IPL Sublamination in Chicken Retinal Spheroids Is Initiated via Müller Cells and Cholinergic Differentiation, and Is Disrupted by NMDA Signaling

Abstract

Purpose: Reaggregates from E6 embryonic chicken retina exhibit areas corresponding to an inner plexiform layer (IPL), which presents an ideal in vitro model to test conditions and constraints of cholinergic and glutamatergic network formation, providing a basis for retinal tissue engineering. Here, we show that ipl formation is regulated by cholinergic starburst amacrine cells (SACs), a glial scaffold and by L-glutamate.

Methods: Rosetted spheroids were cultured in absence or presence of 0.2 to 0.4 mM L-glutamate and analyzed by immuno- and enzyme histochemistry, proliferation, and apoptosis assays.

Results: After 2 days in vitro (div), ipl formation was announced by acetylcholinesterase⁺ (AChE) and choline acetyltransferase⁺ (ChAT) cells. Individual vimentin⁺ or transitin⁺ Müller glial cell precursors (MCPs) in ipl centers coexpressed ChAT. Comparable to in vivo, pairwise arranged ChAT⁺ SACs formed two laminar subbands. Projections of calretinin⁺ amacrine cells (ACs) into ipl associated with MCP processes. In L-glutamate-, or NMDA-treated spheroids ipls were disrupted, including loss of SACs and MCs; coincubation with NMDA receptor inhibitor MK-801 prevented these effects. Also, many Pax6⁺ cells, comprising most ACs, were lost, while rho4D2⁺ rod photoreceptors were increased. Cell proliferation was slightly increased, while apoptosis remained unaffected.

Conclusions: This demonstrated: (1) a far-advanced differentiation of an IPL in retinal spheroids, as never described before; (2) ipl sublamination was initiated by cholinergic precursor cells, which—functioning as “ipl founder cells”—(3) gave rise to neurons and glial cells; (4) these SACs and MCPs together organized ipl formation; and (5) this process was counteracted by NMDA-dependent glutamate actions.

The recent proliferation of 3D spheroid and organoid technologies promises imminent biomedical progress in transplantation medicine and for pharmacological testing.¹ In order to introduce applicable 3D tissue models, a detailed understanding of molecular, cellular, and histologic processes of tissue formation is necessary. In particular, applicability of retinal organoids from induced pluripotent stem cells (iPSCs) as humanized pharmacologic assay systems will depend on their reliable recapitulation of normal development.^2–7 Retinal spheroids from chicken embryonic retina will be instrumental to promote this field, since production of multiple distinct types of retinal spheroids from dissociated embryonic precursor cells has been achieved. For example, one type presents an incomplete lamination (called “rosetted spheroids”), and a second type achieves a complete laminar tissue organization (“stratospheroids”).^8–10 In fact, stratospheroids demonstrated the first complete laminar retinal reconstruction from dispersed cells, providing proof-of-principle that retinal tissue engineering from stem cells is achievable.¹¹ The course from reaggregation until about 8 to 10 days in vitro (div) corresponds well with the period of normal retinal lamination in ovo. In rosetted spheroids, dispersed cells from E6 embryonic chicken retinae reaggregate in rotation culture and present histotypical features of a normal developing retina (Fig. 1). Excluding cells in nonorganized zones (noz), most cells become either organized within rosettes containing photoreceptor precursors (ros), or, within areas homologous to the retina's inner plexiform layer (IPL; here called “ipl”). Retinal ganglion cells (GCs) do not differentiate (they die within the first 2 days IV). Precursors of Müller glial cells (MCPs) also play major roles for in vitro retinal histogenesis (Figs. 2, 3).^12,13 This corresponds with convincing reports that Müller glial cells can function as source for retinal tissue regeneration after its damage. This finding supports hopes that MCs could help to overcome the normally highly restricted regenerative capacity of vertebrate retinae, and thus become applicable in retinal tissue engineering (further in Discussion).^14–17 

Figure 1

 

Histotypic structures in a rosetted spheroid from embryonic chicken retina. (a) Sytox-stained cryosection of a div6 spheroid presenting photoreceptor rosettes (ros), INL-like nuclear areas (inl), large IPL-like areas (ipl) and interspersed noz; (b) section of a div6 spheroid presenting two major building blocks ipl/inl and onl rosettes, triple-stained by the photoreceptor (PR) precursor marker visinin (vis, green), the Müller cell precursor marker vimentin (red) and DAPI (blue); (c) a first PR rosette is forming already at div2; (d) SV2 expression (green; DAPI, blue) shows synaptogenesis in ipls of div6 spheroid; (e) phase contrast view of div6 spheroids in culture dish. Sytox staining in (a) is presented in pseudocolor indicating different fluorescence intensity (yellow = high; blue = low). Scale bars: 50 μm (a), 20 μm (b, c), 10 μm (d), 200 μm (e).

Figure 2

 

Cholinergic ChAT⁺ cells (a–f; green) and Müller cell precursors (g–i, green) are detected from onset of IPL formation. (a–f) ChAT⁺ cells (green) localize at and within forming ipl areas of rosetted spheroids. From div 4 onward (d–f), ChAT⁺ cells undergo orderly arrangement, and become interconnected via unilateral processes, which eventually establish an IPL subband (arrows in [f]). (g–i) Lower: MCP are stained by vimentin (g) or transitin (h) in center of div 2 ipl, where they form an early fiber meshwork. (i) Differentiating Müller cells express glutamine synthetase (green) only from div7 onwards. Blue in all, SYTOX staining; scale bars: 50 μm (a–c, e), 15 μm (d, f, g–i).

Figure 3

 

ChAT⁺ cells ([b, f], red) and Müller cell precursors (c, g, k, l; green; merged in d, h) are derived from same precursor cell. Individual ChAT⁺ cells ([b, f]; red) in center of a forming ipl (div3, a–d, div5, e–h) strongly coexpress vimentin ([c, d], green) and transitin ([g, h], green). Scale bars: 15 μm. (i, j) Spatial organization of individual sytox-stained cells in pseudocolor within a larger ipl space (i); note their circular arrangement ([i], stippled square; [j], stippled circle); scale on left indicates relative pseudocolor values for sytox fluorescence. (k, l) Typically, ChAT⁺ cells (red) remain in direct association with MCP cell bodies ([k, l], stippled circles), stained by vimentin ([k], green) or transitin ([l], green). Scale bars: 15 μm (a–j).

The high reproducibility of avian spheroid formation renders these models applicable as developmental assay systems to analyze genetic (by gene knockout) or molecular (environmental) effects on tissue development (e.g., by growth factors or stress). Morphologic and histologic changes of spheroids became easily trackable by applying simple phase contrast microscopy of whole spheroids in combination with isocontour imaging using Image-J software (http://imagej.nih.gov/ij/; provided in the public domain by the National Institutes of Health, Bethesda, MD, USA).^18,19 Moreover, retinal spheroids were introduced for automated high-throughput pharmacologic and electrophysiologic analyses.^20,21 Cholinergic, followed by glutamatergic actions exert the earliest and strongest effects on retinal development.²² Therefore, herein we investigated cholinergic mechanisms during development of retinal rosetted spheroids, and possible deleterious interferences through excitotoxic glutamatergic actions. 

Stress on cells and tissues can originate from many diverse causes (e.g., induced by physical, toxic, or radiation impacts). Often the dose makes the difference, as is the case with L-glutamate. Glutamate is the major excitatory neurotransmitter in the adult vertebrate retina and plays major roles during neuronal development.^23–27 In particular, a spontaneous glutamatergic activity wave follows transient cholinergic waves; their correlated interplay dictates early circuit formation in the IPL of several vertebrate species.^28,29 Glutamate binds to metabo- or ionotropic types of receptors. NMDA (N-methyl-D-aspartate) receptors are ionotropic, require coactivation by glycine, and regulate developmental, physiologic, and diseased states.^30–32 In retina, all amacrine cells (ACs) possess ionotropic glutamate receptors. Activation of NMDA or non-NMDA receptors releases acetylcholine (ACh) from the only cholinergic retinal cell type, called starburst amacrine cells (SAC). Their excessive activation leads to an overshoot of Ca²⁺-influx and eventually induces cell death of SACs.^21,33,34 After binding of glutamate postsynaptically in a mature retina, Müller glial cells (MCs) take up glutamate to terminate excitatory neurotransmission (glutamate/glutamine shuttle), a process, which again is regulated by NMDA receptors.³⁵ Müller cells have been shown to play major roles in inflammatory and stress responses,^36–38 counteracting excitotoxic effects of glutamate at too high concentrations.^39–42 In support of this scheme, the gliotoxic glutamate analogue DL-α-aminoadipate (AAA) induced a strong inflammatory response of Müller cells in explants from the chick embryonic retina, disturbing IPL differentiation.⁴³ Notably, Müller cells from adult mice were dedifferentiated to a progenitor cell state at subtoxic levels of L-glutamate or AAA.¹⁷ 

In contrast to glutamate, cholinergic mechanisms can inhibit stress responses via a process known as cholinergic anti-inflammatory pathway (CAIP), postulating that ACh can be secreted at inflammation sites to exert anti-inflammatory actions.^44–47 Not only to achieve synaptic, but also nonsynaptic actions of the neurotransmitter ACh, the ACh-synthesizing and -degrading enzymes (acetyltransferase, ChAT; acetylcholinesterase, AChE, respectively) are most relevant.^48,49 In retina, cholinergic differentiation occurs very early and initiates differentiation of the IPL.^22,28 SACs, the only cholinergic cells in mature retina, are reliably characterized by their expression of ChAT. Their major two subtypes (SAC I and II; Fig. 4b; Supplementary Fig. 1h)^50,51 are located closely on either side of the inner plexiform layer (IPL), and project to two cholinergic synaptic sublaminae of the IPL (“a” and “d” subbands, or OFF and ON, respectively).^52–54 A third type of SACs, located further inside the INL, projects deep into IPL (cf. Fig. 5a'). The establishment of SACs depends on muscarinic cholinergic signaling,⁵¹ but establishment of the SAC circuit depends also on nicotinic activities between neighboring SACs. Both IPL sublaminae can be easily detected by ChAT and by AChE staining. Sublamina “a” in chicken stains more strongly for ChAT, while sublamina “d” presents more AChE.⁵¹ Like in other parts of the brain, AChE in chick retina is expressed shortly after final mitosis of neuroepithelial cells and represents the first cells to begin differentiation at E3/4.^52,53 Remarkably, ChAT⁺ cells in chick retina are derived from a larger pool of AChE expressing precursor cells and then develop into cholinergic ACs.⁵⁵ Thus, cells presenting cholinergic traits are the first cells to develop in retina.^{22,50,56–61} This conclusion is also strongly supported by spontaneous spike activity waves running over the retina, the first of which is cholinergic, replaced by a glutamatergic one.^22,29 Therefore, SACs affect later differentiating cells and their network formation. 

Figure 4

 

Initiation of an ipl network by cholinergic cells in a div5 spheroid (a), which is followed by formation of a calretinin⁺ network (c). AChE⁺ cells (red in [a]) are located at border of inl/ipl; type I starburst amacrine cells (ChAT⁺ SACs, green in [a]) are located close by, some of which send processes into a first sublamina (white arrows). Several SACs II are found in center of ipl, forming pairs with their outer SACs I (note white labels of SACs); some SACs II coexpress AChE ([a], yellow). (b) In vivo staining of ChAT⁺ starburst amacrine cells (SACs; green). Pair-wise arrangement of SACs of an E10 chick retina section, projecting into subbands a and d, respectively. AChE (red) stains amacrine cells in INL and GCs in GCL.⁵¹ (c) Calretinin⁺ amacrine cells ([c], green) in inl of a div7 spheroid project into ipl space and form at least 3 parallel circular sublaminae. Note that these remain restricted to the space between pairs of ChAT⁺ cells ([c], red; one pair is indicated by white “SAC” labels). Scale bars: 20 μm (a, b), 10 μm (c).

Figure 5

 

Far advanced ipl network formation as initiated by MCPs and SACs (a, d–f) is directing later differentiating calretinin⁺ neurons (CR in [b, e, f]). (a) Shows ipl network formation as represented by organized MCP processes ([a], vimentin, red) and pairwise-arrangement of ChAT⁺ SACs ([a], green) in a div6 spheroid (note cell pair of 1/1′). Note radial, but also lateral side processes of MCPs, as well as onset of subband a formation ([a], stippled circle). ([a'], insert) Three different types of SACs can be distinguished by ChAT immunostaining: (i) cells with their cell bodies located at inl/ipl border (insert, stars, outside of stippled circle), (ii) cells located internally within ipl (insert, triangles, inside of stippled circle), and (iii) a small number of cells located further outside of inl with thin long processes into ipl (insert, arrow). (b) SACs belong to a small subpopulation of calretinin⁺ ACs (red; ChAT/CR double-stained cells, yellow). While only a few of them are located on inl/ipl border, most displaced SACs in ipl center are CR⁺. (c) Islet-1 (red) and calretinin (CR, green) plus DAPI triple-stained ipl area. Note long CR processes reaching Islet-1+ cells in center of ipl. (d–f) CR⁺ processes closely associate with Vim⁺ MCP processes ([e], red, detail in [f]); note that CR⁺ cells appear only from div 4 onward (e), following patterning of MCP processes ([d]; cf., also Fig. 2h). Scale bars: 15 μm (a–e), 5 μm (f).

To date, it has remained unclear to what degree an IPL could differentiate in vitro from dispersed cells, and whether similar events could direct its formation. Using chicken rosetted spheroids, we here show that i) ipl areas in spheroids can achieve a double-subbanded network structure, including different types of cholinergic amacrine cells (SACs), that glutamate deletes most SACs and MC processes, and that these effects are mediated via NMDA receptors, thereby strongly inhibiting ipl histogenesis. Thus, formation of ipls in 3D retinal spheroids presents an ideal in vitro model to test conditions and constraints of cholinergic and glutamatergic network formation, and also provides a basis for promoting retinal tissue engineering. 

Materials and Methods

Production of Retinal Spheroids From Embryonic Chicken Eyes

Fertilized eggs of White leghorn chicken were purchased from a local hatchery and incubated at 37°C in a humidified chamber until 6 days of embryonic development. Spheroid production was performed as detailed earlier.¹⁸ Briefly, central parts from 30 to 40 retinas of 6-day-old chicken embryos were dissociated by tryptic digestion (0.05 mg/mL trypsin; Worthington Biochemicals, Remagen, Germany). Remaining cell clusters were mechanically dissociated in presence of 0.5 mg/mL nuclease (DNase I; Worthington Biochemicals); 2 × 10⁶ cells/2 mL aggregation medium (AM = DMEM, 10% FCS, 2% chicken serum, 1% L-glutamine, and 0.15% penicillin/streptomycin; all from Gibco, Berlin, Germany) were cultured in 35-mm dishes (approx. 2 retinas/dish; for each following treatment, at least three parallel dishes were cultured) on a gyratory shaker in an incubator (37°C, 95% humidity, 5% CO₂). Spheroids were supplemented with 0.2 or 0.4 mM L-glutamic acid (Sigma-Aldrich Corp., Deisenhofen, Germany), or with control vehicle throughout the treatment periods. NMDA was applied at 0.4 mM, and the NMDA receptor blocker MK-801 at 100 nM (Dizocilpine; Sigma-Aldrich Corp.). Medium and supplements were replaced every second day. Spheroids were cultured up to 15 days in vitro (div). Chicken embryos were treated according to ARVO ethical standards. 

Fixation and Cryosectioning

Rosetted spheroids were harvested at indicated days and fixed in 4% paraformaldehyde (Merck, Darmstadt, Germany) for 30 minutes at RT, and preserved until sectioning in 25% sucrose at 4°C. Fixed spheroids were transferred to a solution (Tissue-Tek; Sakura Finetek Europe B.V., Alphen aan den Rijn, The Netherlands) and 10- to 14-μm thick sections were cut on a cryostat (Microm, Heidelberg, Germany) at −28°C, and mounted on frost-free or gelatin-coated glass slides. The sections were stored at −20°C until further use. For BrdU-incorporation experiments, 25 μM BrdU (Sigma Aldrich Corp.) was added 16 hours before the spheroids were harvested and fixed as described before. Whole eyes for in vivo studies were also fixed in 4% paraformaldehyde overnight at 4°C, immersed in 25% sucrose and sectioned as whole mount eyes, as mentioned above. 

Immunocytochemistry

Cryosections of spheroids were subjected to immunostaining in order to compare the development and differentiation of different retinal cell types. Prior to staining procedures, the frozen sections were dried on a heating plate at 37°C. The area around the sections was marked with a greasy liquid blocker and the slides were pre-incubated in blocking solution (PBS, 3% BSA and 0.1% Triton X-100) for at least 30 minutes. A total of 100 μL of primary antibodies, which were diluted in the blocking solution, were applied on sections inside the greasy area and then incubated (see Supplementary Table 1). In case of double staining, one polyclonal and one monoclonal antibody were mixed and applied together. The slides were washed thrice with PBS and then subjected to 100 μL of fluorescent-conjugated secondary antibodies for 75 minutes at RT. After three PBS washes, the sections were treated with DAPI, or with 500 nM SYTOX orange in PBS (Invitrogen) for 3 minutes in order to stain cell nuclei. The slides were rinsed in distilled water after a final wash in PBS, dried on a heating plate, and embedded in Kaiser's glycerol gelatin (Merck, Darmstadt, Germany). The details of primary and secondary antibodies, their dilutions, and sources used in this study are given in Supplementary Table S1. Sources of antibodies: 3D10 (gift from Karl Tsim, Hong Kong), rho4D2 (gift from David Hicks, Strasbourg), anti-Pax6 (Developmental Studies Hybridoma Bank), glutamine synthetase (BD Biosciences), CERN 901 (gift from William J. DeGrip, The Netherlands), secondary antibodies Cy2 and Cy3 (Dianova). 

Cholinesterase Staining

Acetylcholinesterase staining on cryosections was performed by the Karnovsky-Roots technique.⁶² In brief, the frozen sections were dried on a heating plate, incubated twice for 15 minutes in Tris-maleate buffer and treated with the reaction buffer (0.1 M tris-maleate buffer, 0.1 M sodium citrate, 30 mM copper sulphate, and 5 mM potassium hexacyanoferrate). The reaction was performed for 1 hour at 37°C with 6 mM acetylthiocholine iodide as substrate and 0.1 mM iso-OMPA (tetra-isopropyl-pyrophosphoramide) to inhibit BChE. The reaction produced a brown insoluble precipitate (Hatchett's brown) at the venue of enzymatic activity, which was viewed under light microscope. AChE activity was determined by the Ellman method with acetylthiocholine iodide at 1.5 mM plus 0.1 mM iso-OMPA at 412 nm (SERVA, Heidelberg, Germany; not shown).⁶³ Protein content was determined by the Lowry method using BSA as standard.⁶⁴ 

TUNEL Cell Death Assay

Apoptosis was detected by terminal deoxynucleotidyl-transferase dUTP nick end labeling (TUNEL), using the In Situ Death Detection Kit (Roche), according to manufacturer's instructions. 

Microscopy, Imaging, and Statistics

All experiments have been performed at least 5-fold, revealing qualitatively identical results. More than 10 glutamate treatment series for IHC, five experimental series under NMDA treatment, and NMDA plus MK-801 were performed. Immunostained in vivo and spheroid sections were subjected to fluorescent microscopy (Axiophot; Carl Zeiss, Jena, Germany), which was connected to an automated digital camera (Axiocam). Imaging was done for each fluorescent channel separately and processed through extended focus imager, and all the pictures from different channels were combined as overlays and stored in zvi image format by a software program (Axiovision Basis; Carl Zeiss Meditec). SYTOX orange-stained sections were taken with a confocal microscope (TCS SP5; Leica Microsystems, Wetzlar, Germany) using the same pinhole, gain and offset settings for all sections from one experiment. Pictures were analyzed using ImageJ 1.38x. Microscopic cell counting of labeled cells (e.g., BrdU, TUNEL, etc.) was performed manually within 50 μm² grids of 5 to 8 arbitrarily chosen areas of individual spheroids. Sections were chosen from 6 to 9 individual spheroids derived from 3 parallel dishes/treatment. Determination of sectional sizes and/or areal sizes of ipls was executed as described earlier.¹⁸ P values were obtained by t-test. Briefly, with the above-mentioned software, the circumferences of individual spheroids—and therein, of individual ipl areas—were marked and measured. Then the total area size of spheroid sections, and/or of individual ipl areas were calculated (e.g., Fig. 6D). Note that due to L-glutamate treatment, many ipl areas were not only diminished in size, but often had not formed at all (e.g., cf. Figs. 7c, 7d). Therefore, this procedure tends to underestimate the actual L-glutamate effect. 

Figure 6

 

Glutamate supplementation disrupts ipl formation and decreases AChE expression in rosetted spheroids (A, D). (A) time-dependent increase of histochemical AChE activity on sections of 2 to 8 div spheroids (a–d), revealing the formation of ipl areas outlined by AChE⁺ future amacrine cells, a process that is inhibited by treatment with 0.4 mM glutamate (e–h); bar, 100 μm; (B) details of AChE expression in a div7 ipl (cf. Fig. 4a), as compared with ([C], bar, 50 μm) in vivo AChE expression in an E10 chicken retina; bar, 100 μm; (D) decrease of relative ipl areal sizes in presence of 0.2 (red) and 0.4 mM (green) L-glutamate as compared with control spheroids raised in absence of L-glutamate (blue). Note maximal ipl size reached at div 8/9, a time when L-glutamate effect is also most pronounced. Results are presented as mean ± SEM, calculated by determining ipl sizes from six to nine individual spheroids from three dishes, for each section counting 5 to 8 ipl areas (see Materials and Methods). Note that results from div 5, 7, 9, 13 and those from div 6, 8, and 10 are obtained from two different experiments (n = 1 for both series of measurements, therefore no statistical analysis was carried out), whereby for the latter experiments only 0.4 mM L-glutamate was tested. Note also that this analysis does not take into account that total number of ipl areas decreases in presence of L-glutamate (cf. Fig. 6A, div 6 and 8), further strengthening the deleterious effect of L-glutamate on ipl formation.

Figure 7

 

L-glutamate (0.4 mM, right frames) disturbs ipl structure, deletes amacrine cells, including cholinergic amacrine cells (SACs) and increases rod photoreceptors. (a, b) Pax6⁺ amacrine cells forming a compact cell ring around ipl areas in controls ([a], yellow circle) are nearly absent in glutamate treated spheroids; instead, few Pax6⁺ cells remain in core ipls ([b], yellow circles). (c, d) Axonin-1 (red), calretinin (CR, green) and DAPI triple-stained spheroid sections of div8. Note ipls and axonin-1 staining ([c], circle) are absent in glutamate-treated spheroid (d). (e, f) AChE staining (red) of ACs and ChAT (green) staining for SACs at div6; note, SAC is almost completely lost (arrow; blue, DAPI. Note in [e] details of ipl sublamina network formation (cf. Fig. 4a). Lateral processes of type I SACs (ChAT⁺, green) establish ipl subband a (within stippled white circle). Several SACs II in center of ipl show signs of forming a second ipl subband d (white arrow); stippled white arrows at inl/ipl border point to close neighborhood of type I SACs and AChE⁺ cells. (g, h) Number of rho4D2⁺ rod photoreceptors is much increased in presence of glutamate (h), while PR rosettes are absent (cf. in [g], full circle; dashed circle shows ipl area). Scale bars: 30 μm (a–d), 15 μm (e–h).

Results

IPL Formation in the 3D Model Rosetted Spheroids

Rosetted retinal spheroids present convenient in vitro culture models to analyze retinal development and disturbances thereof (e.g., by excitotoxic glutamate stress). Figure 1a presents a cryosection of a 6-day-old (div6) rosetted spheroid, originating from dispersed E6 retinal cells. The sytox-stained spheroid was dominated by two major histotypic structures, internal rosettes containing photoreceptor (PR) precursors (Fig. 1a, “ros”; note pseudo-color for sytox presents high fluorescence in yellowish, and lower sytox appears in blue; PR precursors are stained green by visinin in Figs. 1b–d), and large matrix-filled areas, corresponding to an IPL (called “ipl”). Outside of these two major constituents, many cells were diffusely distributed in nonorganized zones (“noz”). To further clarify the spatial relation between these two major areas, Figure 1b presents a photoreceptor rosette and an ipl area, triple-stained by visinin (green), vimentin (red), and DAPI (blue). Both major areas contact each other closely. Cells surrounding the ipl represent an inl (note their bright sytox in a), and PR precursors form a one cell layer-wide onl around the PR rosette. It should be noted that consistently a few single cells were found in ipl spaces (Fig. 1a; white arrow in Fig. 1b; see further below); ipl spaces were filled by a matrix of vimentin⁺ processes. Their golden-yellow staining (Fig. 1b) indicates that they were double-labeled by vimentin and visinin. Expression of the synaptic marker SV2 demonstrated ongoing synaptogenesis in ipls of 6 div spheroids (Fig. 1d). It is evident that visinin⁺ cells were found not only in rosettes, but were also numerous outside in noz. First PR rosettes emerged already in div2 reaggregates, as depicted in Figure 1c (arrow). By div4, PR rosettes had grown larger, now presenting an inner open lumen. Figure 1e provides a phase contrast overview of a rosetted spheroid culture at 6 div. The ipl areas appeared somewhat translucent, and at this stage many of them were bulging out from the spheroid surface. 

Cholinergic Differentiation Marks Onset of Ipl Formation

Rosetted spheroid differentiation was followed by immunostaining for ChAT from div2 to div6, with ChAT representing the archetypal marker for cholinergic cells (Fig. 2a–f; cf. also for AChE in Figs. 5b, 7A). At div2, irregular open spaces emerged in the early cellular reaggregate. Few cells were ChAT⁺, which were located closely at, or within these open spaces (Fig. 2a). By div3, the emerging ipl spaces had rounded, and most of them internally presented a few ChAT⁺ cells. At div4, ipl spaces had considerably enlarged, presenting more ChAT⁺ cells, which began to arrange in an inner circle (Figs. 2c, details in 2d, 3i, 3j; Supplementary Fig. 1g). Notably, all ChAT⁺ cells at this time were inside of an ipl. By div6, ChAT⁺ cells in ipl spaces had increased in number (Figs. 2e, 2f). Many of them were now located at the border to the inl; their unipolar processes toward the inside spread bilaterally and connected with neighboring ChAT⁺ cells (Fig. 2f, arrows, details in Figs. 4a, 5a). Thus, from its onset ipl formation was accompanied by ChAT expressing cells (note difference of AChE expression, cf. Figs. 4a, Supplementary Fig. 1g with 6a, 6b and “Discussion”). 

MCPs Form an Initial Ipl Fibrous Meshwork and Coexpress ChAT

Concomitantly with ChAT expression, the Müller cell precursor markers vimentin and transitin were detectable in forming ipl spaces. Remarkably, at div3, individual cells within ipl spaces expressed large amounts of vimentin and presented only short processes (Fig. 2g). Transitin (Fig. 2h), an MCP marker preceding vimentin,⁶⁵ was heavily expressed in a few cells inside an ipl, which sent out long processes into inl areas. Glutamine synthetase (GS) staining documented that these cells were not yet mature MCs, since its expression occurred only at div7 (Fig. 2i). Surprisingly, double staining of vimentin (Figs. 3a–d, at div3) or transitin (Figs. 3e–h, at div5) with ChAT revealed co-expression of both MCP markers with ChAT⁺ cells. Notably, only those MCPs that strongly expressed vimentin or transitin, were double-labelled by ChAT. Vice versa, by far not all ChAT⁺ cells were MCPs (see red cells in Fig. 3d, 3h). However, even at later stages of ipl differentiation, some ChAT⁺ cells remained closely associated with MCP cell bodies or their processes (stippled circles in Fig. 3k, 3l). Thus, at the onset of ipl formation in vitro, a small population of MCPs coexpressed ChAT. 

IPL Differentiation in Spheroids Is Far Advanced, Including Orderly MCP Matrices, Pairwise SACs and Subband Formation

In vivo, retinal lamination in the chick embryo occurs roughly between E5 and E11 (cf. Supplementary Figs. 1b, 1d, 1f, 1h). A row of cells in the middle of the IPL migrating toward the GCL represent displaced amacrine cells.^50,66,67 By E11, processes of cholinergic amacrine cells (SACs; stained green for ChAT in Fig. 4b; Supplementary Fig. 1h) have formed two synaptic subbands, or sublaminae a and d (a, d in Fig. 4b). As ipl differentiation proceeded in rosetted spheroids, an organized pattern of cholinergic cell arrangement emerged (cf. Figs. 3i, 3j). By div5-6, many AChE⁺ cells were found at inl/ipl border (Fig. 4a, red; cf. also Figs. 6a, 6b). Only a few ChAT⁺ cells were close by, but many of them were found located in inner ipl with a majority of them coexpressing AChE (Fig. 4a, yellow). ChAT⁺ SACs at the inl/ipl border projected towards the inner ipl (Fig. 4a, white arrows). As differentiation proceeded, ChAT⁺ side processes established an outer and an inner ring of ipl subbands (inside of stippled circle of Fig. 5a). Thus, the pattern of AChE⁺ and ChAT⁺ cell arrangements can easily be compared with that of the in vivo retina (cf., Figs. 6b, 6c; Supplementary Figs. 1g, 1h).⁵⁵ Double staining of ChAT with vimentin demonstrated how processes of ChAT⁺ and vimentin⁺ cells run closely in parallel, and that ChAT⁺ cells repeatedly can be found pair-wise, one at inl/ipl border, and the other inside of the outer subband a (Fig. 5a, see stippled circle, and cell pair of 1/1′). In fact, ChAT immunostaining distinguished three different subtypes of SACs (Fig. 5a', red): those with their cell bodies located at inl/ipl border (Fig. 5a', stars), those located internally in ipl (Fig. 5a', triangles), and a small number of cells located further inside of inl with thin long processes into ipl (Fig. 5a', arrow). Further, a subpopulation of cholinergic SACs can be marked by calretinin (CR). As Figure 5b shows, calretinin (CR) and ChAT coexpressing cells were arranged pairwise (numbered pairs in Fig. 5b), while cells expressing only CR are located on inl/ipl border (Figs. 5c–f). Originating there, CR⁺ processes typically targeted strongly stained Islet-1⁺ cells in center of ipl (weaker stained Islet-1⁺ cells were found in inl; Fig. 5c). At div4 only few CR⁺ cells were found at the ipl periphery (Fig. 5d), indicating that CR⁺ amacrine cells differentiated after ChAT⁺ amacrine cells (cf. Fig. 4c). At div6, their long processes projecting into the ipl were closely attached to vimentin⁺ processes (Fig. 5e, detail in 5f), indicating that CR processes were guided by MCPs. During the same period (div3-6), MCP processes became numerous and spatially organized. By div6, many vimentin⁺ processes spread in parallel from the ipl center into neighboring inl areas (Figs. 5a, 5e); at the very ipl center, several MCP processes bent perpendicularly (Fig. 5e, near dashed circle). 

Glutamate Inhibits Cholinergic Differentiation AChE and Ipl Formation

Figures 6a and 6b demonstrate strong histochemical expression of AChE in amacrine cells at the ipl/inl border. Expression of AChE enzyme activity was detectable during the entire period of spheroid formation (2–8 div in Fig. 6A, controls). Small islands of AChE⁺ cells indicated the formation of ipl areas. By div 4 through 6 circles of AChE⁺ cells surrounded ipl spaces, each of which became filled by a ring of AChE activity (Fig. 6A at 6 and 8 div, 6b, see also Fig. 8A), announcing the formation of cholinergic synaptic ipl subbands (cf., Figs. 4a; for in vivo comparison, see Fig. 6c). In presence of 0.4 mM glutamate, most of the outer ring of AChE⁺ cells was missing; cells of AChE activity remained focused to island centers. At later stages, AChE⁺ islands became smaller and fewer (Fig. 6a, 8 div). A quantitative analysis of ipl aerial sizes in control and glutamate-treated spheroids supported the notion that glutamate acted most prominently on ipl structures (Fig. 6d). During div 5-13, ipl sizes in controls remained large with peak sizes between div 7 and 9, while in presence of both 0.2 and 0.4 mM glutamate ipl areas were decreased in size by more than 75% (Fig. 6d). It should be noted that this analysis does not take into account that many ipls do not form at all in presence of L-glutamate (cf. Figs. 6a, 6h, 7d). Therefore, the deleterious action of L-glutamate on ipl formation is even stronger than indicated by the quantification in Figure 6d. 

Figure 8

 

Effects of glutamate are simulated by NMDA (A), and are counteracted by MK801 (B), an inhibitor of the ionotropic NMDA receptor. Frames show whole mount spheroid morphologies in (a–c, j, k); markers for AChE (brown, [d–f, l, m]), for Pax6 (red) and CERN901 (green) in (g–i, n, o), DAPI (blue). Scale bars: in (A): 150 μm (a–c, j, k), 100 μm (for d–i; l–o).

Glutamate Destructs MCs in Spheroids, Deletes Many Amacrine Cells Including all SACs, and Increases Numbers of Photoreceptors

When spheroids were cultured in presence of 0.4 mM L-glutamate, ipl areas were strongly disorganized, or absent at all (e.g., the histology of the entire spheroid was strongly affected; Figs. 7b, 7d, 7f, 7h). In inl a great loss of amacrine cells stained by Pax6 antibody at div8 was noted (Fig. 7b versus 7a). Predominantly ACs of compact inl cell rings (Fig. 7a, dashed circle) were deleted by glutamate. Hence, Pax6⁺ cells were only left in nonorganized zones, or, in inner cores of ipls (Fig. 7b, dashed circles). A strong expression of axonin-1 (a marker for neuritic processes in IPL) was absent in glutamate-treated spheroids, as were calretinin⁺ cells (Fig. 7d versus 7c). A similarly pronounced glutamate effect was found for ChAT⁺ cells in the inl and ipl. If ipl areas could be detected at all in glutamate-treated spheroids, ChAT⁺ cells could no longer be found (at 6 div in Fig. 7f versus 7e, green cells; cf. also with Fig. 4a), while quite many AChE⁺ cells at the ipl/inl border had survived glutamate treatment (cf., red cells in Fig. 7f versus 7e). This demonstrates that L-glutamate completely wiped out all ChAT⁺ SACs. At the same time, glutamate treatment strongly increased the relative number of photoreceptors, as shown for rod precursors by rho4D2 staining (Fig. 7h versus 7g). In treated spheroids, no organized ipl areas and only few if any PR rosettes could be found (cf. Fig. 7h with 7g, closed circle). Taken together, glutamate deleted many ACs within the inl and their ipl processes as demonstrated by Pax6, axonin-1, AChE and CR stainings (not all shown), in particular deleting all SAC-I, while the number of rod photoreceptors increased. In effect, histogenesis in rosetted spheroids was completely disrupted by glutamate treatment. 

Glutamate Acts on Spheroids via Ionotropic NMDA Receptors

The effects of glutamate on spheroids were mediated via ionotropic NMDA receptors (Fig. 8). NMDA induced similar morphologic effects as glutamate (e.g., NMDA completely inhibited bulging of ipl areas; Fig. 8A, a–c). When spheroids were incubated in presence of 0.4 mM glutamate together with MK801, a noncompetitive antagonist of the ionotropic NMDA receptor (Dizocilpine),⁶⁸ bulging of ipl areas was fully retained, presenting spheroid morphologies, which were identical with controls (Fig. 8B, 8k versus 8j). These findings were fully supported by AChE stainings (Figs. 8d–f, 8l, 8m) and by applying Pax6 and CERN-901 antibodies as markers for amacrine cells and rod photoreceptors, respectively (Figs. 8g–i, 8n, 8o). Therefore, all effects of glutamate on rosetted spheroids were exerted at the level of ionotropic NMDA receptors. 

Glutamate Stimulates Early Cell Proliferation Without Affecting Cell Death

Processes of lamina formation in rosetted spheroids occur between div 2 and 7, roughly corresponding to embryonic days E6-12. DAPI staining showed that untreated controls presented large ipl areas including peripheral ipl bulges at div 6 and 8 (Figs. 9a, 9e, 9i, 9m), which were not detectable, or, were much smaller in glutamate-treated samples (Figs. 9c, 9g, 9k, 9o). During this period apoptosis in spheroids was minor (Fig. 9A, 9b, 9f; controls). Apoptosis was not affected at the applied concentration of glutamate, as was shown by TUNEL assays after 6 and 8 div (Figs. 9A, 9d, 9h). At div6, cell proliferation was higher in glutamate-treated spheroids than in controls, as shown by BrdU studies (Fig. 9B, cf. controls). Quantification of BrdU versus DAPI stained cells established that 0.4 mM of glutamate stimulated cell proliferation at 6 div, but had no effect at 8 div (Fig. 9C). These analyses showed that the applied doses of glutamate treatment did not affect cell death and thus viability of spheroids. Instead, L-glutamate led to a proliferative response during a sensitive period of spheroid growth. 

Figure 9

 

L-glutamate stimulates cell proliferation at div6 (A, B), but does not affect apoptosis (C). (A) DAPI- (blue) and BrdU-stained (red) sections of div6 (a–d) and div8 (e–h) in absence (a, b, e, f) and presence of 0.4 mM glutamate (c, d, g, h). (B) Quantification of BrdU-stained cells/areal size in relation to total cells/areal size. Note stimulation of proliferation by glutamate at div6, but not anymore at div8. Results are presented as mean ± SEM calculated by counting the number of BrdU⁺ cells in seven 50 μm²–sized areas per sphere section (see “Materials and Methods”). P values were obtained by t-test comparing BrdU ratios from 6 to 7 individual spheroids (n = 3). *P < 0.05. (C) DAPI (blue), TUNEL-stained sections (green) of div6 (a, b, e, f) and div8 (c, d, g, h) in absence (a, c, e, g) and presence of 0.4 mM glutamate (b, d, f, h). Scale bars: 150 μm (A), 100 μm (C).

Discussion

An understanding of 3D tissue formation at all possible biological levels (molecular, cellular, histologic) is essential to promote organoid technologies.^1–7 Although details have been revealed from studies on spheroid and organotypic explant^69,70 models of embryonic avian, rodent or fish origins, that can be used to further optimize human retinal organoid production.^8–10 In particular, avian retinal rosetted spheroids with their clear distinction between photoreceptor rosettes and IPL-like areas (ipl) represent a suitable model where specific aspects of formation of parts of an outer, or an inner retina can be readily studied. Part I of this study demonstrated that IPL differentiation in rosetted spheroids can become far advanced, including MCP fiber organization, pairwise arrangement of cholinergic SACs, onset of formation of two IPL subbands a and d, and synaptogenesis. Such a high degree of network formation, originating from dispersed cells in vitro, has never been reported before for any retinal spheroid or organoid system. Thus, MCPs in close association with cholinergic SACs are involved in the initiation of ipl formation. In the second part, we set out to analyze excitotoxic stress effects on ipl formation, which revealed pronounced deleterious glutamatergic effects on SACs and ipl formation. 

Cholinergic Müller Cell Precursors as IPL Founder Cells?

Soon after primary reaggregation of chicken retinal cells, both rosette and ipl formation can be observed from div 1 to 2 onward.⁸ Along with earliest ipl formation, AChE⁺, ChAT⁺, CR⁺ cells and the MCP markers vimentin or transitin were detected first at the interface and then inside these spaces. Albeit in normal retina Müller glial cells represent the last cell type to differentiate,^59,61 during spheroid formation their precursors (MCPs) are present and functionally relevant from early on.^12,13 It should be noted that not all MCs are born late, as Cajal had already noted,^70–72 and under regeneration or stress paradigms, MCs in most vertebrates have the capacity to dedifferentiate into stem cells, which then can produce cells of glial or neuronal lineage.^{14–17,36,73–77} Moreover, a molecular distinction between neuroepithelial precursors in general and MCPs remains vague.^74–76 Indeed, it came as a complete surprise that the first MCPs co-expressed ChAT, since ChAT is an accepted specific neuronal marker for SACs, while MCPs are not expected to express this enzyme. Whether cholinergic MCPs exist also in vivo and whether they can be considered as “IPL founder cells” presents a novel issue, which needs further investigations. 

These early cholinergic ChAT⁺/vimentin⁺ cells do not include (or produce) GCs, since ipl core cells were Islet-1⁺, ChAT⁺, AChE⁺, but did not express the GC marker Brn-3b (unpublished). Therefore, all ipl core cells represent exclusively MCP/SACs. In fact, GCs in rosetted spheroids are very rare,⁸ or do not survive at all as the retinal tissues are cut off from the optic fiber tract which is a main source of BDNF necessary for their survival.⁷⁸ Thus, ipl network formation was completely independent from presence of GCs, as had been reported for retinal explants.⁴³ 

Stainings of various cell types in rosetted spheroids revealed an amazing degree of ipl network formation. Not only SACs, but also amacrine cells expressing Pax6, AChE, calretinin (CR), islet-1 and others were arranged circularly around ipls (Figs. 4, 5, 7, 8), projecting orderly into the inflating ipl space. Led by MCP processes and pairwise placement of SACs, ChAT⁺ processes managed to form sublaminae at several levels of the ipl width (Figs. 4a, 5a, 7a). To our knowledge, such a high degree of in vitro network formation has never been described before for any retinal reaggregation (organoid) system. Further use of this model can help unravel the question how tissue self-organization is regulated from the state of dispersed cells until establishment of complex neuronal networks. 

Glutamatergic Stress on Cholinergic MCPs

After normal reaggregation of retinal cells in L-glutamate-treated spheroids, sorting out of AChE⁺ cell islands was detectable at div2, while ipl islands remained smaller than in controls (Fig. 6A). Glutamate disturbed the very onset of in vitro histogenesis.^12,13 Possibly precursors of SACs and MCs were both contributing to ipl formation in a specific manner. In particular, following glutamate treatment ChAT⁺ cells had almost completely disappeared, as also observed in vivo.^22,33 Similarly, many Pax6⁺ (Fig. 7), ChAT⁺ and AChE⁺ cells and neuropil were absent by div 6. For instance, axonin-1, a marker of growing AC neurites into ipl, and the synaptic marker SV2 misses in L-glutamate-treated spheroids; the same was true for Islet-1. These proteins are transiently coexpressed with ChAT in some SACs⁵¹; these cells were also deleted in ipl (Fig. 7, Supplementary Fig. 1). 

Glutamatergic Effects on Spheroid Structure via Müller Precursor Cells

The capacity for retinal regeneration after tissue damage is very limited.^14–16 However, several reports have demonstrated that in a damaged retina, Müller glial cells can be a source for regenerating neurons.^{14–17,76,77} For instance, while a toxic dose of NMDA led to damage of the retinal tissue in 1-week-old chicken, at the same token MCs dedifferentiated to a precursor state, and began to proliferate and eventually developed into retinal neurons.¹⁵ Their precursor state was demonstrated by genetic profiling.⁷⁹ Similar findings were reported for rodent retinae.¹⁶ With a seminal in vivo and in vitro study on Müller cells from adult mice, Takeda and coworkers¹⁷ demonstrated that a subtoxic dose of L-glutamate as well as AAA stimulated MCs to a proliferating precursor state. Their descendant newly produced cells migrated into the ONL and differentiated into photoreceptors.¹⁷ 

Our results as presented here are in agreement with these earlier reports. L-glutamate inhibited the formation of photoreceptor rosettes, but at the same time strongly increased the relative number of photoreceptor precursors in nonorganized areas (Fig. 7h). Extended research on formation of PR rosettes in rosetted spheroids has demonstrated that conditions of culturing, in particular, supplementation with different growth factors (e.g., effects of FGF-2, GDNF, PEDF influenced the relative number of PRs), the relative ratio of rods and several types of cones and the size of rosettes. Thus, the entire histologic structure of spheroids depends largely on the chosen culture conditions.^14,80–85 

Notably, Müller cells decisively coregulate distinct effects of growth factors on the organization of histologic structure(s) within retinal spheroids, including formation of PR rosettes and IPL-like areas (“ipls”), or, even of completely laminar spheroids.^8,11–13 For instance, addition of the gliotoxin AAA, a structural analogue of L-glutamate, destabilized MCs in stratified retinal spheroids, and thus led to a complete disorganization of the in vitro retinal tissue.⁸⁶ Similarly, in a retinal explant model application of increasing amounts of the gliotoxin L-AAA induced remarkable changes, initiated by distinct changes of MC morphology.⁴³ As consequences of AAA action, differentiation of the inner retina was delayed, while that of rods in the ONL was accelerated. Eventually, a complete loss of the ONL was observed. All AAA effects on retinal in vitro histology were strongly counter-acted by PEDF. Notably, AAA inhibited production of PEDF, indicating a glutamatergic induction of retina-relevant growth factors within the RPE.⁴³ The present findings fully support our earlier findings from spheroid and explant studies. In particular, this study demonstrated that the glutamatergic effects are exerted via NMDA receptors. 

Glutamate-Inducing MC Responses Versus Direct Effects on Cholinergic SACs?

Here we focused on glutamate actions on cholinergic cells. Indeed, the pronounced effects on AChE- and ChAT-expressing cells occurred early. Although AChE is expressed earlier than ChAT both in vivo and in vitro (this study and Ref. 51), glutamate had a greater effect on ChAT⁺ cells. This is remarkable, since in vivo ChAT⁺ cells (both SACs I and II) originate from AChE⁺ precursor cells.⁵⁵ During the first days of spheroid culture a high AChE background level was detectable in all cells (both in control and in glutamate-treated samples). Therefore, we conclude that glutamate must have acted either directly or indirectly (e.g., via MCPs) on differentiating SACs, but not on all AChE⁺ cells. Physiologic IPL maturation is driven by spontaneous and transient cholinergic waves, which then are overridden by glutamatergic waves.²⁹ It is of note that in newborn ferret retina glutamate abolished nearly all cholinergic ACs (plus α-ganglion cells); however, this treatment did not disturb further sublamination of the inner plexiform layer.⁶⁰ It appears likely that in postnatal ferrets, the role of SACs for IPL structuring is already dispensable. In rosetted spheroids, the observed later decrease of AChE in presence of glutamate could be a consequence of re-adjusting the level of acetylcholine (ACh) in the system (for morphogenetic roles of ACh, see^49,57,87,88). We have earlier postulated that a graded distribution of acetylcholine (ACh), - due to specific localizations of ChAT and AChE -, over the entire width of the retina could direct ipl network formation (see, “firewall hypothesis”^51,53). Notably, after glutamate treatment ipl areas were nearly void of ChAT⁺ cells, but still were surrounded by a significant number of AChE⁺ cells (Fig. 6A, 7b), certainly providing an ipl space devoid of ACh. Hence, the absence of any ACh in ipls could disturb network formation in spheroids, deserving further in vivo analyses. 

NMDA Receptors Mediate Glutamate Effects in Spheroids

Glutamate acted on rosetted spheroids via ionotropic NMDA receptors, since NMDA simulated all glutamate effects, and MK-801 together with NMDA prevented all NMDA-induced effects. A major function of mature MCs is their removal of glutamate from synaptic clefts (glutamate/glutamine shuttle), for which not only ionotropic NMDA receptors and self-stimulation by glutamate are required in MCs,^7,35 but also the enzyme glutamine synthetase (GS). Since GS staining was absent in MC precursors, this shuttle cannot have contributed to the observed glutamate effects at div2. Ionotropic NMDA receptors are expressed by both MCs and SACs,^57,77 and quite likely also by cholinergic MCPs. 

Conclusions

In conclusion, rosetted spheroids from chicken embryonic retina achieve a high degree of IPL network features, as has never been achieved before in any other retinal organoid model. It is suggested that a cholinergic “IPL founder cell” represents the precursor for both MCPs and ChAT-expressing SACs, which then together exerted a decisive role on IPL network formation. L-glutamate interfered with in vitro retinal histogenesis by acting primarily (or exclusively) via ionotropic NMDA receptors in precursors of MCs and of SACs. Since morphologic and histologic changes by glutamate are easily trackable by simple software, rosetted spheroids remain attractive models for automated 3D assay applications. 

Acknowledgments

The authors thank Jutta Huhn, Ulrike Hoppe, and Meike Stotz-Reimers for expert technical assistance; Lynda Wright (Madison, WI, USA) for critically reading the manuscript; Miles Epstein (Madison, WI, USA) for the gift of a ChAT antibody; and Karl Tsim (Hongkong) for his AChE antibody 3D10. 

Supported by DFG La 379/12-1, DAAD exchange grant, ESA/DLR project AO08-IBER-16. The authors alone are responsible for the content and writing of the paper 

Disclosure: G. Bachmann, None; F. Frohns, None; G. Thangaraj, None; A. Bausch, None; P.G. Layer, None