Phagocytosis of Photoreceptor Outer Segments by Transplanted Human Neural Stem Cells as a Neuroprotective Mechanism in Retinal Degeneration

Abstract

Purpose.: Transplantation of human central nervous system stem cells (HuCNS-SC) into the subretinal space of Royal College of Surgeons (RCS) rats preserves photoreceptors and visual function. To explore possible mechanism(s) of action underlying this neuroprotective effect, we performed a detailed morphologic and ultrastructure analysis of HuCNS-SC transplanted retinas.

Methods.: The HuCNS-SC were transplanted into the subretinal space of RCS rats. Histologic examination of the transplanted retinas was performed by light and electron microscopy. Areas of the retina adjacent to HuCNS-SC graft (treated regions) were analyzed and compared to control sections obtained from the same retina, but distant from the transplant site (untreated regions).

Results.: The HuCNS-SC were detected as a layer of STEM 121 immunopositive cells in the subretinal space. In treated regions, preserved photoreceptor nuclei, as well as inner and outer segments were identified readily. In contrast, classic signs of degeneration were observed in the untreated regions. Interestingly, detailed ultrastructure analysis revealed a striking preservation of the photoreceptor–bipolar–horizontal cell synaptic contacts in the outer plexiform layer (OPL) of treated areas, in stark contrast with untreated areas. Finally, the presence of phagosomes and vesicles exhibiting the lamellar structure of outer segments also was detected within the cytosol of HuCNS-SC, indicating that these cells have phagocytic capacity in vivo.

Conclusions.: This study reveals the novel finding that preservation of specialized synaptic contacts between photoreceptors and second order neurons, as well as phagocytosis of photoreceptor outer segments, are potential mechanism(s) of HuCNS-SC transplantation, mediating functional rescue in retinal degeneration.

Introduction

Degenerative diseases of the retina affect millions of people worldwide, causing permanent and progressive vision loss. Age-related macular degeneration (AMD) and retinitis pigmentosa (RP) comprise the largest patient populations; however, very limited treatment options currently are available. Fortunately, a number of animal models are available that replicate or simulate the disease process in humans. A commonly used model is the Royal College of Surgeons (RCS) rat. This rat carries a mutation in the Mertk gene ¹ in RPE cells that largely eliminates the RPE cells' natural function of phagocytosing shed photoreceptor outer segments (OS). The resulting accumulation of toxic debris in the subretinal space causes progressive death of photoreceptors and consequent vision loss. ^2,3 Photoreceptor degeneration in this model secondarily affects horizontal and ganglion cells, leading to significant structural abnormalities. ^{4 –6}  

There are two potential treatment paradigms in diseases of photoreceptor loss: neuroprotection or neuroreplacement. While replacing lost photoreceptors with new donor photoreceptors or photoreceptor progenitor cells is extremely attractive for treating late stage degeneration, there are significant obstacles to overcome, including precise donor cell integration, secondary synaptic rewiring, and vascular changes that complicate a replacement approach. ⁷ Preventing the death of susceptible photoreceptors, that is, neuroprotection, may be a more feasible short-term approach. Transplantation of cells derived from exogenous sources into the subretinal space of young RCS rats have been shown previously to limit photoreceptor death and maintain vision. ^{8 –11} The most effective cell types from a neuroprotection standpoint fall into two strategic categories: cells that provide neurotrophic support, which has been shown to sustain photoreceptor survival, ^9,12,13 and cells that provide mechanical support, such as RPE cells, which are believed primarily to replace the lost phagocytotic function of the host RPE. ^{14 –18}  

The use of stem cells for the treatment of photoreceptor degeneration may address cell replacement and neuroprotection. One particularly exciting stem cell is the human central nervous system stem cell (HuCNS-SC), already in clinical testing for metabolic and dysmyelinating CNS diseases, as well as spinal cord injury. ^19,20 Recently, HuCNS-SC were transplanted into RCS rats in a study that detailed the rescue of optokinetic tracking-based eyesight and of cone photoreceptors that subserve photopic visual acuity. ²¹ The study also described the postinjection migration pattern of transplanted cells, which concentrated at one location immediately after transplantation and then migrated outward radially in the subretinal space over the following weeks. A particularly interesting finding was that, in the area of retina with grafted HuCNS-SC, the typical debris zone, consisting of photoreceptor OS and other materials, was absent; however, away from the graft, the debris zone was present. Preliminary in vitro studies of the potential mechanism of action revealed the HuCNS-SC phagocytosed fluorescently-labeled photoreceptor OS. Based on these preliminary results, our study was designed to examine whether the donor cells phagocytose photoreceptor OS in the RCS rat as a potential mechanism of photoreceptor rescue and vision maintenance. Specifically, we examined the phagocytic capability of two donor cell lines (DCL): one used in previous published studies ²¹ and a good manufacturing practice/good tissue practice (GMP/GTP) compliant DCL currently being used in a Food and Drug Administration (FDA)-authorized phase I/II clinical trial in AMD with geographic atrophy (GA; Clinicaltrials.gov identifier NCT01632527). 

Materials and Methods

Isolation and Expansion of HuCNS-SC

The HuCNS-SC were generated as described previously. ^22,23 Briefly, donated second trimester (16–20 gestation weeks) human brain tissue was dissected, treated enzymatically, and labeled with CD133 and CD24 antibodies. The CD133⁺CD24^−/lo target population was sorted aseptically using a BD Vantage (San Jose, CA) flow cytometer. Sorted cells were cultured as a neurosphere suspension in a chemically-defined, serum-free culture medium composed of X-VIVO (Lonza, Walkersville, MD) 15 medium supplemented with N2, heparin, N-acetyl cysteine (NAC), basic fibroblast growth factor (FGF2), epidermal growth factor (EGF), and leukemia inhibitory factor (LIF) at a density of 10⁵ cells/ml. When neurosphere size reached 200 to 250 μm, cultures were passaged by collagenase treatment and replated in the same medium. 

For GMP manufacturing compliant operations, the research process was developed and qualified further. With the GMP process, a master cell bank (MCB) was created and cryopreserved at an early passage. Working cell banks (WCB) were generated from the MCB and were cryopreserved at later passages. Products for transplant were generated from thawed WCB vials. For DCLs used for clinical trials, all tissue collection was performed in compliance with GTP, the National Organ Transplant Act, and the Uniform Anatomical Gift Act that govern the acquisition, transfer, and use of human tissue, and all tissue processing, manufacturing and testing steps are performed in compliance with applicable GMP and GTP regulations. 

Transplantation

Two-day cultures of HuCNS-SC were formulated at a density of 5 × 10⁴ cells/μL in injection buffer (X-VIVO containing NAC). Five postnatal day 21 (P21) pigmented dystrophic RCS rats received a bilateral injection of 2 μL of a suspension containing approximately 1 × 10⁵ cells into the subretinal space, as described previously. ^{14,24 –26} There were no signs of retinal damage or vascular distress immediately following injection, and all animals completed the study. Animals received daily intraperitoneal injections of dexamethasone (1.6 mg/kg) for 2 weeks following transplantation to control acute inflammatory responses, and also were maintained on oral cyclosporine A administered in the drinking water from 1 day before transplantation until the day of sacrifice. Animal procedures adhered to the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research, and were conducted with approval and under the supervision of the Institutional Animal Care and Use Committee (IACUC) at StemCells, Inc. (Newark, CA). 

Histologic Processing and Microscopy

The first part of the tissue processing and immunostaining was conducted in California. Rats were euthanized at ages P60 and P90 by CO₂ asphyxiation, and the eyes fixed by immersion in 2% paraformaldehyde and 0.05% glutaraldehyde in 0.1 M sodium phosphate buffer (PB) at pH 7.4 for 30 minutes. The corneas and vitreous bodies were removed, and the eye cups further fixed for 90 minutes in the same paraformaldehyde-glutaraldehyde fixative, followed by a 1-hour fixation in 2% paraformaldehyde. After fixation, radial cuts were made to flatten the eye cups and the retinas were dissected away from the RPE layer; this was followed by a 1-hour wash in 0.1 M PB and a 15-minute incubation in 1% sodium borohydride. Endogenous peroxidase activity was blocked by a 15-minute incubation in 1.5% hydrogen peroxide (H₂O₂). After several rinses in 0.1 M PB, retinas were cryoprotected in 15% sucrose for 30 minutes, 20% sucrose for 1 hour, and 30% sucrose overnight. The next day, retinas were put through a freeze-thaw procedure involving dipping in liquid nitrogen–cooled isopentane for a few seconds. After thawing in 30% sucrose, retinas were washed in 0.1 M PB. Immunostaining of the whole mount retinas was performed as follows: the retinas were incubated in 10% normal horse serum in 0.1 M PB for 1 hour at 4°C, then transferred, without washing into primary antibody, STEM121 (Stem Cell Sciences, Cambridge, UK), at a concentration of 0.5 μg/mL in 0.1 M PB for 4 to 5 days at 4°C under agitation. After further washes in PB, the retinas were incubated with biotinylated horse antimouse secondary antibody (IgG; Vector Laboratories, Inc., Burlingame, CA) diluted 1:100 in 0.1 M PB, for 2 days at 4°C. The tissue was washed before transferring to a solution of ABC (avidin-biotin complex; Vector Laboratories, Inc.) in 0.1 M PB for 2 days. Finally, the retinas were washed in PB and preincubated under agitation in the dark with 3,38-diaminobenzidine tetrahydrocholoride (DAB; Vector Laboratories, Inc.) for 15 minutes and further incubated with fresh DAB solution with 0.01% H₂O₂. The DAB reaction was stopped by washes in distilled water. The stained retinas were flattened in a sandwich of filter paper and coverslips, and shipped to the University of Alicante. The tissues were postfixed in 4% paraformaldehyde and 2% glutaraldehyde in 0.1 M sodium phosphate buffer, pH 7.4 for 1 hour and then in 1% osmium tetroxide (OsO₄) in 0.1 M PB for 1 hour. Retinas mounted in the sandwich were dehydrated in a graded series of alcohols, and embedded in Epon. The embedded retinas were sectioned serially for electron microscopy on an MT5000 ultramicrotome using a diamond knife. Semi-thick sections were stained with 1% Toluidine blue. The ultra-thin series of sections were collected on Formvar-supported slot grids, stained with uranyl acetate and lead citrate, and washed. The grids were viewed and serial sections were photographed under a JEM-1011 electron microscope (JEOL Ltd., Tokyo, Japan). The area of the retina distant from the graft was used as a matching negative control for each eye. 

Results

Light Microscopy

The HuCNS-SC cells were transplanted in the subretinal space, between the RPE and the outer nuclear layer (ONL). Before immunostaining, the neural retina was separated from the RPE (and remaining eye cup), thereby exposing the graft. The HuCNS-SC cells were detected in all processed flat mounts as a brown circular graft on the retina's outer surface. Semithin retinal sections generated from these flat mounts were stained with toluidine blue and studied under the light microscope. The HuCNS-SC grafted retinal regions (treated) as well as regions of the same retina, but distant from the HuCNS-SC graft (untreated), were studied. In this study, we did not examine sham-injected retinas, as prior studies with HuCNS-SC ²¹ and other cells ^10,11,27,28 have shown consistently that, although sham injections induce a small and localized impact on ONL thickness, this effect is transient and sham-treated retinas are mostly indistinguishable from untreated ones. Furthermore, it has been demonstrated that by P60, photoreceptor inner segments (ISs) and OSs are absent from sham-injected retinas. ²⁷ In all eyes examined at P60 and P90, human cells were detected in the subretinal space, that is, contiguous with the photoreceptor OS, forming a layer of STEM121 immunopositive (dark-brown) cells (Figs. 1A, 1C, arrows). All major retinal layers were present in the grafted region. Of particular interest was the preserved structure of the photoreceptor layer. Photoreceptor nuclei, as well as photoreceptor IS and OS, were readily and consistently identified (Figs. 1A, 1C, arrowheads). In every retina section examined that contained HuCNS-SC cells, we found 7 to 9 and 4 to 6 rows of photoreceptor nuclei at P60 and P90, respectively (Figs. 1A, 1C), compared to 10 cells deep in normal retinas. There were no major differences in the degree of photoreceptor preservation in retinas injected with either one of the DCLs tested, which also is in agreement with their comparable effect on visual acuity (see Supplementary Fig. S1). In contrast, classic signs of photoreceptor degeneration were observed in the untreated regions: rod OS and IS were missing, and a debris zone formed contiguous to the photoreceptor layer (Figs. 1B, 1D). Approximately 3 rows of photoreceptor nuclei were identified in these regions at P60 (Fig. 1B), and only a few photoreceptor nuclei remained at P90 (Fig. 1D). Also, the thickness of the outer plexiform layer (OPL) appeared reduced in untreated regions (Figs. 1B, 1D), a novel finding not reported previously to our knowledge. 

Figure 1

 

The HuCNS-SC engraft in the subretinal space of RCS rats and preserve photoreceptors. Light microscopy of toluidine blue–stained semithin retina sections across the HuCNS-SC graft area (A, C) and in control areas of the same eyes, but distal from graft site (B, D) at two time points: P60 (A, B) and P90 (C, D). A uniform layer of human cells stained with the human-specific marker STEM121 was detected on top of OS in the temporal area of the retina, where they were injected ([A, C], arrows), at time points. The ONL adjacent to the HuCNS-SC graft is well preserved, with multiple rows of photoreceptor nuclei. At right (B, D) shows cross-sections of the same retinas, but in the area distal to the HuCNS-SC graft. In this region, the ONL is much thinner and many photoreceptors have pyknotic nuclei. INL, inner nuclear layer; IPL, inner plexiform layer; GCL, ganglion cell layer. Scale bars: 20 μm.

Electron Microscopy (EM)

General Observations.

Ultrathin retina sections were examined by transmission EM. We found that cellular and subcellular structures were preserved better in HuCNS-SC–treated regions than in untreated regions. We will first describe the major differences between treated and untreated retinal regions, and then will proceed to describe in detail the ultrastructural changes linked to specific retinal processes. 

As mentioned above, the major difference between treated and untreated regions lies in the number of photoreceptor rows present in the ONL at both ages. At the EM level, all the photoreceptor nuclei in treated regions had normal chromatin at P60 (Fig. 2A) and at P90 (Fig. 2C), whereas the nuclei in untreated regions had condensed pyknotic chromatin, characteristic of apoptotic cells at P60 (Fig. 2B, arrows) and P90 (Fig. 2D, arrow). The EM further confirmed the differences in OPL thickness observed with light microscopy. The OPL of treated retinal regions (Figs. 2A, 2C, brackets) was significantly wider than the OPL in untreated regions (Figs. 2B, 2D, brackets). This suggested that the transplanted HuCNS-SC may help preserve the synaptic contacts between photoreceptors, and horizontal and bipolar cells that occur in the OPL. 

Figure 2

 

EM reveals normal structure of the outer retina adjacent to HuCNS-SC grafts. HuCNS-SC transplantation (A, C, E) preserves the normal structure of the outer retina compared to untreated regions (B, D, F). Brackets highlight the increased OPL thickness in HuCNS-SC treated versus untreated retinas. A clear and continuous adherens junction between Müller cells and photoreceptors could be observed in the treated regions ([A, C, E], arrowheads), but not in untreated regions ([B, D, F], arrowheads). Arrows indicate apoptotic nuclei in area distal to the HuCNS-SC graft (B, D). OML, outer limiting layer. Scale bars: (A–D) 10 μm, (E, F) 1 μm.

The HuCNS-SC transplantation also preserved the outer limiting membrane (OLM); a clear and continuous adherens junction between Müller cells and photoreceptors could be observed in the treated regions at P60 and P90 (Figs. 2A, 2C, 2E, arrowheads). In untreated regions at P60, the OLM was difficult to identify except for a few and large junctions (Figs. 2B, 2F, arrowheads), whereas at P90 it was impossible to find such adherens junctions (Fig. 2D), indicating the possibility that the relation between Müller cells and photoreceptor cells was disrupted in this zone. 

The STEM121 immunolabeling electron signal was evident in human cells adjacent to the photoreceptor layer (Figs. 3B, 4A, 4B). The HuCNS-SC cells were easily identifiable by the black peroxidase precipitate surrounding their nuclei (Figs. 3B, 4A, 4B). Away from the regions in which human cells were identified, a thick layer of debris had formed on top of the remaining photoreceptors (Figs. 2C, 3A, debris zone [DZ]), and some of these neurons demonstrated apoptotic nuclei (Fig. 3D). A high magnification view of the DZ shows an accumulation of compacted and disorganized membranes consisting largely of photoreceptor IS and OS membranes (Fig. 3C), resulting from the phagocytic defect that characterizes the RPE cells in this model. In contrast, no DZ could be identified in the treated regions, suggesting that the transplanted HuCNS-SC may have phagocytic function. 

Figure 3

 

The typical DZ present in the untreated RCS rat retina is absent in regions adjacent to HuCNS-SC grafts. The DZ adjacent to the photoreceptor cell bodies in areas distant from the HuCNS-SC-transplant (A, C). High magnification of DZ (C) showed that it is composed of compacted and disorganized photoreceptor membranes. Below this DZ, the remaining photoreceptor nuclei present apoptotic appearance ([D], arrows). In contrast, in the transplanted region where HuCNS-SC ([B], dark peroxidase stain, arrows) are positioned on top of photoreceptor OS (B), a DZ is absent. HC, human cell. Scale bars: (A) 10 μm, (B, D) 5 μm, (C) 2 μm.

Figure 4

 

The HuCNS-SC phagocytose photoreceptor outer segments. Immunostaining with the STEM121 antibody identifies human cells ([A], dark peroxidase stain) in the subretinal space, that is, above the photoreceptors. In the transplant region, photoreceptors exhibit long OS ([A], arrowheads) and groups of outer segment discs are seen being phagocytosed by HuCNS-SC ([A], arrows). High magnification of the boxed area in (A) shows the engulfed membranous discs in the cytoplasm of human cells ([B], arrows). Numerous phagosomes with lamellar structure and in different stages of digestion in lysosomes were found inside the human cells (C, D). Scale bars: (A) 5 μm, (B, C) 2 μm, (D) 1 μm.

Phagocytosis of Outer Segments.

The stacks of membranous discs comprising the OSs of the photoreceptors are renewed constantly: they are pinched off at the tips and engulfed by the apical processes of the RPE in a diurnal cycle, while new discs are added continuously at the base of the OS. The discarded, spent discs accumulate in the RPE cells as phagosomes, which subsequently are broken down in the lysosomes. ²⁹ The Mertk mutation in RCS rats prevents RPE cells from performing this phagocytic activity at a rate essential for the photoreceptor's viability. To examine whether HuCNS-SC were able to restore this process in the RCS rat, we carefully examined the ultrastructure of individual human cells. 

At P60 and P90, our analysis revealed the presence of phagosomes and vesicles at different stages of digestion within the cytosol of the human cells, indicating that these cells have phagocytic capacity (Figs. 4A, 4B). Figure 4A shows three nuclei (N) of human cells surrounded by phagosomes. On occasion, groups of discs could be recognized readily as they were engulfed and internalized by endocytosis (Fig. 4A, arrows). An enlarged view of this area is presented in Figure 4B, highlighting a large vesicle filled with photoreceptor OS discs as it is being engulfed into the cytoplasm of the human cell. Phagosomes with lamellar structure and others in different stages of digestion in lysosomes can be observed inside the human cells (Figs. 4B–D). Well preserved IS and OS are recognized above the photoreceptor cell bodies (Fig. 4A, arrowheads). These data unequivocally demonstrated that HuCNS-SC exhibit in vivo phagocytic activity, thereby restoring an important function normally provided by the RPE, but absent in the RCS rat. 

Photoreceptor Morphology

We then analyzed the ultrastructural changes in photoreceptor OS and IS, and axon terminals. Both OS and IS were identified readily in retinal regions contiguous to HuCNS-SC (Figs. 5A–C). Some OS were slightly shorter than those exhibited in normal nondystrophic rats and others were not perfectly aligned; however, they all seemed to exhibit a normal ultrastructure and parallel-oriented discs (Figs. 4A, 5A–D). We could not identify OS in untreated regions of the retina (not shown). Normal IS of normal rods and cones are filled with long thin mitochondria, and this also was detected in treated regions (Fig. 5D, arrowheads). The OS is linked to the cell body by the IS through a thin bridge, called the connecting cilium. In the treated regions, a cilium joining the IS and OS was observed commonly (Figs. 5D, 5E, arrows). A high magnification view shows a rod photoreceptor with the connecting cilium (Fig. 5E, arrow) between the IS and a short OS that has been phagocytosed by a human cell (Fig. 5E, black precipitate, arrowheads). In addition, photoreceptors in the treated regions exhibited a thick striated filament or rootlet (Fig. 5F, arrowheads) that originated from the basal bodies (Fig. 5F, arrows) below the connecting cilium. In contrast, we were unable to identify any of these photoreceptor structures in untreated regions at both time points examined. Taken together, these data indicated that HuCNS-SC preserve photoreceptor morphology and ultrastructure. 

Figure 5

 

Photoreceptor morphology in treated regions. Photoreceptor OS were well preserved in treated regions (A–D) where 7 to 8 rows of photoreceptors are evident. Some OS were slightly shorter and others were not perfectly aligned compared to normal OS. Membranous discs in the photoreceptor OS were identified easily (B, C). Photoreceptors in this area exhibited a healthy IS containing a large number of mitochondrion ([D], arrowheads). The connecting cilium between IS and OS was seen commonly in treated regions ([D, E], arrows), and had a normal ultrastructure with rootlet ([F], white arrowheads) and basal bodies ([F], white arrows). Active phagocytosis of an OS by a human cell ([E], arrowheads, black precipitate) also is observed. Scale bars: (A) 10 μm, (B, F) 1 μm, (D, E) 2 μm.

Synaptic Contacts in the OPL.

The observation that the OPL was wider in treated regions prompted a detailed ultrastructural examination of this synapse-rich area. In the OPL, cone pedicles and rod spherules make synaptic connections with various bipolar cell and horizontal cell types. ^{30 –32} Cone pedicles are large, conical, flat end-feet of the cone axon terminal that lie at the outer edge of the OPL. The more numerous rod spherules are small, round enlargements of the axon that contain one large mitochondrion and are located closer to the ONL. In treated regions (Fig. 6), the cone terminals were identified easily by their wide spacing and large size (Figs. 6A, 6C, pink) surrounded by the tightly packed, smaller spherule of rod terminals (Figs. 6A, 6B, yellow). The typical large mitochondrion in rod and the several mitochondria in cone terminals could be identified (Figs. 6A, 6B). At their synapses to bipolar and horizontal cells, cone pedicles and rod spherules exhibited an electron-dense bar structure known as a synaptic ribbon (Figs. 6A–C, arrows). The typical morphology of the ribbon synapse includes a central element that is a dendritic terminal of an invaginating bipolar cell (Figs. 6A–C, green), flanked by two lateral elements, which are the dendritic processes of horizontal cells (Figs. 6A–C, blue). These structures are known as “triads.” Cone and rod terminals are filled with vesicles that carry neurotransmitters to the synaptic ribbon. In untreated regions (Fig. 7), pedicles and spherules presented some degree of disruption, and were difficult to identify. The few pedicles and spherules found had synaptic vesicles and retained their synaptic ribbons; however, these appeared to be free-floating within the terminal, that is, not associated with bipolar and horizontal cell terminals (Figs. 7A–C). Together, these findings showed a loss of cone and rod terminal integrity, and triads in the untreated regions of the retina, accompanied by a progressive degradation of the OPL. By contrast, in treated regions synaptic contacts between photoreceptors and second order neurons had the ultrastructural features typical of a normal retina. 

Figure 6

 

Synaptic contacts in the OPL are well preserved in the treated regions In the HuCNS-SC–treated regions, the synaptic connections between photoreceptors, and bipolar and horizontal cells were numerous, and their typical triad conformation was easily recognized. Synaptic ribbons (arrows) were identified in rod spherules ([A, B], yellow) and cone pedicles ([A, C], pink), and a high concentration of synaptic vesicles was observed consistently. The triads presented a normal organization with a central element, which is the tip of a bipolar cell (green), flanked by two horizontal lateral elements (blue). These complete structures were found in photoreceptor axon terminals in the treated regions. Scale bars: (A) 2 μm, (B, C) 0.5 μm.

Figure 7

 

Normal synaptic contacts in the outer plexiform layer were absent in untreated regions. Scarce spherules ([A, B], yellow) and pedicles ([C], pink) were detected in untreated regions, and appeared in the process of degeneration. When observed, synaptic ribbons appeared to be free-floating and had no associated triad. Scale bars: (A) 2 μm, (B, C) 1 μm.

Discussion

In our study, we established that HuCNS-SC, transplanted in the subretinal space of young RCS rats, efficiently phagocytose photoreceptor OS, thereby providing a functional alternative to the dysfunctional host RPE. These data, obtained by immuno-EM examination of transplanted retinas at P60 and P90, unequivocally confirmed preliminary work describing the ability of HuCNS-SC to phagocytose rod OS in vitro. ²¹ Furthermore, our study details the robust photoreceptor survival in areas of the retina adjacent to HuCNS-SC, and highlights, for the first time to our knowledge, robust preservation of ultrastructure of photoreceptors and synaptic contacts at the OPL level, which may underlie preservation of retinal function in these areas. Finally, our results indicated that the mechanism of photoreceptor preservation involves phagocytosis of shed OS by HuCNS-SC. 

To our knowledge, this is the first time that directed OS phagocytosis by neural stem cells, as reflected by the presence of a large number of phagosomes filled with internalized OS discs inside human cells, has been demonstrated in vivo. The phagocytic function of PNS and CNS cells, such as astroglial, Schwann cells, and neuronal progenitors, has been described previously, but only in the context of synapse elimination/pruning and clearance of apoptotic cells and neural debris. ^{33 –37} The precise lineage identity of the phagocytic cell(s) was not determined in this study, as it is very technically challenging to perform double immuno-EM. Importantly, our previous study showed that the majority of human cells in the subretinal space remained fairly immature. While some human cells expressed doublecortin (DCX), an early neuronal marker, we were unable to detect cells exhibiting the identifying characteristics of mature human astrocytes. ²¹ Therefore, it is likely that, in this study, phagocytic function is carried out by neural stem cells and DCX⁺ neuronal progenitors, and not by astrocytes, since these cells are not present in the graft. Although in vivo phagocytic activity of neural stem cells has not been described before to our knowledge, genes associated with the phagocytic machinery, such as Mertk and Axl, ³⁸ have been found in a HuCNS-SC transcriptome study (StemCells, Inc., unpublished data). We also have reported previously that HuCNS-SC express the αvβ5 integrin, which mediates binding/recognition of OS by RPE. ³⁹ Taken together, these data suggested that the phagocytic process used by RPE cells during the diurnal digestion of shed photoreceptor OS can be restored by neural stem cells and neuronal progenitors in the absence of functionally competent host RPE. Significantly, lentiviral-mediated Mertk gene transfer in young RCS rats leads to long-term preservation of photoreceptors and visual function, ⁴⁰ supporting the notion that restoration of phagocytosis alone may, in fact, strongly underlie the efficacy of HuCNS-SC in the same rat model. Whether HuCNS-SC are capable of carrying out other crucial RPE functions, such as the regeneration of 11-cis-retinal and its transport back to photoreceptors, currently is being investigated. 

In our study, we focused our efforts on the detailed EM characterization of the impact of a HuCNS-SC graft on photoreceptor rescue, rather than examining sham retinas. We ²¹ and others ^10,11,27,28 have shown consistently that sham injections do not lead to a reduction of the DZ and, although a small and localized impact on ONL thickness is observed, this effect is transient, and sham-treated retinas are mostly indistinguishable from untreated ones. Therefore, the transient nature of the sham effect and lack of overall efficacy in sham-injected eyes, contrasting with the long-term efficacy of HuCNS-SC only in areas of the retina adjacent to the graft, very strongly supported that the observed effects on photoreceptor preservation are cell-mediated. 

Previous work ^5,18 described in detail the inner retinal changes that directly result from the loss of photoreceptors in untreated RCS rats as well as after RPE transplantation. Confocal immunohistochemistry of untreated rats revealed a marked reduction of the synaptic contacts in the OPL, and dendrite sprouting by the second order bipolar and horizontal neurons. Distribution of presynaptic and postsynaptic markers in OPL was impaired as early as P21 and progressively lost by P60. In rats transplanted subretinally with an RPE cell line and analyzed at P70, the synaptic contacts between cones/rods and second order neurons appeared more preserved than in untreated rats, and efficacy was proportional to the degree of photoreceptor rescue. The supportive ultrastructure data obtained in our current study takes one crucial step forward from previously reported work, as it demonstrates unequivocally, for the first time to our knowledge, that the anatomic substrates of synaptic connectivity within the OPL, the triad ribbon synapses, are well preserved in areas of the retina adjacent to HuCNS-SC. The various layers of anatomic preservation observed in our study most likely underlie the visual function preservation observed in RCS rats transplanted with HuCNS-SC. 

Interestingly, outside the area of HuCNS-SC engraftment, the retina ultrastructure was very similar to that found in untreated eyes, supporting the view put forth by McGill et al. that efficacy is dependent on the immediate proximity of HuCNS-SC, ²¹ further strengthening the notion that phagocytosis may be an important mechanism of action for these cells. Nonetheless, HuCNS-SC also secrete neurotrophins and growth factors ²¹ that have been shown to exert protective effects on photoreceptors. Therefore, it is likely that the mechanism of photoreceptor preservation could reflect, in part, the release of such factors. The relatively steep reduction in efficacy immediately distal to the graft suggests that the contribution of neuroprotective factor(s) secreted by HuCNS-SC to the overall efficacy may be restricted to a relatively short diffusion range. However, the extensive migratory ability of HuCNS-SC cells within the subretinal space amplifies the area of efficacy from a single injection, a property highly relevant for clinical translation. 

Our study compared the efficacy of HuCNS-SC derived from two distinct DCLs; one used in preclinical studies ²¹ and the other developed for the phase I/II GA AMD trial. No differences in efficacy between the two DCLs were observed, supporting the concept of comparability across DCLs. 

Finally, the preclinical findings in the RCS rat presented here and in previous work ²¹ indicated that the neuroprotective transplantation of HuCNS-SC cells results in the stabilization of photoreceptor degeneration and slowing of progressive visual loss. This is relevant particularly in the setting of the slow rate of changes observed in the natural history of dry AMD compared to other diseases. ⁴¹ These effects potentially could translate into rescue of visual acuity and improvement of visual function in patients, such as peripheral, night, and color vision, as well as enhanced reading speed, facial recognition, mobility, and overall quality of life. A successful therapy for GA AMD either will stabilize or slow the progression of visual loss, possibly also limiting progression to wet AMD, ultimately reducing the physical, emotional, and social burden related to this major blinding disorder. 

Supplementary Materials

pdf S1 

Acknowledgments

Supported by grants from the Spanish Ministry of Economy and Competitiveness (BFU2012-36845), Instituto de Salud Carlos III (RETICS RD12/0034/0010), Organización Nacional de Ciegos Españoles (ONCE), and StemCells, Inc. (NC). 

Disclosure: N. Cuenca, StemCells, Inc. (F); L. Fernández-Sánchez, None; T.J. McGill, StemCells, Inc. (C, R); B. Lu, StemCells, Inc. (C, R); S. Wang, None; R. Lund, None; S. Huhn, StemCells, Inc. (E); A. Capela, StemCells, Inc. (E)