A mutation in transmembrane protein 135 impairs lipid metabolism in mouse eyecups

Aging is a significant factor in the development of age-related diseases but how aging disrupts cellular homeostasis to cause age-related retinal disease is unknown. Here, we further our studies on transmembrane protein 135 (Tmem135), a gene involved in retinal aging, by examining the transcriptomic profiles of wild-type, heterozygous and homozygous Tmem135 mutant posterior eyecup samples through RNA sequencing (RNA-Seq). We found significant gene expression changes in both heterozygous and homozygous Tmem135 mutant mouse eyecups that correlate with visual function deficits. Further analysis revealed that expression of many genes involved in lipid metabolism are changed due to the Tmem135 mutation. Consistent with these changes, we found increased lipid accumulation in mutant Tmem135 eyecup samples. Since mutant Tmem135 mice have similar ocular pathologies as human age-related macular degeneration (AMD) eyes, we compared our homozygous Tmem135 mutant eyecup RNA-Seq dataset with transcriptomic datasets of human AMD donor eyes. We found similar changes in genes involved in lipid metabolism between the homozygous Tmem135 mutant eyecups and AMD donor eyes. Our study suggests that the Tmem135 mutation affects lipid metabolism as similarly observed in human AMD eyes, thus Tmem135 mutant mice can serve as a good model for the role of dysregulated lipid metabolism in AMD.

The most significant risk factor for age-related diseases is aging itself. However, how aging is associated with age-related disease at the molecular level is still largely to be understood. The polygenic nature of age-related neurodegenerative diseases such as age-related macular degeneration (AMD) and Alzheimer's disease makes identification of the molecular pathways involved more challenging 1,2 .Therefore, mouse mutants showing agerelated symptoms are valuable for isolating genetic factors that contribute to aging and age-related diseases.
We identified one such mouse strain harboring a mutation in transmembrane protein 135 (Tmem-135 FUN025/FUN025 ) that displays signs of accelerated aging in the retina as well as pathologies observed in AMD including autofluorescent aggregates in the subretinal space, increased inflammation, and thickened retinal pigmented epithelium (RPE) at a young age 3 . These phenotypes indicated that Tmem135 regulates retinal aging and its impairment results in age-related disease phenotypes. Tmem135 encodes a 52 kilodalton protein with five transmembrane domains that share similarities with the Tim17 protein family, a group of proteins important for mitochondria biogenesis 3,4 . TMEM135 was originally identified as a gene associated with longevity and cold stress in C. elegans 5 . Previous studies also suggested the function of TMEM135 in lipid metabolism 5,6 . In addition, a study using transgenic mice overexpressing the Tmem135 gene (Tmem135 TG) showed that Tmem135 overexpression results in RPE dysmorphogenesis and degeneration 7 . Electron micrographs of the RPE from Tmem135 FUN025/FUN025 mice showed over-fused (elongated) mitochondrial networks while the RPE of Tmem135 TG mice displayed over-fragmented mitochondrial networks relative to controls 3,7 . Despite their morphological differences, both Tmem135 FUN025/FUN025 and Tmem135 TG RPE show reduced mitochondrial function 8 . Taken together, these results strongly indicated that TMEM135 acts as a pro-mitochondrial fission factor and regulates "mitochondrial dynamics, " the collective term for the fusion, fission, and mitophagy events mitochondria undergo to preserve their shape, number and function 9 , which is critical for RPE health. One striking agedependent phenotypic feature of Tmem135 FUN025/FUN025 mice is the robust infiltration of IBA1-and F4/80-positive immune cells into the subretinal space 3 . Accumulation of subretinal immune cells has been associated with all stages of AMD and believed to contribute to the chronic inflammation underlying AMD. The phenotypic similarity between Tmem135 FUN025/FUN025 mice and human AMD motivates us to identify molecular pathways underlying the retinal pathologies in Tmem135 FUN025/FUN025 mice.
In this study, we performed RNA sequencing (RNA-Seq) of the eyecup (RPE/choroid/sclera) of Tmem-135 FUN025 /FUN025 mice compared with heterozygous and wild-type (WT) mice. We found several pathways including cholesterol metabolism, fatty acid metabolism, and steroid metabolic processes were significantly upregulated in Tmem135 FUN025/FUN025 mice. We observed increased lipid accumulations in Tmem135 FUN025/FUN025 eyecups, which confirms the changes in lipid metabolism. Some of these gene expression changes overlap with those previously observed in human AMD-afflicted RPE/choroid samples. Our study suggests that lipid metabolism is a molecular pathway affected by the Tmem135 mutation as observed in human AMD eyes, thus Tmem135 mutant mice can serve as a good model for the role of dysregulated lipid metabolism in AMD.

Results
Heterozygous and homozygous Tmem135 mutant mice display ocular phenotypes. A mutation in Tmem135 effects retinal homeostasis in mice 3,7 . Homozygous Tmem135 mutant (Tmem135 FUN025/FUN025 ) mice show early-onset of aging-associated retinal changes with accelerated progression and pathologies similar to those observed in AMD as early as 2 months of age 3 . At that age, heterozygous Tmem135 mutant (Tmem-135 FUN025/+ ) mice did not show any ocular abnormalities (data not shown). Here, we investigated whether heterozygous Tmem135 mutant (Tmem135 FUN025/+ ) mice display ocular abnormalities at 12 months of age. No distinct morphological changes were observed in Tmem135 FUN025/+ retinal sections compared to WT retinal sections (Fig. 1A). This contrasts the presence of photoreceptor degeneration and subretinal immune cells present in homozygous Tmem135 mutant (Tmem135 FUN025/FUN025 ) retinas (Fig. 1A). We quantitated photoreceptor degeneration in Tmem135 FUN025/FUN025 retinas by normalizing the length of the outer nuclear layer (ONL) to the inner nuclear layer (INL) to calculate the ONL thickness (ONLT) ratio 3 . While there were no ONLT differences between WT and Tmem135 FUN025/+ , the ONLT ratios of both groups were significantly higher than those measured for the Tmem135 FUN025/FUN025 retinas (Fig. 1B). Also, there were no differences in the thickness of WT and Tmem135 FUN025/+ RPE, which is increased in Tmem135 FUN025/FUN025 mice (Fig. 1B). Next, we measured visual function through electroretinography (ERG). 12-month-old Tmem135 FUN025/+ mice had significantly smaller scotopic a-waves than WT mice except at the 30 cd*s/m 2 flash intensity (Fig. 1C). The scotopic b-waves of Tmem135 FUN025/+ mice were significantly decreased at 0.03, 1, 3, and 10 cd*s/m 2 flash intensities relative to WT mice (Fig. 1D). No changes in the photopic a-wave, photopic b-wave, and c-wave were noted between WT and Tmem135 FUN025/+ mice (1E-G). Given that scotopic ERG responses and not photopic ERG responses were changed, rods are sensitive to the heterozygosity of the Tmem135 mutation. The scotopic a-waves and b-waves at all intensities as well as photopic a-wave, photopic b-wave, and c-wave were significantly attenuated between Tmem135 FUN025/FUN025 and WT mice ( Fig. 1C-G) as previously described 3 . Our data indicates both Tmem135 FUN025/+ and Tmem135 FUN025/FUN025 mice have retinal functional abnormalities by 12 months of age, but only Tmem135 FUN025/FUN025 mice have pronounced retinal morphological differences compared to WT mice.

Mutation in Tmem135 results in widespread transcriptional changes.
To identify the origin of the retinal abnormalities due to the Tmem135 mutation, we performed RNA-Seq on eyecups collected from 2.5-month-old WT, Tmem135 FUN025/+ , and Tmem135 FUN025/FUN025 mice. This time point was selected in order to examine transcriptional changes in the early stage of the pathological changes in Tmem135 FUN025/FUN025 mice rather than later stages that would be affected by extensive retinal degeneration ( Fig. 1) 3 . These eyecup tissue preparations include the retinal pigmented epithelium (RPE), a cell-type particularly sensitive to the levels of functional TMEM135 7 and postulated as the primary site affected by the Tmem135 FUN025 mutation 3 . Global transcription profiles were affected in both Tmem135 FUN025/+ and Tmem135 FUN025/FUN025 eyecups compared with WT controls. 1723 genes and 1785 genes were significantly altered (adj.p value < 0.05) in the Tmem135 FUN025/+ and Tmem135 FUN025/FUN025 genotypes relative to WT, respectively. 1112 were commonly affected between the Tmem-135 FUN025/+ and Tmem135 FUN025/FUN025 eyecups compared to WT. There were 611 and 673 genes uniquely changed in Tmem135 FUN025/+ and Tmem135 FUN025/FUN025 eyecups, respectively. We narrowed our RNA-seq analysis on genes with twofold expression differences in Tmem135 FUN025/+ and Tmem135 FUN025/FUN025 eyecups ( Fig. 2A,B). In Tmem135 FUN025/+ eyecups, there were 277 and 418 genes that had twofold upregulation and downregulation compared to WT, respectively ( Fig. 2A). Similarly, there were 333 genes that had twofold upregulation and 433 genes that had twofold downregulation in Tmem135 FUN025/FUN025 eyecups (Fig. 2B).
Next, we conducted gene set enrichment analysis (GSEA) to identify potential pathways which are affected due to the Tmem135 mutation. Several pathways were significantly downregulated (FDR < 0.01) in both Tmem-135 FUN025/+ and Tmem135 FUN025/FUN025 as compared to WT including contractile fiber, sensory perception of light stimulus, ribosome, mitochondrial membrane part, Parkinson disease, and synaptic vesicle cycle (Table 1 and Fig. 2C). On the other hand, several pathways including cholesterol metabolism, fatty acid metabolism, and steroid metabolic processes were significantly upregulated (FDR < 0.01) exclusively in Tmem135 FUN025/FUN025 eyecups relative to WT (Table 1 and Fig. 2D). Importantly, none of the pathways were significantly upregulated (FDR < 0.01) between Tmem135 FUN025/+ and WT eyecups (Table 1). Our results show there are distinct pathways disrupted by the Tmem135 mutation in the murine eyecup.

Correlations between differential gene expression and number of Tmem135 mutant alleles.
We determined if any of the genes with differential expression correlated with the number of Tmem135 mutant alleles. 72 genes had significant correlations (FDR < 0.01) with the Tmem135 mutation status of mice where 35 genes were downregulated and 37 genes were upregulated in Tmem135 FUN025/FUN025 eyecups when compared with WT eyecups (Fig. 4A, B). We used signalling network analysis to identify potential mechanisms which might be affected in a mutation dependent manner. We identified BCL6 corepressor (Bcor) which had a negative correlation score of -0.9. BCOR can repress the transcription of target genes by interacting with histone deacetylases (HDACs) (Fig. 4C) and non-canonical polycomb recessive complex 1 (PRC1) that can downregulate peroxisome proliferator activated receptor gamma (Pparg) expression (Fig. 4D). We also observed a very strong positive correlation of 0.98 for isocitrate dehydrogenase 1 (Idh1) which was significantly increased in Tmem135 FUN025/FUN025 mice. Idh1 expression can be regulated by FOXO transcription factors and sterol regulatory element binding proteins (SREBFs) (Fig. 4E), both of which play important roles in lipid metabolism 10,11 . This analysis reveals a set of genes whose expression is correlated with the amount of functional TMEM135 and connected with transcriptomic pathways disrupted by the Tmem135 mutation.   (Fig. 5B). We prepared RPE flat mounts from 3-month-old and 12-month-old  Lipid metabolic gene expression in eyecups and neural retinas. The RPE is a major tissue involved in lipid metabolism within the retina 12 . Since we detected significant lipid metabolic transcriptional changes from our RNA-seq study, we explored whether these changes were unique to the posterior eyecup containing the RPE. We prepared cDNA from RNA individually isolated from the posterior eyecup and neural retinas of 2.5-month-old Tmem135 FUN025/+ and Tmem135 FUN025/FUN025 mice for quantitative PCR (qPCR). We quantified sterol regulatory element binding transcription factor 2 (Srebf2), Srebf1c, acetyl-CoA carboxylase alpha (Acaca), fatty acid synthase (Fasn), and stearoyl-CoA desaturase 1 (Scd1). We found significant increases in all these genes in Tmem135 FUN025/FUN025 eyecups compared to Tmem135 FUN025/+ eyecups (Fig. 6A). Strikingly, only Scd1 was upregulated in the neural retinas of Tmem135 FUN025/FUN025 mice relative to Tmem135 FUN025/+ neural retinas (Fig. 6B). These data show that Tmem135 mutant-induced increases of lipid metabolic genes are predominantly confined to the posterior eyecup.
Transcriptomic similarities between homozygous Tmem135 mutant eyecups and AMD donor eyes. It is well accepted that dysfunction within the RPE is implicated as a key pathobiological mechanism in AMD [13][14][15] . Since the ocular phenotype of Tmem135 FUN025/FUN025 mice shares similarities with AMD including the presence of subretinal immune cells, neuroinflammation and RPE autofluorescence 3,16 , we were interested in evaluating the eyecup transcriptional profiles of the Tmem135 FUN025/FUN025 mice with those obtained from RPE/ choroid samples from AMD patients. We compared our Tmem135 FUN025/FUN025 versus WT RNA-Seq dataset with previously-published integrated microarray and RNA-Seq datasets (GSE29801 and GSE135092) using RPE/choroid samples from AMD patients [17][18][19] . There were similar genes affected due to the Tmem135 mutation and AMD (Table 2), including 16 upregulated and 9 downregulated genes in common between the Tmem135 FUN025/FUN025 relative to WT RNA-Seq dataset and GSE29801 and GSE135092 datasets ( Table 2 and Fig. 7A). In the GSE29801   Fig. 7A). Furthermore, two upregulated genes (Srebf1 and Lss) and two downregulated genes [GATA binding protein 2 (Gata2) and Cyb5a] were common between Tmem135 FUN025/FUN025 eyecups and advanced AMD ( Table 2 and Fig. 7A). We performed a gene over-representation analysis on the genes in Table 2 and found significant enrichment (FDR < 0.01) of multiple lipid metabolic pathways including cholesterol metabolism, sterol biosynthetic process, sterol metabolic process, lipid biosynthetic process, fatty acid biosynthetic process, and fatty acid metabolism (Table 3). We also investigated the network signaling pathways commonly affected between Tmem135 FUN025/FUN025 mice and AMD. We identified the SREBF1-signaling pathway, a pathway that mediates lipid homeostasis by regulating Pparg expression and lipogenesis (Fig. 7B) 11 , in common between Tmem135 FUN025/FUN025 eyecups and AMD. Similarities between the Tmem135 FUN025/FUN025 mice and AMD transcriptional profiles suggest the Tmem135 FUN025/FUN025 mouse may serve as a valuable model to interrogate the role of lipid metabolism on AMD-like pathology development.  Tmem135 encodes a membrane-bound protein that can localize with mitochondria 3,5 , peroxisomes 22 , and lipid droplets 5 . Our gene enrichment analysis highlighted significant downregulation of mitochondrial membrane and upregulation of lipid metabolism pathways in Tmem135 FUN025/FUN025 eyecups. Mitochondria, peroxisomes, and lipid droplets interact with each other to maintain lipid metabolism within a cell 23 . In particular, cells depend on mitochondria for the complete breakdown of lipids through oxidative phosphorylation 24 . The mitochondria in Tmem135 FUN025/FUN025 RPE appear overly elongated 7 and correlates with increased levels of glycerol 8 , a building block of triglycerides. Triglycerides store fatty acids within the cell for consumption by mitochondria for energy utilization 25 . It has been shown that mitochondria with close proximities to lipid droplets are overly elongated 26 , which could explain the mitochondrial morphological changes in Tmem135 FUN025/FUN025 RPE. Thus, TMEM135 may have a role in lipid metabolism within the RPE, supporting other studies linking the function of TMEM135 to lipid metabolism 5,6 .
The RPE is critical for sustaining lipid homeostasis within the retina, digesting photoreceptor outer segments, ingesting choroid-derived lipoproteins, and transporting lipids out of the retina 12 . It is known that RPE cells express canonical pathways involved in lipid homeostasis including the sterol regulatory element-binding protein (SREBF) pathway in mice and humans 27,28 but may rely on additional mechanisms as suggested by a lack of responsiveness of cholesterol-sensitive genes in wild-type mouse retinas after the consumption of dietary cholesterol 27 . Through our analysis of differential genes that correlated with the number of Tmem135 mutant alleles, we identified BCOR and IDH1 as potential players in RPE lipid homeostasis. BCOR is a co-repressor of BCL6 that mediates BCL6-mediated transcriptional repression 10 . There have been no published reports on a role for BCOR in lipid metabolism, but loss of BCL6 in the liver augments lipid catabolism, lessens highfat-diet induced hepatic steatosis, and reverses fatty acid breakdown and lipid accumulation induced by fasting in peroxisome proliferator activated receptor alpha (Ppara) knockout mice 29 . The decrease in Bcor in the Tmem135 FUN025/FUN025 mouse eyecups may be in response to the increased lipid accumulation such as cholesterol and neutral lipids. IDH1 is an isocitrate dehydrogenase enzyme found in the cytoplasm as well as on peroxisomes that catalyzes the reversible oxidative decarboxylation of isocitrate to alpha-ketoglutarate 30 . Loss-of-function mutations in human IDH1 lead to increased monosaturated fatty acids and increased SCD expression in infiltrating gliomas 31 . Since there is increased eyecup Scd1 expression in Tmem135 FUN025/FUN025 mice and SCD1 is the rate-limiting enzyme for the conversion of saturated fatty acids to monounsaturated fatty acids 32 , the increased Idh1 expression in Tmem135 FUN025/FUN025 mouse eyecups may occur to decrease the amount of monounsaturated fatty acids. Further elucidation of BCOR and IDH1 in Tmem135 FUN025/FUN025 mouse eyecups could lead to insight into lipid metabolism within RPE cells.
Dysregulation of lipid metabolism is strongly associated with AMD development and progression 33 . The association of dysregulated lipid metabolism and AMD comes from genetic and environmental studies. Genetic studies have linked variants with human apolipoprotein E (APOE), hepatic lipase (LIPC), cholesteryl ester transfer protein (CETP), lipoprotein lipase (LPL), and ATP binding cassette subfamily A member 1 (ABCA1) with risk for AMD [34][35][36][37][38][39][40] . The roles of these genes in AMD development and progression are largely unknown. Furthermore, consumption of diets high in fat [41][42][43] , particularly those with high levels of omega-6, monounsaturated, polyunsaturated and trans unsaturated fatty acids 44,45 , increases the likelihood of developing AMD. In contrast, eating diets with a high concentration of fish oils containing omega-3 fatty acids decreases the incidence of AMD [44][45][46][47] . How diets influence retinal homeostasis and modulate the risk for AMD is unclear. To understand the role of these epidemiological risk factors for AMD, it is imperative to study models in which we can gain new insight into AMD disease processes. We have shown in our RNA-Seq study that Tmem135 FUN025/FUN025 mice have altered expression of lipid metabolic genes in their posterior eyecups and lipid accumulation in their RPE. Some of these Table 3. Pathway analysis using overrepresentation analysis (ORA) on genes significantly-affected in both Tmem135 FUN025/FUN025 versus WT eyecup RNA-Seq and GSE29801 and GSE135092 datasets*. *FDR cut-off of 0.01 is used for analysis. Aging is a significant risk factor for a number of diseases including AMD and Alzheimer's disease, but how aging contributes to these diseases is unknown. One strategy to better understand the role of aging in disease development is to study mutant mice with accelerated aging phenotypes and uncover the genetic causes of their accelerated aging phenotypes. Utilizing this strategy, we identified Tmem135 as a gene involved in the aging process in the retina 3 . In the current study, we undertook an RNA-Seq study to further investigate the consequences of the Tmem135 mutation on retinal homeostasis. We found that mutant Tmem135 mice display significant transcriptomic changes but one significant pathway that contained a number of differentially-expressed genes is lipid metabolism. Some of these genes are also altered in transcriptomic profiles of AMD-diagnosed RPE/ choroid specimens. Studying Tmem135 mutant mice may provide insight into how aging effects lipid metabolism and contributes to diseases such as AMD.

Methods
Mice. Generation of Tmem135 FUN025 / + and Tmem135 FUN025/FUN025 mice on the C57BL/6 J background was previously described 3,7 . WT mice on the C57BL/6 J background were used as controls for these experiments. All animals were housed in the same animal facility at the University of Wisconsin-Madison under the same environmental conditions. Both male and female mice that were 2.5 months, 3 months and 12 months of age were used in this study. Details of mice used in each figure of this study are provided in Supplemental Table 1.
No mice used in this study were positive for the Ped6b rd1 or the Crb1 rd8 mutation 50,51 . All experiments performed in this study were in accordance with the National Institute of Health Guide for the Care and Use of Laboratory Animals and authorized by the Animal Care and Use Committee at the University of Wisconsin-Madison. The results and methods in this study are reported in accordance to the ARRIVE guidelines.
Histology. Eyes were fixed in 2% PFA and 2% glutaraldehyde overnight at 4 °C and submitted to the University of Wisconsin-Madison's Translational Research Initiatives in Pathology (TRIP) core for paraffin processing. Sections were stained with hematoxylin and eosin (H&E) to visualize retinal layers and imaged using an Axio Imager 2 microscope (Carl Zeiss MicroImaging, White Plains, NY) at a 40X magnification. Outer nuclear layer (ONL) and inner nuclear layer (INL) thickness were measured using ImageJ (NIH, Bethesda, MD) on a single section from each eye. Measurements were started from the optic nerve head and taken at seven consecutive 300 micron intervals.

Electroretinography (ERG).
Mice were dark-adapted overnight before ERG recording and then anesthetized with a cocktail of ketamine (80 mg/kg) and xylazine (16 mg/kg) diluted in phosphate buffered saline (PBS). Pupils were dilated with 1% tropicamide, and a drop of sterile 2.5% hypromellose ophthalmic solution (Goniovisc, HUB pharmaceuticals LLC, CA) was applied to the cornea for recording electrodes. ERGs were recorded using the Espion system (Diagnosys LLC, MA). Scotopic recordings were obtained from dark-adapted animals, and the eyes were exposed to a sequential increment of flash intensities (0.1 to 30 cd.s/m 2 ) for 300 ms with a 2 s interval between the flashes. For the acquisition of c-wave, the eyes were flashed with light intensities of 2.5 and 25 cd.s/m 2 for 4 s. Photopic recordings were performed after the light adaptation of mice at 30 cd.m 2 background and during the photopic recordings. The flash intensities used for the photopic recordings were 0.78, 1, 2.25, 5 and 10 cd.s/m 2 . Only the 10 cd.s/m 2 recordings were used in this study. ERG components were measured using the Espion software (Diagnosys LLC, MA) and analyzed using Origin2020 (OriginLab Corp., MA).
Bulk-RNA sequencing. Posterior eyecups without the neural retina were collected and pooled from individual 2.5-month-old WT, Tmem135 FUN025/+ , and Tmem135 FUN025/FUN025 mice of both genders between 11:00 and 1:00 PM. Samples were flash frozen and then submitted to GENEWIZ (South Plainfield, NJ, USA) for processing. Total RNA was extracted from eyecups with an RNeasy Plus Universal Mini kit (Qiagen, Hilden, Germany) following the Manufacturer's instructions. Total RNA samples were quantified using a Qubit 2.0 Fluorometer (Life Technologies, Carlsbad, CA, USA). RNA integrity was examined using a TapeStation 4200 (Agilent Technologies, Palo Alto, CA, USA). RNA sequencing libraries were prepared using the NEBNext Ultra RNA Library Prep Kit for Illumina (NEB, Ipswich, MA, USA) following manufacturer's instructions. Briefly, messenger RNAs were first enriched with Oligo(dT) beads. Enriched mRNAs were fragmented for 15 min at 94 °C. First strand and second strand cDNAs were subsequently synthesized. cDNA fragments were end repaired and adenylated at 3' ends, and universal adapters were ligated to cDNA fragments, followed by index addition and library enrichment by limited-cycle PCR. The sequencing libraries were validated on the Agilent TapeStation and quantified by using Qubit 2.0 Fluorometer as well as by quantitative PCR (KAPA Biosystems, Wilmington, MA, USA). The sequencing libraries were clustered on a flowcell. After clustering, the flowcell was loaded on the Illumina HiSeq instrument (4000 or equivalent) according to manufacturer's instructions. The samples were sequenced using a 2 × 150 bp Paired End (PE) configuration. Image analysis and base calling were conducted by the HiSeq Control Software (HCS). Raw sequence data (.bcl files) generated from Illumina HiSeq were converted into fastq files RNA sequencing analysis. Gene expression read counts were analysed using NetworkAnalyst 3.0 52 . M. musculus organism was selected with bulk sequencing analysis workflow. Quality control step involved filtering genes with very high variance across samples. Genes were ranked based on variance and those genes which ranked in the bottom 15% of the percentile were filtered out. Low abundance genes below a threshold were also filtered out. Data was normalized using log 2 counts per million normalization method. Differential gene expression analysis was performed using EdgeR 53 . Gene set enrichment analysis was performed using WebGestalt 54 and different functional databases including Gene Ontology, KEGG, and WikiPathways were used for analysis.
To further restrict the number of gene sets due to overlap of the genes, affinity cluster algorithm 55 was applied. Heatmaps were generated on auto-scaled normalized gene expression data using Euclidean distance measure and ward clustering algorithm using MetaboAnalyst 4.0 56 . Signalling pathway analysis was conducted on differentially expressed genes using the SIGNOR 2.0 database 57 . A comparison of differentially expressed genes in different groups with human age-related macular degeneration samples was conducted by comparing the fold change (up or down) of differentially expressed genes identified from integrated microarray (GSE29801 dataset) 17 and RNA-seq analysis (GSE135092 dataset) 18 of AMD human donor eyes compared to non-AMD control donor eyes. AMD donor eyes used to generate the GSE29801 dataset were independently classified into distinct AMD categories based on the Age-Related Eye Disease Study (AREDS) and Rotterdam grading scales 17 whereas AMD donor eyes graded as having category 4 disease were used to produce the GSE135092 dataset 18 . In total, the GSE29801 dataset was comprised of expression data from 31 control eyes and 37 AMD-afflicted eyes 17  Quantitative real-time PCR. Eyecups and neural retinas from mice were isolated after PBS perfusion, flash frozen, and kept at − 80 °C. RNA was isolated using a RNeasy lipid tissue mini kit (#74804, Qiagen, Germantown, MD). Concentrations of RNA were calculated using a Nanodrop 2000 UV-Vis Spectrophotometer (Thermo Fisher Scientific, Waltham, MA). Equal RNA concentrations were aliquoted to make cDNA using an Oligo d(T)18 primer (#S1316S, NEB, Ipswich, MA) and the ProtoScript II Reverse Transcriptase (#M0368L, NEB, Ipswich, MA). Triplicate reactions for each gene was run in the Roche Lightcycler 480 system using cDNA, 200 nmol/L of each primer, and Lightcycler 480 SYBR Green I Supermix (#507203180, Roche). Primer sequences used in this study are described in Supplemental Table 2. Relative expression was normalized to the ribosomal protein lateral stalk subunit P0 (Rlpl0) using the quantitative 2 −ΔΔCT method 58 .