Introduction

Vision in vertebrates is mediated by the eyes, complex sensory organs that originate from three distinct embryonic tissues: the neuroectoderm, surface ectoderm, and mesenchyme1. The retina, the neural component of the eye, develops from the embryonic neural tube, similarly to the brain and spinal cord. During embryogenesis, the diencephalic region of the neural tube, which expresses key eye-field transcription factors, expands bilaterally to form the optic vesicles, which give rise to the neuronal structures of the eye2,3.

The vertebrate retina comprises seven major cell types, organised into three nuclear layers separated by two synaptic (plexiform) layers2. Among these are rods and cones, the photoreceptors responsible for capturing light and initiating signal transduction. Retinal ganglion cells project axons to brain targets via the optic nerve. Bipolar, horizontal, and amacrine cells refine and relay visual information to ganglion cells. Müller glial cells support retinal architecture and homeostasis. In addition, astrocytes, microglia, and cells of the retinal vasculature contribute to the overall retinal microenvironment2. Vascular endothelial cells and pericytes, for example, form the inner blood–retinal barrier, regulate capillary blood flow, and contribute to angiogenesis4,5.

The field of eye proteomics, particularly retinal proteomics, is rapidly evolving, offering crucial insights into the protein dynamics underlying ocular function and disease6,7,8,9. Proteomic analyses have significantly contributed to elucidating molecular mechanisms involved in the pathogenesis, diagnosis, and treatment of ocular disorders, as exemplified by the launch of the Human Eye Proteome Project in 20137. Furthermore, due to its shared cellular and molecular features with the central nervous system, the retina serves as a valuable model for studying neurodegenerative conditions such as Alzheimer’s disease9.

Environmental stressors are now recognised as important risk factors for ocular diseases, and air pollution, toxic metals, and climate-related factors have been linked to conditions such as cataracts, age-related macular degeneration, uveitis, and dry eye syndrome10. So, retinal proteomics also plays a key role in evaluating ocular toxicity by assessing drug-induced and environmental impacts on retinal health, changes that can ultimately result in vision loss11.

The zebrafish (Danio rerio), a freshwater cyprinid, has emerged as a valuable model organism owing to its ease of maintenance, low cost, fully annotated genome, high genetic homology with humans, and transparent embryos12 that facilitate detailed molecular analyses, such as tissue-specific proteomic profiling. These features, combined with advanced colour vision and a strong reliance on sight, make zebrafish particularly suitable for ophthalmological studies1. Their retinas are anatomically and functionally comparable to those of humans, including a similar abundance of cone photoreceptors. Eye development is also rapid, with the retina reaching functional maturity by around 72 h post-fertilisation (hpf), when robust visual responsiveness becomes evident1. Notably, light sensitivity emerges much earlier: non-visual photoreceptive cells can detect light from 5 hpf, the pineal organ is photosensitive by 24 hpf, and retinal photoreception develops between 48 and 72 hpf13.

Zebrafish are increasingly employed as models for ocular toxicity, with studies conducted in both larvae and adults14. A wide range of contaminants including endocrine disruptors, flame retardants, pharmaceuticals, microplastics, agrochemicals, and heavy metals has been shown to impair retinal structure, induce oxidative stress, and alter the expression of phototransduction-related genes14,15. Recent studies further highlight the ecological sensitivity of zebrafish ocular tissues to environmental stressors. Polystyrene nanoplastics accumulate in the developing eye and rapidly induce oxidative stress, apoptosis, and dysregulation of genes involved in retinal development and antioxidant defence, ultimately impairing visual performance in larvae16. Similarly, environmentally relevant nitrate and nitrite exposure disrupts retinal morphology, alters the expression of key apoptotic and antioxidant markers, and diminishes visually guided behaviours in adults17. Together, these findings indicate that the zebrafish retina acts as a highly responsive environmental sensor capable of detecting and integrating diverse chemical challenges. This ocular sensitivity provides strong ecological and translational justification for investigating whether other environmental variables, such as sustained temperature changes, may likewise affect retinal homeostasis.

A growing body of research shows that zebrafish naturally inhabit environments characterised by substantial thermal variability, making them both ecologically and physiologically suitable models for studying the consequences of temperature fluctuations on biological systems. These findings highlight the sensitivity of the visual system to environmental challenges and support the rationale for exploring temperature-induced effects on retinal proteostasis.

Our research group has been investigating the effects of temperature on the zebrafish nervous system for several years. Zebrafish were selected as a model for thermal stress studies due to their broad thermal tolerance, with viable temperature ranges spanning from 6 °C to 43 °C, and tolerable limits between 4.5 °C and 42 °C18,19,20. This remarkable thermal adaptability reflects the natural habitat of zebrafish in South Asia, where daily temperature fluctuations of approximately 5 °C and seasonal shift from around 6 °C in winter to over 38 °C in summer are common21,22. Zebrafish typically inhabit brooks, lakes, ponds, and rice fields23,24, where water temperatures generally range from 16.5 °C to 34.0°C25. These ecological characteristics make zebrafish an ideal model for investigating the biological effects of prolonged exposure to environmentally relevant elevated and low temperatures25.

Our previous studies have shown that both acute (4-day) and chronic (21-day) exposure of adult zebrafish to elevated (34 °C) and low (18 °C) temperatures result in significant changes in brain proteome composition and behaviour, compared to fish maintained at 26 °C26,27,28,29. These temperatures were selected based on the work of Vergauwen et al.30, as they fall within the species’ viable thermal range and represent ecologically realistic cold and warm conditions. The control temperature of 26 °C was chosen to reflect the zebrafish thermal preferendum18.

The zebrafish brain response to altered temperature involves a complex network of molecular events, including regulation of neurotrophins such as Brain-Derived Neurotrophic Factor (BDNF)31 and alterations in the brain lipidome32. Together, these findings support the neurotoxic effects of prolonged exposure to both warm and cold temperatures in adult zebrafish, highlighting the broader consequences of thermal stress on neural homeostasis and function33,34.

In this context, the zebrafish eye offers a compelling model for studying temperature-dependent proteomic changes. Unlike many mammalian models, adult zebrafish rely heavily on vision and possess cone-rich retinas that closely resemble those of humans, thereby enabling translationally relevant investigations.

Although the zebrafish is gaining prominence in vision research, direct investigations into how chronic thermal stress affects the adult eye proteome remain limited.

This study aims to fill this knowledge gap by examining whether prolonged exposure to constant but ecologically relevant elevated (34 °C), control (26 °C), and low (18 °C) temperatures modifies the proteomic landscape of the zebrafish eye.

Building on previous findings showing differential effects of warm and cold temperatures on the zebrafish brain at the levels of proteome, lipidome, and behaviour, we hypothesise that similar thermal conditions may also impact retinal neural function. These effects could influence visual capacity, with distinct outcomes under low and high temperature regimes. Such findings may provide a basis for understanding how thermal stress influences eye health in ectotherms, while offering insights into conserved molecular mechanisms of ocular neurotoxicity with potential relevance for human disease.

Results and discussion

Proteomic effects of constant thermal conditions in the zebrafish eye

The eyes of adult zebrafish maintained for 21 days at 34 °C, 26 °C, and 18 °C were analysed using a shotgun proteomic approach to assess the impact of the exposure to elevated and low temperatures on protein expression. A total of 685 proteins were detected at 34 °C, 683 at 26 °C, and 691 at 18 °C (Fig. 1, Supplementary Tables S1–S3). As shown in the Venn diagram (Fig. 1A), only a limited number of proteins were exclusively expressed at each temperature (15 at 34 °C, 9 at 26 °C, and 12 at 18 °C). However, a Principal Component Analysis (PCA) demonstrated clear clustering by temperature, with the most pronounced separation observed at 34 °C (Fig. 1B).

Fig. 1
figure 1

Proteomic analysis of the zebrafish eye at three different temperatures. A shotgun proteomics approach was applied to the eye of adult zebrafish acclimated for 21 days to three thermal conditions: elevated temperature (34 °C), control temperature (26 °C), and low temperature (18 °C). (A) Venn diagram comparing the proteomic datasets across the three temperature conditions. PrG: proteins identified by LC–MS/MS analysis. (B) PCA of the three proteomes with confidence ellipses. (C) Venn diagram of all the proteins identified in pairwise comparisons: 34 °C vs. 26 °C, 18 °C vs. 26 °C, and 34 °C vs. 18 °C.

Full size image

Multivariate analysis of variance (MANOVA), applied as an exploratory analysis to identify general proteomic patterns among groups, indicates differences across temperature groups (Wilks’ λ = 0.0748, F(1740, 5236) = 4.134, p < 1 × 10⁻³⁰⁰; Pillai’s trace = 1.51, F(1740, 5241) = 3.055, p = 1.41 × 10⁻²⁰⁷) in keeping with the proteomic results.

Specific analyses were carried out by comparing 34 °C vs. 26° C and 18 °C vs. 26 °C, with 26 °C kept as the control temperature, as well as by comparing 34 °C vs. 18 °C (Fig. 1C). The lists of proteins differently expressed are provided in the supplementary material (Supplementary Tables S4–S6) reporting the common proteins in the pairwise comparisons whose differential expression is statistically significant according to Student’s test and also the proteins exclusively expressed at each of the temperatures tested. The proteins listed in Tables S4-S6 were further analysed using an Ingenuity Pathway Analysis (IPA) (Figs. 2 and 3) and Cytoscape (Fig. 4) to identify potential patterns and networks affected by elevated and low temperature.

Fig. 2
figure 2

Canonical pathway analysis of differentially expressed proteins at elevated (34 °C) and low (18 °C) temperatures compared to the control (26 °C) using IPA. Proteins differentially expressed in the 34 °C vs. 26 °C (A) and 18 °C vs. 26 °C (B) comparisons were analysed using IPA to identify all the most significantly altered canonical pathways. Proteins were considered differentially expressed if they were either uniquely present in one condition or exhibited a statistically significant difference in expression (Student’s test FDR ≤ 0.05). For clarity, only a subset of pathways are labelled in each volcano plot. The Z-score indicates the predicted activation or inhibition state of a canonical pathway: with positive scores (orange) indicating activation and negative scores (blue) indicating inhibition. The legend provides details on symbol colours and sizes. Volcano plot of the canonical pathways in the 34 °C vs. 26 °C (A) and in 18 °C vs. 26 °C (B) comparison, with Z-scores plotted against -log(p-values).

Full size image
Fig. 3
figure 3

Bioinformatic analysis of differentially expressed proteins at high temperature (34 °C) compared to the control (26 °C) using IPA. (A) Merged network analysis combining the neurological network and the lipid metabolism network. Proteins identified in our dataset are coloured in green. The legend provides details on all other symbols. (B) Graphical summary providing a high-level overview of the major biological themes and most significant entities predicted by IPA for differentially expressed proteins in the 34 °C vs. 26 °C comparison.

Full size image
Fig. 4
figure 4

Bioinformatic analysis of differentially expressed proteins at elevated temperature (34 °C) compared to the control (26 °C) using Cytoscape. A bioinformatic analysis was conducted using CLUEGO software (Cytoscape release 3.8.2) to cluster enriched annotation groups of biological processes, pathways, and networks within the set of differentially expressed proteins in the 34 °C vs. 26 °C comparison. Proteins were considered differentially expressed if they were either uniquely present in one condition or exhibited a statistically significant difference in expression (Student’s test FDR ≤ 0.05). Functional grouping of these proteins was then performed using Fisher’s exact test (p-value ≤ 0.05, with a minimum of three counts per category). (A) Proteins that were upregulated or uniquely expressed at 34 °C. (B) Proteins that were downregulated at 34 °C or uniquely expressed at 26 °C.

Full size image

High temperature (34 °C vs. 26 °C)

At 34 °C, a complex pattern of molecular alterations emerged, affecting multiple pathways in a highly specific manner (Figs. 2A and 3A,B; Table 1).

Table 1 Top five most significantly altered pathways in the zebrafish eye at three different temperatures. An Ingenuity Pathway Analysis (IPA) was performed on the eye of adult zebrafish acclimated for 21 days to three thermal conditions: elevated temperature (34 °C), control temperature (26 °C), and low temperature (18 °C). The table lists the five most significantly altered canonical pathways identified by IPA in each pairwise comparison. Proteins were considered differentially expressed if they were either uniquely present in one condition or exhibited a statistically significant difference in expression (Student’s test FDR≤ ≤ 0.05). The significance value represents the probability that the association between a given dataset and a canonical pathway occurred by random chance. P-value = B-H adjusted p value.
Full size table

Heat-induced stress signalling and translational reprogramming

Protein kinase R (PKR), a ubiquitous intracellular stress sensor implicated in the heat-shock response and broader cellular stress signalling35,36,37, was activated together with the Eukaryotic Initiation Factor 2 (IF2) and sirtuin signalling, reflecting an adaptive attempt to maintain homeostasis (Figs. 2A and 3B).

Sirtuins, a conserved family of histone deacetylases, are pivotal regulators of cellular stress responses through protein acetylation38,39. Seven isoforms are strongly expressed in the zebrafish retina40, where they protect against oxidative and genotoxic stress via transcriptional regulation and protein deacetylation41,42,43,44,45. In particular, SIRT1 protects against NMDA-induced retinal damage when activated by resveratrol40. Notably, SIRT1 interacts with and deacetylates eIF2α, thereby constraining its phosphorylation46,47,48. Since eIF2α phosphorylation is central to the integrated stress response49, attenuating protein synthesis and promoting adaptation50, our finding of increased PKR signalling50 is consistent with activation of this axis. Importantly, PKR–eIF2α signalling has been linked to stress granule formation and neurodegeneration51,52. The concurrent up-regulation of translation initiation and elongation pathways suggests Integrated Stress Response-driven translational reprogramming53,54,55 (Fig. 2A).

Synaptic signalling and neuronal connectivity

The concomitant up-regulation of SNARE signalling and synaptogenesis may reflect an adaptive attempt to sustain neurotransmission and maintain photoreceptor–retinal circuitry under heat stress (Fig. 2A). Indeed, in rodents SNAP-25, a core SNARE complex protein, modulates retinal wave activity and retinogeniculate circuit refinement during development56. However, the retina’s exceptionally high metabolic demand suggests that persistent activation of these pathways is energetically burdensome and may lead to aberrant synaptic remodelling, ultimately compromising synaptic efficiency and retinal integrity57.

Mitochondrial impairment and metabolic shift

Mitochondrial-related processes—including oxidative phosphorylation, electron transport, ATP synthesis and heat production by uncoupling proteins, the TCA cycle, and mitochondria protein import—were consistently down-regulated, pointing to impaired mitochondrial activity (Fig. 2A; Table 1). Sirtuins modulation may also be involved as they are known regulators of mitochondrial homeostasis58, with SIRT3 in particular supporting oxidative phosphorylation and redox balance59,60,61. Causal network analysis of the proteins differentially expressed in the 34 °C vs. 26 °C (Supplementary Table S5) identified highly connected networks associated with retinal and neurological disease, lipid metabolism, and molecular transport. Merged networks are shown in Fig. 3A, while Fig. 3B summarises the principal processes altered at 34 °C. Cytoscape confirmed up-regulation of glycolysis, lipid biosynthesis, and hypoxia response and protein folding (Fig. 4A), alongside down-regulation of oxidative phosphorylation, and energy production (Fig. 4B). Thus, impaired mitochondrial function likely reduces energy availability, while increased glycolysis may act as a compensatory mechanism. Similar heat-induced metabolic shifts towards glycolysis have been reported in other systems62,63,64. Reduced expression of precursor-generating pathways further supports the idea of heat-driven metabolic reprogramming in ocular cells. In parallel, lipid biosynthetic processes were up-regulated, indicating altered lipid composition in ocular tissue. This aligns with lipidomic analyses of the zebrafish brain at 34 °C, which revealed major shifts in lipid content32, suggesting lipid metabolism is more strongly affected by heat than by cold exposure.

In the context of ocular physiology, suppression of oxidative phosphorylation is expected to have direct consequences for ATP-dependent visual processes. The vertebrate retina is one of the most metabolically demanding tissues in the body, with photoreceptors relying predominantly on mitochondrial ATP to sustain phototransduction, maintain ion gradients through Na⁺/K⁺-ATPases, and support the continuous renewal of outer-segment discs65,66. Mitochondria located in the ellipsoid region of photoreceptors are particularly critical for maintaining the high ATP flux required for dark current maintenance and cyclic nucleotide turnover67,68,69.

Similarly, synaptic transmission within the retina, especially at ribbon synapses, requires substantial ATP to fuel vesicle cycling, neurotransmitter release, and Ca²⁺ homeostasis70. A heat-induced reduction in mitochondrial energy production is therefore likely to compromise these essential processes, forcing retinal cells to rely more heavily on glycolysis. Although increased glycolytic activity may provide partial metabolic compensation, glycolysis alone is generally insufficient to meet the extremely high energetic demands of photoreceptors and retinal interneurons71.

This energetic imbalance may contribute to the modulation of retinal protein markers observed at 34 °C. Together, these considerations support the hypothesis that heat-induced mitochondrial dysfunction disrupts ATP availability, potentially affecting phototransduction efficiency, synaptic signalling, and overall retinal performance under elevated temperature.

Compromised phototransduction and visual processing

Proteins associated with visual perception and phototransduction were markedly down-regulated (Figs. 2A, 3B and 4B; Table 1; Supplementary Fig. S1), suggesting that prolonged heat exposure compromises retinal visual processing. This effect is likely attributable to mitochondrial impairment and the consequent reduction in energy availability induced by thermal stress. Strikingly, proteomic analyses in patients with Alzheimer’s disease have likewise revealed suppression of photoreceptor activity, oxidative phosphorylation, and mitochondrial function9. Such parallels strengthen the hypothesis that heat stress perturbs ocular physiology in a manner reminiscent of neurodegenerative pathology.

Defective proteostasis and neurodegeneration-like alterations

Proteomic analysis revealed an up-regulation of chaperone-mediated protein folding (Fig. 3A), together with a concomitant reduction in global protein folding capacity and in aggrephagy (Fig. 2A), a selective autophagy pathway responsible for the clearance of misfolded proteins. Impaired aggrephagy has been associated with reduced proteostatic capacity and the accumulation of protein aggregates in neurodegenerative contexts72,73. Indeed, defective aggrephagy is a recognised hallmark of proteinopathies such as Alzheimer’s and Huntington’s disease74,75.

A possible cellular scenario suggested by these results is that, although cells attempt to counteract proteotoxic stress by enhancing chaperone activity, their overall proteostatic machinery remains insufficient. In this context, misfolded proteins may be more prone to persist, while their clearance via selective autophagy is reduced. Such an imbalance in protein homeostasis could represent a critical mechanism by which heat stress affects ocular neurons.

From molecular changes to behavioural outcomes

Although the observed molecular alterations at 34 °C suggest impaired eye function, they do not directly demonstrate reduced vision. Behavioural changes previously reported in zebrafish exposed to 34 °C for 21 days, including increased boldness and reduced anxiety-like responses27,28, may reflect brain-driven processes but could also be influenced by impaired retinal input. Reduced phototransduction efficiency might compromise environmental perception and contribute to these behavioural outcomes. Our findings highlight variations in the ocular proteome, including proteins of lens, cornea, and retina, therefore further studies explicitly assessing retinal function and visual performance are warranted.

Low temperature (18 °C vs. 26 °C)

The five most affected pathways at 18 °C were translation elongation, translation initiation, lysine catabolism, rRNA processing, and mitochondrial dysfunction (Table 1).

Translational control and RNA processing

At 18 °C, the IPA revealed strong activation of translation and rRNA processing pathways (Fig. 2A; Table 1), consistent with the role of translational control in thermal adaptation76,77. Adjustments of ribosome activity and RNA pools are common in ectothermic organisms adapting to energetic constraints. PCA analysis (Fig. 1B) further confirmed a global reorganisation of protein expression under cold exposure.

Stress signalling and mitochondrial function

Unlike the response observed at 34 °C, sirtuin signalling was inhibited at low temperature (Fig. 2B; Table 1). Given the established roles of sirtuins in genome stability, metabolic adaptation, and mitochondrial stress responses78,79, as well as their documented contribution to thermal tolerance in marine ectotherms such as the blue mussels Mytilus galloprovincialis and Mytilus trossulus80, this down-regulation suggests that cold exposure demands fewer stress-protective mechanisms than heat stress.

At 18 °C temperature, the mitochondrial dysfunction pathway was inhibited exhibiting a negative Z-score (− 0.45), and no significant alterations in oxidative phosphorylation, electron transport, or ATP synthesis. This indicates that mitochondrial integrity is better preserved at low temperatures than under high-temperature stress. Such a response is consistent with the reduced metabolic rate characteristic of cold conditions and supports the view that mitochondrial plasticity underpins cold tolerance in fish81,82.

Global proteomic organisation

Unlike the 34 °C vs. 26 °C comparison, the IPA did not generate a graphical summary for the 18 °C vs. 26 °C dataset (Fig. 3B), likely due to too few high-confidence entities. This supports the view that heat drives the most extensive proteomic reorganisation, while cold elicits a subtler but distinct response. Cytoscape analysis corroborated this as no significant pathway associations were found at 18 °C (Fig. 4A,B), whereas 34 °C induced strong up-regulation of protein folding, glycolysis, lipid biosynthesis, and hypoxia response (Fig. 4A), along with down-regulation of visual perception, phototransduction, oxidative phosphorylation, and energy production (Fig. 4B).

Direct comparison of extreme conditions (34 °C vs. 18 °C)

The 34 °C vs. 26 °C and 18 °C vs. 26 °C comparisons revealed distinct effects of elevated and low temperatures on the zebrafish eye. Collectively, these findings show that chronic heat stress induces major metabolic and functional remodelling, whereas cold stress primarily results in subtler adjustments of protein synthesis and mitochondrial activity. The separation along PCA component 1 (Fig. 1B) further confirms that 34 °C has a stronger effect on ocular function and vision than 18 °C.

To further clarify the differences between heat- and cold-induced responses, a direct comparison of the two extreme conditions was performed (34 °C vs. 18 °C). The IPA identified proteins exclusively expressed at one temperature or significantly altered between the two (Supplementary Table S6).

At 34 °C, mitochondrial-related processes were markedly impaired, accompanied by strong inhibition of phototransduction and visual phototransduction pathways, as well as ATP synthesis, heat production by uncoupling proteins, and oxidative phosphorylation (Table 1; Supplementary Fig. S2A). These results suggest that heat exposure compromises mitochondrial efficiency and visual capacity more severely than cold exposure.

Conversely, several pathways were significantly upregulated at 34 °C relative to 18 °C, including sirtuin signalling, mitochondrial dysfunction, heat stress response, collagen biosynthesis, extracellular matrix organisation, and BAG2 signalling (Table 1; Supplementary Fig. S2A,B). This pattern mirrors that observed in the 34 °C vs. 26 °C comparison, reinforcing the notion that heat stress elicits robust adaptive responses, extensive mitochondrial impairment, and profound extracellular remodelling.

Overall, the 34 °C vs. 18 °C comparison underscores the sharper impact of heat relative to cold exposure on the zebrafish eye proteome. While cold conditions induce moderate reorganisation of protein synthesis and stress-related pathways, heat stress severely disrupts mitochondrial processes and visual phototransduction, posing a greater challenge to ocular function and cellular homeostasis. Importantly, comparing the two extremes enables a clear separation of heat- versus cold-specific effects, rather than merely assessing deviations from the control temperature (26 °C). Although such abrupt shifts are uncommon in natural habitats, they provide a critical framework for assessing the boundaries of thermal tolerance and for disentangling the relative contributions of cold and heat stress to ocular physiology.

Shared and distinct proteomic responses to thermal conditions in the zebrafish brain and eye

Although the present study focused on the zebrafish eye, we compared our proteomic findings with previously published brain data obtained under identical thermal conditions (34 °C, 26 °C, and 18 °C for 21 days)28. This qualitative analysis was not intended to imply a direct functional equivalence between the two tissues, but rather to explore whether shared molecular signatures of thermal sensitivity might emerge across neural structures of common embryonic origin. The retina derives from the neural tube and is composed of neurons and glial cells, thus representing a peripheral extension of the central nervous system2,83. On this basis, we hypothesised that both structures could display overlapping responses to thermal stress while retaining tissue-specific traits.

To address this, we re-analysed the published brain dataset28 applying the same IPA pipeline and summarising the five most significantly altered canonical pathways under each pairwise thermal comparison (Table 2). We then compared these results with those obtained for the eye (Table 1), focusing on recurring or divergent pathway-level responses between the two tissues. This qualitative comparison highlighted a set of common molecular alterations despite tissue-specific signatures.

Table 2 Top five most significantly altered pathways in the zebrafish brain at three different temperatures. the IPA was performed on the brain of adult zebrafish acclimated for 21 days to three thermal conditions: elevated temperature (34 °C), control temperature (26 °C), low high temperature (18 °C), based on previously published data27 data28. The table lists the five most significantly altered canonical pathways identified by a new IPA performed on the dataset from27 from28. Proteins were considered differentially expressed if they were either uniquely present in one condition or exhibited a statistically significant difference in expression (Student’s test FDR≤ ≤ 0.05). The significance value represents the probability that the association between a given dataset and a canonical pathway occurred by random chance. P-value = B-H adjusted p value.
Full size table

Elevated temperature responses

Exposure to elevated temperature elicited broadly convergent molecular disturbances in both brain and eye (Tables 1 and 2; Fig. 5). Upregulation of the sirtuin signalling pathway, often associated with stress coping mechanisms and neuroprotection84, occurred in parallel with downregulation of mitochondrial oxidative phosphorylation, electron transport, and ATP synthesis—hallmarks of mitochondrial dysfunction under thermal stress85. These findings indicate a general decline in energy metabolism and cellular efficiency in response to sustained heat exposure.

Fig. 5
figure 5

Key alterations in protein expression in the brain and eye of adult zebrafish exposed to elevated (34 °C, orange) or low (18 °C, blue) temperatures compared with the control (26 °C). The figure summarizes and compares the most significantly enriched pathways identified in the zebrafish eye (Table 1; Fig. 2) and brain (Table 2) after 21 days of acclimation, based on IPA. An upward arrow (↑) denotes IPA-predicted pathway activation, while a downward arrow (↓) denotes inhibition. Zero means no predicted activation/inhibition of the stastistically altered pathways. Single arrows indicate absolute Z-scores > 0 and < 1, whereas double arrows (↑↑/↓↓) represent absolute Z-scores ≥ 1.

Full size image

In the eye, such alterations were accompanied by a reduction in proteins involved in visual phototransduction, pointing to a possible impairment of retinal physiology and visual capacity. In the brain, previously reported findings28 highlighted downregulation of synaptic signalling, neurotransmitter secretion, and key regulatory cascades (e.g., Ras, PDGF, FGF, Integrin, EGF), all central to neuronal plasticity and intercellular communication. Taken together, these parallel responses across two neural tissues reinforce the view that thermal stress impairs core cellular processes such as mitochondrial integrity and proteostasis, while tissue-specific alterations suggest additional vulnerability of eye physiology.

Low temperature responses

By contrast, responses to cold exposure (18 °C) diverged between the two tissues. While sirtuin signalling was upregulated in the brain, it was downregulated in the eye; markers of mitochondrial dysfunction were increased in the brain but decreased in the retina. These opposite trends suggest that low temperature triggers distinct adaptive strategies in different neural tissues rather than a uniform neuroprotective or neurotoxic response.

Thermal stress, neural resilience, and implications

Overall, our findings indicate that both elevated and reduced constant temperatures can exert neurotoxic effects, albeit through distinct, tissue-specific mechanisms. In the eye, sustained heat stress triggers mitochondrial, proteostatic, and neuronal stress responses, with potential implications for neurodegenerative-like conditions. Conversely, cold exposure does not appear to impair visual phototransduction and may involve different molecular pathways.

Further work is needed to determine whether the observed proteomic differences reflect altered protein expression, protein degradation, or both, as well as to establish the long-term consequences of these changes for neural health in zebrafish and other ectothermic species. Additional studies are also required to test whether the proteomic alterations detected in the brain and eye revert to the control condition when zebrafish exposed to 34 °C or 18 °C for 21 days are subsequently returned to 26 °C.

This aspect is particularly relevant given the known resilience and regenerative capacity of the zebrafish retina86,87,88. Such investigations would clarify whether the proteomic signatures indicative of neurotoxic effects are reversible, and whether different levels of resilience to thermal stress exist between the brain and the eye.

From an ecological perspective, temperature-driven proteomic remodelling in the eye suggests that naturally occurring exposure to low or elevated temperatures within the viable thermal range of zebrafish may compromise visual performance, ultimately affecting behaviour and survival.

At the same time, parallels with molecular disturbances associated with human retinal and neurodegenerative disorders emphasise the translational value of the zebrafish as a model for investigating conserved mechanisms of ocular neurotoxicity.

Conclusions

This study demonstrates that a 21-day exposure to constant elevated (34 °C) and low (18 °C) temperatures, relative to the control (26 °C), significantly alters the proteomic profile of the adult zebrafish eye. Both treatments modulated proteins essential for retinal function, but in clearly distinct ways, confirming that the ocular response to thermal stress is temperature-dependent. Notably, elevated temperature induced proteomic signatures closely resembling those previously described in the zebrafish brain, including the up-regulation of sirtuin signalling and the down-regulation of oxidative phosphorylation, electron transport, and ATP-synthesis pathways. Within the eye, these mitochondrial impairments were accompanied by a robust reduction in visual phototransduction proteins, suggesting a pronounced neurotoxic impact of elevated temperature on eye function. In direct relation to the study’s guiding hypothesis, these findings provide strong evidence that chronic thermal exposure differentially disrupts ocular homeostasis, with heat imposing a substantially greater metabolic and functional burden than cold. The marked mitochondrial dysfunction, phototransduction deficits, and activation of stress-related pathways observed at 34 °C—contrasting with the subtler translational and RNA-processing adjustments detected at 18 °C—indicate that elevated temperature is the dominant driver of thermal stress–induced neurotoxic alterations in the zebrafish eye. Taken together, our findings reveal that the adult zebrafish eye is highly sensitive to sustained deviation from the preferred thermal range. By uncovering molecular pathways that are differentially affected underm warm and cold conditions, this work broadens current understanding of temperature-induced neurotoxicity in ectothermic vertebrates and highlights the particular vulnerability of the visual system in the context of global warming.

Methods

Subjects and housing conditions

A total of 96 adult wild-type zebrafish (Danio rerio), with an equal male-to-female ratio (1:1) and a mean weight of 0.4 g, were housed in the zebrafish facility at the University of Bologna. The fish were randomly allocated to three identical glass tanks (40 × 30 × 30 cm; width × depth × height), with 32 individuals per tank at a density of 1 fish per litre.

All fish were maintained at a constant temperature of 26 ± 1 °C (control condition) for 10 days to allow acclimatisation. Throughout the experiment, the tanks were kept under a regulated 14:10 h light/dark photoperiod, and the fish were fed three times daily (10 AM, 2 PM, and 6 PM) using automatic feeders (Eden 90, Eden Water Paradise, Germany). A commercial dry granular diet (TropiGranMIX, Dajanapet, Czech Republic) was provided, ensuring a balanced intake according to fish appetite.

Water quality and tank maintenance

Each tank was equipped with an external filtration system (Eden 511 h, Eden s.r.l., Vicenza, Italy), maintaining a constant flow of filtered water (600 L/h). Water was continuously aerated with aquarium aerators (Sicce AIRlight, SICCE s.r.l., Vicenza, Italy; 3300 cc/min, 200 L/h). Temperature was controlled with digital thermostats (Eden 430, Eden s.r.l., Vicenza, Italy), connected to heating coils (Eden 415, 230 V, 50/60 Hz, 80 W, Eden s.r.l., Vicenza, Italy), and supplemented by a cooling system. Daily measurements were performed using calibrated hand thermometers.

The chemical and physical characteristics of the tank water were monitored at least twice per week using a Sera Aqua-Test Box Kit (Sera GmbH, Heinsberg, Germany) and an eSHa Aqua Quick Test (eSHa Lab, Maastricht, The Netherlands). No differences were observed between the treatment and control tanks. Total ammonia (NH₃/NH₄⁺) and nitrite (NO₂⁻) were consistently below detection limits (0.05 mg/L). Partial water changes were performed whenever nitrate concentrations approached 25 mg/L. The water had a stable pH of 7, conductivity values ranging between 400 and 500 µS/cm, total hardness between 7 and 14 °dH, and carbonate hardness (dKH) of 6. Phosphate (PO₄³⁻) concentrations remained below 1 mg/L, and neither copper (Cu) nor chlorine (Cl₂) was detected. Overall, water quality parameters remained constant across tanks and conditions. Uneaten food and faeces were removed from the tanks at least three times per week to prevent organic build-up.

Thermal treatment

Following acclimatisation, zebrafish were exposed to constant elevated or low temperatures. Tank water was gradually adjusted from 26 °C to either 34 °C (elevated temperature) or 18 °C (low temperature) over a 72-hour period, with controlled steps of ± 2 °C every 16 h. Fish were then maintained at their respective constant temperatures (34 ± 1 °C, 26 ± 1 °C or 18 ± 1 °C) for 21 consecutive days. Each tank was assigned to one temperature condition only, and thus represents a single experimental unit (i.e., one biological replicate per group) which is a limitation of our study warranting careful interpretation of our findings. These thermal regimes were selected based on prior studies, as they fall within the thermal tolerance limits of zebrafish and mimic ecologically relevant high and low environmental conditions30, as discussed in the Introduction section.

The three tanks were placed side by side, and standardised for internal setup and environmental parameters, including lighting (14:10 h light/dark cycle), filtration, aeration, and heating. Fish were fed three times daily (10:00 AM, 2:00 PM, and 6:00 PM) with the same commercial diet, delivered via automatic feeders. To accommodate potential differences in metabolic rate between temperature groups, fish were allowed to feed ad libitum, ensuring they could meet their own nutritional demands, which may vary under different thermal conditions. Behavioural observations were performed daily for five minutes by trained observers to monitor welfare indicators.

Sample collection

At the end of the 21-day exposure, fish were euthanised with tricaine methanesulphonate (MS-222, 300 mg/L), followed by decapitation to collect both eyes. For each tank (i.e., each temperature group), four pools of eyes were prepared, each consisting of both eyes from eight randomly selected fish (n = 8 fish per pool, 4 pools per tank).

Because fish were drawn from the same mixed-sex population with an approximately 1:1 sex ratio, and selection for pooling was fully random, each pool is expected to contain a balanced representation of males and females. Body mass was not recorded at the time of sampling because individuals displayed homogeneous body size at baseline, and differences in final body mass are expected under thermal challenges due to temperature-dependent modulation of metabolic rate, even when fish are allowed to feed ad libitum. Moreover, previous studies investigating zebrafish responses to altered thermal regimes28 have similarly not included post-exposure body size measurements, reflecting a common methodological approach in this experimental context.

These pools represent biological variation at the individual level but originate from the same tank and therefore do not constitute independent biological replicates. Throughout the experimental period, no mortality or overt signs of poor health were observed in any of the groups. No macroscopic morphological anomalies were observed in the eyes under any experimental condition.

To minimise potential time-of-day effects on the eye proteome, samples were collected over a five-hour period (10 AM–15 PM) by alternating individuals from the different temperature groups. This sampling strategy ensured a homogeneous distribution of samples across the time window and reduced the likelihood of systematic bias linked to circadian or time-specific variations. Tissues were immediately immersed in PBS containing protease inhibitors (Roche, Basel, Switzerland), snap-frozen on dry ice, and stored at −80 °C for proteomic analysis.

To ensure blinding at different stages of the experiment, the procedures were carried out by independent researchers at separate institutions. Animal housing, thermal exposure, and tissue collection were performed by M.T. and F.F. at the University of Bologna. Samples were anonymised with codes that did not reveal treatment allocation and transferred to another institution. Proteomic analysis was conducted in a blinded manner by E.M., S.N., F.G.S., J.G., and G.T., who remained unaware of the experimental groups throughout the analysis.

Ethics declarations

All procedures were conducted in accordance with ethical guidelines and were approved by the institutional Animal Care Committee and the Italian Ministry of Health (protocol number 290/2017-PR). The study is reported in compliance with the ARRIVE guidelines. No experimental subject died during housing or during the thermal treatments.

Proteomic analysis

Proteomic analysis of zebrafish eyes was performed after 21 days of constant temperature exposure. Eye samples, consisting of pools from eight fish were analysed using a shotgun label-free proteomic approach for the identification and quantification of expressed proteins.

Each eye sample was homogenised using a Potter homogeniser in 100 µl of extraction buffer (8 M urea, 20 mM Hepes pH 8, with protease inhibitors Complete Mini), on ice, at full speed for 10 s. The homogenate was centrifuged at 15,000 rpm for 10 min at 4 °C to remove unhomogenised tissue and large cellular debris. The pellet was discarded, and the protein concentration of the supernatant was determined using the Bradford assay (Sigma-Aldrich, Milan, Italy) with bovine serum albumin as the standard89.

Prior to proteolysis, proteins were reduced with 13 mM dithioerythritol (DTE; 15 min at 50 °C) and alkylated with 26 mM iodoacetamide (IAA; 30 min at room temperature, in the dark). Protein digestion was carried out using sequence-grade trypsin (Promega, Milan, Italy) for 16 h at 37 °C with a protein to enzyme ratio of 20:1.

The proteolytic digests were desalted using Zip-Tip C18 (Millipore, Milan, Italy) before mass spectrometric (MS) analysis90.

NanoHPLC coupled to MS/MS analysis was performed on a Dionex Ultimate 3000 HPLC system with an EASY-SprayTM 2 μm 15 cm × 150 μm capillary column packed with 2 μm C18 100 Å particles, connected to a Q-Exactive Plus Orbitrap (Thermo Fisher Scientific, Waltham, MA, USA). MS spectra were acquired over an m/z range of 350–2000 Da at a resolution of 70,000, operating in data dependent mode. HCD MS/MS spectra were collected at a resolution of 17,500 for the 10 most abundant ions in each MS scan using a normalised collision energy of 35%, and an isolation window of 3 m/z. Rejection of + 1 and unassigned charge states were enabled91. Data acquisition was controlled by Xcalibur (Thermo Fisher Scientific, Monza (MI), Italy)92. The instrument was externally calibrated prior to each analytical sequence. A HeLa protein digest (Thermo Fisher Scientific) was injected as a quality control to monitor mass accuracy, chromatographic stability and signal intensity.

Parallel Reaction Monitoring (PRM) analysis was performed to validate selected proteins of functional relevance within key pathways identified by enrichment analysis, namely NADH dehydrogenase [ubiquinone] flavoprotein 1 (ndufv1), Cytochrome c oxidase subunit 6 C (cox6c), L-lactate dehydrogenase A chain (Idha), Long-wavelength-sensitive-1 cone opsin (opn1lw1), Opsin 1 (cone pigments), short-wave-sensitive 2 (opn1sw2), and S-arrestin (saga).

For PRM validation, at least two peptides were selected for each target protein, and nine transitions were monitored per peptide using Skyline software (https://skyline.ms/project/home/software/Skyline/begin.view, version 25.1.0.142). MS² scans were triggered by an inclusion list from either ndufv1, cox6c, Idha, opn1lw1, opn1sw2, and saga peptides, without time scheduling. Data were acquired at a resolution of 35,000 with an AGC target value of 2 × 105, a maximum injection time of 100 ms, and an isolation window of 1.5 m/z. PRM raw data files were processed in Skyline against a protein sequence database of the six target proteins. Search parameters included trypsin digestion (maximum one missed cleavage sites), a minimum peptide length of six amino acids, and a mass tolerance of 20 ppm for precursor ions and 100 ppm for fragment ions. Carbamidomethylation of cysteine was set as a fixed modification, while deamidation of asparagine (Asn) and glutamine (Gln) was set as a variable modification. No internal standards were used, and the total peak area of each peptide was extracted for comparison across samples. Representative extracted-ion chromatograms (XICs) for each target protein are shown in Supplementary Fig. S3–S4, and representative MS² spectra are reported in Supplementary Fig. S5-S6.

Statistical and bioinformatic analysis of the proteomic data

MS spectra were searched against the Danio rerio UniProt sequence database (release 2023, 47,579 entries) using MaxQuant (version 1.6.1.0). The following parameters were used: initial maximum allowed mass deviation of 10 ppm for monoisotopic precursor ions and 0.5 Da for MS/MS peaks, trypsin enzyme specificity, a maximum of two missed cleavages, carbamidomethyl cysteine as fixed modification, N-terminal acetylation, methionine oxidation and asparagine/glutamine deamidation as variable modifications. False protein identification rate (5%) was estimated by searching MS/MS spectra against the corresponding reversed-sequence (decoy) database. Minimum required peptide length was set to 7 amino acids and minimum number of unique peptide supporting protein identification was set to 1. Quantification in MaxQuant was performed using the built-in LFQ algorithms based on extracted ion intensity of precursor ions. The coefficient of variation (CV%) of LFQ (Label-Free Quantification) intensities among technical replicates was below 12%, confirming good quantitative reproducibility across runs.

Four pools of eyes (from eight animals each) were analysed per temperature group (elevated: 34 °C; control; 26 °C; low: 18 °C). Only proteins detected in at least three out of four pools per group were included in statistical analyses, and no missing-value imputation was applied.

Statistical analyses of Max Quant results were performed using the Perseus software module (version 1.5.5.3). Principal component analysis-based multiconfidence ellipse analysis, PCA, was carried out on quantitative data related to proteins expressed at the three different thermal treatments using the SRplot web tool (https://www.bioinformatics.com.cn/srplot). Prior to multivariate analysis of variance (MANOVA), data distribution was evaluated using the Shapiro-Wilk test, which indicated deviations from perfect normality (W = 0.63–0.68), an expected feature of proteomic datasets with wide dynamic ranges. Multivariate analysis of variance (MANOVA) was conducted using Paleontological Statistics (PAST) software package93. Focusing on specific comparisons, namely 34 °C vs. 26 °C and 18 °C vs. 26 °C, proteins were considered differentially expressed if they were exclusively present in one condition or showed significant t-test difference (Student’s test FDR ≤ 0.05). No additional fold-change threshold was applied.

Bioinformatic analyses were carried out by CLUEGO software (Cytoscape release 3.8.2)94, and an Ingenuity Pathway Analysis (IPA, release 2024, QIAGEN Inc., https://www.qiagenbioinformatics.com/products/ingenuitypathway-analysis) to cluster enriched annotation groups of biological processes, pathways, and networks within the set of identified proteins using Danio rerio as data base. Functional grouping was based on Fischer’s exact test (p-value ≤ 0.05 and at least 3 counts per category) in GLUEGO and P-value (Benjamini-Hochberg adjusted p-value) ≤ 0.05 (-log (p-value) cutoff of 1.3) in IPA.