Gene expression changes in the retina following subretinal injection of human neural progenitor cells into a rodent model for retinal degeneration

Derivation, maintenance, and transplantation of hNPCs

Human neural progenitor cells (hNPCs) isolated from fetal cortical brain tissue were obtained with institutional review board approval and in accordance with the National Institutes of Health guidelines for the collection of such tissues. hNPCs were cultured as neurospheres in Stemline Neural Stem Cell Expansion Medium (Sigma-Aldrich, St. Louis, MO) supplemented with 20 ng/ml epidermal growth factor (Sigma-Aldrich) and 10 ng/ml leukemia inhibitor factor (Millipore, Billerica, MA) as previously described [12]. Neurospheres of hNPCs at passage 23–25 were dissociated into a single cell suspension by incubation at 37 °C for 10 min with Accutase (Sigma-Aldrich) followed by trypsin inhibitor (Sigma-Aldrich) for 5 min and DNase (Sigma-Aldrich) for 10 min with gentle trituration in PBS (1X; 120 mM NaCl, 20 mM KCl, 10 mM NaPO₄, 5 mM KPO₄, pH 7.4; Life Technologies, Paisley, UK). An injection of 4×10⁴ cells/eye in 2 µl of balanced salt solution (BSS) cell carrying medium (Alcon, Fort Worth, TX) was delivered to the subretinal space through a glass micropipette, as previously described [12]. Rats receiving a sham surgery were injected with BSS cell carrying medium alone.

Animals

Retinal degenerate RCS rats (n = 8) received a subretinal injection of hNPCs into one eye (RCS^hNPCs) and either sham surgery (RCS^sham; n = 4) or no treatment (n = 4) in the fellow eye at postnatal day 21 (P21). Long Evans (LE) rats (n = 3) received subretinal sham surgery (LE^sham) of cell carrying media into one eye, and the fellow eye received no treatment at P21. All rats received dexamethasone intraperitoneal injections for 2 weeks (2.5 mg/kg/day) following subretinal injection and cyclosporine A (Novartis, Basel, Switzerland) in the drinking water (210 mg/l) until the rats were euthanized at P60 following the functional studies. This study adhered to the Association for Research in Vision and Ophthalmology Statement for the Use of Animals in Ophthalmic and Vision research and was conducted with the approval of the Institutional Animal Care and Use Committee at Cedars-Sinai Medical Center.

Visual function testing

The optokinetic response (OKR) testing apparatus comprises a rotating cylinder displaying a vertical sine wave grating presented in virtual three-dimensional space on four computer monitors arranged in a square. Unrestrained rats were placed on a platform in the center of the square and tracked the grating with reflexive head movements. Visual acuity was quantified by increasing the spatial frequency of the grating until an OKR could no longer be elicited. Statistical analyses were performed with one-way ANOVA (ANOVA) and Newman-Keuls multiple comparison test using GraphPad Prism 5.01 (GraphPad Software, La Jolla, CA). Data were expressed as mean ± standard error of the mean (SEM), and a p value of less than 0.05 was considered statistically significant.

Tissue extraction and preparation

Eyes were enucleated from rats at age P60–64. For histology staining, eyes were fixed with 4% paraformaldehyde (Sigma-Aldrich) for 1 h followed by increasing sucrose before embedding in optimum cutting temperature (OCT) compound (Sakura Finetek, Torrance, CA) and storage at −80 °C. Horizontal sections were cut on a Leica CM1850 cryostat microtome (Leica Biosystems, Nussloch, Germany) at 10 μm per section. Four sections (50 μm apart)/slide were collected in five series, and every fifth slide was used for cresyl violet staining and imaged with a Leica DM6000B transmitted light microscope (Leica Microsystems, Wetzlar, Germany). For the RNA extraction experiments, the rats were euthanized with CO₂ followed by bilateral pneumothorax, and the eyes were briefly cleansed with ethanol before they were removed. Neural retinas were removed from the eye cup and oriented so that only the hemisphere containing the injected cells was isolated, flash frozen with liquid nitrogen, and stored at −80 °C until RNA extraction.

RNA isolation

Total RNA was extracted from individual halved neural retinas using the Purelink RNA Mini Kit (Life Technologies) according to the manufacturer’s instructions. RNA quality was assessed using a NanoDrop ND-1000 spectrophotometer (NanoDrop Technologies, Wilmington, DE) and the Agilent 2100 system (Agilent Technologies, Santa Clara, CA).

RNA-sequencing (RNA-seq)

Total RNA (1 μg) was poly-A selected using Dynabeads® Oligo(dT)25 (Life Technologies). cDNA libraries were constructed with barcoded primers using Ion Total RNA-Seq Kit v2 (Life Technologies), and were multiplexed and amplified onto ion sphere particles (ISPs) using Ion PI™ Template OT2 200 Kit v3 (Life Technologies). Libraries were sequenced to an average depth of 10 million reads/sample using the Ion PI™ Sequencing 200 Kit v3 (Life Technologies) by the Genomics Core at Cedars-Sinai Medical Center.

Bioinformatic analyses

Raw reads were filtered and trimmed with the FASTX toolkit and then were aligned to the rat reference genome rn5. Fragment per kilobase of gene per million reads sequenced (FPKM) values were calculated for 26,407 genes with Cufflinks 2.0.8 software [29]. Genes with sample FPKMs equaling 0 were excluded from further analyses. Sequence data were deposited in GEO per MIAME standards (accession number: GSE70600) [30,31]. All FPKM values were increased with an addition of 1, log₂ transformed, and hierarchical cluster analysis of gene expression was performed with Cluster and TreeView software [32]. Differentially expressed genes were determined using a two-tailed t test, and were then corrected with calculation of the q value with the Benjamini-Hochberg method. Genes with significant expression differences with a false-discovery rate below 5% (q<0.05) were used for further analyses. Differential gene profiles (RCS^sham versus LE^sham and RCS^hNPCs versus RCS^sham) were compared for matching genes. Rescue genes were defined as those common between upregulated RCS^sham versus LE^sham and downregulated RCS^hNPCs versus RCS^sham or downregulated RCS^sham versus LE^sham and upregulated RCS^hNPCs versus RCS^sham. To determine the degree to which the genes were rescued, the fold changes between the two sets were then added to yield a fold change difference and were sorted based on values closest to 0. Functional annotation clustering analysis of the differentially expressed gene lists (RCS^sham versus LE^sham up- and downregulated and RCS^hNPCs versus RCS^sham up- and downregulated) were performed by submission to the Database for Annotation, Visualization and Integrated Discovery (DAVID v6.7) [33,34], and Gene Ontology (GO) term significance was accepted at Benjamini-Hochberg <0.001. Pie charts were generated using Microsoft Excel using the activation scores. Biologic processes and cellular components were determined by submission of the aforementioned differentially expressed gene lists to the Gene Ontology enRIchment anaLysis and visuaLizAtion tool (GOrilla) [35,36] for GO term analysis, and subsequently submitted to REViGO online software for visualization [37]. For canonical pathway analysis, differentially expressed genes from each gene list were submitted to the Ingenuity Pathway Analysis (IPA) Spring 2015 Release software (QIAGEN, Redwood City, CA), and significance was accepted with a –log Fisher’s exact test p≥1.3 and a z score of ≥2 or ≤–2.

qRT-PCR

The top five characterized rescue genes from the gene list in the square in Figure 3 and the top six rescue genes from the genes in the circle in Figure 3 were further analyzed with quantitative real-time polymerase chain reaction (qRT-PCR) for expression validation. Total RNA was extracted from isolated halved neural retina samples of three biologic replicates each of LE^sham, RCS^sham, and RCS^hNPCs as described above. RCS^sham and RCS^hNPCs retinas were taken from the same animals, corresponding to the left and right eyes, respectively. cDNA synthesis was performed using the High-Capacity cDNA Reverse Transcription Kit (Life Technologies) according to the manufacturer’s instructions. The qRT-PCR analyses were performed using 50 ng cDNA on 96-well plates (Applied Biosystems, Paisley, UK), and run in technical duplicates on a 7500 Real-Time PCR System (Applied Biosystems). The ΔΔCt method was used to calculate fold changes, using ActB as a housekeeping standard and the LE^sham sample as the calibrator. Primers were designed with Primer3 online software [38,39] and selected to amplify fragments with 150–250 bp and with a Tm of approximately 60 °C (Table 1). Primers were designed for the following gene targets: Actb (NM_031144), Amigo2 (NM_182816), Cdc42ep5 (NM_001108469), Cdh22 (NM_019161), Htr1f (NM_021857), Pax4 (NM_031799), Pdc (NM_012872), Rp1 (NM_001195676), Sebox (NM_023951), Ubald1 (NM_001007668), and Ypel1 (XM_002727914). To ensure accurate qRT-PCR expression patterns, primers were also designed and used for rhodopsin (NM_033441), which is expressed in photoreceptors.

Preservation of photoreceptors and visual function following hNPC treatment

At P21, wild-type LE rats received sham surgery (LE^sham), while retinal degenerate RCS rats received sham surgery (RCS^sham) in one eye and a subretinal injection of hNPCs (RCS^hNPCs) into the fellow eye. To ensure functional benefit, the OKR test measured the relative visual acuity of the LE^sham, RCS^sham, and RCS^hNPCs eyes at age P60. The OKR measurements for the LE^sham eyes were 0.6350 ± 0.0020 c/d, compared to a significant decrease to 0.3245 ± 0.0047 c/d in the RCS^sham eyes (p<0.001) and a subsequent significant increase to 0.4503 ± 0.0064 c/d in RCS^hNPCs eyes (p<0.001; Figure 1A). Following OKR analysis, eyes were enucleated and fixed for histological analysis. To confirm that hNPCs aided in photoreceptor survival, retinal cross sections were stained with cresyl violet dye (Figure 1B–D). The LE^sham eyes had approximately ten layers of photoreceptor cells located in the outer nuclear layer (ONL; Figure 1B), whereas the number of photoreceptor cells decreased to only three layers in the RCS^sham eyes (Figure 1C). In the RCS^hNPCs eyes, transplanted hNPCs survived in the subretinal space and were able to preserve approximately five to six cell layers of photoreceptors (Figure 1D). Similar to previous studies [10-12], the hNPCs were able to preserve visual function and aid in photoreceptor survival.

Analysis of global gene expression

To examine the overall gene expression in the host retinal tissue, retinal RNA was isolated (n = 2 for each of LE^sham, RCS^sham, and RCS^hNPCs), and RNA-seq was performed. Expression levels of 26,407 RefSeq protein-coding genes using the FPKM values was determined. A complete list of genes is registered at GEO (accession GSE70600). Only genes with expression of FPKM >0 were included for analysis. The total number of expressed genes for each sample was similar, with an average number of 18,254 (69%) in LE^sham, 18,662 (71%) in RCS^sham, and 18,627 (71%) in RCS^hNPCs (Table 2). To compare the similarity of the global gene expression profiles of the different samples, average linkage hierarchical cluster analysis was performed (Figure 2). The LE^sham samples clearly separated from the RCS samples, suggesting that there is a distinct difference in gene expression between wild-type and degenerating retinas. These data are in agreement with other transcriptomic studies of animal models for retinal degeneration [24,25]. Additionally, the RCS^hNPCs samples segregated from the RCS^sham samples indicating that distinguishable gene expression changes follow injection of hNPCs.

Differential gene expression in retinal degenerative RCS rats

To investigate gene expression changes with retinal degeneration, computational analysis of the differential gene expression between RCS^sham and LE^sham was performed. Genes were considered to be differentially expressed with an FDR <5% (q <0.05). A total of 1,215 differentially expressed genes were identified in the RCS^sham versus LE^sham comparison (Table 1; Appendix 1). Of these genes, 852 (70%) genes had increased expression in the RCS^sham samples (Figure 3, red ellipse), and 82 genes (10%) were uncharacterized. The top five genes with the greatest fold changes included Mir671, Lcn2, Cd74, Gfap, and Cebpd. Lcn2, Cd74, Gfap, and Cebpd have all been shown to be increased with retinal degeneration [15,19,25,40-45], confirming retinal degeneration is detectable at the molecular level. Additionally, Mir671, Lcn2, Cd74, and Cebpd play roles in the immune response to macrophages and/or microglia [45-48], suggesting that there is an increase in macrophage/microglia activity with retinal degeneration.

Of the 1,215 differentially expressed genes in the RCS^sham versus LE^sham comparison, 363 (30%) genes had decreased expression in the RCS^sham samples (Figure 3 green ellipse), and 29 (8%) were uncharacterized. The top five genes with the greatest fold change were Optc, Gnat1, Hk2, Lig4, and Nrl. Gnat1, Hk2, and Nrl are expressed in photoreceptors [46-53], and Rho (rhodopsin) was also greatly decreased with a fold change of −6.4, indicating that there is a significant decrease in photoreceptor-specific genes in the RCS^sham samples. This corroborates with the loss of photoreceptors in the retinal histology (Figure 1C). Optc, Gnat1, Hk2, Lig4, and Nrl have either been implicated in human retinal degenerative diseases [54-58] or are used in animal models for vision disorders [24,50,52], further establishing that retinal degeneration is discernible on the gene expression level.

Differential gene expression due to hNPC-induced retinal preservation in RCS rats

Computational analysis of the differential gene expression between RCS^hNPCs and RCS^sham was performed. A total of 283 differentially expressed genes were identified in the RCS^hNPCs versus RCS^sham comparison (Table 1; Appendix 2). A total of 51 (18%) genes had increased expression (Figure 3, yellow ellipse), and 21 (41%) of these genes are uncharacterized, suggesting that many of the genes expressed following hNPC treatment are currently unidentifiable. The top genes with greatest fold changes were Mir671, ENSRNOG00000049107, Pde4d, Anxa9, and Mir770. As previously described, Mir671 is expressed in response to macrophage/microglia activity but also plays a role in regulating extracellular matrix production [59]. Downregulation of Pde4d and Anxa9 expression is observed in retinal degeneration [24,60,61], and the increase in expression in this data set indicates that there is less retinal degeneration in RCS^hNPCs. Mir770 is ubiquitously expressed in the mouse eye at P60 [62], and this may be further elucidated by looking at the target genes of this microRNA.

Of the 283 differentially expressed genes between RCS^hNPCs and RCS^sham, 232 (82%) genes had decreased expression in the RCS^hNPCs samples (Figure 3, blue ellipse). Similar to the upregulated RCS^hNPCs versus RCS^sham gene set, a high percentage of genes (79; 34%) are uncharacterized. The top five most downregulated genes include Cryaa, Tomm6, ENSRNOG00000050736, Crybb3, and Cryba1. The expression of crystallin genes, such as Cryaa, Crybb3, and Cryba1, is altered following retinal trauma and may play a protective role [63-68]. Decreased crystallin expression in RCS^hNPCs may be due to increased photoreceptor survival and less need for endogenous retinal neuroprotection. Tomm6 encodes a protein that is part of the TOM complex of the mitochondrial membrane [69], and altered expression of TOM complex proteins has been detected in patients with diabetic retinopathies [70]. Collectively, the downregulation of these genes suggests that there is less retinal injury following treatment with hNPCs.

Functional, biologic, and cellular component analyses of differential gene expression sets

To identify functional processes that are affected, gene profiles were submitted to DAVID [33,34]. Of the 852 upregulated differentially expressed genes in the RCS^sham versus LE^sham comparison, ten functional processes were identified (Figure 4A). These genes were heavily associated with immune and inflammatory responses, suggesting that there is an increase in the immune response in the RCS^sham samples. The remaining functional processes participate in motility, suggesting there is an increase in the transportation of proteins across the cell and in cellular movement. Functional analysis of the 363 downregulated genes from the RCS^sham versus LE^sham comparison identified four significantly enriched pathways (Figure 4B). The downregulation of these processes (visual perception, phototransduction, photoreceptor cell development and differentiation, and detection of visible light) further confirm that retinal degeneration is detectable on a molecular basis in the RCS^sham samples. To determine the affected functional processes with treatment of hNPCs, the differentially expressed gene sets from the RCS^hNPCs versus RCS^sham comparisons were submitted to DAVID. The 51 upregulated genes yielded two processes (protein biosynthesis and cytosol; Figure 4C), and no processes were enriched using the downregulated genes. The lack of processes may be due to the small number of genes submitted to DAVID or the high percentage of uncharacterized genes in the RCS^hNPCs versus RCS^sham comparison.

To determine gene relation to biologic processes, GO terms from the differential gene expression profiles were submitted to REViGO. Similar to the functional analysis by DAVID, the RCS^sham versus LE^sham upregulated biologic processes mainly included immune system regulation (Figure 5A). This analysis further demonstrates that the immune response is heightened in RCS^sham. The downregulated differentially expressed gene list from RCS^sham versus LE^sham procured only one biologic process (Figure 5B). Cation transport is important for the directed movement of molecules between or within cells, and the downregulation of this process may be linked to the decreases of photoreceptor cells and in phototransduction. To identify biologic processes that are affected following treatment with hNPCs, GO terms from the RCS^hNPCs versus RCS^sham differentially expressed gene lists were submitted to REViGO. Three biologic processes were determined from the upregulated gene list (Figure 5C). These processes participate in cell adhesion and regulation of morphogenesis, which could be due to the increase in rescued photoreceptors. The genes from the downregulated gene set could not be classified into any significantly enriched biologic processes.

The differential gene lists were also analyzed for cellular component ontology, which describes where the gene product is located in the cell. Submission of GO terms from the RCS^sham versus LE^sham upregulated gene list identified five cellular components (Figure 6A). These components suggest that the abundance of immune cells is increased in the RCS^sham samples. The downregulated gene list from RCS^sham versus LE^sham yielded one cellular component (cilium; Figure 6B). Photoreceptor outer segments are characterized as specialized sensory cilia [71], and RCS^sham have a defect in phagocytosing outer segments [26]. Additionally, inherited retinal degenerative diseases can be linked to cilia mutations [72], further demonstrating that retinal degeneration can be detected at the molecular level in RCS^sham. No significant cellular components were established in either of the RCS^hNPCs versus RCS^sham comparisons.

Rescue gene expression in RCS^hNPCs

RCS^hNPCs were shown to have increased visual function and photoreceptor cell survival (Figure 1). To analyze genes that may be aiding in these processes, computational analysis of the differential gene sets between the comparisons was conducted, and rescue genes were identified (Figure 3, circle). Fifty-five genes common to the upregulated RCS^sham versus LE^sham and downregulated RCS^hNPCs versus RCS^sham gene lists were identified (Figure 3, square, Appendix 3). To determine genes with RCS^hNPCs expression similar to that of LE^sham, the fold change difference (FCD) was calculated. FCD was defined as ((RCS^sham versus LE^sham fold change) + (RCS^hNPCs versus RCS^sham fold change). The top characterized genes with an FCD closest to 0 are Ubald1 (FCD 0.02) Sebox (FCD −0.02), Cdh22 (FCD −0.02), Amigo2 (FCD −0.03), and Cdc42ep5 (FCD 0.04). To date, there are no publications on the expression of Ubald1 or Sebox in the retina. Amigo2 is expressed in the rat retina [73], and Cdh22 is expressed in the developing mouse brain [74], but little is known about their biologic relevance. Cdc42 is important for tissue organization during retinal development, and loss of Cdc42 results in retinal degeneration [75,76]. Cdc42ep5 encodes an effector protein that binds to Cdc42 to negatively regulate its function [77,78]. The decrease in Cdc42ep5 expression in RCS^hNPCs may allow for more Cdc42 expression and subsequent retinal preservation.

To further evaluate rescue genes, the downregulated RCS^sham versus LE^sham and upregulated RCS^hNPCs versus RCS^sham were compared. A total of 13 genes were identified (Figure 3, circle), and the top rescue genes with an FCD closest to 0 are Htr1f (FCD −0.14), Ypel1 (FCD −0.17), Pdc (FCD −0.36), Glb1l2 (FCD 0.47), and Pax4 (FCD 0.50; Appendix 4). Htr1f was found to have higher expression in the temporal retina than in the macular retina of human patients [79]. Ypel1 may play a role in regulation of cell morphology [80] and/or in cell division [81], but no expression analysis in the retina has been performed. Pdc (phosducin) is highly expressed in photoreceptors [82,83], suggesting that there is a significant increase in photoreceptor gene expression in RCS^hNPCs. Glb1l2 is ubiquitously expressed in the eye, including the retina, and may play a role in retinal cell homeostasis [84]. Pax4 is expressed in photoreceptors [85] and can stimulate expression of the rod-derived cone viability factor for photoreceptor survival [86,87]. Collectively, the upregulation of these rescue genes indicate that there is an increase in photoreceptor gene expression following treatment with hNPCs.

Validation of differential gene expression

qRT-PCR analysis was performed on the genes with an FCD closest to 0 from each of the lists of rescued genes. Of the rescued genes that were upregulated in the RCS^sham versus LE^sham and downregulated in the RCS^hNPCs versus RCS^sham sets (Appendix 3), one of the five genes (Cdc42ep5) followed similar gene expression patterns as seen in the RNA-seq expression (Figure 7A). As described previously, the downregulation of Cdc42ep5 expression in RCS^hNPCs may be due to photoreceptor preservation. Of the rescued genes that were downregulated in the RCS^sham versus LE^sham and upregulated in the RCS^hNPCs versus RCS^sham sets (Appendix 4), all six genes (Htr1f, Ype1l, Pdc, Glb1l2, Pax4, and Rp1) followed similar expression patterns as those detected in the RNA-seq analysis (Figure 7B). To confirm the histological analysis, the photoreceptor gene rhodopsin (Rho) was also included in qRT-PCR analysis.

Pathway analysis of rescued genes following hNPC transplantation

To determine which pathways may be rescued following transplantation of hNPCs into the RCS rat, gene lists were uploaded into the Ingenuity Pathway Analysis (IPA) software. All of the genes that were significantly upregulated in the RCS^sham versus LE^sham and downregulated in the RCS^hNPCs versus RCS^sham gene profiles were compared, and three pathways were found to coincide (Table 3). Integrin signaling is involved in promotion of inflammation [88] and the uptake of apoptotic cells by macrophages and microglia [89]. The second affected pathway, phospholipase C signaling, is important for efficient phagocytosis [90], which is one function of macrophages and microglia. The third pathway, Rho Family GTPase signaling, promotes phagocytic engulfment [91]. The importance of the three pathways coincides with the functional, biologic, and cellular components analyses that indicated that there is an increased immune response in RCS^sham (Figure 4, Figure 5). The three affected pathways all play roles in the phagocytic response, suggesting that the host RCS^sham retina is infiltrated with macrophages and microglia. Fewer macrophages and microglia may be due to less degenerative materials or less stress on photoreceptors targeted for phagocytosis, since photoreceptors are preserved with hNPC treatment. No pathways were identified in the downregulated RCS^sham versus LE^sham and upregulated RCS^hNPCs versus RCS^sham comparison.

Decrease in abundance of macrophages and microglia following hNPC treatment

Comparison of the pathways affected by the gene sets suggest phagocytosis signaling as the common biologic process by which all three pathways participate. To determine whether the presence of macrophages and microglia contributes to the overall gene expression changes, immunofluorescent staining was performed (Figure 8). A greater amount of positive staining for Iba1, a marker of macrophages and microglia, was detected in the RCS^sham retina (Figure 8B) compared to the LE^sham retina (Figure 8A). Treatment with hNPCs decreased the amount of positive staining in areas with rescued photoreceptors (Figure 8C), similar to the expression patterns detected in the LE^sham retina. Similar to RCS^sham expression, areas away from the grafted region of the treated retina had an increase in macrophages and microglia (Figure 8D), suggesting that the decrease in the abundance of macrophages and microglia is directly due to the presence of hNPCs.