Abstract Introduction Retinopathy of prematurity (ROP) is one of the leading causes of babies’ visual impairment and blindness. There is no effective prevention and treatment of ROP so far, and the shared genetic and developmental similarities among the brain, kidneys and retina may offer novel potential therapeutic approaches to ROP. Objectives The aim of this study is to explore a correlation of ROP patients and the renal, eye tissue of the mouse model of oxygen-induced retinopathy (OIR). Methods and analysis We used renal and vitreous untargeted/targeted metabolomics in OIR to conduct our study. Network association analysis and machine learning were performed with the above results and previous studies: retinal-targeted metabolomics of OIR and human blood-targeted metabolomics of ROP. Results OIR results in retinal neovascularisation and renal injury. Nine canonical signalling pathways were enriched, which are involved in the initiation and progression of pathologic retinal neovascularisation. Arginine biosynthesis emerged as a common pathway across renal, vitreous, retinal and blood metabolomics, suggesting its potential as a predictive biomarker and therapeutic target for ROP and neonatal kidney injury. Conclusion The presence of renal injury-related indicators may assist in diagnosing retinal neovascular diseases such as ROP. Arginine biosynthesis is the best common pathway of kidney-untargeted OIR metabolomics, vitreous- and retina-targeted OIR metabolomics and blood-targeted metabolomics of ROP, indicating that arginine biosynthesis might be the common pathway of ROP and neonatal kidney injury. Keywords: Retina, Neovascularisation, Experimental & animal models __________________________________________________________________ WHAT IS ALREADY KNOWN ON THIS TOPIC * Retinopathy of prematurity (ROP) is a major cause of visual impairment in preterm infants and may be associated with renal injury due to potential shared developmental and vascular mechanisms. Previous studies have highlighted potential metabolic links between the retina and kidney, but the underlying pathways remain unclear. WHAT THIS STUDY ADDS * This study revealed that arginine metabolism and the tricarboxylic acid cycle are disrupted in both the retina and kidneys of oxygen-induced retinopathy (OIR) mice. Key metabolites, such as shikimic acid and pimelic acid, were identified as potential biomarkers, suggesting systemic metabolic dysregulation connects retinal and renal pathologies in OIR. HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY * These findings provide a foundation for using renal metabolic indicators to aid in the diagnosis and monitoring of ROP. The study also underscores the importance of targeting shared metabolic pathways, such as arginine biosynthesis, to develop novel therapeutic strategies for preterm infants. Introduction Managing oxygen levels in preterm infants is critical, but hyperoxia can harm organs such as the eyes, kidneys, brain and intestines.[47]^1 Retinopathy of prematurity (ROP), a condition caused by abnormal blood vessel development in the retina, can result in retinal detachment and blindness. Each year, approximately 15 million preterm infants are born with underdeveloped retinal blood vessels, which fail to grow normally outside the uterus.[48]^2 Oxygen supplementation in these infants may disrupt normal growth factor delivery, leading to abnormal retinal vessel development and metabolic imbalances.[49]^3 An acute increase in oxygen tension stimulates apoptosis of vascular endothelial cells (ECs) and may cause vaso-obliteration during phase I by generating reactive oxygen species (ROS). In phase II, the vaso-occluded retinas experience hypoxic/ischaemic stress, which drives pathological neovascularisation.[50]^4 Recent advances have identified several novel biomarkers that could enhance our understanding of ROP and aid in its diagnosis. These include metabolites, cytokines, growth factors, non-coding RNAs, oxidative stress (OS) biomarkers and gut microbiota.[51]^5 Among these, growth factors such as IGF (insulin-like growth factor)-1 and IGFBP (IGF binding proteins) have drawn significant attention due to their critical roles in retinal vascular development. Clinical studies have linked low IGF-1 levels with an increased risk of severe ROP.[52]^6 However, attempts to reduce ROP severity through early IGF-1 supplementation have shown limited success, as a recent trial failed to meet its primary endpoint,[53]^7 suggesting the need for further research into optimal dosage and timing. Understanding how metabolic disturbances caused by hyperglycaemia, hyperoxia or hypoxia affect neonatal retinal development may reveal new avenues for early prevention and management of ROP. ROP and kidney damage may share similar clinical and biological mechanisms. In preterm infants, OS after birth leads to metabolic disturbances and kidney injury.[54]^8 The presence of free radicals and insufficient antioxidant protection make the body more vulnerable to damage, especially in the retina, lungs, brain and intestines.[55]^9 This imbalance has been identified as a key factor in ‘free radical-related diseases’ of preterm infants, including ROP and kidney injury, with which neovascularisation may share a common underlying cause.[56]^10 11 Additionally, conditions like age-related macular degeneration have been shown to share risk factors with kidney disease and diabetes.[57]^12 This suggests that similar biological processes may link diabetic retinopathy (DR) and kidney disease, as inflammation and blood vessel growth in DR are associated with kidney injury.[58]^13 OS is also a critical factor in the progression of ROP. The oxygen-induced retinopathy (OIR) model is widely used to study ROP because it closely mimics the pathological processes of retinal neovascularisation and hyperoxia-induced injury seen in preterm infants. Notably, hyperoxia in the OIR model also leads to kidney damage, reflecting the impaired nephrogenesis observed in preterm infants.[59]^14 Recent studies have shown genetic and developmental similarities between the kidney and retina,[60]^15 and both organs share similar embryological origins.[61]^16 There is growing evidence that ROP in infants often occurs alongside kidney damage, and the severity of kidney injury correlates with the severity of ROP. A recent clinical study by Eroglu et al found a significant association between proteinuria and ROP. The study showed that 71.4% of patients with type I or II ROP and 91.7% of those with type I ROP had proteinuria. Additionally, preterm infants with proteinuria and a gestational age of ≤32 weeks were four times more likely to develop severe ROP.[62]^17 Research by Nakagawa et al revealed that OIR mice suffer kidney damage during development, providing further evidence of a link between kidney and retinal diseases in preterm infants.[63]^14 Despite progress in understanding ROP, the exact mechanisms underlying the disease remain unclear, and current diagnostic and treatment options are limited. Many studies face challenges in replication, and there has been little research on how abnormal kidney metabolism may induce retinal neovascularisation in neonates. To improve the prevention and treatment of ROP, exploring new therapeutic markers is essential. By identifying additional indicators such as blood metabolism, kidney function and alternative disease pathways, researchers can better anticipate the progression of ROP and develop interventions earlier. This approach could also lead to more accurate diagnostic methods. In conclusion, this study hypothesises a potential link between kidney metabolism and ROP/OIR metabolism. Understanding this relationship could serve as the foundation for future research and improve our overall understanding of ROP. Using metabolomics analysis, we examined ROP-related phenotypes by comparing the kidneys, vitreous humor and retinas of mice, as well as human blood samples. By studying kidney metabolomics in OIR mice, we aim to deepen our understanding of the relationship between kidney function and ROP, which could lead to more effective research and clinical applications in the future. Materials and methods Establishment of a mouse OIR model All animal procedures adhered to the ARVO (Association for Research in Vision and Ophthalmology) Statement for the Use of Animals in Ophthalmic and Vision Research and were approved by the Institutional Animal Care and Use Committee at Shenzhen Eye Hospital. The OIR mouse model was developed based on the protocol by Connor and Smith et al,[64]^18 and in line with our previous publications.[65]^19 20 Briefly, C57BL6/J timed-pregnant mice (Vital River Laboratory Animal, Beijing, China) were housed at the Shenzhen Eye Hospital Laboratory Animal Resource Center. In this model, 7-day-old mice were placed with their nursing mothers in a 75% oxygen environment for 5 days. On postnatal day 12 (P12), the mice were returned to room air until postnatal day 17 (P17), at which point they were euthanised. The OIR group consisted of three distinct litters with three, three and two pups, respectively, with one pup lost due to maternal cannibalism during the experiment. To confirm the success of the OIR model, retinas from both the OIR and wild-type (WT) room air groups were flat-mounted and stained with isolectin B4 (IB4), an endothelial-specific marker (Vector Laboratories, Burlingame, California, USA). Retinal vasculature was then photographed using a DMI4000B fluorescent microscope (Leica, Wetzlar, Germany). Histopathological validation of kidney in OIR We embedded OIR (P17) and WT (P17) kidneys in paraffin after dehydrating them in ethanol and serially embedding them in 4% paraformaldehyde. H&E was applied to 5 m tissue sections. In this study, we examined and graded tubular and glomerular injury patterns. There are several types of tubular injury, including dilation, atrophy, vacuolisation, sloughing of the tubular epithelium and thickening of the tubular basement membrane. All slides were examined by light microscope (×36.5 magnification). The assessment of kidney damage through histopathology was conducted based on particular standards, which involved randomly choosing three fields of view at an initial magnification of ×36.5. Two pathologists independently performed histological evaluations. The glomerular diameter measurements were modified according to the guidelines provided by Toledo-Rodriguez et al.[66]^21 H&E-stained images were analysed digitally using ImageJ 1.53 software, and glomeruli grid points were quantified. A scale bar was used to measure the glomerular diameter, with mean size calculated from the largest and smallest diameters within three fields of view in both the middle cortex and juxtamedullary zones. Tubular injury was evaluated based on dilatation, atrophy, vacuolation, degeneration, sloughing of tubular epithelial cells and thickening of the tubular basement membrane, with scoring applied only to proximal tubules. The scoring system ranged from 0 to 5: 0 indicating no injury, 1 indicating ≤10% of tubules injured, 2 indicating 10%–25%, 3 indicating 26%–50%, 4 indicating 51%–75% and 5 indicating >75% injury.[67]^22 Metabolic profiling and UPLC-MS/MS Kidneys were excised and immediately processed for metabolomics extraction. Frozen eyes were dissected to isolate vitreous. The parameters of vitreous samples in the targeted metabolomics and kidney preparation were as in prior studies.[68]^23 Metabolomic data project ID: LM2021-23621. The non-targeted metabolomics analysis of 5022 compounds in 16 eligible subjects was performed by the UPLC (Ultra-Performance Liquid Chromotography)-MS/MS (Mass Spectrometry) method. The original data, code and statistical values are described in detail in the [69]online supplemental materials. Multivariate data analysis and pathway enrichment analysis All the multivariate data analysis was conducted using SIMCA (V.14.1; Umetrics, Umeå, Sweden). The method of orthogonal partial least squares-discriminant analysis (OPLS-DA) model was described in a previous study.[70]^24 Pathway analysis and statistical analysis were undertaken using MetaboAnalyst 5.0 software,[71]^25 which was similar to that previously described. To select the most discriminant pathways, we chose levels of impact and extremely significant differences (impact >0.2, p<0.05, respectively). Venn networks[72]^21 were assembled using an online Venn diagram builder from [73]http://ehbio.com/test/venn/#/. Besides, in the box plot, the raw data were transformed to Z scores using the mean and SD. Analogous analysis Our method used feature selection through regularized logistic regression models, including Least Absolute Shrinkage and Selection Operator (LASSO)—a technique that applies L1 regularization to shrink coefficients and inherently perform feature selection by driving some coefficients to zero—alongside Elastic-Net (EN) regression and Adaptive LASSO, all implemented in Python 3.7.0. Analysis data from renal untargeted metabolomics of OIR, vitreous-targeted metabolomics of OIR, retina-targeted metabolomics of OIR by Zhou et al[74]^26 and blood-targeted metabolomics of ROP infants by Yang et al.[75]^23 Venn network analysis was applied to VIP metabolites selected from the OIR retina, vitreous, kidney metabolomics and ROP blood metabolomics studies (Data Supplement). Statistical analysis Normality was tested using the Shapiro-Wilk test. Normally distributed variables were presented as mean±SD; otherwise, data were expressed as median and mean ranks. The Student’s t-test and the Wilcoxon test were then used to compare variables that were normally and abnormally distributed, respectively, between OIR and WT groups. Following Z-score transformation and hierarchical clustering analysis were performed. Differences were considered significant when p<0.05, as well as those approaching significance (0.050.05, n=16). (D) Kidney injury scores were significantly higher in the OIR group compared with WT (p<0.0001, n=8), indicating substantial kidney damage. Body weight and kidney injury scores were analysed using a Student’s t-test. Data are presented as mean±SD, where N represents the number of mice used in the experiment. OIR, oxygen-induced retinopathy; WT, wild-type. [80]Figure 2 [81]Open in a new tab Renal metabolic profile of OIR OPLS-DA model of kidney and vitreous The OPLS-DA scatter plot revealed a clear clustering of samples between the R-OIR (renal-OIR) group and the R-WT (renal-WT) group ([82]online supplemental figure 1). Strong discriminative power and good stability are also exhibited by this model based on its R2Ycum (0.998) and Q2cum (0.899). A 200-time permutation analysis validated the OPLS-DA model as well ([83]online supplemental figure 2A). Outlier detection was carried out using DModX and Hotelling’s T2 tests ([84]online supplemental figure 2B,C). Similarly robust associations were observed for the vitreous group, respectively, the V-OIR (vitreous-OIR) group and the V-WT (vitreous-WT) group ([85]online supplemental figures 3 and 4 and Supplementary Data). Metabolomic pathway analysis Ranks were based on the contribution plots of metabolites. Among the top 131 VIP metabolites (VIP >2), the discriminant variables were selected based on their higher VIP scores, and in-depth analyses were performed on these metabolites ([86]online supplemental figure 5). We used MetaboAnalyst 5.0 to analyse pathway enrichment, and significant enrichment was observed in nine canonical signalling pathways (Impact >0.2, p<0.05) ([87]online supplemental figure 6 and table S2). Screening for specific metabolites Feature selection To enhance statistical power and identify the most predictive criteria, three feature selection methods—LASSO, EN regression and adaptive LASSO—were used alongside a Volcano plot ([88]online supplemental figure 7) to select key metabolites for Sankey diagram pathway interaction analysis. Each method selected the 15 most correlated variables, and the intersection of these was identified ([89]online supplemental figure 8, [90]figure 3). Two common variables, PC (14:1(9Z)/20:2(11Z,14Z)) and Shikimic acid, showed the highest correlation with dependent variables ([91]online supplemental figure 9 and table S3). Figure 3. A Sankey diagram reveals the possible association between the metabolic pathways with statistically significant differences and the top 15 feature importance of different feature selections. Data represent metabolite changes in kidney tissue of OIR and control groups and were clustered based on significant metabolites (p<0.05). The heatmaps on the left present the top nine metabolic pathways with statistically significant differences (Impact >0.2, p<0.05). The closer p is to 0, the redder the colour is, whereas white represents p=0.05. The above metabolites on the right are possibly involved in the corresponding metabolism. Data were analysed using a Student’s t-test. OIR, oxygen-induced retinopathy. CoA, coenzyme A. [92]Figure 3 [93]Open in a new tab Specific metabolites and interaction analysis For the interaction analysis, we focused on metabolites linked to enriched pathways, identifying potential biomarkers. Nicotyrine and N-gamma-glutamyl-S-propylcysteine were associated with tryptophan metabolism, arginine biosynthesis and alanine, aspartate and glutamate metabolism. Alanine and L-histidine were involved in alanine and beta-alanine metabolism, while ADP ribose, L-3-hydroxykynurenine, quinoxaline-2-carboxylic acid, guanosine, isoguanosine and guanine were linked to tryptophan, purine and histidine metabolism ([94]figure 3). Notably, Shikimic acid and pimelic acid, identified through an overlap analysis of the top 15 metabolites selected by the feature selection methods and Volcano plot, were strongly associated with the tricarboxylic acid (TCA) cycle and lipid metabolism. These metabolites showed the highest Pearson correlation in distinguishing OIR and were proposed as highly specific biomarkers for OIR ([95]online supplemental table S3). Additionally, L-histidine, isoguanosine, guanosine, nicotyrine, quinoxaline-2-carboxylic acid, L-3-hydroxykynurenine, N-gamma-glutamyl-S-propylcysteine, alanine and ADP ribose were identified as potential biomarkers of OIR ([96]online supplemental figure 10). To further explore the connection between OIR, kidney injury and metabolomics, a correlation analysis was performed linking kidney injury scores to these biomarkers. Nicotyrine and quinoxaline-2-carboxylic acid increased with the severity of kidney injury, while N-gamma-glutamyl-S-propylcysteine initially increased, then decreased as kidney injury worsened, peaking at a score of 1.5. The other markers showed a decline with increasing kidney injury scores ([97]figure 4). This analysis provides deeper insights into the altered renal metabolomics in OIR mice, and the interrelationships among these biomarkers will be discussed further in the next section. Figure 4. Non-linear regression analysis assessed the correlation between injury scores and potential OIR biomarkers. Results are based on kidney tissue analysis. Although shikimic acid, pimelic acid and others were expected to vary with injury scores, the analysis did not confirm a linear relationship. OIR, oxygen-induced retinopathy. [98]Figure 4 [99]Open in a new tab Cross-correlations of OIR and ROP metabolomics Arginine biosynthesis and D-glutamine and D-glutamate metabolism pathways were identified as significantly enriched in ROP-related metabolomics studies in both human and mouse samples (blue node, [100]figure 5). These pathways were consistent across ROP metabolism and different tissues in the OIR study, suggesting their critical role in ROP research. Arginine and proline metabolism, alanine, aspartate and glutamate metabolism and histidine metabolism were enriched in mouse samples (brown node, [101]figure 5). Purine metabolism, pantothenate and CoA biosynthesis, tryptophan metabolism and beta-alanine metabolism are uniquely enriched pathways in the kidney metabolism of OIR mouse ([102]figure 5). Venn network analysis showed that tryptophan, arginine, ornithine, creatine and histidine are relevant to arginine biosynthesis and play a major discriminant role in it ([103]online supplemental figure 11). Creatine was used to be defined as a biomarker in human plasma analysis in the past study.[104]^24 To note, tryptophan was found in the above metabolomics, which corroborates its interaction with the arginine metabolic pathway. Figure 5. Venn network of the enriched pathways. Through comparative analysis of this study and previous studies, a total of 14 different metabolic pathways with significant changes were enriched (10, 9, 9 and 6 pathways in OIR retina, vitreous targeted metabolomics, OIR kidney untargeted metabolomics, and ROP blood targeted metabolomics, respectively). See supplementary data for details. The larger the node, the more the count of pathways. OIR, oxygen-induced retinopathy; ROP, retinopathy of prematurity. [105]Figure 5 [106]Open in a new tab Discussion Through a comparative analysis of kidney metabolomics in the OIR and WT groups, we obtained a significant renal metabolomics profile. Since the OIR model is a well-established model for retinal neovascularisation, we also analysed the potential relationship between renal metabolomics and pathological retinal neovascularisation. It is likely that kidney injury and ROP share a common underlying vascular pathological mechanism. Evaluating how alterations in renal metabolism impact ROP is highly relevant, as the disease closely mirrors the OIR model. However, whether changes in renal metabolic markers can serve as indicators for ocular diseases needs further investigation. In the following discussion, we explore possible pathways related to OIR/ROP based on metabolic findings. Arginine and ROS metabolism play a crucial role in the development of neovascularisation. Arginase, a ureohydrolase enzyme, catalyses the conversion of L-arginine into L-ornithine and urea. The upregulation of arginase has been associated with inflammation, OS and peripheral vascular dysfunction.[107]^27 Arginase has already been recognised as a therapeutic target for cardiovascular and central nervous system diseases, including stroke and ischaemic retinopathy, due to its role in mediating retinal neurovascular injury and targeting early neovascularisation processes.[108]^28 29 In our study, we observed that arginine and ornithine levels were elevated in the R-OIR group, while nitric oxide (NO) and cyclic guanosine monophosphate (cGMP) levels were relatively reduced. This imbalance may be related to enhanced arginase activity. Excessive L-ornithine can damage vasculature, induce neurotoxicity and even promote tumour growth.[109]^28 Arginase is a key regulator of NO production, competing with endothelial NO synthase (eNOS) for the L-arginine substrate, thereby inhibiting NO formation. NO, as an endothelium-derived molecule, is crucial for maintaining vascular homeostasis. Reduced NO production may contribute to pathological neovascularisation. Moreover, NO activates soluble guanylate cyclase, which regulates vascular tone by producing cGMP, a key secondary messenger.[110]^30 Consistent with previous findings, reduced NO synthesis in ECs impairs angiogenesis and ROS levels and alters tube formation, NO and cGMP levels[111]^31 ([112]online supplemental figure 12). Furthermore, increased arginase activity leads to eNOS uncoupling, reducing NO formation and subsequently increasing ROS production.[113]^32 Excess ROS, such as superoxide anions (O2−) and hydrogen peroxide (H2O2), are toxic when present in high quantities. However, at physiological levels, ROS are critical for redox signalling and contribute to biological processes like cell proliferation, migration, differentiation and gene expression. Accumulating evidence indicates that ROS produced by ECs and other cell types, such as vascular smooth muscle cells and bone marrow cells, can also stimulate angiogenesis.[114]^33 Several studies have linked retinal neovascularisation to abnormal arginine metabolism. Lu et al discovered arginine metabolism abnormalities in the serum of OIR mice.[115]^34 Excessive arginase activity reduces the L-arginine supply available for NOS, leading to the production of superoxide instead of NO, which drives pathological angiogenesis.[116]^35 A metabonomics study on patients with proliferative DR revealed dysregulation in arginine- and citrulline-related pathways.[117]^36 Arginase activation causes neurovascular damage by uncoupling NO, inducing polyamine oxidation, and increasing glutamate production, all of which contribute to OS and the progression of ischaemic retinopathies like DR, ROP and retinal vein occlusion.[118]^37 Moreover, short-term exposure to hypoxia/reoxygenation, whether in OIR or ROP, induces ROS production and OS, contributing to pathological neovascularisation or tumour vascularisation. ROS is generated through several pathways, including the mitochondrial electron transport chain, cytochrome P450, xanthine oxidase and uncoupled NOS.[119]^38 While ROS can cause tissue damage, they also play a role in promoting tissue repair through angiogenesis. Elevated ROS levels are implicated in pathological retinal neovascularisation. Therefore, disruptions in arginine and ROS metabolism pathways may be closely associated with increased pathological neovascularisation in ROP. In comparison with the R-WT group, we found that TCA cycle metabolites, such as shikimic acid, were decreased in the R-OIR group, suggesting disturbances in the Warburg effect, which generates key intermediates such as pyruvate and lactate, with less glucose metabolism channelled into the TCA cycle. Simultaneously, our findings revealed an increase in pyruvate-related metabolites, indicating an upregulation of aerobic glycolysis. This metabolic shift could lead to reduced conversion of pyruvate to acetyl-CoA, thereby limiting the substrate supply for the TCA cycle ([120]online supplemental figure 12). This phenomenon aligns with the vigorous proliferation of retinal neovascularisation in OIR ECs.[121]^39 In the redox homeostasis pathway, glutamate and glutathione(GSH) levels were significantly higher in the R-OIR group compared with the R-WT group ([122]online supplemental figure 12). We hypothesise that the R-OIR group experiences increased OS, which may promote the maintenance of GSH redox status.[123]^40 These observations suggest that glycometabolism and redox homeostasis are altered in OIR mice, leading to the accumulation of metabolites like nicotinamide adenine dinucleotide phosphate(NADPH) and GSH and reduced levels of metabolites involved in the phosphate pentose pathway and TCA cycles. These changes may be closely linked to the onset and progression of ROP. Combined with feature selection ([124]online supplemental figure 8), shikimic acid and PC(14:1(9Z)/20:2(11Z,14Z) emerged as potential biomarkers capable of distinguishing between the two groups. Although further validation is required, these biomarkers could have important clinical implications and may be useful for preliminary screening of insidious and difficult-to-diagnose cases of ROP. Limitations and outlook This study’s small sample size may affect the accuracy of the findings, but the OPLS-DA model demonstrated significant differences between the OIR and WT groups, showing robustness. A limitation is the lack of longitudinal assessments, such as mice weight gain and the correlation of retinal neovascularisation severity, which impacts result comprehensiveness. Another one is the relatively small litter size, which may have influenced the neovascularisation outcomes. Maternal stress and cannibalism of pups after returning to room air were contributing factors. In future experiments, we will consider increasing litter size or using surrogate nursing dams housed in room air to ensure consistent pup development and body weight.[125]^18 41 Additionally, the body weight of the pups was relatively high (9–9.5 g at P17), and signs of neovascular regression may have already occurred, resulting in a smaller neovascular area. As a result, the retinal samples may not strictly represent the stage of maximal retinal neovascularisation, which typically occurs at P17 in the OIR model. This factor may have influenced the interpretation of our findings. Future studies will focus on optimising experimental conditions to better capture the maximal neovascularisation stage. Although the body weight and neovascular area do not strictly match the standard OIR model, a comparison of Zhou et al’s and Tomita et al’s findings revealed significant consistency in metabolomic profiles.[126]^42 Despite potential differences in the number of differential metabolites, the overall direction of changes, functional classifications and enriched pathways remains similar, supporting the robustness of our findings. Several biomarkers have been associated with ROP; however, the speed and practicality of detection methods are crucial for premature infants. Future research should incorporate bioinformatics, genomics and proteomics to identify more ROP-related genes and improve diagnosis and treatment strategies. To our knowledge, this is the first study to simultaneously analyse differential metabolites in OIR kidney and vitreous using liquid chromatography-mass spectrometry (LC-MS) and compare them with previous retinal and blood metabolomics. The findings highlight arginine metabolism as crucial for ROP development. Continued exploration of these pathways could lead to targeted interventions that improve outcomes for preterm infants at risk of ROP. Conclusions This study demonstrated consistent alterations in renal metabolism that align with metabolomics analyses across other tissues (vitreous, retina and blood) and platforms. The results suggest that OIR may lead to both kidney damage and retinal neovascularisation. Analysing indicators of renal injury could aid in diagnosing retinal neovascular diseases such as ROP. Our findings suggest that arginine biosynthesis plays a crucial role in the development of both neonatal kidney injury and ROP. Furthermore, shikimic acid and pimelic acid may serve as highly specific biomarkers for ROP. In summary, this study identified key metabolites and pathways within a kidney-retina axis that may protect against pathological neovascularisation. Our results suggest a functional interaction between renal and ocular metabolomics, including those of the retina and vitreous. Targeting arginine biosynthesis could provide valuable insights for managing ROP and improving outcomes for affected infants. Supplementary material online supplemental file 1 [127]bmjophth-10-1-s001.pdf^ (2.3MB, pdf) DOI: 10.1136/bmjophth-2024-001955 Acknowledgements