Arachidonic acid metabolism regulates the development of retinopathy of prematurity among preterm infants

Abstract

Extremely preterm infants are at risk of developing retinopathy of prematurity (ROP), characterized by neovascularization and neuroinflammation leading to blindness. Polyunsaturated fatty acid (PUFA) supplementation is recommended in preterm infants to lower the risk of ROP, however, with no significant improvement in visual acuity. Reasonably, this could be as a result of the non-consideration of PUFA metabolizing enzymes. We hypothesize that abnormal metabolism of the arachidonic acid (AA) pathway may contribute to severe stages of ROP. The present study investigated the AA-metabolizing enzymes in ROP pathogenesis by a targeted gene expression analysis of blood (severe ROP = 70, No/Mild = 56), placenta (preterm placenta = 6, full term placenta = 3), and human primary retinal cell cultures and further confirmed at the protein level by performing IHC in sections of ROP retina. The lipid metabolites were identified by LC–MS in the vitreous humor (VH; severe ROP = 15, control = 15). Prostaglandins D2 (p = 0.02), leukotrienes B5 (p = 0.0001), 11,12-epoxyeicosatrienoic acid (p = 0.01), and lipid-metabolizing enzymes of the AA pathway such as CYP1B1, CYP2C8, COX2, and ALOX15 were significantly upregulated while EPHX2 was significantly (0.04) downregulated in ROP cases. Genes involved in hypoxic stress, angiogenesis, and apoptosis showed increased expression in ROP. An increase in the metabolic intermediates generated from the AA metabolism pathway further confirmed the role of these enzymes in ROP, while metabolites for EPHX2 activity were low in abundance. Inflammatory lipid intermediates were higher compared to anti-inflammatory lipids in VH and showed an association with enzyme activity. Both the placenta of preterm infants who developed ROP and hypoxic retinal cultures showed a reduced expression of EPHX2. These findings suggested a strong involvement of EPHX2 in regulating retinal neovascularization and inflammation. The study results underscore the role of arachidonic acid metabolism in the development of ROP and as a potential target for preventing vision loss among preterm-born infants.

1. Introduction

Retinopathy of prematurity (ROP) is a major ocular complication of preterm birth among children and is characterized prominently by abnormal blood vascularization and neurodegeneration in the retina, that eventually leads to fibrovascular proliferation and retinal detachment, causing total vision loss very early in life. Approximately 10%–18% of preterm infants born annually in high-income countries and 40% in low- and middle-income countries develop ROP, of which about 18% have significant vision loss (Blencowe et al., 2013; Rivera et al., 2017; Sanghi et al., 2018). Preterm infants, once diagnosed with plus ROP (characterized by the development of abnormal blood vessels on a partial avascular retina), are treated with laser and/or anti-VEGF (Vascular Endothelial Growth Factor) therapy (Hartnett, 2015). Currently available treatment options for ROP mainly aim at preventing the progression of the disease with some known side effects and show variable responses to retinal maturation and functions. Besides, preterm infants who develop ROP are at a higher risk of developing secondary ocular complications, including retinal detachment, myopia (nearsightedness), strabismus (crossed eyes), amblyopia (lazy eyes), and glaucoma. Many studies attempted to demonstrate the underlying cellular and molecular mechanisms in the pathogenesis of ROP, but it is still not fully understood (Kaur et al., 2022; Rivera et al., 2017). Moreover, it is yet unclear why only a subset of mild ROP cases progress to severe ROP while in others the disease spontaneously regresses to mature retina and the role of maternal factors. Mounting evidence thus far has indicated a complex interplay between metabolism and pathways regulating hypoxia, inflammation, and angiogenesis (Chen & Smith, 2007; Fu et al., 2022; Hartnett & Lane, 2013; Patnaik et al., 2021).

The deficiency of ω-3 polyunsaturated fatty acids (PUFAs), particularly DHA (docosahexaenoic acid), has been reported as a potential contributor to ROP pathology (Fu et al., 2022; Lapillonne & Moltu, 2016). Several randomized control trials performed on PUFA supplementation for preterm infants showed a trend toward benefit in mitigating severe forms of ROP; however, there was no change observed after a prolonged follow-up for the visual acuity at the age of 2.5 years (Liu et al., 2013; Welch et al., 2007). Plausibly, this could be as a result of the non-consideration of PUFA metabolizing enzymes and PUFA metabolism. The macular layer in the retina is rich in long-chain PUFAs such as arachidonic acids (AA), DHA, eicosapentaenoic acid (EPA), etc., which are prone to frequent lipid peroxidation and structural modification leading to the loss of neurons (Wang et al., 2019). The role of PUFAs and lipid-metabolizing enzymes such as cyclooxygenases, lipoxygenases, and hydrolases has been widely studied in cancer and cardiovascular diseases; however, their involvement in ROP has been less explored. An increased activity of COXs (Cyclooxygenases) and LOXs (Lipoxygenases) on AA generates prostaglandins (PGs), thromboxane A2 (TXA2), leukotrienes (LTs), and hydroxy eicosatetraenoic acid (HETEs), which are known to contribute to inflammation and hypertension (Snodgrass & Brune, 2019; Zemski Berry et al., 2014). Further, AA is also acted upon by a group of CYPs (Cytochrome P450) enzymes (ω-hydroxylases and epoxygenase pathways) generating HETEs and eicosatetraenoic acid (EETs), which regulate some cellular processes of carcinogenesis and progression, including cell proliferation, survival, angiogenesis, invasion, and metastasis. The EETs are mainly metabolized by soluble epoxide hydrolase (sEH/EPHX2) to the corresponding diols or dihydroxyeicosatrienoic acids (DHETs). Hypoxia is extensively explored as a major factor in the development and progression of ROP and has a significant association with abnormal fatty acid metabolism. However, the role of PUFA-metabolizing enzymes in regulating angiogenesis has not been explored in human patients. Recently, a few studies using oxygen-induced retinopathy mouse models have shown the prominent role of EPHX2, a major enzyme in the AA metabolism, in astrocyte and endothelial cell proliferation, leading to new vessel growth in the retina (Hu et al., 2019). Further, the reports on the role of EPHX2 are variable, and it is unclear if they confer protection or contribute to abnormal angiogenesis under hyperoxia/hypoxia. Based on the available literature and preliminary data on global gene expression data from ROP infants, we hypothesized that loss of EPHX2 activity in the growing retina alters PUFA metabolism, leading to abnormal angiogenesis via Notch signaling. Thus, the present study is an attempt to systematically understand the role of lipid-metabolizing enzymes in the AA pathway in contributing to abnormal angiogenesis and neuronal loss in ROP. This was achieved by integrating targeted transcriptomics and fatty acid metabolomics in premature infants. The study also attempts to understand the interplay between cellular and signaling pathways regulating lipid metabolism, inflammation, angiogenesis, and neurodegeneration in ROP development and progression by analyzing multiple tissue types such as blood, vitreous humor (VH), preterm retina, and placenta. Aberrant fatty acid metabolism among preterm-born infants may be a key event in the development and progression of ROP, and, thereby, could offer a major therapeutic option for ROP.

2. Materials and Methods

2.1. Study subjects

The study protocol adhered to the principles of the Declaration of Helsinki and was approved by the Institutional Review Board (LEC-BHR-P-11-22-959) for VH, retina, and blood (LEC-BHR-01-20-380 for placental samples) of the L V Prasad Eye Institute (LVPEI), Hyderabad. Preterm babies referred for further management from the neonatal intensive care units across the city to the LVPEI were enrolled in the study. The study cohort for blood samples comprised 126 preterm infants of GA ≤35 weeks and/or BW ≤1700 g with severe ROP (n = 70) and no/mild ROP (n = 56). The study cohort for VH samples comprised 15 preterm babies with severe ROP and 15 controls with GA ≥35 weeks and/or BW ≥2500 with congenital cataracts. The inclusion and exclusion criteria for placenta sample collection are provided in Table T2 and are also mentioned in detail in another manuscript (Kaur et al. bioRxiv 2023.02.13.528236). The included preterm infants were thoroughly examined by an ophthalmologist on the 21st day of life for retinal vascular status as part of the routine screening done by the retina specialists of LVPEI. Based upon the examination, the preterm infants were further divided into preterm without ROP and preterm with ROP groups. The detailed demographic and clinical history of all the infants/patients recruited for the study was documented in a pre-designed proforma, and written informed consent was obtained from their guardians prior to the sample collection.

2.1.1. Criteria applied for the diagnosis and progression of ROP

Diagnosis and categorization of ROP cases were done based on severity (stages 1–5), location (zones I, II, and III), amount of disease (clock hours), and presence or absence of “plus” disease following ICROP guidelines. Severe ROP includes progressive disease and a highly vascularized retina, which requires prompt treatment. The no/mild ROP cases include less severe disease, which does not require any treatment and has no neo-vascularization changes. The preterm infants with no/mild ROP and no neo-vascularization changes in the retina were considered control, while infants with stage 4 and stage 5 plus who had severe neo-vascular changes and/or retinal detachment were categorized as severe ROP in this study. All infants enrolled in the study were regularly followed for disease status until the disease completely regressed. Infants on steroid treatment and with a history of anti-VEGF were excluded. In addition to blood samples and VH, placenta samples from preterm and full-term-born infants (at the time of delivery) and retina tissue (at the time of evisceration) were also collected from a subset of patients for understanding the lipid metabolism in utero and in severe stages of ROP, respectively. Infants included for placenta analysis were thoroughly examined and followed up for at least six months for the development of ROP/normal vascularization.

2.2. Sample collection

Venous blood (0.5–1.5 mL) was collected from the preterm infants diagnosed with severe ROP and controls by venipuncture. Placental tissue (decidua representing the maternal–fetal interface) was collected based on gestational age as per study criteria and categorized into preterm (n = 6) and full-term (n = 3). RNA was extracted from the blood and placental samples using Trizol and an automated cDNA extraction platform following the manufacturer's guidelines. Likewise, for metabolomics studies, the VH samples (400 μL) were collected from preterm infants with stage IV and V ROP (n = 15) who had undergone vitrectomy as a part of their routine clinical management. All the infants enrolled in this study had a minimum of 6 h of fasting prior to the surgery to normalize the effect of external nutrition on their metabolic profile (Toms & Rai, 2019). The controls for the metabolomics studies included infants with congenital cataracts (<6 months of age) who underwent partial vitrectomy as part of their surgical management (n = 15). The mean age at the VH sample collection was non-significant to minimize its maturation time post-birth. Retina from patients with ROP and an age-matched control with anterior staphyloma was collected during evisceration.

2.3. Materials

All the primers for semi-quantitative PCR were synthesized by Bioserve Biotechnologies (India) Pvt Ltd. The rabbit polyclonal anti-body against EPHX2 (RRID:AB_10263545) was purchased from R&D Biosystems Inc. (RRID:SCR_006140), and the mouse monoclonal antibody anti-Glial fibrillary acidic protein (GFAP; RRID:AB_2314537) was purchased from Cell Signaling Technology (RRID:SCR_002071) for immunohistochemistry (IHC). Secondary antibodies for IHC were from LI-Cor Biosciences. All other chemicals (unless otherwise specified) were from Sigma-Aldrich (RRID:SCR_008988).

2.4. Reverse transcription and semi-quantitative PCR

RNA from blood samples of severe ROP and no/mild ROP was extracted by the QIAamp RNA Blood Mini kit method (QIAGEN (RRID:SCR_008539); Hilden, Germany). The RNA from the placenta and human primary retinal cell culture was extracted using TRIzol. The quantity and quality of extracted RNA were assessed by Thermo NanoDrop (concentration range = 100–1000 mg/μL, A260/230 = 2.0 ± 0.05). cDNA was prepared using the iScript cDNA synthesis kit (Bio-Rad Laboratories (RRID:SCR_008426), CA, USA). Semi-quantitative PCR was carried out using the specific primers for AA metabolism genes (CYP1B1, CYP2C8, COX2, ALOX15, and EPHX2), angiogenic genes (VEGF165/189, VEGFR2, DLL4, PSEN1, APH1B, and NOTCH1), hypoxia-inducible factor gene (HIF1A), and cell death marker genes (CASP3 and CASP8). All primers were designed using Primer3 software (RRID:SCR_003139). The expression of each gene was normalized to that of β-actin as an endogenous control (Table S1).

2.5. IHC and hematoxylin and eosin (H&E) staining of retina

To determine the expression of EPHX2 and GFAP in ROP eyes from preterm infants, retinal tissue from ROP and an age-matched control were collected at the time of evisceration. The retinal tissues were washed with 1X PBS thrice. The sclera, choroid, retinal pigment epithelium (RPE), Buch’s membrane, along with vessels in these layers, was removed and the neural retina was immersed in 4% formalin for 24 h. The tissue was later embedded in a paraffin block and stored until use. The tissue section of 10 μm from the paraffin block was made on a charged glass slide and deparaffinized. Sections were cut, air dried, and stained with H&E (Fischer et al., 2008) to understand their morphology. The antigen retrieval was performed using ammonium acetate buffer, followed by brief heating. For whole-mount IHC, the retina was fixed in 4% paraformaldehyde (PFA) for 20 min at room temperature and washed with PBS buffer thrice. After isolation, tissues were blocked and permeabilized in 2% BSA and 0.5% Triton X-100 at room temperature for 1 h. The primary antibodies for EPHX2 (1:300) and GFAP (1:300) were diluted in 2% BSA and kept at 4°C overnight. Afterward, Alexa Fluor–coupled secondary antibodies (1:300) were used. Cell nuclei were visualized with DAPI (1:200; Molecular Probes, Thermo Fisher Scientific USA). Human skin was used as a positive control for EPHX2 (SF2) (Naeem et al., 2024). All high-resolution images were visualized and captured with a confocal microscope (Carl Zeiss AG, Oberkochen, Germany; Scale: 20 μm).

2.6. Preparation of human mixed retinal cell cultures and hypoxia induction

The detailed human mixed retinal cell cultures were established from cadaveric donor retinas, as previously explained (Shahulhameed et al., 2020; Institutional Review Board: LEC 02-14-029). Briefly, retinal tissue was chopped into small pieces with the help of a sterile surgical blade and then washed with 1X PBS. The dissociated pieces of retina were subjected to trypsinization using 1X trypsin EDTA (0.25%) for 15 min at 37°C. The reaction was stopped by adding a complete DMEM (Dulbecco's Modified Eagle's Medium) containing 10% serum and 1% antibiotics, followed by resuspending these in 2 mL of PBS and gently mixing. The tubes containing resuspended cells were centrifuged, and the supernatant was removed to obtain a clear cell pellet, which was again suspended in DMEM. The suspended cells were then seeded in a sterile tissue culture grade T-75 mm flask and left undisturbed under standard cell culture conditions for 7 days, with the medium changing every third day. The newly grown cells were seeded (n = 15 000/well) on a glass coverslip and allowed to grow for 70%–80% confluency; 150 μM of CoCl₂ was used for 24 h in serum-deprived medium to induce hypoxia. The unexposed cells in a serum free medium (24 h) were used as a control. After hypoxia treatment, the cells were isolated for RNA extraction.

2.7. Metabolomics analysis

Global VH metabolome data for ROP and controls were generated in our lab using the previously described protocol (Kumar et al., 2023; Patnaik et al., 2019). The VH samples collected during surgery were kept on ice and immediately brought to the lab, and metabolite extraction was performed within 30 min, snap-frozen in liquid N₂, and then stored at −80°C until further use. All the samples were run in triplicate, and the metabolites were considered for analysis only if they were present in more than 80% of all the samples and replicates to maintain quality control. Further, we have performed baseline correction with the blank prior to injecting the sample and running the blank in between the runs. The mass spectrometry was calibrated, and tuning was done frequently between runs to achieve adequate quality control for the data generated. The abundance value underwent a log₂ transformation. The dataset was further mined for lipid-specific metabolites with a 15-ppm difference for annotation using XCMS (RRID:SCR_015538), and identified in the Human Metabolome Data Bank (HMDB; RRID:SCR_007712). Metabolites generated by the activity of AA pathway genes were mapped, and a predictive analysis for the EPHX2 gene and its interactions with AA-derived metabolite {LC–MS (liquid chromatography-mass spectrometry) data} was performed using the MetScape-Cytoscape analysis tool.

2.8. Statistical analysis

The data are represented in terms of the mean and the standard error mean (SEM). The significance of the differences between the groups was analyzed using the two-tailed Student's T-test. The qPCR expression was determined using cycle threshold (C_t) values calculated for each test gene obtained for each sample using SDS2.3 software, and the mean fold change was calculated using 2^−ΔΔCt. We have calculated the relative mRNA expression based on fold change by using individual ΔC_t of the control and severe ROP following recent published articles (Diebold et al., 2019; Harshitha & Arunraj, 2021). To study the variance among the categories, the degree of freedom, t-value, F-value, and actual p-value were calculated. For IHC quantification, we utilized ImageJ Fiji software. Lipid metabolites were annotated using the Metlin XCMS online tool in both positive and negative ion adduct mode (http://metlin.scripps.edu). The post-hoc power analysis for determining the effect size for this study was performed using a dichotomous endpoint and two independent sample groups based on the incidence of severe ROP in India (https://clincalc.com/stats/Power.aspx) and showed a score of 92.5%. The data were not assessed for normality, and no test for outliers was conducted. Statistical significance was considered at p ≤ 0.05. Statistical analyses were carried out using GraphPad Prism version 9.0.0 (RRID:SCR_002798).

3. Results

3.1. Study cohort

A total of 126 preterm infants were recruited and analyzed for gene expression profiling. Table 1 provides the demographic details for all the study subjects. The blood sample cohort showed severe ROP infants had significantly lower gestational age and birth weight as compared to those with no/mild ROP. The study population for the metabolomics study from VH samples included 30 infants (ROP: 15; controls: 15).

Table 1. Demographic details of the patients recruited.

Variable	Severe ROP	Control/mild ROP	p-value
(a) Blood
Total infants, n	70	56	-
Male, n (%)	37 (52.8)	33 (58.9)	-
Female, n (%)	33 (47.2)	23 (41.1)	-
Mean BW (in grams) SD	1309.1 398.17	1514.4 415.6	0.00623
Mean GA (in months) SD	29.08 2.03	32.10 2.20	<0.00001
(b) VH
Total infants, n	15	15	-
Male, n (%)	7 (46.7)	9 (60)	-
Female, n (%)	8 (53.3)	6 (40)	-
Age (in months) at the time of surgery SD	3.83 1.12	3.4 1.7	n.s.

Note: (a) Details of the patients and controls recruited for gene expression profiling in blood. (b) Details of the patients and controls recruited for the metabolite profiling in the VH.

Abbreviations: BW, birth weight; GA, gestational age; n.s, non-significant; ROP, retinopathy of prematurity; SD, standard deviation.

3.1.1. Targeted gene expression profiling among severe ROP and control cases shows significant alterations in the expression of the enzymes involved in AA metabolism

Taking a cue from our earlier results obtained from a previous published study on proteomics and genomics analysis of ROP (Rathi et al., 2017) and to have a detailed understanding of the role of AA metabolism in ROP pathogenesis, we first compared the expression of key genes in this pathway, including COX2, ALOX15, CYP1B1, CYP2C8, and EPHX2, among preterm infants with severe ROP and control (Figure 1). Four of the prominent genes/enzymes involved in AA metabolism, including COX2, ALOX15, CYP1B1, and CYP2C8, were significantly upregulated (2.5-fold, p = 0.05; 3.9-fold, p = 0.03; 5.8-fold, p = 0.007; and 2.0-fold, p = 0.05, respectively), while EPHX2 was significantly downregulated (0.56-fold, p = 0.04; Figure 1).

Figure 1.

Arachidonic acid metabolism: Classification of different metabolites formed from AA by different enzymatic pathways. The cyclooxygenase and lipoxygenase pathways are mediated by the activity of COXs forming prostaglandins and thromboxanes and LOXs forming HETEs, HPETEs, and lipoxins, respectively. Epoxygenase and ω-hydroxylase pathways are mediated by the activity of CYPs, forming classes of EETs, HETEs, and HPETEs, respectively. EPHX2 acts on the epoxygenase-generated metabolites to form dihydroxy alcohols (n-DHETs). (a) qRT-PCR analysis of transcripts for COX2 (D_f = 42; t-value = 2.01; n = 21 preterm babies with severe ROP; n = 23 preterm babies with no/mild ROP; p = 0.0503) (b) qRT-PCR analysis of transcripts for CYP2C8 (D_f = 59; t-value = 2.01; n = 25 preterm babies with severe ROP; n = 36 preterm babies with no/mild ROP; p = 0.0504); CYP1B1 (D_f = 60; t-value = 2.80; n = 26 preterm babies with severe ROP; n = 36 preterm babies with no/mild ROP; p = 0.0072); and qRT-PCR analysis of transcripts for EPHX2 (D_f = 124; t-value = 2.07; n = 70 preterm babies with severe ROP; n = 56 preterm babies with no/mild ROP; p = 0.0415). (c) ALOX15 expression (D_f = 42; t-value = 2.16; n = 21 preterm babies with severe ROP; n = 23 preterm babies with no/mild ROP; p = 0.0503). All the graphs are representative of experimental triplicates. β-Actin was used as the normalization control for qRT-PCR. Statistical significance is represented as * and ** for p ≤ 0.05 and p ≤ 0.01, respectively.

3.1.2. Reduced expression of EPHX2 further dysregulates angiogenic genes via upregulating notch signaling

EPHX2 is known to regulate Notch signaling, which in turn regulates angiogenesis and cell death. Therefore, to confirm the reduced expression of EPHX2 and its implication in ROP, we compared the quantitative expression of angiogenic genes, VEGF165, VEGF189, PSEN1, APH1B, DLL4, and NOTCH1 in preterm infants with severe ROP and with no/mild ROP (Figure 2). Pro-angiogenic genes, including VEGF165 and VEGF189 and its receptor VEGFR2 showed increased expression (6.0-fold, p = 0.001; 5.7-fold, p = 0.02; 14.5-fold, p = 0.007). The γ-secretase component, which cleaves the notch receptor, APH1B, was significantly upregulated (2.3-fold, p = 0.035, respectively), while PSEN1 was significantly (0.4-fold, p = 0.04) downregulated. DLL4, which is a ligand for notch activation and negatively regulates VEGF, did not show any change. NOTCH1, which acts as a regulator of angiogenesis, was also significantly (3.0-fold, p = 0.003) upregulated.

Figure 2. mRNA expression levels of angiogenic genes in blood.

(a) qRT-PCR analysis of transcripts for VEGF165 (D_f = 49; t-value = 2.24; n = 18 preterm babies with severe ROP; n = 33 preterm babies with no/mild ROP; p = 0.0265; (b) VEGF189 (D_f = 48; t-value = 3.42; n = 17 preterm babies with severe ROP; n = 33 preterm babies with no/mild ROP; p = 0.00163); (c) VEGFR2 (D_f = 66; t-value = 3.42; n = 30 preterm babies with severe ROP; n = 38 preterm babies with no/mild ROP; p = 0.00163) expression; (d) qRT-PCR analysis of transcripts for PSEN1 (D_f = 50; t-value = 2.07; n = 29 preterm babies with severe ROP; n = 23 preterm babies with no/mild ROP; p = 0.0474); (e) APH1B (D_f = 54; t-value = 2.19; n = 26 preterm babies with severe ROP; n = 30 preterm babies with no/mild ROP; p = 0.0355) expression (f) qRT-PCR analysis of transcripts for DLL4 (D_f = 66; t-value = 0.634; n = 38 preterm babies with severe ROP; n = 30 preterm babies with no/mild ROP; p = 0.5308) (g) qRT-PCR analysis of transcripts for NOTCH1 (D_f = 54; t-value = 3.08; n = 26 preterm babies with severe ROP; n = 30 preterm babies with no/mild ROP; p = 0.0039). All the graphs are representatives of experimental triplicates. β-Actin was used as the normalization control for qRT-PCR. Statistical significance is represented as * and ** for p ≤ 0.05 and p ≤ 0.01, respectively.

The gene expression of apoptotic genes, CASP3 and CASP8 was assessed among severe ROP and control samples (Figure 3). Both CASP3 and CASP8 were significantly upregulated (2.2-fold, p = 0.01- and 1.5-fold, p = 0.004). Hypoxia-inducible factor 1A (HIF1A) was significantly increased in cases (8.4-fold, p = 0.0009). These results clearly suggest that a reduced expression of EPHX2 affects its major activity of converting EETs to DHETs, which might further contribute to neovascularization via notch signaling. Also, increased EETs would further increase VEGF.

Figure 3. mRNA expression levels of hypoxia-inducible factor and apoptotic genes in blood.

(a) qRT-PCR analysis of transcripts for HIF1A (D_f = 66; t-value = 3.58; n = 30 preterm babies with severe ROP; n = 38 preterm babies with no/mild ROP; p = 0.00095) expression; (b) qRT-PCR analysis of transcripts for CASP3 (D_f = 56; t-value = 2.51; n = 27 preterm babies with severe ROP; n = 31 preterm babies with no/ mild ROP; p = 0.0156); (c) CASP8 (D_f = 53; t-value = 3.05; n = 25 preterm babies with severe ROP; n = 30 preterm babies with no/mild ROP; p = 0.00423) expression. All the graphs are representative of experimental triplicates. β-Actin was used as the normalization control for qRT-PCR. Statistical significance is represented as * and ** for p ≤ 0.05 and p ≤ 0.01, respectively.

3.2. Functional implications of differential expression of lipid-metabolizing enzymes in the eyes of ROP patients

Next, to assess the functional effect of differential expression of lipid-metabolizing enzymes on retinal pathology, their activity was estimated from the metabolites generated in the VH samples of ROP patients and controls. These data generated in our lab are being analyzed to identify the key mechanisms involved in ROP pathogenesis (a manuscript submitted to another journal). Of the 1031 total identified metabolites in the VH, lipids constituted approximately 40% of the total differentially expressed metabolites among cases and controls. A total of 69 metabolites involved in fatty acid metabolic pathways showed differential expression between ROP and control (Figure 4a, Table 2).

Figure 4. AA metabolites found in VH and lipid metabolite interactions with its genes.

(a) Abundance of fatty acid metabolites in the VH. The graphical representation of the abundance of metabolites detected in the VH of severe ROP compared to control. The corresponding x-axis represents the Metline ID for the metabolites detected. n = 15 preterm babies with severe ROP; n = 15 full-term babies with no ROP. (b) Network mapping of EPHX2 and lipid-derived metabolites: the EPHX2 gene showing strong association with CYPs and other lipid-metabolizing enzymes. (c) Network mapping of CYP450 and lipid-derived metabolites: Different CYP450 family gene interactions with its metabolites. The network map shows the interactions among different metabolites and their associations with the genes mapped. All the graphs are representatives of biological duplicates. All graphs were prepared using GraphPad Prism version 9.0.0. Statistical significance is represented as *, **, and *** for p ≤ 0.05, p ≤ 0.01, and p ≤ 0.001, respectively.

Table 2.

List of metabolites and their corresponding METLIN IDs observed against peak is observed in the VH. XCMS-derived METLIN ID annotated against their retention time and with 15 ppm error was used for the analysis, and the identified features were both analyzed in positive mode and negative mode.

METLIN ID	Retention time	m/z	Metabolites	Enzyme involved	Fold change	p-value	t-value	Degree of freedom (Df)
M341T34	27.06	335.22	(15S)-Prostaglandin A2; Prostaglandin B2; Prostaglandin-C2; Delta-12-Prostaglandin J2; AA (hydroperoxide); Leukotriene B5	LOXs/COXs	1.69	<0.001	5.42	28
M351T30	29.59	351.25	9-deoxy-9-methylene-PGE2; 11-deoxy-11-methylene-PGD2; PGA1 methyl ester; 5-HETE methyl ester	COXs/ non-enzymatic	1.74	0.02	5.73	23
M353T23	23.21	353.22	13,14-Dihydro-15-keto-PGD2, DG [PGE2]; 8-iso-15-keto-PGF2a; Prostaglandin E2; 13,14-Dihydro-15-keto-PGE2; Lipoxin B4	LOXs/COXs	4.32	<0.0001	10.88	27
M331T21	23.21	331.25	Trihydroxy-9-octadecenoic acid; 9,10,13-TriHOME; 9,12,13-TriHOME	Non-enzymatic	1.3	n.s.	1.51	28
M366T31	31.43	366.32	Prostaglandin F1a-d9	COXs	10.15	<0.001	7.55	25
M287T33	21.41	283.11	11,12-Epoxyeicosatrienoic acid (EETs)	CYPs	1.71	<0.01	6.30	25
M196T4	33.35	287.22	Hexadecanedioic acid	Non-enzymatic	3.68	<0.001	1.95	26
M339T27	26.85	339.25	5,6-Dihydroxy-8,11,14-eicosatrienoic acid; 8,9-dihydroxy-5,11,14-eicosatrienoic acid (DHETs)	EPHX2/sEH	−2.75	0.002	3.64	27

p-value and fold change are denoted in the respective columns.

3.2.1. Combined analysis of the EPHX2 gene and metabolic interactions

To delineate the underlying mechanisms involving EPHX2 in the pathogenesis of ROP, as shown above, an interactive map for the EPHX2 gene and epoxy-derived metabolites was constructed using Metscape (Cytoscape plugin). The EPHX2 gene has both direct and indirect interactions with several pro-inflammatory metabolites and CYP genes (CYP1B1, CYP2C8, etc.). Most of these interactions are involved in the different lipid metabolic pathways, namely AA metabolism, di-unsaturated fatty acid beta-oxidation, leukotriene metabolism, linoleate metabolism, putative anti-inflammatory metabolite formation from EETs, and xenobiotic metabolism pathways (Figure 4b,c).

3.3. Expression of EPHX2 is modulated by hypoxia in vitro and in utero

We next assessed if the reduced expression of the EPHX2 is triggered or regulated by hypoxia by exposing primary retinal cells to hypoxia (Figure 5). The expression of EPHX2 gene in human primary mixed retinal culture showed a downregulation under hypoxia as compared to normal oxygen-supplemented cultures (0.11-fold, p-value = 0.002). Further, since the placenta provides oxygen and nutrition in utero, we tested if compromised growth/anatomical changes in the placenta leading to ischemia could modulate the EPHX2 expression, which may predispose these infants to develop ROP postnatally. Compared to the placenta from a full-term delivery, expression of EPHX2 was reduced in the placenta from preterm delivery who did not develop ROP (Figure 5); however, it was significantly reduced for those who developed ROP (fold change = 0.16; p-value = 0.00003).

Figure 5. Effect of hypoxia and prematurity on EPHX2 expression.

(a) qRT-PCR analysis of transcripts for EPHX2 in human primary retinal culture under hypoxia. The hypoxia was induced using 150 μM CoCl₂, and the control was the primary culture exposed to normal room temperature oxygenation (D_f = 4; t-value = 21.05; p = 0.0031). For each independent experimental (3 biological set) set for human primary retinal culture, 15 000 cells/well were seeded. (b) qRT-PCR analysis of transcripts for EPHX2 full-term placenta, n = 3; preterm placenta with no-ROP, n = 3; and preterm placenta from the infants who later developed ROP, n = 3. (One-way ANOVA; F(2, 6) = 40.28; p = 0.0003). All the graphs are representative of experimental triplicates. β-Actin was used as the normalization control for qRT-PCR. Statistical significance is represented as *, **, and *** for p ≤ 0.05, p ≤ 0.01, and p ≤ 0.001, respectively.

3.4. Reduced EPHX2 expression in the severe stage of ROP retina corroborates its role in ROP pathogenesis

Next, we assessed the retinal structural and morphology on H&E staining in the retina from a control and severe stage of a human preterm infant, which showed proper retinal layers in the control retina as compared to the ROP retina. The H&E section of the ROP retina showed loss of lamination (arrangements of photoreceptors, neurons, and other cell types in different layers), gliosis with more fibrotic cells, and vascular proliferation. Clearly, the retinal layers in ROP sections were indistinguishable compared to the controls. These phenotypic changes could be attributed to the altered metabolite levels and high inflammation. Lastly, we tested if the altered expression of EPHX2 could potentially lead to the progression of ROP by checking its expression in the astrocytes and other glial cells of retinal sections. An intense positive staining of the astrocyte/glial marker GFAP was seen in the ROP retina, which was relatively higher than that seen in the control retina. EPHX2 showed no expression in the retina in ROP subjects with very few neuronal cells. Among the control retina sections, a higher expression of EPHX2 was observed near the blood vessel region (yellow arrow) with astrocytic cells expressed in GCL and NFL (Figure 6).

Figure 6.

(a) H&E staining of the control retinal panel showing different retinal layers and blood vessels. (Scale: 200 μm; Magnification: 20×; Microscope: EVOS M5000 Imaging System) (b) EPHX2 and GFAP expression in the control retina showing expression of GFAP in the blood vessels lining, near the NFL and GCL toward the VH, while EPHX2 expression is observed at the IPL and blood vessels lining. (c) Zoom-out section of the control retina showing the expression of EPHX2 colocalized with GFAP in the blood vessels lining. (d) H&E staining of the ROP retina shows no distinct retinal layer and loss of different retinal cells. (Scale: 200 μm; Magnification: 20×; Microscope: EVOS M5000 Imaging System) (e) EPHX2 and GFAP expression in the retina of ROP infants. The image shows an almost lack of expression of EPHX2, while significantly higher expression of GFAP across the retina. (f) Zoom-out section of the ROP retina showing very high expression of GFAP in the blood vessels lining but no expression of EPHX2 (g) Quantification of EPHX2 (D_f = 5; t-value = 2.56; n = 3 for ROP retina; n = 4 for control retina; p = 0.05005) (h) Quantification of GFAP (D_f = 5; t-value = 5.91; n = 3 for ROP retina; n = 4 for control retina; p = 0.0101). ELM – External Limiting Membrane; ONL – Outer Nuclear Layer; OPL – Outer Plexiform Layer; INL – Inner Nuclear Layer; IPL – Inner Plexiform Layer; GCL – Ganglion Cell Layer; NFL – Nerve Fiber Layer; BV– Blood Vessels. The arrow shows blood vessels lining. The experiments were performed thrice with a similar protocol. A retinal section with no primary antibody was used as a negative control for normalization and human skin as a positive control for EPHX2 (Figure SF2). Scale: 20 μm; Magnification: 64X; Microscope: Carl Zeiss AG. Statistical significance is represented as * and ** for p ≤ 0.05 and p ≤ 0.01, respectively.

4. Discussion

Multiple key factors, including oxidative stress, inflammation, neovascularization, and apoptosis, are known to affect the overall development and maturation of the retina in preterm-born infants and contribute to ROP pathogenesis (Hartnett & Lane, 2013; Rivera et al., 2017). In this study, we have shown that these factors are modulated by several lipid-metabolizing enzymes and angiogenic genes. The expression of genes coding for the major enzymes in the AA pathway, COX2, ALOX5, CYP2C8, and CYP1B1 showed a significant upregulation at severe stages, along with their increased activity as reflected by the increased metabolite levels of inflammatory prostanoids such as PGs, Prostaglandin B2, Thbx, LTs, Leukotriene B5, Lipoxin B4, 11,12-EETs, and HETEs (first-order metabolites of AA pathway) in the ROP VH (Table 2; Figure 4). Previously, it was shown that EETs metabolism upon the induction of COX-2 tends to be proangiogenic in endothelial cells and might be mediating its proangiogenic activity through the VEGFR-2/3 signaling pathway by positive feedback of VEGF (Cheranov et al., 2008; Rand et al., 2019; Webler et al., 2008). The results of the present study confirm the same among ROP and no-ROP preterm infants.

Further, the present study also demonstrated a reduced EPHX2 expression yielding low PUFA-derived diols (5,6 DHET and 8,9 DHET; Table 2). EPHX2-derived diols are required to protect astrocytes from mitochondrial-mediated apoptosis to prevent the development of ROP (Hu et al., 2014; Hu et al., 2019). Studies by Hu et al. also suggested that supplementation of 19,20-DHDP, but not the parent epoxide, was able to rescue the defective angiogenesis in sEH^−/− mice. Interestingly, among diabetic retinopathy (DR) cases, the inhibition of sEH was reported to prevent or delay the DR progression (Hu et al., 2017). It is well known that sEH is a bifunctional enzyme (N-terminal has phosphatase activity and C-terminal has epoxygenase activity), and its expression can be altered under phase I and phase II of ROP development (Harris & Hammock, 2013; Morisseau & Hammock, 2013). The N-terminal activity of sEH negatively regulates VEGF expression via nitric oxide synthase (Hou et al., 2012). An increased PGE2 (Table 2), as seen in this study, might be involved in mTORC activation leading to VEGF production, causing cell proliferation and angiogenesis, as shown in previous reports (Dufour et al., 2014; Wang et al., 2021). Likewise, based on the published reports by other groups (Luo et al., 2016; Nguyen et al., 2016; Spector & Norris, 2007), we speculate that increased EETs and HETEs formed by the activity of increased CYP1B1 and CYP2C8 may cause endothelial-mesenchymal transition leading to fibrosis as seen in the ROP retina (Figure 6). Further, exploring the angiogenic and VEGF pathways showed that the alterations in AA metabolites corroborate nicely with aggressive vascularization, as seen among severe ROP patients. We also found a significant upregulation of HIF1A (which regulates VEGF signaling) among severe ROP and controls. Hypoxic insults resulted in increased HIF1A expression in microglial cells, causing changes in cellular morphology and cell death (Dhyani et al., 2023). Concurrently, VEGF165, a pathogenic isoform of VEGF, was increased in comparison to VEGF189 required for normal vascularization (Zhou et al., 2007). VEGF is known to activate and bind to different receptors to mediate its various effects. Binding of VEGF165 and VEGFR-2 was shown to be associated with pathologic features in a rat retinopathy model (Hartnett, 2015). This suggested that increased VEGF165 binding to increased VEGFR2 (Figure 2) may have significant implications in the vascular phases of human ROP. The results of the present study clearly demonstrated the same among ROP and no-ROP preterm infants. A high VEGF level increases several other inflammatory and proangiogenic factors, such as cytokines and ECM proteases, as revealed in our earlier findings (Patnaik et al., 2021; Rathi et al., 2017). Delta-like ligand 4 (Dll4) is induced by VEGF as a negative regulator of angiogenic sprouting; however, we did not see any significant change in DLL4 expression. This could be explained based on the study by Lobov et al. (2007), which demonstrated that DLL4, which is known to act as feedback or “brake” for increasing VEGF, does not seem to regulate its expression at severe ROP (Lobov et al., 2007). Besides, Notch signaling may directly and/or indirectly regulate angiogenesis through crosstalk with VEFGRs (Garcia & Kandel, 2012; Hellström et al., 2007). DLL4, which is a ligand for the notch receptor, is cleaved by γ-secretases and ECM proteases (ADAM10); however, both act independently of each other (Hartmann et al., 2002; Hu et al., 2019; Sorensen & Conner, 2010). Further, other evidence from OIR models also showed a similar finding that DLL4 levels were not changed by fatty acid supplementation. However, the Notch cleavage is mediated by PSEN1 of γ-secretase activity (Hu et al., 2014). Differential expression of PSEN1 and APH1B (both being components of γ-secretases; Figure 2) contributes to the cleavage of the notch, thereby increasing angiogenesis (Brunkan & Goate, 2005; Kopan & Ilagan, 2009). Notch, being a developmental pathway regulator, may be compromised among ROP infants because of an incomplete developmental cycle, as corroborated by relatively high-notch signaling. Increased hypoxia activates Notch-responsive promoters and increases expression of Notch-direct downstream genes (Gustafsson et al., 2005), which further strengthens the findings of our study.

Prematurity and oxidative insults are key factors in ROP progression and development, which were also assessed using in vitro human retinal primary cell culture models and placenta. Under both circumstances, EPHX2 was seen downregulated. Maternal factors such as implantation, pregnancy, and labor are known to change metabolic reprogramming, which could trigger an early onset of ROP, with the additional impact of oxidative insults (Szczuko et al., 2020). Placental genomics and RNA expression profiles are known to mediate genetic associations with complex health traits and diseases (Bhattacharya et al., 2022; Rasmussen et al., 2022). A detailed study by our group shows the involvement of maternal–fetal factors, placental histopathology, and molecular alterations related to hypoxia, inflammation, and complement activation at the maternal–fetal interface in preterm ROP placentas (Kaur et al., 2023). Retinas from ROP infants showed increased fibrosis, gliosis, cell proliferation, and an increased number of blood vessels (Figure 6). The different layers of the retina could not be distinguished in ROP infants because of a marked loss of photoreceptors and ganglion cells compared to the age-matched control retina. Further, EPHX2 was significantly downregulated in ROP retina, validating our findings on placental and blood gene expression at severe ROP. We also observed that there was no significant change in EPHX2 expression (fold change: 1.02; p-value: 0.39; Figure S1) when the cohort was grouped on the basis of birth weight (birth weight: ≤1200 g; >1200 g). However, we believe post-birth weight gain among infants would play a significant role in its expression, which could not be studied because of the unavailability of the desired data from follow-ups. Our study cohort included infants with different treatments and duration of treatment, which could have an impact on the study results; however, this will be confirmed by further functional validation in an appropriate animal model. Further validation of the gene expression across different stages of ROP would have been ideal for studying a severity-dependent change in lipid-metabolizing enzymes and their activity; however, this was not possible because of the non-availability of VH samples from mild ROP. Astrocytes and other glial cells express EPHX2 in the retina, thus performing a comparative expression of EPHX2 across different cell types among cases and controls by immunostaining could have helped in distinguishing its diseased vs normal state, but this too could not be done because of the non-availability of the retina in normal pediatric donors. Nevertheless, notwithstanding these limitations, the study still very clearly demonstrated the involvement of fatty acid (PUFAs)-metabolizing enzymes in ROP, from placental to retinal changes. A major limitation of the present study is that the absolute quantification of most of the metabolites could not be performed as LC-MS/MS fragmentation could not be curated. However, adducts were analyzed in both positive and negative modes to exclude isobaric lipids. The absolute levels of some other prominent lipid metabolites could not be elucidated, as these did not qualify the threshold of 15 ppm mass tolerance. These are warranted in future studies to clearly establish their role in ROP pathogenesis.

5. Conclusions

In summary, a systematic correlation of the expression of lipid-metabolizing enzymes in the AA pathway and their activity with the genes involved in angiogenesis as seen in severe ROP patients (present study) confirmed that altered lipid metabolism leading to pro-inflammatory conditions could be a potential mediator of neovascularization in the eyes of infants with severe ROP (Figure 7). The findings of this study would aid in discovering lipid-based biomarkers and therapeutics for prognostic testing and detecting newer drug targets for efficient management of ROP. This study can further be extended to study the response of mothers to ω-3 fatty acid supplementation during the prenatal and early postnatal stages to identify the infants who would benefit from this treatment. A recent clinical study showed improved visual acuity among ROP infants with ω-3 fatty acid supplementation (Lundgren et al., 2023). Identifying and treating the disease at an early onset stage would certainly help in the appropriate and effective management of the disease and its progression.

Figure 7. Schematic diagram of the mechanism of the proposed hypothesis of AA-metabolizing enzymes and its potential effect on angiogenesis, cell death, and inflammation.

Data Availability Statement

All data generated or analyzed during this study are included in this published article and is available upon request. The Preprint of this article was posted on BioRXiv, May 16th 2022; https://www.biorxiv.org/content/10.1101/2022.05.13.491711v1.full.pdf.

Abbreviations

AA

arachidonic acid

ALOX

activator of LOX

APH1B

Aph-1 homolog B

BV

blood vessels

CASP

caspases

COXs

cyclooxygenases

CYP1B1

cytochrome P450 family 1 subfamily B member 1

CYP2C8

cytochrome P450 family 2 subfamily C member 8

DHA

docosahexaenoic acid

DLL4

delta-like ligand4

DMEM

Dulbecco’s Modified Eagle’s Medium

EETs

eicosatetraenoic acid

ELM

external limiting membrane

EPA

eicosapentaenoic acid

EPHX2

epoxide hydrolase2

GCL

ganglion cell layer

GFAP

glial fibrillary acidic protein

H&E

hematoxylin and eosin

HETEs

hydroxy eicosatetraenoic acids

HIF1A

hypoxia-inducible factor 1 alpha

IHC

immuno-histochemistry

INL

inner nuclear layer

IPL

inner plexiform layer

LA

linoleic acid

LCMS

liquid chromatography mass spectrometry

LOXs

lipoxygenases

LTs

leukotrienes

NFL

nerve fiber layer

OIR

oxygen-induced retinopathy

ONL

outer nuclear layer

OPL

outer plexiform layer

PBS

phosphate buffer saline

PGs

prostaglandins

PSEN1

presenilin 1

PUFAs

polyunsaturated fatty acids

ROP

retinopathy of prematurity

RPE

retinal pigment epithelium

RRID

research resource identifiers

TXA2

thromboxane A2

VEGF

vascular endothelial growth factor

VEGFR

vascular endothelial growth factor receptor

VH

vitreous humor