Defining the mouse RPE transcriptome

We and others previously described numerous gene expression changes in the retina following detachment from the RPE [30,31,32]. However, the overall transcriptional response of the RPE after retinal detachment has not been examined. We hypothesized that retinal detachment would lead to considerable gene expression changes in RPE that has been separated from the neural retina and is no longer in close association with PR outer segments. Previous studies showed that the simultaneous RPE isolation and RNA stabilization (SRIRS) method of Xin-Zhao and co-workers [19] provided highly enriched RPE RNA from the mouse eyecup. This method has the advantage of obtaining highly enriched RPE RNA and minimizing the chances for transcriptomic changes during isolation. Preliminary studies indicated that in our hands the SRIRS method provided RNA preparations greatly enriched in mouse RPE RNA (Supplemental Figure S1). This was demonstrated by qRT-PCR analysis showing enrichment of the RPE marker gene Rpe65 (47-fold RPE > retina, p < 0.0001), as well as de-enrichment of rod PR-specific mRNA Rho (0.1317-fold, RPE < retina, p < 0.0001) the endothelial cell-specific mRNAs cadherin 5 (Cdh5) mRNA (0.034-fold, RPE < retina, p = 0.015) and platelet and endothelial cell adhesion molecule 1 (Pecam1, 0.045-fold, RPE < retina, p < 0.0001) mRNA (Supplemental Figures S1B-F). Rhd5 and Pecam1 are expressed in endothelial cells in both retinal vasculature and the choriocapillaris [33, 34]. Because expression of Collagen Type I Alpha 1 Chain is particularly high in choroidal stromal cells [33], the levels of Col1a1 mRNA were also compared in the RPE SRIRS samples and RNA isolated from choroid/sclera tissue. This showed 187-fold higher level of Col1a1 mRNA in choroid/sclera than in the RPE SRIRS preparation (RPE < choroid/sclera, p < 0.0001) (Supplemental Figure S1G), further ruling out any appreciable contamination from choroid RNA sources. We therefore used this method to isolate RPE RNA from age and sex (male) matched naïve C57BL/6J mouse eyecups and from eyes with partially detached retinas at 1 dprd and 7 dprd, all harvested at the same time of day (from 9:00 am to 11:00 am). Naïve eyes were chosen as the control group because the effects of RD on gene expression in the contralateral (fellow) eye have not been defined. Instead, we compared the RD groups to matched naïve (Nv) RPE samples, which allowed us to also define a normal mouse RPE transcriptome. Corresponding retinas were obtained from these same eyes and subjected to RNA-Seq analysis (reported elsewhere).

To define mouse RPE transcriptome based upon highly expressed mRNAs, mean transcripts per million (TPM) of mapped sequence reads for the Nv group were used to derive a list of 2671 genes with mean TPM ≥ 45, which represented the top 1/8th of 21,558 total detected genes in the Nv samples (Supplemental Table S1A). This emulates the strategy employed by Bergen’s group, who defined RPE signature genes as the top 10% of highly expressed transcripts [35]. The slightly less restrictive top 12.5% cutoff value was chosen because several previously published RPE signature genes were excluded when a top 10% cutoff was applied.

Although the SRIRS method to isolate RPE RNA resulted in considerable enrichment of RPE RNA, retinal tissue can carry over into the eyecup preparation. An initial KEGG pathway enrichment analysis of the 2671 genes (TPM ≥ 45) in the Nv mouse RPE preparations showed significant enrichment (FDR = 3.7E-10) of 18 ‘Phototransduction’ KEGG pathway genes, including PR-expressed genes. To correct for this, the Nv RPE transcriptome was compared to that of retina samples obtained from the same naïve eyes. DESeq2 analysis identified 4454 retina DEG with significant (padj ≤ 0.05) 2-fold greater transcript abundance in Nv retina than in Nv RPE. A Venn comparison identified 225 of these retina > RPE DEG as present in the RPE transcriptome (Fig. 1A, Supplemental Data Table S1A); no known RPE marker or signature genes were included in this intersection of sets and the majority represented photoreceptor-specific genes, with the most significantly enriched KEGG pathway being ‘Phototransduction’ (14 genes, FDR = 2.1E-18) and the most significantly enriched GO biological process group being ‘Visual Perception’ (28 genes, FDR = 2.7E-20) (Supplemental Tables S1C and S1D). The 225 genes were therefore excluded from the RPE transcriptome, resulting in mouse RPE transcriptome containing 2446 genes highly expressed mouse RPE genes and representing 11.4% of all genes mapped in Nv RPE samples (Supplemental Data Table S1F). This is considered the mouse RPE transcriptome.

Several previous studies have endeavored to defined RPE signature genes. The 2446-gene mouse RPE transcriptome was compared to three previously published RPE signature gene lists which were derived from: (1) human fetal and adult RPE tissue [36], (2) a mouse laser captured RPE transcriptome compared to human RPE cell data sets [35], and (3) a mouse RPE cell cluster transcriptome derived from single cell sequencing [33]. Venn comparison with a list of 627 combined RPE signature genes from these three publications identified 277 common genes included in the present mouse RPE transcriptome at least one of these previous publications (Fig. 1B, Supplemental Data Table S1G). Comparing to the individual RPE gene lists, 64% (223 of 349) of Lehmann mouse RPE signature genes are in common with the present transcriptome. In contrast, only approximately 30% of the Bennis (83 of 272) and Strunnikova (46 or 154) human RPE signature genes are shared with the present mouse RPE transcriptome (Supplemental Data Table S1G). Importantly, only 14 genes (Bmp4, Crim1, Degs1, Gja1, Itgav, Mfap3l, Pdpn, Ptgds, Rbp1, Rnf13, Rpe65, Slc4a2, Sulf1 and Ttr) are common to the present mouse RPE transcriptome and all three RPE signature gene lists (Fig. 1C). This is not surprising, given that just 19 genes are common to all three published RPE signature gene lists (Supplemental Data Table S1G). The 5 common RPE signature genes that are not within the present RPE transcriptome are: Sema3c (TPM = 30.5), Rragd (TPM = 19.0), Cdh3 (TPM = 15.8), Lhx2 (not mapped) and Srfp5 (not mapped). Other notable RPE genes that are in one or two of the published RPE signature gene lists but are missing from the present mouse RPE transcriptome are: Ins2 (TPM = 11.2), Mertk (TPM = 9.4) and Best1 (TPM = 1.3).

Fig. 1

Analysis of the mouse RPE transcriptome. (A) Venn diagram showing comparison of RPE DEGs with TPM greater than 45 (total = 2671) and retina DEG with 2-fold greater transcript abundance in retina than the RPE padj ≤ 0.05 (total = 4454), and an overlap of 225 genes. (B) Venn diagram showing comparison of the 2446 gene mouse RPE transcriptome identified in the present study and the 627 RPE signature genes identified in three previous publications [33, 35, 36], and an intersection of 277 genes. (C) Venn diagram showing comparison of the RPE transcriptome identified in the present study and the lists identified in the works of Bennis, Lehmann and Strunnikova [33, 35, 36], with 14 genes common to the mouse RPE transcriptome and all three RPE signature gene lists. (D) Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis of 2446 gene naïve RPE transcriptome. (E) Gene ontology (GO) analysis to identify Biological Processes in the naïve RPE transcriptome. (F) Relative mRNA expression of Gsto1 mRNA in naïve RPE, retina and choroid using qRT-PCR, normalized to 18 S ribosomal RNA (Rn18s). Bar graphs represent mean ± SEM. Statistical analysis was performed using one-way ANOVA with repeated measures followed by Tukey’s post hoc test. *p < 0.05; ****p < 0.0001. (G) Immunofluorescence analysis of retinal sections from C57BL6/J mice and human showing cell nuclei (blue), GSTO1 (green), the RPE cell marker RPE65 (red). The right quadrants of each composite image represent the merged overlay of the green, red, and blue channels. Images were acquired using the Leica STELLARIS 8 FALCON Confocal Microscope, 40x objective. Scale bar = 50 μm

Analysis of the mouse RPE transcriptome

Enrichment analysis was performed with the 2446-gene mouse RPE transcriptome (Fig. 1D and E, Supplemental Table S1H-K). The enriched pathways and GO groups identified well-known features and functions of the RPE. Genes in the KEGG pathways ‘Phagosome’ (45 genes, FDR = 9.1E-06) and ‘Lysosome’ (41 genes, FDR = 4.8E-07) are enriched. In keeping with pigmentation of RPE, the ‘Melanosome’ cellular component GO group genes are highly enriched (56 genes, 5.0E-21, Supplemental Table S1I). Regarding maintenance of outer blood-retinal barrier (oBRB), the transcriptome contains 52 ‘Tight Junction’ related genes (FDR = 5.1E-09), including several claudin genes (Cldn3, Cldn4, Cldn7 and Cldn23) and zonula occludens genes (Tjp2 and Tjp3). Several other claudin mRNAs were mapped, but with TPM < 45. Notably, despite its published importance to RPE barrier function [3], Cldn19 (TPM = 10.1) did not make it into our mouse RPE transcriptome, nor into any of the published RPE signature gene lists.

Surprisingly, enrichment analysis of the mouse RPE transcriptome revealed several KEGG groups related to ‘Pathways of Neurodegeneration’ (189 genes, FDR = 1.2E-43), which include the most significantly enriched pathway, ‘Parkinson Disease’ (155 genes, FDR = 1.3E-65) (Fig. 1D). This pathway encompasses genes included in other enriched KEGG pathways, including ‘Oxidative Phosphorylation’ (99 genes, FDR = 1.7E-58), ‘Protein Processing in the Endoplasmic Reticulum’ (66 genes, FDR = 1.4E-16), ‘Ubiquitin-mediated Proteolysis’ (41 genes, FDR = 1.2E-05) and ‘Proteosome’ (36 genes, FDR = 9.1E-23). In addition, ‘Ribosome’ genes, encoding large and small ribosomal subunits, are very significantly enriched (102 genes, FDR = 1.3E-60). Similarly, analysis of enrichment of biological process GO groups identified highly significant enrichment of several groups related to ‘Translation’ (209 genes, FDR = 3.3E-36), including the most significantly enriched ‘Peptide Metabolic Process’ (262 genes, FDR = 4.5E-44) (Fig. 1E). In keeping with protein synthesis, enriched cellular component GO groups include many related to ribosomes and their components (Supplemental Data Table S1I) and many enriched GO molecular function groups are related to ribosome structure and function, regulation of mRNA translation and protein processing (Supplemental Data Table S1J). There are also a relatively large number of enriched cellular component genes (377, 7.3E-26) encoding ‘Endoplasmic Reticulum’ proteins (Supplemental Data Table S1I) suggesting that many plasma membrane-targeted or secreted proteins may be produced by mouse RPE.

Many genes in the mouse RPE transcriptome are metabolism-related, with 336 genes in the KEGG ‘Metabolic Pathways’ group (FDR = 1.8E-23) (Fig. 1D). Given that RPE metabolism is highly oxidative to allow utilization of lactate produced by PR and fatty acids derived from PR outer segments [37], it is not surprising that several highly enriched GO biological process groups were also related to ‘Cellular Respiration’ (98 genes, FDR = 2.5E-31), ‘Oxidative Phosphorylation’ (74 genes, FDR = 4.8E-34) and ‘Electron Transport Chain’ (74 genes, FDR = 4.7E-28) (Fig. 1E). Similarly, several highly enrich GO molecular function groups were related to electron transport chain and oxidative phosphorylation (Fig. 1E, Supplemental Data Table S1J). Enriched cellular component GO groups also include several related to mitochondria and respiration (Supplemental Data Table S1I). In keeping with the recent demonstration that peroxisomal β-oxidation is essential for RPE lysosomal function and digestion of very long chain polyunsaturated fatty acids present in PR OS [38], the analysis identified significant enrichment of genes in the KEGG ‘Fatty Acid Degradation’ pathway (21 genes, FDR = 1.5E-06), the GO biological process group ‘Fatty Acid Oxidation’ (34 genes, FDR = 4.9E-05), the KEGG ‘Peroxisome’ pathway (24 genes, FDR = 2.3E-03) and the GO cellular component ‘Peroxisome’ group (34 genes, FDR = 1.8E-03).

Several other highly expressed genes in the mouse RPE transcriptome support established RPE functions. For example, in keeping with the RPE’s apical import of PR-produced lactate and export of β-hydoxybutyrate produced by fatty acid oxidation, Slc16a1 mRNA, encoding the monocarboxylate transporter MCT1, is relatively highly expressed (TPM = 89.9). Interestingly, the abundance of Slc16a8 mRNA (TPM = 167.2), encoding MCT3, which is thought to transport excess lactate out of the RPE at the basolateral side [37], exceeds that of Slc16a1.

Enrichment analyses also identified the KEGG pathway ‘Glutathione Metabolism’ (FDR = 2.0E-05) and the GO biological process group ‘Glutathione Metabolic Process’ (FDR = 1.1E-05) (Supplemental Table S1G). These groups contain several glutathione peroxidase genes (Gpx1, Gpx3 and Gpx4) and several glutathione S-transferase genes (Gsta1, Gsta2, Gsta3, Gsta4, Gstm1, Gstm2, Gsto1, Gstp1 and Mgst1). Remarkably, Gsto1 (glutathione S-transferase omega 1) has the most mapped sequences in the mouse RPE transcriptome, with a TPM = 24336.4. Although GSTO1 protein expression was reported for a wide range of human cells and tissues [39], in porcine corneal epithelium [40] and mouse cone photoreceptors [41], to the best of our knowledge GSTO1 expression by RPE has not been described. We validated the high expression of Gsto1 mRNA in naïve RPE using qRT-PCR (Fig. 1F). Gsto1 mRNA expression in RPE was much higher than in retina (177-fold, p < 0.0001) and choroid (3.5-fold, p-value < 0.001). Because Gsto1 was reported to be relatively highly expressed in liver [39, 42], mRNA levels in mouse liver and RPE RNA were also compared. Surprisingly, Gsto1 mRNA expression in RPE was found to be 811-fold higher in RPE than in liver (p < 0.001), when using 18S ribosomal RNA as internal control mRNA (data not shown). Thus, qRT-PCR confirmed the RNA-Seq results suggesting exceptionally high Gsto1 mRNA expression in the mouse RPE.

Immunofluorescence (IF) in ocular sections, using a gene knockout-validated antibody to GSTO1 protein and a well-characterized antibody to RPE65, showed co-localization, suggesting GSTO1 protein expression in mouse RPE (Fig. 1G). However, the analysis did not detect GSTO1 IF in the cone PR, as previously published [41]. We also found that the anti-GSTO1 antibody co-localized with anti-RPE65 in the human RPE, with no apparent expression in human PR (Fig. 1G). Similarly, in rat GSTO1 IF co-localized with RPE65 and no signal was apparent in PR (Supplement Figure S2). In contrast, although predicted to bind pig GSTO1, the antibody showed minimal co-localization with RPE65 in pig and rhesus monkey sections. Rather, in pig and monkey most GSTO1 IF signal was located basal to the RPE, in regions corresponding to the Bruch’s membrane and the choriocapillaris (Supplement Figure S2).

Temporal effects of retinal detachment on RPE gene expression

A previous study using in situ hybridization found that the mouse RPE exhibits altered levels of mRNAs for several genes after RD at 4 dprd [18]. To comprehensively examine the RPE’s response to detachment, and to define the earlier response, we used RNA sequencing to examine gene expression changes at 1 and 7 dprd. Detachments representing approximately 50% of the retina were produced, and no efforts were made to separate the RPE in the detached region from those still associated with the retina. Thus, the transcriptomes represent a mixture of RPE under RD conditions, both with and without physical association with the neural retina. For transcriptomic comparisons 1 dprd (n = 6) and 7 dprd (n = 6) RD groups were each compared to a single Nv control group (n = 5). DEG were defined as those with a baseMean ≥ 19.5, 0.67 ≥ FC ≥ 1.5 (-0.585 ≥ Log2FC ≥ 0.585) and Padj≤0.05, with baseMean defined as the average of normalized counts for all samples, relative to size factors. The analysis showed an extensive acute response to RD at 1 dprd, with 2293 total DEG identified – 1334 upregulated and 959 downregulated DEG. Comparison of 7 dprd group to the Nv group made obvious that these transcriptomic changes had diminished; only 18 significant DEG were identified at 7 dprd, with 6 upregulated and 12 downregulated genes. The DEG groups are shown as organized into those that were significantly upregulated at both times post RD (Up/Up, 3 genes, Fig. 2B), significantly up only at 1 dprd (Up/NSC, 1328 genes, Fig. 2C), significantly up only at 7 dprd (NSC/Up, 3 genes, Fig. 2D), significantly down only at 1 dprd (Down/NSC. 947-genes, Fig. 2E), and significantly downregulated at both times post RD (Down/Down, 12 genes, Fig. 2F).

Fig. 2

Temporal gene and protein expression changes after retinal detachment and sequencing validation. (A) Heatmap showing significantly upregulated (FC ≥ 1.5, padj≤0.05, red) and downregulated (FC ≤ 0.67, padj≤0.05, blue) DEGs 1 and 7 dprd. (B-F) Relative expression values of DEG grouped by temporal expression. (B) Log2FC of all DEGs significantly upregulated at both 1 and 7 dprd (Up/Up). (C) Log2FC of DEGs significantly upregulated only at 1 dprd (Up/NSC). Gray indicates DEG that were not significantly changed (NSC). (D) Log2FC of DEGs significantly upregulated only at 7 dprd (NSC/Up). (E) Log2FC of DEGs significantly downregulated only at 1 dprd (Down/NSC). (F) Log2FC of DEGs significantly downregulated at both 1 and 7 dprd (Down/Down). (G-K) QRT-PCR of temporal DEG expression changes after retinal detachment for RNA-Seq validation. Relative mRNA expression of (G) Grem1, (H) S100a8, (I) Serpine3, (J) Rpe65, and (K) Tsc22d3, obtained from isolated naïve RPE (n = 6), as well RD RPE at 1 dprd (n = 6) and 7 dprd (n = 6). Bar graphs represent mean ± SEM. Statistical analysis was performed using one-way ANOVA with repeated measures followed by Tukey’s post hoc test. *p < 0.05; **p < 0.01; ***p < 0.001, ****p < 0.0001

To validate the RNA-Seq DEG, mRNA expression of genes from each DEG temporal group was tested in a validation set of RNAs obtained from a separate set of detachments using qRT-PCR. Grem1 mRNA, an example of an Up/Up DEG, was found to be upregulated 17-fold at 1 dprd (p ≤ 0.001) and trended up 6.5-fold at 7 dprd (p = 0.25) (Fig. 2G). An Up/NSC DEG, S100a8 mRNA, was increased by 51-fold at 1 dprd (p < 0.0001) and not increased at 7 dprd (p = 1.0) (Fig. 2H). The NSC/Up DEG Serpine3 mRNA was not change at 1 dprd (p = 0.999) but was significantly increased 18-fold at 7 dprd (p < 0.0001) (Fig. 2I), consistent with the RNA-Seq results showing it to be the most highly increased mRNA in the long-detached RPE. However, using the two commercially available antibodies to SERPINE3 that we were able to identify, we were unable to convincingly demonstrate increased SERPINE3 protein expression in the detached RPE (Supplemental Figure S4). The Down/NSC DEG, Rpe65 was significantly downregulated (0.41-fold, p < 0.01) at 1 dprd and trended down at 7 dprd (0.55-fold, p = 0.08). The Down/Down DEG, Tsc22d3 was found to not be downregulated at 1 dprd (p = 0.963), but was significantly downregulated at 7 dprd (0.66-fold, p < 0.05). Additional DEG were validated by qRT-PCR in this sample set (Supplemental Figure S3) and largely validated results from RNA-Seq.

Lipocalin-2, which among its many functions can act as a inducer of chemokine expression [43,44,45], was identified as one of the DEG most highly responsive to RD, with a 42.9-fold increase at 1 dprd in the RNA-Seq analysis. We therefore chose to validate Lcn2 mRNA results and examine LCN2 protein expression. Samples obtained from a germline Lcn2 KO mouse were used as a negative control. QRT-PCR confirmed a 40-fold increase in Lcn2 mRNA (p < 0.0001), which reverted to normal at 7 dprd (Fig. 3A). Western blotting showed that LCN2 protein in the RPE/choroid was highly increased at 1 dprd versus Nv (9.7-fold, p < 0.01), and significantly elevated at 7 dprd (1.7-fold, p < 0.01) (Fig. 3B). IF of ocular sections revealed that LCN2 protein was nearly undetectable in naïve B6 mouse ocular sections (Fig. 3C). At 1 dprd, LCN2 protein IF was intense on the apical surface of the RPE, which was confirmed by co-localization with the apical RPE marker Ezrin (Fig. 3C). Interestingly, LCN2 protein IF was even more apparent in RPE at 7 dprd and was also located intracellularly and on the basolateral surface in some regions, as demonstrated by a broader distribution and co-localization with TMEM98, a transmembrane protein located at both the apical and basolateral surfaces of the RPE [46]. We also noted intense LCN2 IF in the retina at 1 and 7 dprd (data not shown). Importantly, antibody binding was highly specific, for sections from the Lcn2 KO mouse exhibited no anti-LCN2 IF background, even in sections from 1 to 7 dprd (data not shown).

Fig. 3

Temporal LCN2 expression changes after retinal detachment for sequencing validation. (A) QRT-PCR analysis of Lcn2 mRNA levels in naïve and detached RPE at 1 dprd and 7 dprd (n = 6/group). (B) Representative immunoblot of LCN2 protein in RPE/choroid lysates from Lcn2 knockout mice (Lcn2−/−), as well as RPE/choroid lysates from wildtype C57BL6/J mice under naïve, 1 dprd and 7 dprd conditions. Graph shows normalized results for 6 independent blots (n = 6 mice/group) showing mean ± SEM. Statistical analysis was performed using one-way ANOVA with repeated measures followed by Dunnett’s post hoc test. **p < 0.01. (C) IF co-localization of LCN2 (magenta) with Ezrin (green) and TMEM98 (red). Ezrin is an actin binding protein located in the apical microvilli of RPE cells, while TMEM98 is a transmembrane protein located at both the apical and basolateral surfaces of the RPE [46]. Ocular sections from C57BL6/J mice under naïve, 1 dprd and 7 dprd conditions and a section from Lcn2 KO mice with detached retina at 7 dprd. Staining for cell nuclei (blue) is shown in the merged images (right column) and magnified regions (boxes). Images were acquired using the Leica STELLARIS 8 FALCON Confocal Microscope, 40x objective

C4b, which plays an essential role in both classical and lectin complement pathways [47], was one of few DEG upregulated at 7 dprd. A significant (p < 0.0001) increase in C4b mRNA of 6.7-fold at 7 dprd was validated by qRT-PCR in a repeat set of samples (Fig. 4A). C4 protein expression was thus examined by western blotting and IF in ocular sections. Western blotting of soluble ocular fluid (vitreous, retinal soluble protein and subretinal fluid) with an often-cited mAb to C4 protein showed no increase at 1 dprd and a significant (p < 0.05) 2.3-fold increase at 7 dprd (Fig. 4B). In IF analyses, anti-C4 serum IF was colocalized with anti-RPE65 IF in the RPE, but the serum also produced staining of the neuroretina, even in naïve samples (Fig. 4C). Anti-C4 immunolabeling with the mAb showed intense IF on Iba1+ subretinal immune cells (Fig. 4D), suggesting that C4 protein may accumulate on or in microglia and perhaps monocyte-derived macrophages in the subretinal space.

Fig. 4

C4 protein localization after retinal detachment for sequencing validation. (A) QRT-PCR analysis of C4b mRNA levels in naïve and detached RPE at 1 dprd and 7 dprd. (B) Representative immunoblot of complement C4 in the soluble fluid fraction from C57BL6/J mouse eyes under naïve, contralateral (fellow at 7 dprd), and detached (7 dprd) conditions with protein samples separated under non-reducing conditions. Blots are representative of 3 independent animals in each condition. Bar graphs represent mean ± SEM. Statistical analysis was performed using one-way ANOVA with repeated measures followed by Dunnett’s post hoc test. *p < 0.05; ****p < 0.0001. (C) IF analysis of ocular sections from C57BL6/J mice under naïve, 1 dprd and 7 dprd conditions. Cell nuclei staining (blue), and IF of anti-RPE65 mAb (green) and anti-C4 serum (red) are shown. (D) IF analysis of C4 protein in retinal and subretinal Iba-1+ immune cells in ocular sections from C57BL6/J mice under naïve, 1 dprd and 7 dprd conditions. IF of anti-C4 mAb (green), anti-Iba1 mAb (red, a microglia and monocyte/macrophage marker [48]). Staining cell nuclei (blue) is shown in the merged images (right column). Images were acquired using the Leica STELLARIS 8 FALCON Confocal Microscope, 40x objective

Analysis of differentially expressed genes after one day of retinal detachment

Previous studies have shown that epithelial barriers can react to injury and infection by taking on an innate immune phenotype to provide local immunity [49]. DESeq2 comparison of RNA-Seq reads from naïve RPE (n = 5) and RPE with retinal detachments harvested at 1 dprd (n = 6) identified 2297 DEG with a baseMean expression cutoff of 19.5, 0.67 ≥ FC ≥ 1.5 and padj≤0.05 (Supplemental Table S2B). We hypothesized that the RPE would react to RD by rapidly increasing its barrier function and mounting a sterile innate immune response. To test this, we examined DEG upregulated at 1 dprd and used KEGG pathway and GO group enrichment analysis to characterize them. Upregulated DEG were defined as those with FC ≥ 1.5 (Log2FC ≥ 0.585) and Padj≤0.05, which yields 1334 upregulated DEG (Supplemental Table S2C). Within this list are 18 genes (1.3%) that are expressed at greater levels in retina than RPE. Because of the small fraction they represent, and because none are genes associated with phototransduction, these 18 DEG were not excluded from the analysis of DEG.

KEGG pathway enrichment was examined in the 1334 upregulated DEG at 1dprd. Several neurodegeneration-related groups are significantly enriched, including ‘Prion Disease Amyotrophic Lateral Sclerosis’, ‘Parkinson Disease’, ‘Huntington Disease’, ‘Pathways of Neurodegeneration’ and ‘Alzheimer Disease’ are highly enriched (Fig. 5A). ‘Prion Disease’ is the most significantly enriched KEGG pathway (FDR = 9.0E-14). Although neurodegeneration seems incongruous, it should be noted that, the KEGG neurodegeneration groups are dominated by genes in pathways related to oxidative phosphorylation, mitochondria-initiated apoptosis, ER-stress associated degradation, and proteosome genes (Supplemental Table S2E). Indeed, KEGG groups related to oxidative phosphorylation (‘Oxidative Phosphorylation’, ‘Thermogenesis’, ‘Chemical Carcinogenesis’, ‘Diabetic Cardiomyopathy’, and ‘Non-Alcoholic Fatty Liver Disease’) and phagosome-related groups (‘Phagosome’, ‘Rheumatoid Arthritis’, and ‘Tuberculosis’) are also significantly enriched. Other highly enriched KEGG pathway groups include: a cell cycle group (‘Cell Cycle’ and ‘P53 Signaling Pathway’), a DNA replication/repair group (‘DNA Replication’ and ‘Nucleotide Excision Repair’), a ‘Drug and Pyrimidine Metabolism’ group, and a protein synthesis and secretion group (‘Ribosome’, ‘Protein Export’ and ‘Protein Processing in the Endoplasmic Reticulum’). Of note is that more than 10% of the upregulated DEG (148 genes) are included in the KEGG ‘Metabolic Pathways’ group. Thus, the KEGG enrichment analysis suggests that the RPE mounts a stress response following RD that in many ways resembles those associated with neurodegenerative diseases.

KEGG enrichment analysis also identified pathways associated with infection (‘Influenza A’, ‘Epstein-Barr Virus Infection’, ‘Viral Carcinogenesis’, ‘Measles’, and ‘NOD-like Receptor Signaling’ pathways), as well as ‘Viral Protein Interaction with Cytokine and Cytokine Receptors’, ‘Cytokine/Chemokine Signaling’ and ‘Antigen Processing and Presentation’ pathways (Supplemental Table S2E) suggesting the RPE mounts an innate immune response to retinal detachment. Likewise, analysis of enrichment of biological process GO groups in the upregulated DEG at 1 dprd revealed many groups characterized by innate immune defense responses (Fig. 5A, Supplemental Table S2F). These include the most significantly enriched group, ‘Defense Response to Other Organism’ (FDR = 2.2E-13). Other groups in this category include: ‘Innate Immune Response’, ‘Immune Response’, ‘Response to Cytokine’, and ‘Response to Virus’. In addition, innate immune response groups include ‘Response to Interferon − Beta’, ‘Response to Interferon − Gamma’, and ‘Positive Regulation of Tumor Necrosis Factor Superfamily Cytokine Production’. As with KEGG pathways, several enriched GO biological process groups are also related to cell cycle and cell division, and to protein catabolic processes (Supplemental Table S2F). Thus, as predicted the analysis suggests the RPE’s transcriptional response to RD involves upregulation of genes involved with an innate immune defense response, along with the seemingly contradictory processes of cell growth and protein degradation.

Analysis of cellular component gene ontology groups in the upregulated DEG at 1 dprd (Fig. 5A, Supplemental Table S2G) revealed that the most highly enriched group is ‘Mitochondrial Protein-containing Complex’ (FDR = 4.4E-22). Nearly 18% (238) of the 1334 up-regulated DEG at 1 dprd encode ‘Components of Mitochondrion’ (FDR = 2.1E-09). These include 34 ‘Mitochondrial Ribosomal Subunits’ (FDR = 1.6E-10), which may reflect increased synthesis of electron transport chain proteins that are translated in the mitochondria [