Comprehensive transcriptomic analysis reveals canonical and novel pathways modulated by nanoceria in mammalian retinal degeneration

Abstract

Retinal neurodegenerative diseases such as Age-related Macular Degeneration (AMD) and Retinitis Pigmentosa cause irreversible vision loss due to the limited regenerative capacity of the mammalian retina. Cerium oxide nanoparticles (nanoceria) are emerging therapeutics against oxidative stress and inflammation, major drivers of photoreceptor degeneration, and have demonstrated morphological and functional neuroprotection in preclinical models. However, the genome-wide transcriptional mechanisms underlying these effects remain incompletely characterized. We performed retinal transcriptomic analysis in a rat AMD model induced by intense light and treated intravitreally with nanoceria. Six groups were analyzed: control, light damage, vehicle, nanoceria, vehicle + light damage, and nanoceria + light damage. Light damage activated inflammatory and apoptotic programs, with upregulation of cytokines (Tnf, Il6, Il1b, Ccl2) and downregulation of photoreceptor genes (Rho, Pde6a/b, Gnat1). Nanoceria treatment counteracted these effects, suppressing pro-inflammatory mediators, restoring antioxidative genes (Nfe2l2, Gclc, Sod2), and enhancing neuroprotective factors (Bdnf, Cntf, Ngf). Pathway analyses revealed inhibition of TNF/NF-κB/IL-17 signaling and activation of PI3K-Akt, JAK-STAT, and neurotrophin pathways. Unexpectedly, nanoceria also modulated amino acid and insulin metabolism (Ass1, Cps1, Insr, Irs1, Slc2a4) and reactivated transcription factors (Ascl1, Sox2, Notch1) typically silent in adult retina. Our findings highlight nanoceria as a multifunctional therapeutic that mitigates retinal degeneration by coordinating oxidative, inflammatory, and regenerative responses. Together with prior morphological and functional validations, these results support the translational potential of nanoceria for treating retinal neurodegenerative diseases.

Introduction

Retinal neurodegenerative diseases represent a leading cause of blindness worldwide, with age-related macular degeneration (AMD) and retinitis pigmentosa (RP) affecting millions of patients¹. Unlike lower vertebrates such as zebrafish that possess remarkable retinal regenerative capacity through Müller glia reprogramming, the adult mammalian retina has minimal intrinsic regenerative potential, making photoreceptors loss irreversible². The pathophysiology of retinal degeneration involves complex interactions between oxidative stress, chronic inflammation, and metabolic dysfunction^3,4. Acute high-intensity light damage (LD) in rodents reproduces key features of photoreceptor degeneration observed in AMD, driven by oxidative stress, inflammatory activation, and apoptosis⁵.

Based on our knowledge, cerium oxide nanoparticles (nanoceria) have emerged as promising therapeutics due to their unique redox-switching properties between Ce³⁺/Ce⁴⁺ oxidation states, conferring catalytic, self-regenerating free-radical scavenging activity^6,7. Previous studies demonstrated that intravitreal nanoceria reduce the ROS accumulation, temper microgliosis, preserve photoreceptors, and protect the retinal pigment epithelium in LD paradigms^3,4,6,7,8,9. However, the genome-wide molecular programs engaged by nanoceria in the degenerating mammalian retina remain incompletely defined.

Here we perform bulk RNA-seq in a rat LD model with and without intravitreal nanoceria to systematically map nanoceria-responsive gene networks. We test whether nanoceria not only blunt canonical oxidative-stress and inflammatory pathways but also modulate metabolic and regeneration-linked circuits relevant to photoreceptor survival. By integrating differential expression and pathway enrichment analyses, we provide a transcriptomic framework for nanoceria’s mode of action and nominate testable targets for neuroprotective therapy in AMD-like injury⁵. We hypothesize that nanoceria not only mitigate inflammatory and oxidative injury but also create a permissive transcriptional environment that may enable partial reactivation of latent regenerative programs in the adult retina.

Results

Light damage activates inflammatory/apoptotic programs and suppresses photoreceptor identity

Principal component analysis (PCA) of log₂-transformed TPM-normalized RNA-seq data revealed clear transcriptomic segregation among all six experimental groups (Fig. 1A, B, Supplementary Table S1). The first principal component (PC1), explaining 25.4% of total variance, primarily captured the transcriptional impact of acute light damage, with light-damaged samples (LD, VEH + LD, NANO + LD) clustering distinctly from non-damaged groups (CTRL, VEH, NANO) along this axis (one-way ANOVA: F = 3.26, p = 0.0075). The second principal component (PC2, explaining 14.0% of variance) reflected the treatment effect of nanoceria, effectively separating nanoceria-treated samples (NANO, NANO + LD) from vehicle-treated or untreated controls (F = 4.48, p = 0.0007). The third principal component (PC3, 9.2% variance) captured residual biological variability and further differentiated treatment responses, particularly within the light-damaged cohort (F = 1.81, p = 0.113).

Fig. 1

Principal component analysis (PCA) of transcriptomic profiles across experimental groups. (A) PCA scatter plot (PC1 vs. PC2) of variance-stabilized RNA-seq data (TPM-normalized) reveals clear transcriptomic segregation among the six experimental groups: control (CTRL, black circles), light damage (LD, red circles), vehicle injection (VEH, gray circles), nanoceria treatment (NANO, blue circles), vehicle + light damage (VEH + LD, orange circles), and nanoceria + light damage (NANO + LD, green circles). Each point represents an individual biological replicate (n = 5–7 per group). Shaded ellipses represent 95% confidence intervals for each group. PC1 (explaining 25.4% of total variance) primarily captures the effect of light damage, separating light-damaged samples (LD, VEH + LD, NANO + LD) from non-damaged groups (CTRL, VEH, NANO). PC2 (explaining 14.0% of variance) reflects the nanoceria treatment effect, distinguishing nanoceria-treated samples (NANO, NANO + LD) from their vehicle-treated or untreated counterparts. (B) PCA scatter plot (PC1 vs. PC3) provides additional resolution of transcriptomic relationships. PC3 (explaining 9.2% of variance) captures residual biological variability and further separates treatment responses within experimental groups. (C–E) Violin plots showing the distribution of sample scores along PC1, PC2, and PC3, respectively, highlighting tight clustering of biological replicates within groups and statistically significant separation between experimental conditions (one-way ANOVA: PC1, F = 3.26, p = 0.0075; PC2, F = 4.48, p = 0.0007; PC3, F = 1.81, p = 0.113). PCA was performed on variance-stabilized transformation (log2-transformed TPM values) of normalized count data. Individual data points (jitter) represent biological replicates. Statistical significance assessed by one-way ANOVA with Tukey’s post-hoc test. Total samples: n = 34 (CTRL = 5, VEH = 5, LD = 7, VEH + LD = 6, NANO = 5, NANO + LD = 6).

Importantly, biological replicates within each experimental group clustered tightly (n = 5–7 per group; total n = 34), indicating high technical reproducibility and robust biological consistency across samples (Fig. 1C–E). Vehicle injection alone (VEH) did not substantially alter the transcriptomic profile relative to untreated controls (CTRL), as evidenced by their close proximity and overlapping confidence ellipses in PCA space. Notably, the NANO + LD group occupied an intermediate position between LD and control groups along PC1, suggesting that nanoceria treatment partially reversed the transcriptional signature induced by light damage. These global transcriptomic patterns are fully consistent with subsequent differential expression analyses, which demonstrate that nanoceria modulate both baseline retinal homeostasis (NANO vs. CTRL) and injury-induced transcriptional programs (NANO + LD vs. LD).

Light damage activates inflammatory/apoptotic programs and suppresses photoreceptor identity

Acute high‑intensity light exposure (LD) produced a broad transcriptional shift versus controls (CTRL), with hundreds of DEGs (e.g., ~ 920 at FDR < 0.05; 666 up/254 down). Upregulated genes encompassed cytokines/chemokines (Tnf, Il6, Il1b, Ccl2/Ccl3/Ccl4), gliosis/microglial markers (Gfap, Aif1), and stress/apoptosis mediators (Bax, Casp3), while photoreceptor and visual‑cycle transcripts were prominently reduced (Rho, Pde6a/b, Gnat1, Cnga1; transcriptional regulators Crx, Nrl, Nr2e3). Enrichment analyses highlighted TNF/NF‑κB and p53/apoptosis among upregulated pathways, with strong depletion of phototransduction and retinol/visual‑cycle terms. These data establish the canonical oxidative‑inflammatory injury signature and loss of photoreceptor identity in LD (Fig. 2; Table 1; Fig. S1).

Fig. 2

Light damage (LD) activates inflammatory/apoptotic programs and suppresses phototransduction. (A) Volcano plot for LD vs. CTRL (FDR < 0.05) highlighting upregulation of cytokines/chemokines (e.g., Tnf, Il6, Il1b, Ccl2/Ccl3/Ccl4) and gliosis markers, with strong downregulation of photoreceptor/visual‑cycle transcripts (e.g., Rho, Pde6a/b, Gnat1, Cnga1; Crx, Nrl, Nr2e3). Log₂ fold-change values are apeGLM-shrunken estimates. Red/blue points indicate significantly upregulated/downregulated genes, respectively. (B) Heatmap of top DEGs illustrates clear CTRL vs. LD segregation. (C, D) Summary enrichment (GO/KEGG) indicates upregulation of inflammatory response/TNF–NF‑κB/apoptosis and downregulation of phototransduction/retinol metabolism; full enrichment in Fig. S1. Differential expression determined with DESeq2, using Wald test and Benjamini–Hochberg correction. Significance defined at adjusted p (FDR) < 0.05. Sample size: n = 6 biological replicates per group.

Table 1 Consolidated “Top DEGs” for the two pivotal contrasts. Top up‑ and down‑regulated genes for LD vs. CTRL and NANO‑LD vs. LD (ranked by |log₂FC|; FDR < 0.05). Complete DEG lists for all five contrasts are provided in supplementary tables S2–S6.

Intravitreal vehicle injection exerts minimal transcriptomic impact

Vehicle (VEH) alone induced only minor changes relative to CTRL (dozens of DEGs with small effect sizes and no significant GO/KEGG categories after correction), indicating the injection procedure/vehicle is largely transcriptionally inert at the assayed time point. Similarly, VEH‑LD vs. LD showed only subtle and inconsistent differences, confirming that injection does not confound the LD injury signature (Supplementary Fig. S2–S3; Supplementary Tables S4–S5).

In uninjured retina, nanoceria prime a cytoprotective mitochondrial/antioxidant state

Nanoceria treatment in healthy eyes (NANO vs. CTRL) reprogrammed the transcriptome toward oxidative‑stress resilience and metabolic competence. Antioxidant master regulator Nfe2l2 (Nrf2) and canonical targets (Hmox1, Nqo1, Gclc, Sod2, Gpx family, Prdx1) were induced, alongside mitochondrial/OXPHOS and bioenergetic genes (Idh3a/b, Cpt1a, Acat2, Ppargc1a). In parallel, inflammatory and apoptotic transcripts (Tnf, Il1b, Nlrp3, Casp3, Bax) were reduced. These data support a “preventive” effect whereby nanoceria elevate antioxidant capacity and mitochondrial organization while dampening basal inflammatory tone (Fig. S4–S5; Supplementary Table S5).

In injured retina, nanoceria reverse degeneration signatures and enhance survival/trophic programs

In LD eyes, nanoceria (NANO‑LD) broadly opposed the injury signature relative to LD alone. Damage‑induced inflammatory mediators and gliosis markers (e.g., Tnf, Il6, Il1b, Ccl2, Nfkbia, Gfap, Aif1) were strongly reduced, while photoreceptor transcripts (Rho, Pde6b) were preserved/partially restored. Nanoceria concomitantly induced neuroprotective and trophic factors (Bdnf, Cntf, Ngf, Sirt1) and regeneration‑linked regulators (Ascl1, Sox2, Notch1, Wnt2b), indicating transcriptional reversal of injury-induced pathways. Pathway analyses showed suppression of TNF/NF‑κB/IL‑17 and activation of PI3K–Akt, JAK–STAT, neurotrophin signaling, with strengthened oxidative phosphorylation and mitochondrial organization (Fig. 3; Table 2; Supplementary Fig. S6, Supplementary Table S6). Direct comparison against vehicle in LD (NANO‑LD vs. VEH‑LD) confirmed these nanoceria‑specific effects (Fig. 4). Together, these results indicate that nanoceria both quell inflammatory/apoptotic cascades and engage pro‑survival networks that promote transcriptomic recovery.

Fig. 3

In injured retina, nanoceria oppose LD‑induced signatures and enhance pro‑survival networks. (A) Volcano for NANO‑LD vs. LD (FDR < 0.05) shows downregulation of inflammatory mediators (Tnf, Il6, Il1b, Ccl2, Nfkbia) and gliosis markers (Gfap, Aif1) with preservation/recovery of photoreceptor transcripts (Rho, Pde6b). Log₂ fold-change values are apeGLM-shrunken estimates. Red/blue points indicate significantly upregulated/downregulated genes, respectively. (B) Heatmap of representative DEGs. (C, D) Pathway summaries indicate suppression of TNF/NF‑κB/IL‑17 and activation of PI3K–Akt, JAK–STAT, neurotrophin signaling, alongside strengthened oxidative phosphorylation and mitochondrial organization (full enrichment: Fig. S6). Differential expression determined with DESeq2, using Wald test and Benjamini–Hochberg correction. Significance defined at adjusted p (FDR) < 0.05. Sample size: n = 6 biological replicates per group.

Table 2 Summary of differential expression and enriched pathways across Comparisons. For each comparison, the table reports the total number of significant differentially expressed genes (DEGs) with the count of up- (↑) and down-regulated (↓) genes, alongside the key significantly enriched gene ontology (GO) terms and KEGG pathways.

Fig. 4

Nanoceria vs. vehicle in LD directly demonstrates nanoceria‑specific effects. Compact pathway bar/dot plot for NANO‑LD vs. VEH‑LD emphasizes stronger suppression of inflammatory pathways and enhanced activation of survival/metabolic programs under nanoceria compared with saline. Full volcano/heatmap and enrichment details are in Fig. S7. Differential expression determined with DESeq2, using Wald test and Benjamini–Hochberg correction. Significance defined at adjusted p (FDR) < 0.05. Sample size: n = 6 biological replicates per group.

Nanoceria uncover “non‑canonical” axes: amino‑acid/urea‑cycle and insulin/glucose signaling

Beyond antioxidant and inflammatory pathways, nanoceria modulated amino‑acid metabolism/urea‑cycle genes (e.g., Ass1, Cps1, Otc), reversing their LD‑associated repression and elevating expression toward or above baseline, consistent with enhanced nitrogen handling and anaplerotic support. In parallel, nanoceria increased insulin/IGF signaling components (Insr, Irs1) and the insulin‑responsive glucose transporter Slc2a4 (Glut4), suggesting improved glucose utilization and coupling to PI3K–Akt survival signaling in stressed retina. These axes were not appreciably engaged by vehicle and represent novel, testable mechanisms by which nanoceria may stabilize retinal bioenergetics (Supplementary Fig. S7; Supplementary Table S7).

qRT‑PCR validates key signatures

Targeted qRT‑PCR corroborated RNA‑seq trends: Il6 rose after LD and decreased with nanoceria, whereas Sod2 and Nfe2l2 increased with LD and remained elevated under nanoceria, consistent with reinforced antioxidant defense. RNA‑seq vs. qPCR values showed strong concordance (e.g., Pearson’s r ≈ 0.90–0.97 across genes; p < 0.05), supporting dataset robustness (Fig. 5).

Fig. 5

Correlation between RNA-Seq and qRT-PCR expression for Il6, Sod2, and Nrf2. Scatter plots show a strong concordance between transcriptomic and qPCR data across treatment groups. Expression values from RNA-Seq (TPM) are plotted against corresponding qPCR fold changes. All genes demonstrated robust and statistically significant linear correlations (R² > 0.80; p < 0.05), confirming the consistency of differential expression profiles between the two methodologies. n = 6 animals per group, reactions in technical triplicate.

Discussions

The current transcriptomic analysis reveals that nanoceria suppress inflammatory and oxidative stress pathways while enhancing antioxidant, metabolic, and regenerative gene networks. These findings extend previous histological and functional evidence^{8,9,10,11,12,13} by providing mechanistic insight into nanoceria’s multifaceted neuroprotective action.

Beyond their catalytic antioxidant properties, our findings suggest that nanoceria modulate the retinal microenvironment at the transcriptional level in a way that suppresses injury while simultaneously enabling reactivation of genes typically associated with developmental or regenerative processes.

Rather than acting as passive ROS scavengers^14,15, nanoceria appear to engage upstream transcriptional programs that recondition the degenerating retina and shift the balance toward resilience and repair. Their long-lasting antioxidant activity and favorable safety profile in preclinical models^12,16,17 further support their potential as next-generation, multifunctional therapeutics for retinal disease.

Among the most unexpected and clinically relevant findings, we observed that nanoceria modulate several non-canonical metabolic pathways, including amino acid catabolism, the urea cycle, and insulin/glucose signaling. Genes such as Ass1 and Cps1, which are involved in nitrogen detoxification and arginine metabolism, were suppressed by light damage and restored by nanoceria, suggesting that metabolic reprogramming is a critical aspect of retinal stress adaptation^18,19.

In parallel, nanoceria upregulated Insr, Irs1, and Slc2a4, pointing to reactivation of insulin signaling—a pathway increasingly recognized in retinal degeneration²⁰. These metabolic effects may improve glucose utilization and support bioenergetic demands through the PI3K–Akt axis while alleviating inflammatory-metabolic stress crosstalk^21,22.

These transcriptional changes have not previously been associated with nanoceria and position them as modulators of cellular metabolism in addition to their well-known redox properties. Addressing metabolic dysfunction alongside oxidative stress may represent a promising multi-targeted strategy for retinal neuroprotection.

Collectively, these properties reinforce the therapeutic value of nanoceria as a next-generation, long-acting neuroprotective nanomedicine for retinal disease.

A principal mechanism by which nanoceria exert neuroprotection in the retina is through attenuation of oxidative stress and suppression of inflammatory signaling cascades. Retinal degenerative diseases—including age-related macular degeneration (AMD), retinitis pigmentosa (RP), and diabetic retinopathy—are driven by chronic oxidative damage and persistent low-grade inflammation that promote photoreceptor apoptosis and retinal pigment epithelium (RPE) dysfunction²³. Nanoceria, via redox-active Ce³⁺/Ce⁴⁺ cycling, act as self-regenerating antioxidants capable of neutralizing superoxide, hydrogen peroxide, and peroxynitrite^24,25. Consistent with our earlier immunohistochemical findings, which demonstrated reduced Iba1-positive microglial activation and decreased inflammatory markers^8,10, the present transcriptomic analysis reveals that nanoceria markedly downregulates proinflammatory genes (Tnf, Il1b, Il6, Ccl2) in light-damaged retinas, further supporting its anti-inflammatory mode of action. At the signaling level, we observed decreased expression of Tlr4 and Myd88, consistent with inhibition of the Toll-like receptor (TLR)–mediated NF-κB pathway. This pathway is a well-established trigger of inflammatory transcriptional programs and microglial activation in retinal degeneration. Supporting this, previous studies demonstrated that cerium oxide nanoparticles attenuate TLR4/NF-κB activation in brain inflammation models, reducing cytokine production and microgliosis²⁶. In retinal tissue specifically, Badia et al. ²⁷ demonstrated that CeO₂-NPs reduce microglial activation and suppress TNF-α expression following oxidative stress. These anti-inflammatory effects appear selective and non-immunosuppressive. Instead of silencing broad immune responses, nanoceria specifically interrupt pathological inflammatory cycles while preserving beneficial immune functions such as phagocytosis and matrix remodeling. For instance, in our dataset, Timd4 and Ctse—genes involved in apoptotic cell clearance and lysosomal remodeling—were upregulated, indicating enhanced homeostatic responses. Nanoceria’s antioxidant action further complements this immunomodulatory profile. By lowering ROS levels, nanoceria suppress upstream NF-κB activators such as oxidized mitochondrial DNA and lipid peroxidation products²⁴. We also observed activation of Nrf2 signaling (Nfe2l2), a master regulator of redox balance. Nanoceria-treated retinas exhibited significant upregulation of canonical Nrf2 target genes: Hmox1, Gclc, Nqo1, and Sod2, aligning with prior evidence of Nrf2-mediated neuroprotection by cerium nanoparticles^16,25. This coordinated suppression of TLR4–MyD88–NF-κB and enhancement of Nrf2-driven antioxidant defense supports a dual mechanism by which nanoceria modulate inflammation and oxidative injury in the retina. Such combined action may be especially advantageous in retinal diseases where these pathways are interlinked in a self-amplifying loop. Our transcriptomic data reinforce this model. Chemokine genes Ccl2, Ccl3, and Cxcl10, which are known drivers of neuroinflammation in light-damaged retinas, were significantly downregulated by nanoceria. Notably, Ccl2 is a key mediator of microglial recruitment and photoreceptor apoptosis²⁸, while Ccl3 and Cxcl10 participate in chemokine-driven neurodegeneration²⁹. Proinflammatory cytokines Il6 and Tnf followed expected dynamics: induced after photic injury and suppressed by nanoceria. These cytokines are well-characterized effectors of gliosis and retinal damage. The glial stress marker Gfap, upregulated during Müller cell activation, was also reduced by nanoceria, indicating mitigation of reactive gliosis²⁸. Conversely, the neuroprotective factor Fgf2 remained elevated in nanoceria-treated groups, in line with its role in photoreceptor survival³⁰. Downregulation of Aif1 (IBA-1), a microglial activation marker, further supports nanoceria’s capacity to suppress innate immune activation³. These transcriptomic signatures strongly support the view that nanoceria modulate the oxidative stress–inflammation axis in a precise and therapeutically favorable manner, reinforcing their potential as neuroprotectants in retinal disease.

Nanoceria neuroprotective activity extends beyond antioxidative effects and includes direct modulation of apoptosis, stress response, and pro-survival pathways. Multiple transcriptomic and qPCR studies have demonstrated that nanoceria treatment downregulates pro-apoptotic genes such as Casp3, Bax, and Il1b, while promoting expression of pro-survival factors including Bcl2, Sirt1, and neurotrophic genes like Bdnf, Cntf, and Ngf³¹. These changes correlate with delayed photoreceptor cell death and preservation of retinal architecture observed in histological analyses. In the present study, we observed significant suppression of Casp3, Il1b, Tnf, and Nlrp3 following nanoceria treatment in the LD model, both in direct comparisons and pathway enrichment. These effects were accompanied by the activation of Nrf2 (Nfe2l2) and its downstream targets Hmox1, Gclc, and Nqo1, key components of the antioxidant defense system. The role of Nrf2 in protecting against oxidative and proteotoxic stress in neurons and glia has been firmly established, and it represents a viable therapeutic target in AMD and Parkinson’s disease³². Importantly, nanoceria-mediated activation of Nrf2 has also been associated with improved mitochondrial quality control. In dopaminergic neurons, CeO₂ nanoparticles reduce mitochondrial fragmentation by inhibiting DRP1 hyperactivation and peroxynitrite formation³¹. Consistently, prior ultrastructural analyses revealed preserved mitochondrial morphology in nanoceria-treated retinas, contrasting with the disrupted, swollen mitochondria observed in untreated LD eyes^33,34. Transcriptomic analysis further revealed upregulation of stress-responsive transcription factors Atf3 and Sirt1, both associated with cellular recovery following oxidative and metabolic insults. Sirt1, in particular, is known to support mitochondrial biogenesis, energy homeostasis, and antioxidant defense by interacting with FoxO and PGC-1α pathways³². The observed increase in Sirt1 expression in treated animals may synergize with Nrf2 to promote neuroprotection, especially in the high-energy-demand environment of photoreceptors. An unexpected finding was the upregulation of regeneration-associated transcription factors such as Ascl1, Sox2, and Notch1 in nanoceria-treated retinas. These factors are typically quiescent in the adult mammalian retina but are central to the Müller glia reprogramming network in regenerative species like zebrafish³⁵. Their re-expression in this context suggests that nanoceria may induce a transcriptional “priming” effect, rendering the retina more responsive to potential regenerative cues. Although histological analysis did not reveal active regeneration or Müller cell proliferation, the transcriptional activation of these genes is noteworthy and may offer a window of opportunity for synergistic therapeutic interventions (e.g., gene therapy or trophic factor delivery). Taken together, our results underscore nanoceria’s multifaceted action on survival and stress-response mechanisms. This includes suppression of inflammation and apoptosis, enhancement of antioxidant and mitochondrial defense via Nrf2 and Sirt1, and activation of latent regenerative programs. These concerted transcriptomic changes shift the degenerating retina from a trajectory of degeneration toward stabilization and potentially recovery.

One of the most compelling emerging mechanisms by which nanoceria exert neuroprotective effects is their ability to stabilize autophagy and preserve mitochondrial function—two closely linked processes essential for retinal health. Chronic oxidative stress, a hallmark of retinal degenerative diseases such as AMD and RP, disrupts autophagic flux and promotes mitochondrial dysfunction, leading to bioenergetic collapse and cell death, particularly in metabolically demanding tissues like the retina and RPE³⁶. Recent studies have highlighted nanoceria role in preserving autophagic homeostasis. For example, in a light-damage-induced retinal degeneration model, intravitreal nanoceria treatment prevented abnormal accumulation of LC3B-II and p62 proteins in RPE cells, indicative of restored autophagic flux. Notably, nanoceria blocked aberrant nuclear localization of LC3B—a feature associated with impaired autophagosome maturation and epithelial-mesenchymal transition (EMT) in degenerating RPE³⁷. By maintaining normal autophagy, nanoceria preserved RPE integrity and prevented cell death. Complementing these molecular findings, ultrastructural evidence from previous studies demonstrates that nanoceria-treated retinas maintain mitochondrial integrity after injury³⁸. These observations align with prior work in neurodegenerative models where nanoceria localized to mitochondria and prevented excessive fission by inhibiting DRP1 hyperactivation, a key step in the mitochondrial apoptotic cascade³⁹. Mechanistically, nanoceria appear to exert these effects through multiple routes:

• (1) Direct ROS scavenging, reducing mitochondrial oxidative burden;

• (2) Suppression of inflammation that can secondarily impair mitophagy;

• (3) Stabilization of Nrf2 signaling, which indirectly supports mitochondrial biogenesis and autophagy-related gene expression (e.g., PINK1, Parkin)⁴⁰.

Indeed, transcriptomic data from our model show that genes involved in mitochondrial dynamics—Opa1, Mfn1, and Tomm20—were better preserved in nanoceria-treated retinas, suggesting maintenance of mitochondrial fusion and protein import systems. This preservation of mitochondrial and lysosomal health is significant. In AMD, defective autophagy and impaired mitophagy contribute to RPE degeneration, lipofuscin accumulation, and drusen formation. Therapies that restore autophagic flux and mitochondrial clearance—such as rapamycin or metformin—have shown benefit in animal models, but often with systemic side effects. Nanoceria, by contrast, provide a local, long-acting alternative that appears to accomplish similar outcomes through redox modulation and transcriptional reprogramming without systemic toxicity. In sum, the impact of nanoceria on autophagy and mitochondrial maintenance adds a crucial dimension to their neuroprotective repertoire. By targeting fundamental organelle-level processes that govern cell survival, nanoceria go beyond antioxidant action and engage with the core bioenergetic and proteostatic machinery of retinal cells. This capacity to stabilize intracellular quality control systems, particularly under stress, likely underlies the long-term structural and functional protection observed in treated retinas.

Taken together, our data demonstrate that nanoceria exert a multi-dimensional effect on the degenerating retina by attenuating inflammatory mediators (e.g., Tnf, Il1b, Il6, Ccl2), preserving neuroprotective and metabolic gene expression (Bdnf, Cntf, Sirt1, Nfe2l2), and reactivating transcriptional programs typically silent in the adult retina. In addition to the expected antioxidant and anti-inflammatory responses, we observed novel transcriptional signatures, including upregulation of amino acid metabolism and urea cycle genes (Cps1, Ass1, Otc), insulin/IGF signaling components (Insr, Irs1, Slc2a4), and regenerative transcription factors (Ascl1, Sox2, Notch1).

These results suggest that nanoceria not only suppress degeneration-associated cascades but also promote a reparative transcriptional landscape. This neuroprotective effect may create a permissive environment for regeneration. A potential limitation of our study is the use of bulk RNA-sequencing, which provides an averaged view of gene expression across all retinal layers and cell types. However, this approach was chosen based on previous literature demonstrating that LD-induced injury is largely confined to the outer retina, particularly the ONL and RPE, and engages a limited number of key cellular players—most notably photoreceptors, Müller glia, and microglia. In this context, bulk RNA-seq is well-suited to detect broad transcriptional programs relevant to damage and repair and has enabled the identification of nanoceria-responsive pathways involved in oxidative stress, metabolism, and regeneration. These results now provide a strong rationale for future single-cell or spatial transcriptomic studies specifically targeting these responsive cell populations.

Conclusions and future directions

Our transcriptomic analysis provides compelling evidence that cerium oxide nanoparticles (nanoceria) elicit a multifaceted neuroprotective response in the retina under both homeostatic and degenerative conditions. In a rat model of light-induced retinal degeneration, nanoceria administration not only attenuated classical markers of oxidative stress and inflammation but also preserved metabolic signaling pathways, restored components of cellular homeostasis, and induced elements of developmental and regenerative transcriptional programs. At the molecular level, nanoceria-treated retinas exhibited reduced expression of genes encoding pro-inflammatory mediators such as TNF-α, IL-6, and CCL2, alongside downregulation of the TLR4–MyD88–NF-κB signaling axis, which is often implicated in retinal and neuroinflammatory pathologies. Concomitantly, we observed the activation of survival-promoting pathways such as PI3K/Akt, neurotrophin signaling, and antioxidant networks under control of Nfe2l2 (Nrf2). Unexpectedly, nanoceria also restored the expression of metabolic regulator genes like Insr and Glut4, and modulated amino acid and nitrogen metabolism (e.g., ASS1 and CPS1), pointing to novel roles in sustaining retinal bioenergetics and limiting toxic metabolic byproducts. Furthermore, nanoceria triggered the expression of genes typically associated with retinal regeneration in lower vertebrates (e.g., Notch1, Sox2, Ascl1), suggesting that even in the adult mammalian retina, nanoceria may partially reawaken latent regenerative programs. While true neurogenesis was not observed histologically, the transcriptional reprogramming indicates a shift toward a reparative retinal environment—a finding with major implications for regenerative therapeutics. From a translational perspective, these results reinforce the broad-spectrum potential of nanoceria as a next-generation therapy for retinal degenerative diseases such as AMD and RP. Their intrinsic regenerative redox cycling enables persistent antioxidant activity, and recent advances in delivery platforms (e.g., eye drops, sustained-release hydrogels, intranasal administration) support their clinical feasibility. Importantly, no overt toxicity has been observed in long-term in vivo models, and multiple independent studies across neurodegenerative systems (retina, brain) confirm their efficacy and safety. In conclusion, nanoceria represent a promising disease-modifying agent capable of intervening at multiple levels of retinal degeneration—quenching oxidative and inflammatory cascades, preserving cellular metabolism, and possibly priming repair pathways. This sequential interplay—first quenching stress and inflammation, then initiating regenerative transcriptional activity—highlights a two-stage therapeutic paradigm that merits further investigation. These insights expand our understanding of nanoceria’s mode of action and lay the groundwork for future preclinical and clinical studies aimed at establishing nanoceria-based interventions as a therapeutic approach in ophthalmology and beyond.

Methods

Animal model and ethical approval

Adult albino Sprague–Dawley rats (≈ 2–3 months old, males) were used for all experiments. Rats were born and raised under dim cyclic lighting (12 h light/12 h dark) at constant temperature (~ 22 °C) and humidity, with diurnal light ≈ 5 lx. Food (standard pellet diet) and water were provided ad libitum. Animals were randomly assigned to experimental groups (n = 6 per group) including untreated controls, light-damaged (LD) only, and nanoceria-treated + LD groups (additional vehicle-injected controls were included for injection procedures). All experiments adhered to the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research and the European Directive 2010/63/EU and were approved by the Italian Ministry of Health (authorization no. 763/2020-PR). The authors complied with the ARRIVE guidelines. All efforts were made to minimize animal suffering and to reduce the number of animals used in accordance with the 3Rs principle. To better clarify the experimental design, the seven experimental groups are summarized in Table 3.

Table 3 Experimental groups and treatment conditions. Summary of the six groups (CTRL, LD, VEH, NANO, VEH‑LD, NANO‑LD; n = 6 each) and interventions (light exposure, intravitreal saline, intravitreal nanoceria).

Then, to induce retinal degeneration and recapitulate the features of AMD, albino rats were exposed to high intensity light for 24 h. The detailed experimental procedures were already reported in previous papers^{3,4,8,9,37,41}.

Nanoceria Preparation and intravitreal injection

Cerium oxide nanoparticles (nanoceria, CeO₂-NPs) were synthesized and characterized as described previously⁴². For in vivo administration, nanoceria were sterile-filtered (0.22 μm) and prepared at 1 mM in sterile 0.9% NaCl. Under anesthesia (intraperitoneal ketamine 100 mg/kg + xylazine 10 mg/kg), rats received a bilateral intravitreal injection of 2 µL of the nanoceria suspension (1 mM) in each eye using a Hamilton microsyringe with a fine gauge needle⁴³. In the vehicle control group, 2 µL of sterile saline was similarly injected. Injections were performed as previously described^8,9,10.

RNA extraction and quality control

Seven days post-light exposure (or at corresponding time points for control animals), rats were euthanized by CO₂ inhalation and eyes were immediately enucleated. Retinas were rapidly dissected from each eye on ice. Total RNA was extracted from retinal tissue using TRIzol™ Reagent (Invitrogen, ThermoFisher Scientific) according to the manufacturer’s instructions. RNA quantification was performed using the Qubit RNA HS Assay Kit (Thermo Fisher Scientific), and the RNA integrity was assessed on an Agilent 2100 Bioanalyzer (Agilent Technologies); only high-quality RNA samples with RNA Integrity Number (RIN) ≥ 7.0 were used for library preparation.

Library preparation and sequencing

For each sample, 1 µg of total RNA was processed for poly(A) + mRNA enrichment. Messenger RNA was isolated using oligo(dT) magnetic beads and converted to strand-specific cDNA. Strand-specific sequencing libraries were prepared using the Watchmaker RNA Library Prep Kit (Twist Bioscience) following the manufacturer’s protocol. Adapter ligation, PCR amplification (with unique dual indexing for multiplexing), and library purification were performed as outlined in the Watchmaker workflow. Post-amplification, library quality and insert size distributions were assessed using a Fragment analyzer, confirming fragment sizes of 250–350 bp. Equimolar pools of indexed libraries were sequenced on an Illumina NovaSeq 6000 system to generate 150 bp paired-end reads, with each retina library sequenced to a mean depth of ~ 80 million reads. Raw sequencing data (FASTQ files) underwent quality control checks, including evaluation of base quality scores, GC content, and adapter contamination; all samples met predefined thresholds for downstream transcriptomic analysis.

RNA-seq data analysis

Primary analysis of RNA-seq data was carried out using a composite bioinformatics pipeline combining a canonical open-source workflow with CLC Genomics Workbench (v25.0.0) analyses. In the canonical pipeline, raw reads were first processed using fastp (v0.20) for adapter trimming and quality filtering⁴⁴. Fastp performed per-read quality pruning and removed Illumina adapter sequences, yielding high-quality clean reads for each sample. Clean reads were then aligned to the Rattus norvegicus reference genome (e.g., Rnor_6.0) using the STAR aligner (v2.7.11) with default parameters for two-pass spliced read mapping⁴⁵. Mapping quality was high, with the majority of reads uniquely aligned to exonic regions of the genome. Gene-level read counts were computed using featureCounts (v2.0) from the Subread package, summarizing aligned reads per gene based on Ensembl gene annotation⁴⁶. The resulting raw count matrix was imported into R (v4.1) for downstream analysis. DESeq2 (v1.48.0) was used to normalize gene counts and identify differentially expressed (DE) genes between experimental groups⁴⁷. The DESeq2 model accounted for biological replicate variability, and shrinkage estimators were applied to dispersions and fold changes. Shrinkage estimation of log₂ fold-change (log₂FC) values was performed using the apeGLM method as implemented in the lfcShrink() function of DESeq2 ⁴⁸. This adaptive shrinkage approach reduces variance in log₂FC estimates for low-count genes while preserving effect sizes for highly expressed genes, thereby improving the reliability of differential expression calls. Volcano plots (Figs. 2A and 3A) were generated using the shrunken log₂FC values, not the unshrunken Maximum Likelihood Estimates (MLE). The wide dispersion of fold-change values observed in these plots reflects the broad transcriptomic response to light damage and nanoceria treatment, with inflammatory genes showing strong upregulation and phototransduction genes exhibiting marked downregulation. Genes with a false discovery rate (FDR)-adjusted p-value < 0.05 were considered significantly differentially expressed. No fold-change cutoff was applied; all genes passing the FDR threshold were retained. Unsupervised sample clustering via principal component analysis (PCA) was performed to evaluate sample relationships and detect potential outliers. PCA results were visualized with violin plots, showing the distribution and density of sample loadings across the first three principal components (PC1–PC3), thereby highlighting transcriptional heterogeneity and treatment-specific clustering. Results from the canonical pipeline were then compared with those obtained via CLC Genomics Workbench, improving the reliability and consistency of DE findings. DE gene profiles were visualized with volcano plots and heatmaps showing the top regulated genes. To interpret transcriptomic alterations functionally, up- and down-regulated gene sets were analyzed for gene ontology (GO) and KEGG pathway enrichment using the clusterProfiler R package⁴⁹. Enrichment testing used an FDR cutoff of 0.05. Significant GO terms (Biological Process, Molecular Function, Cellular Component) and KEGG pathways were visualized using dot plots and enrichment maps, providing insights into nanoceria-modulated biological processes in retinal degeneration. All data visualizations (PCA, heatmaps, volcano plots, enrichment plots) were generated in R using ggplot2 ⁵⁰ or clusterProfiler-native plotting functions.

Real-Time PCR validation of differentially expressed genes

To validate the RNA-seq results, we performed quantitative real-time PCR (qRT-PCR) analysis on a selection of differentially expressed genes (DEGs) identified across experimental groups. Specific primer pairs were designed for key upregulated and downregulated genes, including Nfe2l2 (Nrf2), IL6, and Sod2 (Table 4). Total RNA was extracted from retinal tissue samples using TRIzol Reagent, followed by cDNA synthesis. qRT-PCR reactions were carried out using CFX96 Real Time System (Bio-Rad) and IQ SYBR Green Supermix (Bio-Rad) according to the manufacturer instructions. Each reaction was performed in technical triplicate. Relative expression levels were normalized to the housekeeping gene for actine, and the ΔΔCt method was used to quantify fold changes.

Table 4 Primer sequences used for qRT-PCR. Forward and reverse primer sequences used for quantitative real-time PCR (qRT-PCR) validation of selected differentially expressed genes (Nrf2, IL6, Sod2) and the housekeeping gene (Actin).

Statistical analysis

For non-transcriptomic measurements (e.g., nanoparticle characterization, histological quantification), data are presented as mean ± standard error of the mean (SEM), unless otherwise specified. Statistical analyses were conducted using GraphPad Prism⁵¹ and R ⁵².Group comparisons were performed using appropriate parametric tests. For comparisons involving more than two groups, one-way analysis of variance (ANOVA) followed by Tukey’s post hoc test was applied. For pairwise group comparisons, unpaired two-tailed Student’s t-tests were used. A significance threshold of p < 0.05 was employed for all statistical tests unless otherwise stated. Statistical significance for differential gene expression in RNA-seq data was assessed as described in Sect. 2.5, based on adjusted p-values calculated using the Benjamini-Hochberg false discovery rate (FDR) method.