Neuronal Integration in an Abutting-Retinas Culture System

Abstract

Purpose. Limited integration is consistently observed between subretinal transplants and host retinas. In the current study, an in vitro model system for studying connections forming between two abutting retinas was developed.

Methods. Neuroretinas were dissected from normal wild-type (WT) mice and green fluorescent protein (GFP) transgenic mice (obtained at postnatal days [P]0, P5, or P60), as well as from adult rd mice. Pieces from two different retinas (WT-WT, GFP-WT, GFP-rd) were placed side-by-side (contacting each other at the margins) or overlapping each other in organ cultures for 7 or 12 days. The abutting retinal pieces derived from animals of the same age (P5-P5; P60-P60) or of different ages (P0-P60; P5-P60). Retinal cells and fibers were visualized in wholemount preparations and in cross sections by immunocytochemistry using antibodies against neurofilament (NF+), neuronal nitric oxide synthase (NOS+), and protein kinase C (PKC+) and by GFP fluorescence (GFP+).

Results. In side-by-side pairs (WT-WT, GFP-WT), numerous horizontal cell fibers (NF+) and amacrine cell fibers (NOS+) crossed the interface between the two pieces, forming continuous plexiform layers. In overlapping pairs, NF+, NOS+, and PKC+ fibers displayed parallel plexiform layers, and no crossover of fibers was observed in any of the pair combinations examined (WT-WT, GFP-WT, GFP-rd). Some integration was seen only in small areas where the structure of both retinal pieces was disrupted at the interface.

Conclusions. The results demonstrate the ability of neurites to extend between abutting retinas and to make appropriate target choices when they are placed side-by-side. However, this ability is limited when they overlap each other, similar to that observed in subretinal transplantation.

Embryonic retinal cells transplanted to the brain develop specific connections to retinorecipient nuclei of the host brain and are capable of responding to light. ^{1 2} Like these intracranial grafts, subretinally placed neuroretinal grafts also are able to integrate with the host retina. Several studies have shown that grafted cells can send out neurites into the host retina, irrespective of the morphologic structure of the graft. ^{3 4 5 6 7 8} However, common to all these studies is the observation that the number of bridging fibers is very low. 

Several in vitro studies have established that organotypic cultures of retinal tissue can be reliably used to study retinal development, neurocircuitry, and function, for example, as well as the effect of various agents (reviewed by Seigel ⁹ ). Prompted by the problem of poor graft-host integration, we developed a modified culture system in which the outgrowth of fibers between two retinal pieces could be analyzed. The system consists of two abutting retinal pieces, placed overlapping each other, which is analogous to the in vivo situation of subretinal transplantation, or side by side. Using specific neuronal markers, we examined in wholemount preparations and in transverse sections, whether neuronal fibers can extend from one retinal piece into the abutting piece. Pairs were formed using retinal pieces derived from 5-day-old (P5) mice and cultured for 7 days, thus encompassing a time window (P5-P12) during which, in normal mouse development, substantial outgrowth of retinal cell processes occurs within the synaptic layers, and synaptic maturation is initiated. ¹⁰ We also examined pairs formed between tissue derived from young animals (newborn or 5-day-old animals) and tissue derived from adult animals. In the latter, retinas derived from 60-day-old rd mice were also included, more closely reproducing the in vivo situation of subretinal transplantation. In the rd mouse, a mutation in the gene encoding the β-subunit of a rod-specific enzyme, cGMP-phosphodiesterase, leads to total loss of rod cells by the first 5 weeks of age, leaving for the subsequent 2 to 3 months only a progressively declining number of cone cells. ^{11 12}  

Materials and Methods

Animals and Tissue Culture Preparation

The experiments were conducted with the approval of the local animal experimentation and ethics committee. Animals were handled according to the guidelines on care and use of experimental animals set forth by the Government Committee on Animal Experimentation at the University of Lund and the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. The organ culture condition has been described in detail. ¹³ Retinas were dissected from normal mice (C57BL/6), from GFP mice (harboring a transgene consisting of enhanced GFP [EGFP] cDNA under the control of a chicken β-actin promoter and a cytomegalovirus enhancer), ¹⁴ and from their congenic controls, killed at 0 (P0), 5 (P5), and 60 (P60) days after birth. Retinas from rd mice (C3H/HeA, rd/rd) and from their congenic controls were obtained from 60-day-old animals (P60). Normal animals and congenic controls are referred to as wild-type (WT) throughout. 

After the superior and nasal cornea were marked, the eyes were enucleated under sterile conditions and transferred to a dish containing serum-free medium (R16; Invitrogen-Gibco, Gaithersburg, MD). ¹⁵ Retinas were dissected from the retinal pigment epithelium (RPE) and from hyaloid vessels. Each retina was cut under fresh medium into four pieces along the superior-inferior and the nasal-temporal axes (Fig. 1A) , in such a way that each quadrant could be identified. Retinal pieces from two different animals were put onto a cellulose filter attached to a polyamide grid, either side by side (touching each other at the central margins) or partially overlapping. The pairs formed are depicted in Figures 1B 1C 1D 1E and Table 1 . All retinal pairs were flatmounted with the photoreceptor side on the filter, except for the top pieces of overlapping pairs, which were placed with either the photoreceptor side or the ganglion cell side down. In most overlapping pairs, the overlapping region usually corresponded to less than 50% of the total length of each piece. Pairs including tissue from 5- or 60-day-old animals (P5 or P60) were cultured for 7 days, whereas pairs including tissue from newborn animals (P0) were cultured for 12 days, allowing the pieces derived from young animals (P0 and P5) to reach approximately the same developmental stage (∼P12). All pairs were incubated in 1.6 mL of serum-free medium, changed every 2 days. 

Tissue Preparation

A few preparations (side-by-side pairs and overlapping pairs) were fixed in Bouin for 24 hours, rinsed, dehydrated, and paraffin embedded. Sections (4 μm) were stained with hematoxylin and eosin. For immunohistochemistry, explants were fixed in 4% paraformaldehyde in Sörensen’s buffer (0.1 M; pH 7.2) and kept at room temperature for 2 hours. The tissue was subsequently rinsed and cryoprotected in Sörensen’s buffer containing increasing concentrations of sucrose at 4°C. Some side-by-side pairs were immunoprocessed as wholemounts. The remaining side-by-side pairs and the overlapping pairs were embedded in an albumin-gelatin medium (30 g egg albumin, 3 g gelatin, 100 mL distilled water) and frozen. Cross sections (12 μm) were cut on a cryostat, collected onto gelatin-coated glass slides, and air dried. 

Immunohistochemistry

The wholemounts and cryostat sections were preincubated for 90 minutes with 0.1 M PBS containing 0.25% Triton X-100, 1% BSA (PBTX), and 5% normal serum, followed by incubation for 12 to 24 hours at 4°C with primary antibodies: mouse anti-neurofilament (NF, 200 kDa, 1:1000; Sigma-Aldrich, St. Louis, MO), sheep anti-neuronal nitric oxide synthase (nNOS, 1:4000; a gift from Ian G. Charles and Piers C. Emson, Medical Research Council, Cambridge, UK), rabbit anti-human protein kinase C (PKC, 1:1000; Chemicon, Temecula, CA), and glial fibrillary acidic protein (GFAP, 1:1500; Dako, Glostrup, Denmark). All primary antisera were diluted in PBTX containing 2% normal serum. After rinsing, sections were incubated for 90 minutes with Texas red sulfonyl chloride conjugated to donkey anti-mouse, donkey anti-sheep, or donkey anti-rabbit (1:100; Jackson ImmunoResearch, West Grove, PA). Some cross sections were also processed for the localization of microglial cells by incubation for 60 minutes at room temperature with isolectin GS-IB4 from Griffonia simplicifolia lectin (Alexa Fluor 594 conjugate; Molecular Probes, Eugene, OR). After the staining procedure was completed, sections were rinsed, coverslipped with buffered glycerol containing the antifade agent phenylenediamine, and examined by light epifluorescence and laser scanning confocal microscopy (model 1024; Bio-Rad, Richmond, CA). 

Results

Side-by-Side Retina Cultures

Side-by side pairs were formed from tissue derived from 5-day-old or 60-day-old normal animals (Table 1) . No differences were noted in the morphology of the retinas in GFP-wild type (GFP-WT) cultured pairs compared with WT-WT pairs. The general appearance of abutting P5 retinal pieces (P5-P5) placed side-by-side was also indistinguishable after 7 days in vitro (div) from that of single retinal cultures under the same conditions. The morphologic organization of the explanted retinas was also comparable to that of intact tissue obtained from age-matched littermates (P12), except for a slight overall thinning of the retina, poor development of photoreceptor outer segments, and the presence of rosettes at the edges of the retinal pieces (not shown). Further, no obvious differences could be detected between young (P5-P5) and adult (P60-P60) retinal explants, except that a few more pyknotic cells were observed in adult explants. 

The two pieces were made to touch each other along their central retinal margins (Figs. 1D 2E 4A) . In Figure 2 , sections through planes A and B correspond to more peripheral regions, where the two pieces barely touched due to the circular shape of the retina (Figs. 2A 2B) . Good fusion with continuous inner retinal layers was seen between the two pieces in the areas where the two abutted (Figs. 2C 2D) . The outer nuclear layer (ONL) of the two pieces also fused, but usually formed folds. 

Neuronal cells and processes could be detected in all retinal pieces of WT-WT and GFP-WT pairs by using neuronal markers. The localization of the stained cells and their morphologic characteristics indicated that the markers used labeled the expected cell types. 

In explants obtained from young animals (P5-P5+7div), only horizontal cells and their processes were strongly immunoreactive for 200-kDa NF, whereas in explants obtained from adult animals, ganglion cells and axons also appeared stained, which corresponds to normal observations in vivo. ¹⁶ In wholemount preparations of young pairs (Fig. 3A) , many NF-positive fibers crossed the border in the regions where the two pieces fully touched. A continuous NF-immunolabeled outer plexiform layer (OPL) was created by the bridging fibers, as seen in cross sections, confirming that they originated from horizontal cells (Fig. 3B) . The two retinal pieces of adult pairs (P60-P60+7div) appeared to have fused as well as young pairs. A few NF-immunolabeled ganglion cell axons were visible (Fig. 3C) , in addition to horizontal cell fibers, projecting toward the abutting piece. 

NOS immunoreactive amacrine cells were present in WT-WT pairs (P5-P5+7div) in the inner nuclear layer (INL) with stained fibers extending within the inner plexiform layer (IPL). As with horizontal cell processes, NOS-immunolabeled fibers ran across the two pieces, forming a continuous IPL between the two (Fig. 3D) . 

After protein kinase C (PKC) immunostaining, labeled bipolar cells were detected in the INL, with corresponding labeling of dendrites in the OPL and of axon terminals in the IPL. Unlike the observations with NF and NOS, PKC-labeled fibers were not detected projecting laterally between the two retinal pieces. However, a good fusion of the labeling profiles was observed, so that it was difficult to distinguish the margins of the two pieces at the inner retinal level (Fig. 3E) . 

GFP-WT pairs (P5-P5+7div) were then used to facilitate the identification of the boundaries between the abutting retinal pieces (Fig. 4A) . In cross sections of GFP-WT pairs, similar to observations in WT-WT pairs, NOS-immunolabeled fibers also formed a continuous plexus between the two pieces (Fig. 4B) . In wholemount preparations of such pairs, it was also possible to follow individual NF-labeled fibers crossing the border between WT- and GFP-retinal pieces (Figs. 4C 4D) . 

Figure 4D shows a few GFP-expressing fibers projecting into the WT retinal piece. Such fibers in colocalization studies were found to express a glial cell marker (see later discussion). The latter is in accordance with the observation that in the GFP mouse strain used, mainly photoreceptors, Müller cells, and possibly astrocytes in the nerve fiber layer expressed GFP (Figs. 4E 4F) . 

Overlapping Retina Cultures

Overlapping pairs were formed with tissue derived from animals of the same age or of different ages. Retinas from both normal and rd mice were used in these cultures (Table 1) . In P5-P5+7div overlapping pairs, the general appearance of the retinal pieces was analogous to that in side-by-side pairs, both in WT-WT and GFP-WT explants, provided that the overlapping portion was less than half the size of each piece (Fig. 5) . Disorganization of the retinal structure by folds or rosettes was seen at the edges of the retinal pieces (Figs. 5A 5B 5D 5E) , but seldom in the more central regions (Figs. 5A 5C 5E 5G) . In some cases, where the overlaying retinal piece covered the total length of the piece beneath by more than 50%, rosette formation and disruption of the retinal layers was apparent in the latter (Figs. 5D 5F) . Normal retinal morphology with regular lamination and good adherence between the two pieces were otherwise noted, irrespective of the orientation of the overlaying piece (ONL of the overlaying piece facing the ganglion cell layer [GCL] of the piece beneath [ONL-GCL; Figs. 5A 5B 5C 5D 5F ] or both GCLs facing each other [GCL-GCL; Figs. 5E 5G ]). 

The localization of labeled cells and processes in P5-P5+7div overlapping pairs was analogous to that observed in side-by-side pairs. However, unlike what was observed in the latter, NF-, NOS-, and PKC-labeled fibers appeared unable to cross the borders between the two pieces wherever the structure of the retinas was well preserved. NF-labeled fibers were thus confined to the OPLs, running parallel to each other in both ONL-GCL (Fig. 6A) and GCL-GCL pairs (Fig. 6B) . NOS-labeled fibers were also seen to run within the IPL of each piece without bridging (Fig. 6C) , even when the distance between the two plexuses was very short, as in the GCL-GCL pairs (Fig. 6D) . The same was observed with PKC-labeled fibers, which terminated in their respective IPLs, without extending into the abutting piece (Figs. 6E 6F) . 

At times, PKC-labeled processes grew into the ONL of the overlaying piece (Fig. 7A) , yet without crossing the border between the two pieces. In contrast, bridging between the two retinal pieces occurred in overlapping pairs in association with areas where folds or rosettes had formed in the overlaying piece in ONL-GCL pairs. In theses cases, PKC-labeled cells and fibers were observed between the folds of the ONL or surrounding the rosettes, projecting toward and terminating in the IPL of the piece below (Figs. 7B 7C) . Fusion between the IPLs of the two pieces was observed, occurring at the edges of the overlaying piece, which was often folded. Figures 7D and 7E , show examples of mingling NOS-labeled fibers, both in ONL-GCL and GCL-GCL pairs. It was noted in these and other specimens that in the areas of fusion, the inner margin of the piece lying underneath also appeared to be disrupted. 

In P5-P60+7div and P0-P60+12div overlapping pairs, the overall structure of overlaying retinal pieces derived from adult normal or adult congenic control mice (WT) appeared normal. Labeled profiles were visible in their predicted locations and lamination was preserved (Figs. 8A 8B 9A) . The structure of overlaying retinal pieces derived from adult rd mice was also relatively well preserved, despite the absence of photoreceptor cells (Figs. 8C 8D 8E 8F and 9B , 9C). The structure of the underlaying retinal pieces, which derived from newborn (P0) or from 5-day-old (P5) GFP mice, also appeared normal (Figs. 8 9) . Similar to observations in P5-P5+7div overlapping pairs, no labeled crossing fibers (NF, NOS, or PKC) were observed in P5-P60+7div and P0-P60+12div overlapping retinal pieces (Figs. 8 9) . Crossing fibers were not seen, even if one of the abutting retinal pieces was derived from adult rd mice, despite the fact that in the latter, PKC-labeled bipolar cell bodies were seen next to the interface between the two pieces (Figs. 8C 8D 8E 8F) . 

Glial Markers

As mentioned earlier, a few GFP-expressing fibers extended into WT retinal pieces in side-by side pairs (P5-P5+7div; Fig. 4D ). These fibers were labeled also for glial fibrillary acidic acid (GFAP), as shown in Figure 10B . In contrast, fibers expressing GFP (and GFAP) were never observed projecting into the WT-derived piece in overlapping pairs in the regions where the structure of both pieces was well preserved (Fig. 10E) . 

On the other hand, we found GFP-expressing cells scattered within the WT-derived abutting retinal pieces, at times at a distance from the border, both in side-by-side and overlapping pairs (Figs. 10A 10C 10D 10E) . In both cases, these GFP-expressing cells were found in colocalization studies to label for IB4, the microglial cell marker (Figs. 10C 10D) . 

Discussion

The present study was conducted with the purpose of developing an in vitro system in which bridging of neuronal fibers between two retinal pieces could be analyzed. Some deviations from normal retinal development were observed in the culture condition, but they were not found to limit the usefulness of the system. Previous studies of organotypic cultures of neural retina have shown that the differentiation of most retinal cell types and the synaptic development are comparable to the in vivo condition. ^{13 17 18} A relatively normal morphology and the predicted distribution of neuronal cell markers were observed also in the abutting retinal pieces cultured under the conditions used in the present study. However, similar to previously reported findings for single-retina cultures, a poor development of photoreceptor outer segments was observed, which is believed to result, at least in part, from the absence of RPE. ^{13 18} However, for the purpose of comparison with the situation in in vivo transplantation, in which the RPE is absent between the graft and the host retina, the RPE was not included in the present study. Furthermore, it has been recently demonstrated that the poor development of outer segments does not influence the establishment of the neuronal network. ¹⁸  

As previously observed in single-retina cultures, ^{13 17} rosettes and folds developed at the edges of the explanted retinal pieces. However, they did not appear to influence the attachment of the underlaying piece to the filter or the adhesion between the two pieces in the overlapping areas. In fact, bridging fibers were seen only in pairs in which folds or rosettes formed in the overlaying piece (see later discussion). 

The side-by-side setup was valuable because it revealed extensive neuronal crossovers, demonstrating the potential for neuronal integration between two different retinas. Previous in vitro studies have demonstrated the ability of embryonic ganglion cell axons to grow into and out of intraretinal grafts surrounded by chick or quail host tissues. ¹⁹ By using various neuronal cell markers, we were able to demonstrate that at least horizontal and amacrine cells also have the ability to project to appropriate target areas, not only within the same piece, but also within the abutting retina. Outgrowth of amacrine cell processes has been demonstrated in preparations of dissociated retinal cells cultured onto different cellular adhesion molecule (CAM) substrates or cocultured with heterologous cells expressing CAMs. ^{20 21} However, this, to our knowledge, is the first in vitro study to demonstrate outgrowth from nonprojecting neurons between two retinal pieces. Further examination at the electron microscopic level is necessary to determine whether synaptic contacts are indeed established in these conditions. Nevertheless, our observations give an indication of the high degree of plasticity that can be exhibited by the growing neurites. 

It is therefore noteworthy that neuronal integration was limited when the two pieces overlapped each other in the same culture condition of side-by-side pairs. This observation is in accordance with those from several in vivo studies of subretinal neuroretinal transplantation in which neuronal integration occurred, but was not extensive. ^{3 4 5 6 7 8 22 23 24 25 26} This could be due to several factors, such as the limited plasticity of the grafted cells, the age difference between the grafted cells and the host retina (so that the host environment may not be optimally permissive or attractive to developing graft-derived fibers), immunologic factors, and/or other chemical or physical barriers. Immunologic factors can be ruled out in the present study as a reason for poor integration. A deficiency in plasticity, as such, may also be disregarded in the current experiments, because extensive neurite outgrowth was observed between side-by-side pairs, not only between pieces derived from young animals (P5), but also between retinas derived from adult animals. Integration was not observed in areas where the structure of the retinal pieces was preserved, regardless of whether pieces derived from animals of the same age (P5-P5) or of different ages (P0-P60; P5-P60) were cocultured. This suggests that age mismatch was not what primarily prevented the bridging of fibers through those areas. In the present study, the retinal pieces were cultured for 7 or 12 days, and extensive neurite elongation was seen in side-by-side pairs as early as 7 days. It is therefore unlikely that better integration would occur in overlapping pairs with prolonged culture times as long as the structure of both retinal pieces is well preserved. 

During development, differentiating retinal neurons migrate to their final positions forming a laminar structure. As long as this structure is maintained, neurites will project to specific target areas guided by complex interactions between cell surface and matrix molecules. The radial glial cells of the retina, the Müller cells, have been proposed to play an important role in this process, by providing an orientation scaffold to the migrating cells. ^{27 28 29 30} Disruption of Müller cells by glial cell toxins has for instance been shown to result in a displacement of developing photoreceptor cells into the inner segment area, apparently due to a loss of the integrity of the outer limiting membrane (OLM), which consists of adherens junctions between Müller cells and photoreceptors. ^{31 32} After subretinal transplantation of fragmented retinal tissue, we have previously found bridging of neuronal fibers between the graft and the host only between folds and rosettes produced in the host outer nuclear layer or in areas where the host photoreceptor cell layer was discontinuous or missing at the graft-host border. ^{8 26} These observations have led us to propose that neuronal fibers are only able to project between the graft and the host retina upon loss of integrity of the host outer limiting membrane with accompanying modifications in the physical and chemical properties of Müller cells and/or photoreceptors. ²⁶  

We found in the present study that bridging fibers were not observed, even if retinal pieces derived from adult rd mice (P60) were placed next to pieces derived from young animals (P0 or P5), a situation that more closely compares with that of in vivo subretinal transplantation. At P60 in the rd mouse retina, most rod photoreceptor cells have degenerated, but cones can still be observed. ^{11 12 33 34 35} We have not found rod photoreceptor cells in any of the overlaying rd retinas processed for rhodopsin immunohistochemistry (not illustrated), which does not exclude the possibility that cone photoreceptor cells could still remain. It is possible then that crossing fibers are barred at the level of the OLM, which at P60 in the rd mouse is likely to consist mostly of junctions between Müller cells, but also of a few junctions between Müller cells and cone photoreceptor cells. A physical and/or molecular barrier may be created at this site, e.g., as a result of the activation of Müller cells (induced by the photoreceptor degeneration and by the culturing of the retina), resulting in structural changes and the production of molecules potentially inhibitory for neurite outgrowth, ³⁶ not allowing fibers to penetrate. This notion is in good agreement with previous in vivo studies showing at the electron microscopic level the presence of a glial barrier between subretinal grafts and the host retina. The barrier was found in these cases to be associated with the outer margin of the host retina and coincided with areas of poor graft-host integration. ^{4 24}  

Bridging may be prevented also at the vitreal surface of the underlaying retinal piece. The innermost elements of the postnatal retina include the axons of ganglion cell; Müller cell basal processes, some of which ensheathe the axons; astrocytes, which in mice are predominantly associated with blood vessels ³⁷ ; microglial cells; and a basal lamina (the inner limiting membrane [ILM]), a thin sheet composed of extracellular matrix proteins. ³⁸ Microglial cells are found evenly distributed in the rat retina at birth, and an increase in their number is seen during the first postnatal week. ³⁹ Astrocytes are restricted to the most central retina at birth, and as the retinal vasculature develops, their number increases and they can be seen throughout the retina. ³⁷ On isolation of older postnatal retinas, the optic nerve is sectioned and the blood supply is interrupted, inducing a reactive response perhaps in all glial cell types (Müller cells, astrocytes, microglia). The activation of one or more of these glial cell types could then result in the creation of a barrier to the passage of neurites. The retinal pieces used in the present study originated from newborn mice and 5-day-old mice, and they were taken from all quadrants and included central and peripheral retina. Yet, bridging of fibers was never observed through overlapping areas where the margins of the pieces was undamaged, irrespective of the retinal piece used or the age of the tissue (derived from P0 or P5 mice). 

A role in restricting graft-host integration has been ascribed to the ILM itself. In studies in which retinal tissue was placed epiretinally, it has been shown at the electron microscopic level that nerve cell processes originating in the grafts could enter the host retina only through breaks in the ILM. ⁴⁰ Also, in subretinal transplantation of intact retinal sheets, it has been found that graft-host integration is poor in areas where the equivalent to an ILM is observed in the graft. ^{6 22 23 25} This was seen even with fetal donor tissue, in which, as mentioned earlier, astrocytes are found only near the optic disc and which contains fewer microglial cells than postnatal retinas. ^{37 39} Thus, although we did not use fetal tissue in the present study, these in vivo studies ^{6 22 23 25} indicate that a barrier is formed at the level of the inner margin prenatally. 

It appears thus that structures and/or molecules at the inner and outer surfaces of the retina may prevent fibers from crossing the interface. Accordingly, bridging fibers were seen in the present study only in overlapping pairs when the structure of the outer retina in the overlaying piece and of the inner retina in the piece beneath were disrupted (in ONL-GCL pairs) or when the structure of the inner retina in both pieces was disrupted (in GCL-GCL pairs). In ONL-GCL pairs, fusion was observed in places where folds or rosettes developed in the overlaying piece. In these cases, the inner retinal neurons and fibers of the overlaying piece were displaced toward the piece beneath, bringing them closer to the border between the two. However, the short distance between the plexuses of the two pieces did not alone allow fusion, because in GCL-GCL overlapping pairs, PKC- and NOS-containing fibers running within the respective IPLs appeared in most cases unable to cross the border between the two pieces. In the latter, bridging of fibers was observed only in places where the inner margin of the two pieces was disrupted, which was often seen at the edge of the overlaying piece. 

Another important observation made in this study was that, similar to what was seen with neurites, glial cell fibers (stained with GFAP) were able to extend into the abutting retinal piece in side-by-side pairs but not when the pieces overlap each other, suggesting an interdependence between the outgrowth of neurites and of glial fibers. A close association has been observed between ectopic outgrowth of neurites and that of glial processes in retinas subjected to experimental detachment and in retinas with tapetoretinal degenerations. In both cases, the growing neurites were seen in areas of Müller cell activation, running along their processes. ^{41 42 43} It is, in this context, interesting then to note that microglial cells (identified with IB4) are able to migrate into the abutting retinal piece both in side-by side as well as in overlapping pairs. Activation of microglial cells can be expected in the explants and has been demonstrated in organotypic retinal cultures. ⁴⁴ It thus appears that whatever mechanisms that prevent neurites from projecting into the abutting piece in overlapping pairs also prevent the crossing of glial cell fibers, but not the crossing of microglial cells. 

The abutting-retinas culture system described herein thus provides evidence that although extensive neuronal fusion can occur between two retinas when placed side by side, integration is limited when they overlap each other, which is then analogous to in vivo observations, even in long survival times after subretinal transplantation. Numerous morphologic and functional studies have shown that graft-host integration occurs. ⁴⁵ However, the presence of bridging fibers is often restricted to small areas, and their number is always low. ^{3 4 5 6 7 8 22 23 24 25 26} Similarly, although functional improvement has been demonstrated, ^{46 47 48 49 50 51} it cannot be observed at all times, is only seen in a small number of surgical subjects, or cannot be unequivocally attributed to the existence of graft-host functional connections. ^{50 51 52 53} The functional capacity of a subretinal transplant is certainly determined by a number of different aspects. ⁴⁵ However, it is reasonable to establish that extensive graft-host integration, or as shown in the current study, integration between overlapping retinal pieces, is limited and that the factors involved in this process should be identified and, if possible, should be overcome. The system developed here was useful in simulating the in vivo situation of subretinal transplantation and offers therefore extended possibilities for examining these factors. 

 Supported by the Faculty of Medicine, University of Lund; Swedish Medical Research Council Grant MTRP 12209; Crown Princess Margareta’s Committee for the Blind; the Swedish Society for Medical Research; the K. A. Wallenberg, Crafoord, C. Groschinsky, T. E. Segerfalk, and Syskonen Svensson Foundations; a 2nd O.N.C.E. International Award for New Technologies for the Blind; European Union Grants QLK-CT-2000-00569; QLK6-CT-2001-00385; and the Foundation Fighting Blindness.

 Submitted for publication June 28, 2002; revised December 2, 2002, and February 26, 2003; accepted March 14, 2003.

 Disclosure: Y. Zhang, None; A.R. Caffé, None; S. Azadi, None; T. van Veen, None; B. Ehinger, None; M.T.R. Perez, None

 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.

 

Figure 1.

 

Schematic illustration of the abutting-retinas culture system and culture procedure. (A) Each retina was cut into four pieces (a, b, c, and d) through the superior-inferior and nasal-temporal axes. (B, C) The quadrants of each piece were identified, with a1 corresponding to a2, b1 to b2, c1 to c2, and d1 to d2. (D) Two retinal pieces were placed side by side (touching each other at the central margins), forming the pairs a1-b2, a2-b1, c1-d2, and c2-d1. (E) Two pieces were placed partially overlapping forming pairs of the same designation as in (D).

Table 1.

 

Number of Examined Pairs

	WT-WT			GFP-WT			GFP-rd
		P5–P5 (+7div)	P60–P60 (+7div)	P5–P5 (+7div)	P5–P60 (+7div)	P5–P60 (+7div)	P0–P60 (+12div)
Side by side (ONL down)	24	16	16	—	—	—
Overlapping (ONL-GCL)	12	—	4	9	11	6
Overlapping (GCL-GCL)	9	—	4	—	—	—

Figure 2.

 

Hematoxylin and eosin staining: (A–D) cross sections illustrating the morphology of abutting retinal pieces in WT-WT pairs (P5-P5+7div) when they were placed side by side. (E) Wholemount preparation showing two abutting retinal pieces. (A–C) Sections correspond to the planes designated as A, B, and C in (E). (D) Higher magnification of the area within the box in (C). WT1, WT2, retinal pieces derived from two different WT animals. Scale bars: (A–D) 50 μm; (E) 300 μm.

Figure 3.

 

(A) and (C) Wholemount preparations showing NF-labeled fibers crossing the border between the two abutting retinal pieces in WT-WT pairs placed side by side. Horizontal cell processes appeared immunoreactive in young (P5-P5+7div) retinas (A, confocal image), whereas in adult pairs horizontal cell processes (C, small arrow) and ganglion cell axons appeared labeled (C, large arrow). (B, D–E) Cross sections of (P5-P5+7div) side-by-side WT-WT retinal pairs from different animals. (B) Continuous NF-immunolabeled OPL). (D) NOS immunolabeling showing the same phenomenon at the level of the IPL (arrows). (E) Good integration of the pieces was detected with PKC staining, with an indistinguishable border at the level of the INL and IPL. WT1, WT2, retinal pieces derived from two different WT animals. Scale bars: (A, B, D, E) 50 μm; (C) 200 μm.

Figure 4.

 

(A) Wholemount preparation showing a WT-GFP retinal pair (P5-P5+7div), placed side by side. (B) Cross section showing NOS-immunolabeled fibers at the level of the IPL at the transition between WT- and GFP-derived tissues (arrows). (C, D) Confocal images: NF-labeled fibers in wholemount preparations crossed the border between the WT- and the GFP-derived pieces. GFP-expressing fibers (NF negative) also extended into the WT-derived piece. (E, F) GFP expression (green) at P12 in noncultured retina of GFP mice in photoreceptors in the ONL, in Müller cell bodies (in the INL) and radial processes, and at the level of the GCL. Note that the NOS-immunolabeled displaced amacrine cell in the GCL (red, arrow) did not express GFP. NOS-expressing fibers in the IPL (E, red) and NF-expressing fibers in the OPL (F, red) were also negative for GFP. WT, GFP, retinal pieces derived from the corresponding mouse strains. Scale bars: (A) 300 μm; (B–F) 50 μm.

Figure 5.

 

Hematoxylin and eosin staining of cross sections illustrating the morphology of abutting retinal pieces in WT-WT pairs (P5-P5+7div) when placed overlapping each other. (A–D, F) The ONL of the overlaying piece faces the GCL of the piece beneath (ONL-GCL pairs). (E, G) The GCL of both pieces face each other (GCL-GCL pair). (A, B) At the edges of the retinal pieces, some disruption of the morphology was observed. (B) Higher magnification of the area within the dashed box in (A). (D, F) In pairs in which the overlaying piece covered more than half of the piece beneath, degeneration of the latter was also observed. (F) Higher magnification of the area indicated by the box in (D). Otherwise, good morphology was seen in both pieces in the central regions in ONL-GCL pairs (A, C) and in GCL-GCL pairs (E, G). (C) Higher magnification of the area within the solid box in (A); (G) higher magnification of the area within the solid box in (E). Scale bars: (A, D, E) 100 μm; (B, C, F, G) 50 μm.

Figure 6.

 

Neuronal fibers were not detected crossing the border between two overlapping pieces in WT-WT pairs (P5-P5+7div) when the structure of both pieces appeared normal. (A, C, E) ONL-GGL pairs; (B, D, F) GCL-GGL pairs; (A, B) NF immunolabeling; (C, D) NOS immunolabeling; (E, F) PKC immunolabeling. WT1, WT2, retinal pieces derived from two WT animals. Scale bar, 50 μm.

Figure 7.

 

(A) A sprouting PKC-labeled fiber projected through the ONL of the overlaying piece (arrow). Yet, the fiber did not extend beyond the outer margin of the ONL. (B, C) PKC-labeled cells and fibers appeared between the folds formed in the overlaying piece and intermingled with PKC-labeled fibers of the piece beneath (arrows). Bridging of NOS-immunolabeled fibers also occurred in both the ONL-GCL and GCL-GCL pairs, where the inner margins of both pieces appeared disrupted, often around the edges of the overlaying piece (D, E). WT1, WT2, retinal pieces derived from two WT animals. Scale bars, 50 μm.

Figure 8.

 

No bridging of PKC-labeled fibers (red) was detected in overlapping retinal pairs derived from (A, B) P5 GFP mice and P60 WT mice (P5-P60+7div); (C, D) P5 GFP mice and P60 rd mice (P5-P60+7div); (E, F) P0 GFP mice and P60 rd mice (P0-P60+12div). (A, C, E) GFP fluorescence (green) demarcates the position of the underlaying retinal pieces. WT, rd, and GFP, retinal pieces derived from the corresponding mouse strains. Scale bars, 50 μm.

Figure 9.

 

NOS immunolabeling in overlapping pairs derived from (A) P5 GFP and P60 WT mice (P5-P60+7div); (B) P5 GFP and P60 rd mice (P5-P60+7div); (C) P0 GFP and P60 rd mice (P0-P60+12div). In all cases, NOS-immunolabeled fibers were seen to run in parallel plexuses, without projecting into the abutting retinal piece. WT, rd, and GFP, retinal pieces derived from the corresponding mouse strains. Scale bars, 50 μm.

Figure 10.

 

(A, B) Wholemount preparations showing in more detail GFP-expressing fibers projecting into the WT-derived retinal piece in a side-by-side pair (P5-P5+7div). Several of these GFP-positive fibers also expressed GFAP (B, red). GFP-expressing cells were also present within the WT-derived piece (A); most of these cells also expressed IB4 (C, yellow, cross section). (D) GFP-expressing cells were seen within the WT-derived piece also in overlapping pairs (P5-P5+7div), all of which express IB4 (E). In contrast, no fibers expressing both GFP and GFAP were seen to project into the WT-derived piece. IB4, isolectin GS-IB4; WT, GFP, retinal pieces derived from the corresponding mouse strains. Scale bars, 50 μm.