Prominin-1 Regulates Retinal Pigment Epithelium Homeostasis: Transcriptomic Insights into Degenerative Mechanisms

1. Introduction

Inherited retinal dystrophies (IRDs) are a genetically heterogeneous group of disorders that cause progressive vision loss and, in many cases, irreversible blindness [1,2]. While many IRDs have been traditionally viewed as photoreceptor-centric diseases, increasing evidence suggests that retinal pigment epithelium (RPE) dysfunction plays a critical and possibly primary role in disease progression, particularly in IRDs where the mutated gene is present in both RPE and photoreceptors [3,4]. This distinction is particularly relevant for IRDs associated with Prominin-1 (Prom1, also known as CD133), where clinical features such as RPE atrophy and photoreceptor degeneration closely resemble those seen in atrophic age-related macular degeneration (AMD) [5,6]. Despite this phenotypic overlap, the mechanisms by which Prom1 regulates RPE homeostasis remain poorly understood, representing a significant gap in our understanding of macular disease pathogenesis.

Prom1 is a pentaspan transmembrane glycoprotein widely recognized as a stem cell and cancer stem cell marker [7,8]. Beyond its role in stemness, Prom1 is expressed in differentiated epithelial and non-epithelial cells [9], glial cells [9], and the adult retina [10], suggesting broader physiological functions. In photoreceptors, Prom1 localizes to the base of the outer segments, where it regulates disk morphogenesis and membrane architecture [11,12]. Loss-of-function mutations in Prom1 cause a spectrum of retinal diseases, including autosomal dominant and recessive retinitis pigmentosa [11,13], cone-rod dystrophies [14,15,16], and macular dystrophies such as Stargardt disease type 4 (STGD4) [12,17].

While Prom1-related retinal dystrophies have traditionally been viewed as photoreceptor-centric, emerging evidence suggests a complex pathophysiology involving RPE. In a Prom1-null Xenopus laevis model, CRISPR/Cas9-mediated Prom1 loss led to age-dependent RPE degeneration and subretinal drusenoid-like deposits that preceded photoreceptor loss, challenging the paradigm that Prom1 dysfunction primarily affects photoreceptors [18]. These deposits resembled human drusenoid material, suggesting that RPE pathology is an initiating event in Prom1-associated degeneration.

Stargardt disease 4 (STGD4) shares clinical and pathological features with ABCA4-related Stargardt disease type 1 (STGD1) and the atrophic form of AMD, including central photoreceptor degeneration and RPE atrophy [19,20,21]. Although STGD1 is driven by bisretinoid lipofuscin accumulation due to ABCA4 dysfunction, STGD4 may involve overlapping or distinct mechanisms. Notably, Prom1 mutations such as p.R373C and c.869delG have been associated with parafoveal RPE atrophy, granular mottling, and thinning of the outer retina in patients [5,22,23]. In younger individuals, spectral-domain OCT reveals early RPE/Bruch’s membrane thinning and progression to geographic atrophy (GA) [24]. Longitudinal studies confirm the expansion of GA, profound outer retinal degeneration, and phenotypic variability in patients with the Prom1 R373C mutation [25]. These findings suggest that Prom1 dysfunction may directly impair RPE integrity, contributing to macular degeneration independent of bisretinoid accumulation.

The relationship between Prom1 and ABCA4 is further supported by studies showing additive effects of Prom1 and Abca4 mutations on RPE pathology, including granular mottling and lipofuscin-like deposits [19,22]. Despite these clinical observations, the mechanistic role of Prom1 in RPE biology remains poorly defined. Our own studies have shown that Prom1 is expressed in human RPE and localizes predominantly to the cytoplasm, where it regulates autophagy by modulating mTORC1/2 signaling and autophagosome trafficking [26]. Additionally, we confirmed Prom1 expression in mouse RPE (mRPE) in situ by immunogold electron microscopy and RNAscope assays [27]. We showed that targeted Prom1 knockdown in situ using AAV2/1 vectors induces RPE cell death and photoreceptor degeneration, recapitulating the features of dry AMD [27].

We recently demonstrated that Prom1 loss in mRPE activates mTORC1, suppresses TFEB activity, and induces epithelial-mesenchymal transition (EMT), implicating Prom1-mTORC1-TFEB signaling as a central regulator of RPE homeostasis [28]. Although Prom1-related IRDs have long been considered photoreceptor cell-autonomous pathologies, emerging evidence suggests that Prom1 is a major driver of RPE integrity and function. To address this knowledge gap, we performed transcriptomic profiling of Prom1-deficient mRPE cells to identify molecular pathways disrupted by Prom1 loss. Using bulk RNA sequencing and gene set enrichment analysis (GSEA), we identified dysregulated genes and pathways involved in stress signaling, autophagy, lysosomal function, and epithelial identity—molecular signatures that mirror RPE dysfunction in IRDs and atrophic AMD. These findings support a model in which Prom1 maintains RPE homeostasis in a cell-autonomous manner, suggesting that its loss contributes to retinal degeneration through mechanisms beyond photoreceptor disk morphogenesis.

2. Materials and Methods

2.1. Reagents

Materials purchased include the following: Fetal bovine serum (FBS, Atlanta Biologicals); Enhanced chemiluminescence (ECL) western blot detection system (Perkin Elmer, Inc); Protease/Phosphatase Inhibitor Cocktail (Thermo Fisher Scientific, Waltham, MA); Prom1/CD1330 rabbit polyclonal antibody (OAA100379, Aviva), Prom1 rabbit polyclonal (abcam, ab19898); MertK (Invitrogen, 14-5751-82), ZO-1 monoclonal antibody (Invitrogen, ZO1-1A12, 33-9100) and GREM1 polyclonal antibody (catalog# PA5-119163, Invitrogen), goat amti-rabbit Alexa Fluor 488 (Thermo Fisher, cat#A-11008), donkey anti-mouse Alexa Fluor 647 (Thermo Fisher, cat#A-31571) and Prom1 gRNA (Thermo Fisher Scientific, Waltham, MA).

2.2. Mice and Colony Management

C57/BL6J mice were obtained from the Jackson Laboratory (Bar Harbor, ME, USA) (JAX, stock #000664). Mice were housed, maintained on a 12h light-dark cycle, and provided food and water ad libitum. The Institutional Animal Care and Use Committee of Vanderbilt University Medical Center (VUMC) approved all experiments. All animal procedures followed the guidelines of the Association for Research in Vision and Ophthalmology Statement on the Use of Animals in Ophthalmic and Vision Research. Both male and female mice (6-8 weeks old) were used for this project.

2.3. Cell Culture

Mouse RPE (mRPE) cells were obtained from Rosario Fernandez-Godino (Harvard, MA, USA), as described earlier [29]. Briefly, isolated mRPE cells from 8–12-week-old C57/BL6J mice were pooled and plated on a 24-well plate coated with laminin (10mg/ml). After 72h, cells were allowed to reach 50% confluence and then transduced overnight with Lenti-HPV E6/E7 (10⁶ TU/ml) (ABM cat #G268) with 4mg/ml polybrene. Cells were selected in RPE media containing 5% FBS and puromycin (1mg/ml) for 12 days to generate immortalized mRPE cells. These cells were cultured in culturing medium containing N1 medium supplement (1/100 vol/vol), glutamine (1/100 vol/vol), penicillin-streptomycin (1/100 vol/vol), non-essential amino acid (1/100 vol/vol), hydrocortisone (20mg/ml), taurine (250 mg/l), triiodo-thyronin (0.013mg/l), 5% FBS in alpha-MEM at 37⁰C in 5% CO₂ and media replaced three times a week. The cells were subcultured using trypsin (0.25%) and frozen in 10% FBS with 10% DMSO.

2.4. Generation of Prom1-Deficient mRPE Cells via CRISPR/Cas9

Prom1 was knocked out in mRPE cells by CRISPR/Cas9-mediated gene editing, as described earlier [28]. mRPE cells were cultured in 6-well plates. The cells were allowed to reach 70% confluency, then transduced with 10 μL of the lentiviral Cas9 construct and 2 μL of polybrene. After 24h, the infected cells were subcultured into media containing 1.5mg/ml puromycin and grown to 90% confluency. The cells were selected in media containing 0.5mg/ml puromycin for an additional week. After puromycin selection, the stably expressing Cas9-mRPE cells were transfected with Prom1 (chromosome 5) synthetic guide RNA (gRNA) sequence (5’-CGTTGCTGCAACAAATGCGG-3’) (ThermoFisher Scientific, catalog # A35533) using Lipofectamine CRISPRMAX transfection reagent (ThermoFisher, Cat # CMAX00008) to target the mouse Prom1 gene at exon 5, following the manufacturer’s protocol. Prom1-KO was verified by genomic DNA analysis. Cas9-expressing mRPE cells were transfected with either scrambled or Prom1 gRNA, and these cells were then used to extract genomic DNA using the QIAamp DNA Micro Kit. PCR was performed using the forward primer: 5’-GTGCATACTGGGGTCCTCAC-3’ and reverse primer: 5’-ATCTCCCTGCAACACCCTAA-3’ and Qiagen Taq PCR master mix kit. Genomic sequences were analyzed using BLAST to confirm the WT and different Prom1-KO sequences. We introduced two deletions at separate genomic sites within Prom1, generating two independent Prom1-KO lines, as described earlier [28].

2.5. Mouse RPE Flat Mount Preparation and Immunohistochemistry

After euthanasia, the mouse eyes were enucleated and fixed in neutral buffered formalin for 15 min at room temperature. They were washed twice in PBS, and the anterior segment (cornea and lens) was removed to expose the posterior eyecup. The neural retina was carefully detached, leaving the RPE–choroid–sclera intact. Four radial cuts were made from the periphery toward the optic nerve to flatten the eyecup, as described earlier [30]. The tissue was placed RPE-side up in a dish or on a slide. Samples were permeabilized and blocked in PBS containing 0.5% Triton X-100 and 5% normal donkey serum (Abcam, ab7475) for 1h. Primary antibody diluted in blocking buffer was applied and incubated overnight at 4 °C, followed by PBS washes and secondary antibody incubation for 2h at room temperature. Finally, tissues were mounted in Fluoromount-G medium containing DAPI (Thermo Fisher, 00-4959-52) and imaged using confocal microscopy.

2.6. Mouse Retina Sections and Confocal Imaging

The enucleated mouse eyes were immersed in 4% PFA overnight at 4 °C and rinsed 3 times in PBS. The eyes were immersed in 10% sucrose in PBS for 2-4 h, transferred to 20% sucrose, and finally to 30% sucrose overnight at 4 °C. The eyes were embedded in cryomold, sectioned at 10 μm with a cryostat, mounted on slides, and stored at -80 °C. The cryo slides were taken out of the freezer, warmed for 10 mins at room temperature, washed twice in PBS with 0.5% Triton X-100 for 15 mins, blocked for 2 h at room temperature in 1x PBST with 5% normal donkey serum, incubated with primary antibody overnight in 1X PBS, followed by three washes in 1X PBS for 10 mins, and incubation with secondary antibody in 1X PBST with 5% NDS for 1h at room temperature. Slides were washed in 0.5% TritonX-100 in PBS and coverslipped with Fluoromount-G with DAPI. Confocal images of retina sections and RPE flatmount samples were captured using a Zeiss LSM880 confocal microscope with Zeiss ZEN (black) 2.3 software, applying the following parameters: For retina section images at 20x magnification: Plan-Apochromat 20x/0.8 objective; 405 nm excitation at 1.5% and 488 nm excitation at 1.0%; detection wavelengths from 410-495 nm and 495-630 nm; gain set to 800; offset at 0; pixel dwell time of 1.02 microseconds per pixel; averaged twice; Z-stack of 12 slices covering approximately 25 microns at Nyquist sampling; scale = 0.42x0.42x0.29 microns per pixel. For retina section images at 40x: Plan-Apochromat 40x/1.3 Oil DIC UV-IR objective; 405 nm excitation at 1.5% and 488 nm excitation at 1.5%; detection wavelengths from 410-495 nm and 495-630 nm; gain = 800; offset = 0; pixel dwell time of 1.02 microseconds per pixel; averaged twice; Z-stack of 18 slices over approximately 18 microns at Nyquist sampling; scale = 0.21x0.21x1.07 microns per pixel. For retina section images at 63x: Plan-Apochromat 63x/1.4 Oil DIC objective; 405 nm excitation at 2.2% and 488 nm excitation at 2.4%; detection wavelengths from 410-495 nm and 495-630 nm; gain = 800; offset = 0; pixel dwell time of 1.02 microseconds per pixel; averaged twice; Z-stack of 25 slices over approximately 12 microns at Nyquist sampling; scale = 0.13x0.13x0.53 microns per pixel. For RPE flatmount images at 40x: Plan-Apochromat 40x/1.3 Oil DIC UV-IR objective; 488 nm excitation at 0.7% and 633 nm excitation at 6.5%; detection wavelengths from 493-628 nm and 638-755 nm; gain = 800; offset = 0; pixel dwell time of 2.05 microseconds per pixel; averaged twice; Z-stack of 11 slices over roughly 8 microns at Nyquist sampling; scale = 0.21x0.21x0.82 microns per pixel. Maximum Intensity Projections and orthogonal projections were generated using Zeiss ZEN (blue) 3.8.

2.7. Western Blotting

Cell lysates were prepared using mammalian protein extraction buffer (Cell Signaling Technology, Beverly, MA, USA) and a Halt protease/phosphatase inhibitor cocktail (Thermo Fisher Scientific, Waltham, MA, USA), followed by SDS-PAGE, as described previously [26,31]. Proteins were transferred to Immobilon-PVDF membranes with 0.45μm pore size (Millipore, Bedford, MA, USA) and incubated overnight with primary antibodies at 4 °C in Tris-buffered saline containing 0.1% Tween-20 and 5% nonfat dry milk (Biorad, Hercules, CA, USA). Membranes were subsequently incubated with horseradish peroxidase-conjugated secondary antibodies at room temperature for 1 hour, and the immune complexes were visualized using the ECL detection system (PerkinElmer, Waltham, MA) and the Azure c500 Imaging Biosystem (Dublin, CA, USA). Membranes were stripped and re-probed for actin as a loading control. Representative Western blots for three experiments are shown. Densitometric analysis of all Western blots was performed using ImageJ software (developed by Wayne Rasband, National Institutes of Health, Bethesda, MD, USA; available at http://rsb.info.nih.gov/ij/index.html), as described earlier [26]. Blots used in the figures comply with the digital image and integrity policies. The uncropped full-length blots are available as a Supplementary File.

2.8. Real-Time Quantitative PCR

TRIzol reagent (Thermo Fisher Scientific, Waltham, MA, USA) was used to extract total RNA from WT and Prom1-KO mRPE cells, as described earlier [28]. Total RNA concentrations were quantified by measuring the A260 and A280 absorbance using a NanoDrop spectrophotometer, as described previously [26]. Total RNA (1mg) was reverse-transcribed to cDNA using a kit from Promega (Madison, WI, USA) and following the manufacturer’s instructions. The cDNA was diluted 1:5 with DNase-free water. Real-time qPCR was performed using an Ariamx Real-Time PCR system (Agilent Technologies, Santa Clara, CA, USA) with 2.5ml of the cDNA product in a 25ml reaction mixture containing 1X SYBR® Green Master Mix (Applied Biosystems, Foster City, CA, USA) and 120nM forward and reverse primers. A list of forward and reverse primers used for qPCR analyses of genes in WT vs. Prom1-KO mRPE cells is provided in Table 1.

2.9. Bulk RNA Sequencing and Data Analysis

Total RNA was isolated from retinal pigment epithelium (RPE) cells dissected from wild-type (WT) and Prom1-knockout (KO) mouse RPE using the RNeasy Mini Kit (Qiagen) according to the manufacturer’s protocol. RNA integrity was assessed using the Agilent 2100 Bioanalyzer, and samples with an RNA Integrity Number (RIN) of 8.0 or higher were selected for sequencing to ensure high-quality input. Bulk RNA sequencing was performed by the Vanderbilt Technologies for Advanced Genomics (VANTAGE) core facility at Vanderbilt University Medical Center. Libraries were prepared using the Illumina TruSeq Stranded mRNA Library Prep Kit, and paired-end sequencing (150 bp reads) was conducted on the Illumina NovaSeq 6000 platform. Raw sequencing reads were trimmed for adapter sequences and filtered for quality using TrimGalore v0.6.7 (https://zenodo.org/records/5127899) by Creative Data Solutions, Vanderbilt University. High-quality reads were aligned to the Mus musculus reference genome (mm39) using STAR v2.7.9a [32] with the --quantMode GeneCounts parameter enabled to generate gene-level count matrices. Gene-level counts were normalized and analyzed for differential gene expression (DEG) using DESeq2 v1.36.0 [33]. Genes with fewer than five counts across at least three samples were excluded to reduce background noise. Differentially expressed genes were identified using an adjusted p-value threshold of 0.05 and a log2 fold change >1.2. Gene set enrichment analysis (GSEA) was performed using Hallmark Gene sets from the Molecular Signatures Database (MSigDB v7.5) (https://www.gsea-msigdb.org/gsea/msigdb/). Enriched gene sets were visualized and clustered in Cytoscape (v3.9.1) using the EnrichmentMap and AutoAnnotate plug-ins to generate functional network representations.

2.10. Statistical Analysis

All data were analyzed using GraphPad Prism 9 (GraphPad Software, Inc., San Diego, CA). Data are expressed as mean ± SE. Experiments were repeated three times, with triplicate samples for each. An unpaired 2-tailed Student’s t-test and Bonferroni post-hoc testing were used to assess statistical significance. Unless otherwise stated, *P<0.05, **P<0.01, ***P<0.001, and ****P<0.0001 values were considered significant; ns, not significant.

3. Results

3.1. Prom1 is Expressed in Mouse RPE In Situ

We recently showed Prom1 mRNA localization in mouse RPE in situ using RNAscope and Prom1 protein localization by immunogold electron microscopy [27]. To confirm Prom1 protein expression in mRPE, we performed Immunohistochemistry (IHC) on mouse RPE flat mounts. Our studies show Prom1 staining (green) alongside ZO-1 (red), a tight junction marker in mouse RPE flat mounts. At 40X magnification, Prom1 was predominantly localized in the RPE cytoplasm, with enrichment in nuclear and perinuclear regions and partial association with apical junctions marked by ZO-1 (Figure 1A). At 40X zoom, this pattern was more pronounced, with Prom1 puncta evident near cell borders, cytoplasm, and concentrated around nuclei. The merged images show partial co-localization of Prom1 with ZO-1, which supports its weak association with apical domains (Figure 1A). The presence of Prom1 signal in mRPE cytoplasmic and nuclear compartments suggests potential roles in trafficking or signaling beyond membrane organization. Consistent with these observations, Prom1 was detected (green) in the RPE layer in situ in mouse retinal sections, evidenced by 20X, 40X, and 63X confocal images, confirming its expression outside photoreceptors and providing protein-level evidence of its presence and subcellular organization (Figure 1B-C). The 3D orthogonal reconstruction shows that the Prom1 protein is present within the RPE monolayer, spatially distinct from photoreceptor outer segments. The labeling pattern in the XZ/YZ plane appears punctate cytoplasmic with apical enrichment, consistent with intracellular localization in RPE rather than a junctional or membrane-restricted signal (Figure 1B-C). Western blotting analysis of mRPE cells cultured in vitro confirmed robust Prom1 expression in wild-type mRPE and its absence in Prom1-knockout (KO) samples (using CRISPR/Cas9) (Figure 1D). At the same time, qPCR demonstrated significant Prom1 transcript reduction in Prom1-KO cells (Figure 1E). Together, these results establish Prom1 as a cytoplasmic and junction-associated protein in mRPE, supporting its potential role in maintaining RPE homeostasis.

3.2. Bulk RNA-Sequencing of WT and Prom1-KO mRPE Cells

To investigate the molecular consequences of Prom1 loss in retinal pigment epithelium (RPE), we performed bulk RNA sequencing on wild-type (WT) and Prom1-knockout (KO) mouse RPE cells. Sequencing reads were normalized for differences in sequencing depth and filtered to remove lowly or non-expressed genes, ensuring that downstream analyses focused on biologically relevant transcripts. Genes with fewer than five counts in at least three samples were excluded to minimize noise from sporadically expressed features. Raw counts, normalized counts, and post-filtering values were visualized on a log2 + 1 scale to include zero-count features (Figure 2). As expected, increasing sequencing depth improved feature capture, consistent with coverage principles for bulk RNA-seq. After normalization and filtering, WT and KO samples exhibited comparable distributions, confirming effective correction for sequencing depth and removal of low-abundance genes (Figure 2). This approach provided a robust dataset for differential expression analysis and pathway enrichment, enabling the identification of Prom1-dependent signaling networks.

Principal component analysis (PCA) was performed to visualize variance across WT and Prom1-KO mRPE transcriptomes. Uncorrected PCA revealed clustering by sequencing batch, indicating a strong batch effect (Figure 3A). After applying batch correction with limma, PC1 accounted for 92% of the variance, while PC2 accounted for the residual variability of unknown origin. Importantly, batch correction improved sample clustering by genotype, confirming that biological differences rather than technical artifacts drive the major variance component (Figure 3B). Two Prom1-KO samples (8995-SB-3 and 9262-SB-6) remained outliers after correction, suggesting intrinsic biological heterogeneity or technical anomalies (Figure 3B). These findings validate the need for batch adjustment and confirm that genotype is the dominant source of variance in the dataset. Following the removal of two outlier Prom1-KO samples (8995-SB-3 and 9262-SB-6), PCA demonstrated improved clustering by genotype (Figure 3C). PC1 accounted for 96% of the total variance, indicating that the primary source of variability was the experimental condition rather than technical factors. WT and Prom1-KO samples separated clearly along PC1, confirming robust transcriptional differences between groups. PC2 explained only 3% of the variance, and did not correspond to batch or condition, suggesting minimal residual confounding (Figure 3C). These findings validate the dataset’s data quality and support its suitability for downstream differential expression and pathway analyses.

3.3. Differential Expression and Gene Set Enrichment Analyses

Differential expression analysis revealed a pronounced transcriptional shift, including 15 upregulated and downregulated genes (Figure 4A). The volcano plot summarizing DESeq2 data for WT and Prom1-KO mRPE cells shows a cluster of high-magnitude, high-significance hits on both tails, with prominent upregulated genes with a positive log2 fold change (red), including Gremlin-1 (Grem1), an endogenous BMP antagonist that induces EMT in fetal RPE cells [34]; Serpine 2, a serpin with neurotropic and anti-angiogenic activities secreted by the RPE into the interphotoreceptor matrix toward the neural retina [35]; Interleukin-receptor-like 1(Il1r1), an activator of the complement alternative pathway in RPE cells in macular degeneration [36]; Interleukin 33 (IL33), a key regulator of inflammation and RPE atrophy in AMD [37]; and retinoic acid-induced 14 (Rai14), which is expressed in the RPE and has been shown to promote mTOR-mediated inflammation [38,39] (Figure 4A). The down-regulated genes with a negative log2 fold change (blue) include actin-binding LIM protein family member 1 (Ablim1), which directly binds F-actin to serve as a scaffold for signaling modules of the actin cytoskeleton and governs the formation of dense cortical actin meshwork to prevent mechanical tension-induced blebbing during hTERT-RPE1 migration [40,41]; IGF binding protein 2 (IGFBP2), which plays a significant role in modulating IGF-1 activity in the RPE and its loss induces an early reactive RPE phenotype [42,43]; Bone morphogenetic protein-3 (BMP3), its reduced expression leads to fibrosis; and Osteoglycin (OGN), a proteoglycan with strong expression in the Bruch’s membrane and choroid [44] (Figure 4A). We used gene set enrichment analysis (GSEA) to rank gene lists and identify enriched pathways (Figure 4B). Enrichment scores were calculated by hypergeometric tests and normalized to the size of the gene set. A positive normalized gene enrichment score (NES) indicates that the gene set is upregulated in the experimental group. Preliminary analysis of hallmark gene sets from the Molecular Signature Database shows significant upregulation of hallmark signatures for cell cycle transcription factors (E2F and MYC targets, G2M checkpoint), mTORC1 signaling, unfolded protein response (UPR), reactive oxygen species (ROS) pathway, TNFA signaling via NF-kappaB, DNA repair, and oxidative phosphorylation (Figure 4B). Downregulated pathways of interest include apical junction, bile acid metabolism, and EMT pathways (Figure 4B). Together, the DEG landscape and Hallmark Pathway analysis indicate a shift toward RPE innate immune/stress-response remodeling accompanied by dampening of mitochondrial energy metabolism and cell-cycle progression, providing pathway-level context for the gene-wise changes observed due to the loss of Prom1.

3.4. Heatmap Analysis and Biological Context

The heatmap of the top differentially expressed 100 genes (P-value sorted) shows strong clustering of biological replicates, confirming data reproducibility and condition-specific transcriptional signatures (Figure 5). Prom1-KO samples (9262-SB-5, 9262-SB-4, and 8995-SB-2) cluster distinctly from WT samples, reflecting robust genotype-driven expression differences (Figure 5). Several genes with known or emerging roles in RPE physiology and stress response are prominently represented. For example, MertK, essential for photoreceptor outer segment phagocytosis, is reduced in Prom1-KO mRPE compared to WT, suggesting impaired RPE clearance and homeostasis [45]. Cldn2, involved in tight junction integrity in murine RPE, is downregulated in Prom1-KO mRPE, indicating potential disruption of barrier properties [46]. Tgf-beta receptor 2 (Tgfbr2) maintains RPE barrier integrity, and its reduction suggests loss of homeostatic signaling [47]. Caveolin-1 (Cav1), a scaffolding protein in RPE caveolae that regulates vesicular trafficking and endocytosis, is upregulated, suggesting altered lipid raft dynamics that amplify stress signaling [48]. Map3k5 activates JNK and MAPK pathways under oxidative stress, and its upregulation in Prom1-KO mRPE suggests heightened stress signaling and apoptosis susceptibility [49]. Metabolic regulators such as SLC38A1 and SLC7A11 (cysteine/glutamate transporter) are upregulated in Prom1-KO mRPE, reflecting shifts in energy and redox balance, aligning with suppression of oxidative phosphorylation and lipid metabolism [50]. Upregulation of Il33 and downregulation of Igfbp2 further support activation of innate immunity/inflammatory and downregulation of metabolic signaling, echoing TNFα/NF-κB and IL6–JAK–STAT3 pathway enrichment. Collectively, these gene-level changes support the interpretation that Prom1 loss drives a transition from homeostatic RPE function toward a stress- and inflammatory-related phenotype with metabolic compromise.

3.5. Gene Network Map in Prom1-KO versus WT mRPE

Network mapping of Prom1-KO versus WT mouse RPE transcriptomes identified five highly connected hallmark clusters—MYC targets, E2F targets, G2M checkpoint, and mTORC1 signaling —each reflecting coordinated programs of RPE stress, metabolic reprogramming, and degeneration (Figure 6). The mTORC1 cluster, including Serp1, Map2k3, Slc2a1, and Slc7a11, indicates enhanced nutrient-sensing and stress signaling consistent with mTORC1 hyperactivation and impaired autophagic flux. Within the G2M checkpoint cluster, Aurka, Ccnb2, Plk1, and Slc38a1 are included, indicating altered amino acid transport and mitotic checkpoint activation. The MYC and E2F target cluster includes Bub1b, Cdk1, Npm1, Mcm5, Mcm7, Hmgb3, and Hmgb2, which are associated with glycolytic shift, chromatin remodeling, and proliferative stress (Figure 6). Collectively, these cluster-specific changes reinforce a model where Prom1 loss drives inflammatory signaling, metabolic stress, and impaired autophagy, converging on pathways implicated in RPE dysfunction and retinal disease.

3.6. Validation of Transcriptomic Data

We showed earlier that the loss of Prom1 in mRPE cells leads to EMT [28]. In contrast, our transcriptomic data in this study suggest reduced activation of the Hallmark EMT pathway in Prom1-KO mRPE cells (Figure 4B). To validate mRPE EMT dynamics, we performed gene-level interrogation as a biologically relevant approach. A volcano plot of the Hallmark EMT pathway revealed significant upregulation of EMT-associated genes, including Grem1, Pcolce2, Cxcl2, CD44, and Serpine2, while IGFBP2 and POSTN were notably downregulated (Figure 7A). qPCR confirmed IGFBP2 and POSTN gene downregulation (Figure 7B-C) and elevated expression of Grem1 gene in Prom1-deficient RPE cells (KO22 and KO26.4) (Figure D). Western blot confirmed increased Grem1 protein in Prom1-KO cells; however, densitometric analysis showed a statistically significant increase only in KO26.4, while KO22 exhibited a non-significant (ns) trend (Figure 7E). These findings suggest that Prom1 loss upregulates Grem1 transcription, but protein levels may be influenced by post-transcriptional regulation or by the specific genome region disrupted within the Prom1 gene in each CRISPR-edited Prom1-KO clone (Figure 7E). Although GSEA did not reveal clear enrichment for EMT-related pathways, targeted analysis of individual gene expression changes provides evidence for EMT-like alterations, which is consistent with our earlier observations showing EMT in Prom1-KO mRPE cells [28]. The downregulation of extracellular matrix remodeling (POSTN) and the upregulation of an EMT-promoting gene (Grem1), coupled with the suppression of epithelial-supporting factors (IGFBP2), suggest that Prom1 deficiency may initiate a partial or context-specific EMT program that is not fully captured by canonical pathway-level analyses.

To define how Prom1 loss perturbs RPE function, we examined the DEGs highlighted in the volcano plot across oxidative stress/reactive oxygen species (ROS), unfolded protein response (UPR), phagocytosis, and TNF/NF-kappaB pathways (Figure 8A). Among ROS-related genes, Egfr was significantly downregulated, consistent with prior work showing that oxidative stress reduces Egfr/Erk-Akt signaling and impairs RPE viability; conversely, Egfr-mediated antioxidant rescue restores RPE survival [51,52]. Fibulin-5 (Fbln5) was also decreased; Fbln5 is an ECM protein that limits integrin-driven ROS generation and modulates redox tone – its loss is linked to heightened oxidative stress in vivo, suggesting reduced ECM-based ROS control in the RPE microenvironment [53]. Pax2 downregulation is noteworthy given its established role as redundant with Pax6 in specifying RPE fate during development; reduced Pax2 is therefore compatible with impaired RPE identity/homeostasis under stress [54]. Finally, PINK1 gene expression was significantly reduced, suggesting defective mitophagy in RPE; PINK1 loss drives mitochondrial ROS-Nrf2 signaling, EMT-like changes, and structural abnormalities, all of which can cause RPE degeneration [55]. qPCR confirmed PINK1 was significantly reduced in Prom1-KO mRPE groups, indicating impaired mitochondrial quality control due to loss of Prom1 (Figure 8B).

In contrast, several oxidative-response genes were upregulated (Figure 8A). Nqo1, a canonical Nrf2 target that blunts redox cycling, was increased, consistent with a compensatory antioxidant response to elevated ROS [56]. Map3k5 (ASK1), a redox-sensitive Map3k that activates JNK/p38 under oxidative stress, was elevated, aligning with pro-apoptotic and pro-inflammatory stress signaling in stressed epithelia [57]. Hgf was increased; Hgf signaling in RPE promotes proliferation/motility and disrupts tight/adherens junctions via MET/Akt, indicating junctional remodeling and RPE barrier compromise under stress [58,59]. Ripk3 was also upregulated; RIPK1/3 activity has been implicated in RPE injury by promoting mitochondrial DNA release, AIM2 inflammasome activation, and inflammatory cell death, supporting a necroptosis-prone state in diseased RPE [60]. Sod3 gene levels increased as well; SOD3 protects the ECM from superoxide, and its modulation strongly influences retinal function in vivo, consistent with an extracellular antioxidant countermeasure to oxidative burden [61]. UPR-markers such as Qrich1, a transcriptional effector driving proteostasis under ER-stress, and Ptpn2, which protect epithelia from ER-stress-induced death, were upregulated, suggesting activation of UPR signaling in stressed mRPE [62,63]. Finally, the TNF/NF-kB pathway component was increased in Prom1-KO mRPE. Traf1 is an NF-kB-inducible adapter that modulates inflammatory and pro-survival signaling and is a classic footprint of active TNF/NF-kB circuits [64]. Together, these changes map a coordinated phenotype in Prom1-deficient mRPE – elevated oxidative and ER stress, impaired mitochondrial quality control, junctional/ECM remodeling, and TNF-driven inflammation.

The volcano plot revealed downregulation of the PINK1 gene in Prom1-KO cells (Figure 8A). qPCR confirmed a significant reduction in PINK1 gene expression, indicating impaired mitochondrial quality control and mitophagy signaling in Prom1-deficient mRPE cells (Figure 8B). Of note, the MerTk gene was downregulated (Figure 8A). Consistent with transcriptomic findings, western blotting showed a significant reduction of the MerTK protein in both Prom1-KO22 and -KO26.4 mRPE cells, and qPCR confirmed reduced MerTK gene expression in these cells (Figure 8C-D). Because MerTk is indispensable for daily outer-segment clearance by RPE, reduced MerTK expression provides a mechanistic link to impaired phagocytic flux and secondary photoreceptor stress [65].

qPCR studies confirmed the upregulation of the Slc7a11 gene, involved in RPE oxidative stress response (Figure 8E) [66]; and downregulation of genes, including cytoskeletal remodeling, Ablim1 (Figure 8F) [40]; extracellular matrix organization, OGN (Figure 8G); and retinoic acid synthesis, Aldh1a1 (Figure 8H) [67], in Prom1-KO mRPE cells. These findings show that Prom1 deficiency disrupts multiple pathways critical for RPE homeostasis, including mitochondrial turnover, phagocytosis, oxidative stress regulation, and ECM integrity—mechanisms that may underlie retinal degeneration in Prom1-associated disease.

4. Discussion

Prom1 has long been recognized as a primary driver of cell-autonomous photoreceptor pathology, yet its role in the retinal pigment epithelium (RPE) has largely remained unexplored. Building on our prior observations of Prom1 transcripts in mRPE, this study provides definitive evidence of Prom1 protein expression in mRPE both in situ and in cultured cells, confirming its presence beyond photoreceptors [27]. Although Prom1 expression in the RPE monolayer is quantitatively lower than in the photoreceptor-rich retina, our data demonstrate that Prom1 is not merely incidental; instead, it functions as a major regulator of RPE homeostasis in a cell-autonomous manner [27,28]. Through transcriptomic analysis, we show that Prom1 loss initiates a cascade of stress, metabolic, and structural disruptions that converge on pathways central to RPE integrity. These findings reposition Prom1 from a photoreceptor-centric protein to a critical node in RPE biology, with implications for both inherited retinal dystrophies and atrophic age-related macular degeneration. Loss of Prom1 reprograms mouse RPE cells toward a degenerative state marked by activation of stress and inflammatory molecules (TNF/NF-kB; Map3k5, RIPK3), failure of mitochondrial quality control (reduced PINK1); reduced phagocytic capacity (reduced MerTk), junctional/ECM remodeling compatible with HGF signaling; and broader metabolic rewiring (including mTORC1 activation, Slc7a11 induction, and suppression of bile acid metabolism) together with transcriptional features of incomplete EMT – an array of changes that closely mirror mechanisms implicated in both IRDs and aAMD [68,69].

Loss of Prom1 disrupts multiple RPE survival programs through interconnected mechanisms. Increased mTORC1 signaling in Prom1-KO RPE correlates with impaired RPE autophagy and lysosomal activity, which drive EMT and result in AMD-like pathology [28,70]. Reduced PINK1 compromises mitophagy, allowing damaged mitochondria to accumulate, which elevates ROS and activates retrograde stress signaling that promotes EMT-like changes [71]. In parallel, downregulation of MERTK impairs phagocytosis of photoreceptor outer segments, a process essential for retinal homeostasis; its failure accelerates photoreceptor stress and degeneration [65]. These vulnerabilities are compounded by upregulation of ASK1/MAP3K5 and RIPK3, which amplify oxidative and ER stress responses, driving JNK/p38-mediated apoptosis and necroptosis [60]. Junctional instability observed in Prom1-KO cells is consistent with HGF/MET signaling, which is known to dismantle tight junctions, increase RPE motility, and activate AKT survival pathways [59]. Finally, induction of SLC7A11 reflects an adaptive antioxidant response to counter ferroptosis, underscoring a shift toward stress tolerance rather than restoration of homeostasis [72]. Together, these changes form a coherent degenerative program linking mitochondrial failure, impaired clearance, inflammatory signaling, and barrier breakdown -- hallmarks of RPE dysfunction in AMD and IRDs.

Notably, the Hallmark pathway suppression of bile acid metabolism has important implications: beyond lipid emulsification, bile acids and their receptors FXR and TGR5 act as hormone-like immunometabolic cues that constrain inflammation, support epithelial barrier integrity, and tune energy metabolism across tissues [73]. In the eye, TUDCA/UDCA mitigate oxidative and ER stress, dampen pro-inflammatory cytokines, stabilize barrier properties, and in RPE promote autophagy-mediated cytoprotection, suggesting that loss of bile acid signaling removes an endogenous pro-homeostatic brake in Prom1-deficient RPE and offers an actionable therapeutic axis [74,75]. These metabolite-signaling considerations are consistent with broader lipid metabolic dysfunction observed in multi-omic RPE models and AMD patient serum, reinforcing the clinical relevance of our metabolic signatures [76].

The global architecture of the Prom1-KO transcriptome aligns with previous RPE transcriptomics in human AMD [68,69]. Combined meta-analyses and single-cell atlases highlight inflammatory and proteostatic remodeling, as well as metabolic rewiring (including TNFα/NF-κB activation, UPR/proteostasis engagement, and changes in oxidative metabolism) in AMD regions, paralleling the pathway enrichments observed in Prom1-KO mRPE cells. Longitudinal RNA-seq of RPE during outer-segment processing documents coordinated autophagy, lysosome, and proteostasis programs, aligning with our UPR signatures and supporting a central role for proteostasis failure in disease progression [77]. Genetic activation of mTOR in mouse RPE drives EMT, disruption of autophagy, metabolic imbalance, and aAMD-like pathology [70]. This mirrors the mTORC1/EMT transcriptomic changes in Prom1-KO mRPE, supporting the idea that metabolism and cell-state transitions are closely linked in degenerating RPE [70]. Furthermore, our combination of EMT-effector induction with overall suppression of the EMT hallmark aligns with the emerging concept of incomplete or partial EMT in AMD and RPE injury—loss of epithelial features with partial mesenchymal acquisition rather than full transdifferentiation—reconciling transcriptomic complexity with pathology [78].

5. Conclusions

In summary, our study places RPE-localized Prom1 at a central point connecting vesicular dynamics, mitochondrial quality control, phagocytic and barrier functions, and immunometabolism. The Prom1-dependent network we identify not only aligns with human RPE transcriptomics in AMD and IRDs but also provides a targeted set of mechanistic, actionable entry points for maintaining RPE health across genetic and age-related retinal degeneration.