipette, then centrifuged at 70 g for 5 min at 4°C. The final collected pellets were resuspended in 10% FBS/L-15 medium and cell number was counted. Viability of cells was estimated by trypan blue exclusion. Brain cell suspension yielded 5.2×10 7 viable cells/g per tissue weight with a viability of 92.8 ± 0.06%.

Neuronal and radial glial cells
Studies on NSCs defined these cells as self-renewing CNS cells that can differentiate into any of the three major neural cell lineages, specifically neurons, astrocytes and oligodendrocytes [16,17]. To study the potential role of NPCs and Radial Glia Progenitors (RGP) in fish neurogenesis, the neuroglial primary culture of the fish S. aurata has been used by our laboratory as a teleost CNS model system.
Multilineage development and differentiation of neurons, astrocytes and oligodendrocytes from S. aurata neuroglial cells in primary culture (SaGliPs), previously isolated in our laboratory, were morphologically characterized by inverted phase-contrast microscopy, and analyzed by Motic Images Advanced (V. 3.2) acquisition software (Motic, Seneco, Milan, Italy) [35]. In the panel of Figure   Within the first week after plating, the NGRPs extended their basal process in a radial direction, thus gaining the typical form of RDG rosette like colony (Figures 2b-2d). The contrast phase analysis has been shown that the new cell production occurred in specific mitotically active proliferation zones. Inside the RDG, a typical cell population of neuronal perikaryal appearance started to proliferate directly from the foot of cone-shaped radial cell, thus forming a cluster of highly proliferative cells (Figures 2e, 2e', 2f). From these proliferation zones, the newly formed young neuronal cells are guided, in their migration to the final target areas, by RDG fibers (Figures 3a,3b).
The neurogenic niches with astroglial morphology riches of highly replicating cells were observed during the 2nd week of differentiation (Figures 2 and 3). Fish neurogenic islands (generated by RGCs) showed a greater strong ability to regenerate neural cells, succeeding in a continuous generation of young neurons and RDGc renewal ( Figures  2 and 3).
To determine whether RDGc were mitotically active, we tested their mitotic activity by fluorescence DNA markers (Figures 3e, 4d, 4e, 5e, 5f ', 5h): (1) the UV-light excitable DAPI used as a DNA specific fluorochrome, as it appears to associate with AT clusters in the dsDNA; and (2) the Proliferating Cell Nuclear Antigen (PCNA), the most widely used expression marker of proliferation-associated proteins. PCNA is an antigen detected by antibody to cyclins D, E, A, and B which is specific for G1/2 and S, M phases of the mitotic cycle [40,41].
The morphology and distribution of the radial fiber indicates the type of dividing cells, as a consequence of different cleavage plane during cell cycle division of RDGc (Figures 3a-3f). During the early stage of fiber generation, the basal fiber of RDG remained elongated; it appeared like tired out forming a short radial stub projecting from the soma, that thins distally (Figure 2a'', 2b, 2g). Our results suggest that the vertical cleavage plane result only in neuronal production of daughter cells similar in shape (Figure 3a), while the horizontal cleavages produce two types of asymmetric divisions by generating (1) one new radial glial cell and one neuron progenitor (cell/cell horizontal cleavage) (Figure 3b); (2) one radial fiber and one neuron progenitor (fiber/cell horizontal cleavage) (Figure 3d). Furthermore, the contrast phase and immunofluorescence analysis revealed that young RDGc, during their cell division, generate two different types of daughter cells, one destined to become a long fiber process and the other destined to develop a neuron (Figures 3c, 3c', 3c''). In particular, the end of the fiber grew a thick process that would break, generating a neuron that started to create a line of newborn neurons by symmetric divisions (Figure 3). Such kind of neurogenesis was a peculiar characteristic of the seabream RDG model (Figure 3).
Two populations of glial fibers were apparent, one immunopositive for GFAP, the other for vimentin (Figure 4d). Immunocytochemistry analyses for double-or triple-immunofluorescent labelling were performed to define the status of differentiation of cultured cells. After 2 weeks of culturing, immunofluorescence revealed elongated DAPIlabelled nuclei, in close apposition to GFAP-labelled radial glial fibers (Figure 4d), thus signifying that NPCs grew closely opposed to the RDG fibers (Figure 4e). The GFAP-positive RDG fibers increased near the proliferating site of new neurons. It seems conceivable that these   neurons moved from end to the tip of the radial process by nuclear translocation, as visualized with the DNA marker (DAPI), thus reaching their final target site predominantly by the nucleus migration within the fiber ( Figure 5).
Between the end of the second week and the beginning of the third of culturing, the seabream RDG colony assumed a specialized aspect, with RDG fibers arranged in an interlaced scaffold to which NPCs were closely apposed, possibly for their radial migration towards the final target site. The scaffold was organized with GFAP positive differentiated astrocytes with many filopodia contacting multiple axons (vimentin positive) in order to sustain oligodendrocytes during the myelination process (Figures 4d and 4e). Neurons were apparent when fish CNS cultures were at 2 weeks of culturing, emerging from the tip of RDG fibers ( Figure 5). The morphology and the immunophenotyping of neural cells were displayed by the immunostaining with antineurofilament (NeuN), anti-βIII-tubulin, anti-nestin, anti-vimentin antibodies and the neuron-specific markers for ubiquitin carboxy-terminal hydrolase 1 (UCHL1) ( Figure 5).
We tested neural differentiation of NPCs by examining the expression of maturation protein markers. To characterize the first early events of differentiation, we studied markers of NPCs during the immature status. After 1 week of culture, most immature precursors cells (IPCs) neural structures contained neuroepithelial cells immunoreactive for nestin. Nestin is an intermediate filament protein commonly expressed by many progenitor cells, and here appeared to be highly expressed in high density into neurospheres and IPCs processes ( Figure 5). Most nestin positive cells contained PCNA expression which colocalized with the DAPI counterstaining, thus counting for mitotic figures ( Figure 5). RDGc can produce new generation of young neurons even in the old culture as revealed by PCNA within the inner cluster ( Figures  4d and 4e). PCNA is a cell cycle dependent intranuclear protein, which assists the delta-DNA polymerase during DNA replication, whereby its expression is considered a marker of DNA synthesis, being detectable in G1/S phase [40]. By producing new neurons and guiding them to the growing edge of colony, radial glial cells confirmed their relatively strong ability as in the adult fish brain to regenerate neural tissues.
In a first early event, the basal fibers of RDGc supported the production and migration of new born neurons (Figure 5a-5a'''), around the edge of the RDG rosettes (Figure 5c and 5c'). Whereas during the phase of neuronal differentiation a new circuit formation of horizontal fibers occurred later, displaying an interlaced network, as a mesh woven of axonal fibres closely connected ( Figure 4). Next, to evaluate the transition from IPCs to mature neurons, we tested two neuronal differentiation markers, βIII-tubulin to label immature neurons and NeuN the definite marker of mature neurons. After 14-20 days of differentiation, the cultures were highly enriched in neurons expressing the βIII-tubulin marker ( Figure 5). The colocalization of βIII-tubulin/NeuN was expressed almost by differentiating neurons, while cells immunoreactive for NeuN were cells which completed the transition from immature to mature neurons. All differentiated neurons presented the multilineage neural phenotype, defined as coexpression of the following markers gradient oriented according the developmental status: GFAP, Nestin, Beta III-tubulin, NeuN, UCHL1 ( Figure 5). During long term culturing the density of fibers markedly increased within 3 weeks, and differentiation up to 4 weeks achieved a complete coverage of the astrocyte bed with a massive arborisation of neurites ( Figure 6). The density of vimentin-positive RDG fibers increased over the monolayer as long as the neuritis full-grown ( Figure  6). By analyzing the sequence of events taken in the formation of astroneuronal tissue in vitro, it appears apparent that the formation of the fiber network of RDG is a plastic prerequisite needed to enable the support of growth and migration of the newly generated neuronal cells (Figure 6), as confirmed by the presence of actively dividing nuclei, doublestained with Ab anti-PCNA-FITC and counterstained with DAPI (blue), within the fiber extension (Figure 5a''). Finally, even in the 3 rd week old cultures were still observed mitosis both for newborn neurons migrating along neuronal fibers (UCHL1+/PCNA+/DAPI+) and within the inner rosette-like colony, suggesting the coexistence of differentiating astrocytes and neurons with newly formed cells.

Astroglial cells
Two specific progenitor cells, Glial-Restricted Precursor (GRP) cells and oligodendrocyte/type-2 astrocyte progenitor cells (O-2A/ OPCs) have been studied as distinct glial precursor subtypes restricted to astrocyte development in mammals; they only differed in the differentiation potential and in the timing of their developmental appearance for the astrocyte subtypes produced, type 1 and type 2 astrocytes [42,43].
Neural seabream mixed cultures grew best in SaGliPs-enriched growth factors medium supplemented with serum, and could be maintained for at least 2 months. The astrocytes displayed a heterogeneous phenotype, including a highly branched fibrous astrocytes (type-II astrocytes) and polygonal stellate cells (type-II astrocytes) (Figures 7a, 7b, 7e and Figures 7c and 7d respectively). After 2 weeks, SaGliPs cultured in enriched medium started acquiring features of differentiation, such as the typical appearance of mature astrocyte forming a tight monolayer and generating neurons ( Figure  7 and 8). According the lineage derivation the NGRPs-derived the type-I astrocytes, which firstly formed a confluent layer, and secondly the other cells grew on top; neurons grew in small clusters of RDGc with large bundles of axons. Next, after 3-4 weeks of culturing, neurons formed on the astrocyte bed a dense dendrite and neurite branching radiating outward (Figure 9).  In seabream neural cell cultures, after 2 weeks of culturing, the growth velocity of fish astrocytes seemed to be higher, seabream cells developed quickly, reaching confluent culture within 2 weeks ( Figure  7), fish cells were flat and stellate (Figure 7). The GFAP astroglial lineage marker was searched to highlight the presence of cultured astrocytes using anti-GFAP antibodies. In the fish model a typical astrocyte bed was depicted by the GFAP labelling after 3 weeks of culture.

Oligodendroglial cells
Oligodendrocytes, like other neural cells in the CNS, develop from the neuroepithelial precursors and are one of the last cell types to differentiate in the CNS.
In 1 week old cultures we observed cells with a simple bipolar form identified as O-2A progenitor cells undergoing a morphological transition during the 2nd week of culturing ( Figure 9) to prooligodendrocytes (pro OLIGO), with a more complex and flattened form enriched of multiple processes, accompanied by a change in cell surface antigens through the S100B expression acquisition which is essentially restricted to the oligodendroglial cell lineage (Figure 10b).
We observed, by CF and microscopic time lapse imaging, that proliferative oligodendrocyte progenitor cells converted into mature oligodendrocytes and after 2 weeks oligodendrocytes were able to recognize neurons and wrap their target axons (Figures 9 and 10).
After 3 weeks of culturing, fish oligodendrocytes stopped to divide, completed their differentiation transforming in mature Oligodendrocyte (mOL), and enveloped axons. In panel of Figure 10 we showed all stages of lineage development including proliferation, differentiation and long-term survival of oligodendrocytes up to the ultimate progression into fully-fledged oligodendrocytes expressing myelin products released in membrane sheats (Figures 9 and 10). In Figure 10, mOL showed a round cell body enriched of cytoplasmic granules, with a dense filamentous meshwork, initially enwrapping neuroglial processes in the peripheral areas.

Discussion and Conclusion
Since a long time, studies on NSCs defined as self-renewing CNS cells that can differentiate into any of the three major neural cell lineages, specifically neurons, astrocytes and oligodendrocytes, have been attempt almost in mammals [16,17]. Because of the lack of data on the culture of fish neural cells, with the exception on studies from the adult zebra fish injured brain, we carried out the characterization and lineage differentiation of NPCs from seabream CNS, primarily isolated by our research group and known as SaGliPs [35]. This is the first well characterized in vitro CNS teleost model, getting the chance of producing newly neural cells for long time in culture. The differentiation of neuronal lineages in seabream has never been described before. Previous studies supported the notion that, also in fish, radial glia may be a potential source of new neurons. For instance, while lower vertebrates display the capacity to generate new retinal neurons in response to injury, in teleost new neurons arise mainly from RDG following neural induction, and their subsequent development and differentiation [44].
We detailed in seabream the early stage of RDG generation; firstly the basal cell producing the fiber of Radial Glia (RG) remained elongated, appearing like tired out from the base and forming a short radial stub projecting from the soma. Studies from Miyata et al. [45] demonstrated that, during RDGc division, each neural connected fiber becomes thin but is neither lost nor divided, and it is inherited with an asymmetrical cleavage by one daughter cell. In these divisions were produced a neuron and a progenitor cell; the neuron inherits the pial fiber, and is therefore indistinguishable from the progenitor RDG. Our work confirmed results obtained from such recent in vitro and in vivo studies, examining the neurogenic potential of RDGc by characterizing the morphology and modes of neurogenic division planes as cited by Noctor et al. [33] Chenn and McConnel [47]. Moreover, our findings on the fate of the radial fibers, during RDGs development, revealed that in seabream the young RDGc, throughout their cell division, generate two different types of daughter cells, one destined to become a long fiber process and the other destined to develop a neuron. We found also that the cleavage plane will result only in neuronal production of two daughter cells similar in shape, while the horizontal cleavages will produce two types of asymmetric divisions by generating (1) one new RDGc and one NPC (horizontal cleavage versus cell/cell), and (2) one radial fiber and one NPC(horizontal cleavage versus fiber/cell).
Recently, also in in vivo studies from the zebrafish, RDGc were found to be able to function as multipotent retinal stem cells that responded to the loss of photoreceptors by specifically regenerating the missing neurons [46]. The work of Chenn and McConnell [47] showed that as neurogenesis proceeded, vertical cleavages decreased, while horizontal cleavages increased. This appear very consistent with our finding, in that the vertical cleavages were symmetric divisions producing two similar precursor cells; whereas horizontal cleavages were asymmetric divisions, resulting in the production of one neuron and one RDG precursor cell [48].
In regards to the formation of fiber scaffold, after 2 weeks of culturing, we observed that NPCs grew closely apposed to the RDG fibers, thus providing strong evidence for their radial migration through the scaffold of RDG fibers, until cells reached the regenerating site and final confluence. We observed a characteristic fiber behavior: fibers initiate to grow from the tips of the protrusion of a basal cell, later propagate out of the tip origin, by splitting at the end into two lateral turns [49]. Afterward, they continue to run in two opposite lateral directions to the left and right side, generating a row of new neurons that resemble to a pearl necklace. This behavior is in accordance with the findings of Zupanc et al. [50,51] that observed, in the adult fish, RDGc actively divided to produce new neurons in the injured brain and guided them to the lesion site to regenerate the neural tissue [50,51]. Such evidences were supportive for the role exerted by RDGc in adult teleost brain to be responsible for the strong ability of neural tissue regeneration [52]. The possibility that RDGc might exist as radially oriented cells is absolutely dissimilar in mammals, where RDGc were transformed into multipolar astrocytes in the adult brain, no longer regenerating [52,53].
In seabream brain most dividing NPCs come from RDG progenitors. We demonstrated that RDGc maintained a vimentin-positive radial fiber throughout each stage of cell division, and an asymmetric inheritance of this fiber. Out of directing neuronal migration, radial fibers supported another important feature in the role of being neuronal progenitors; they were found highly mitotically active, and underwent nuclear migration. Radial glial divisions produced both additional radial glia and newly young neurons, while a separate population of neuronal precursors generated almost exclusively neurons.
Our findings overlap with the current understandings of rodent neurogenesis: two distinct types of NPCs were identified. The RG was the first type, being conceivably recognized as neuronal and glial progenitors also by other authors [30,33,45,54]. The island of RDGc behavior as a neuroepithelial niche, giving rise to radial clones of newborn neurons, where neuronal progenitor cells generated rounds of asymmetric division. The second type, known as (IPCs), was recently characterized as neuronal progenitors, generated from RG; it proliferate for one or two mitotic cycles, and produce neurons only [55][56][57]. In fish we found that RDGc generate IPCs, which divide to produce pairs of neurons. This is an intriguing result for the potential use of fish neural stem cells in regenerating nervous tissue [58]. In recent times, GRPs have been so well studied for the development of astrocyte transplantation therapies, highlighting new exciting potentials for regenerative therapies in the injured adult nervous tissue [48,49].
The self-renewing ability of NSCs which can differentiate neuroepithelial precursors into neural cell lineage of oligodendrocytes it has been recently demonstrated [16,17]. In mammals, oligodendrocytes develop from the neuroepithelial precursors that line the so-called VZ. Soon after the migration, they can differentiate into either oligodendrocytes type or "type-2" astrocytes in vitro, depending on the composition of the culture medium. This is the reason for these cells are referred to as O-2A progenitor cells [59,60]. Dissimilarly to other neuronal progenitors, O-2A glial progenitors continue to divide after they leave the VZ, then they undergo a transition to pro-oligodendrocytes. Peculiar modifications characterized their morphology: they changed from a simple bipolar form to a more complex form with multiple processes, accompanied by a change in cell surface antigens [61]. Others Authors found that, in proximity of the site of terminal differentiation, the pro-oligodendrocytes stop to divide and enwrapped axons; here, they ultimate the lineage development by expressing myelin gene products and mature into fully-fledged oligodendrocytes [62]. All the previous discoveries overlapped with our findings; as a result we demonstrated all stages of lineage development including proliferation, differentiation and long-term survival. In the CNS, multiple precursor cells were found to develop both the neuronal precursor cells and glial precursors; among the latter, Glial-Restricted Precursor (GRP) cells and oligodendrocyte/type-2 astrocyte progenitor cells (termed also oligodendrocyte precursor cells, O-2A/ OPCs) represent the two well-studied distinct glial precursor subtypes in mammals [42]. Although GRP cells and O-2A/OPCs were restricted to astrocyte development, they both differed in the differentiation potential, in the timing of developmental appearance and in the astrocyte subtypes produced (type 1 and type 2 astrocytes); they were also able to generate oligodendrocytes [42,43].
In seabream, multiple myelinating areas of single axons were easily identified by using fluorescent antibodies directed against components of myelin (i.e. S100B, MBP). Also, phase contrast analysis gave evidence of axonal myelin shapes, with some loosely wrapped layers and protruding processes. The excellent remyelinating fish ability of mOLs might be a promising approach for repairing the CNS injuries, also in other animals, as a new coming strategy for function recovery of damaged nerves.