Differentiation of soluble aqueous humor metabolites in primary open angle glaucoma and controls

Abstract

We report an analysis of the aqueous humor (AH) metabolome of primary open angle glaucoma (POAG) in comparison to normal controls. The AH samples were obtained from human donors [control (n=35), POAG (n=23)]. The AH samples were subjected to one-dimensional ¹H nuclear magnetic resonance (NMR) analyses on a Bruker Avance 600 MHz instrument with a 1.7mM NMR probe. The same samples were then subjected to isotopic ratio outlier analysis (IROA) using a Q Exactive orbitrap mass spectrometer after chromatography on an Accela 600 HPLC. Clusterfinder Build 3.1.10 was used for identification and quantification based on long-term metabolite matrix standards. In total, 278 metabolites were identified in control samples and 273 in POAG AH. The metabolites identified were fed into previously reported proteome and genome information and the OmicsNet interaction network generator to construct a protein-metabolite interactions network with an embedded protein-protein network. Significant differences in metabolite composition in POAG compared to controls were identified indicating potential protein/gene pathways associated with these metabolites. These results will expand our previous understanding of the impeded AH metabolite composition, provide new insight into the regulation of AH outflow, and likely aid in future AH and trabecular meshwork multi-omics network analyses.

Graphical Abstract

1. Introduction

Glaucoma refers to a group of irreversible blinding optic neuropathies that affect approximately 70 million people worldwide (Quigley and Broman, 2006). Primary open angle glaucoma (POAG) is the most common form of the disease, which is frequently associated with elevated intraocular pressure (IOP). Elevated IOP in glaucomatous eyes is consistent with findings of impeded aqueous humor (AH) outflow due to pathologic changes in the anterior chamber angle. The trabecular meshwork (TM), a filter-like structure in the anterior chamber angle, is thought to be a site of resistance to aqueous humor outflow (Morrison and Acott, 2003). However, resistance to AH outflow may arise distal to the TM, further downstream (Carreon et al., 2017).

Our understanding of the dynamic nature of AH, and its generation and egress from the anterior eye chamber has advanced through several decades (Mark, 2010). AH is actively secreted by the ciliary body (CB) and/or its non-pigmented epithelium (Civan and Macknight, 2004). AH outflow occurs through the conventional pathway in the anterior chamber angle structures that involve TM or through the uveoscleral pathway that involves the CB. The TM has been subjected to extensive investigation to understand how its properties undergo pathologic changes in glaucoma. This includes multi-omics analysis: genomics, proteomics (Bhattacharya et al., 2005), lipidomics, (Aribindi et al., 2013a; (Aribindi et al., 2013b) and metabolomics. Comprehensive analysis of the AH proteome, lipidome and metabolome are expected to provide complementary insight into pathologic changes associated with outflow resistance. Proteomics, lipidomics, (Edwards et al., 2014a; (Edwards et al., 2014b) and metabolomics (Buisset et al., 2019; (Mayordomo-Febrer et al., 2015) studies of AH have been conducted. Due to the large number of isomeric lipids and metabolites, multiple studies of AH lipidomics and metabolomics are desirable. A comparison of CB secretions (Margolis et al., 2018) with AH has suggested that the metabolite composition of AH may likely change as AH passes through the anterior segment as ocular tissues absorb and release metabolites into the AH. This is consistent with emerging localized studies of lipids and metabolites using imaging mass spectrometry (Guerra et al., 2015). As noted above, the possible involvement of distal outflow regions and all tissues of the anterior eye segment towards AH level balance and IOP homeostasis is increasingly being realized.

Non-destructive nuclear magnetic resonance (NMR) analysis of AH followed by destructive analysis using mass spectrometry to yield a greater confidence quantification of metabolites and possible wider identification, remains yet to be done. The metabolome is thought to be the most transient repertoire of all molecular omics, therefore several independent approaches to compare identifications is desirable. This was the rationale behind our use of NMR and mass spectrometry. The isotopic ratio outlier analysis combined with mass spectrometry is a recent method that enables a higher confidence in metabolite identification compared to other mass spectrometric methods, which was the rationale behind using this method. The experimental metabolome can be compared with a previously identified proteome hub for AH and TM to provide insights into our understanding, which is also lacking in the current literature, motivating us to undertake the studies presented here.

2. Materials and Methods

2.1. Human subjects/donors/chemicals

All materials were collected from human donors without identifiers under institutional review board exemption/approval. We acquired 58 aqueous humor (AH) samples from participants at the Veterans Administration (VA) Medical Center (Miami, FL). Patient samples used for ¹H-Nuclear Magnetic Resonance (NMR) and Isotopic Ratio Outlier Analysis (IROA) consisted of individuals with POAG (n=23) and non-POAG controls (n=35). Patient information is presented in Table 1. All the chemicals utilized in this study were used without further purification, unless otherwise stated. Deuterated water (D₂O) and the internal standard D6–4,4-dimethyl-4-silapentane-1-sulfonic acid (DSS) were procured from Cambridge Isotope Laboratories, MA, USA.

Table 1:

Aqueous humor human subjects information

Control
Sample Number	Age	Gender	Right/Left Eye	Ophthalmic Medications	Other Diseases	Other Medications	Cataract Present (Cataract/No Cataract)
N-4		Male	Left	Polymyxin B, Cyclopentolate, Flurbiprofin, Phenylephrine, Lidocaine	Coronary Artery Disease	Fluticosone	Cataract
N-5	75	Male	Right	Cyclopentolate, Flurbiprofen, Phenylephrine, Polymyxin B, Tetracaine, Lidocaine			Cataract
N-12	62	Male	Left	Polymyxin B, Tetracaine, idocaine	Dyslipidemia, Vitamin D Deficiency		Cataract
N-14	64	Male	Right	Ketorolac, Polymyxin B, Cyclopentolate, Flurbiprofen, Gentamicin, Phenylephrine, Tetracaine, Lidocaine	Diabetes Mellitus, Hyperlipidemia, Hypertension		Cataract
N-19	71	Male		Cyclopentolate, Flurbiprofen, Phenylephrine, Polymyxin B, Tetracaine, Lidocaine	Hyperlipidemia		Cataract
N-28	75	Male	Right	Cyclopentolate, Flurbiprofen, Phenylephrine, Polymyxin B, Tetracaine, Lidocaine	Dyslipidemia, Hypertension, Psoriasis	Carvedilol, Ravitiding, Amplodipine	Cataract
N-29	70	Male		Polymyxin B, Cyclopentolate, Flurbiprofin, Phenylephrine, Lidocaine			Cataract
N-30	52	Male	Right	Cyclopentolate, Flurbiprofen, Phenylephrine, Polymyxin B, Tetracaine, Lidocaine	Diabetes Mellitus, Hyperlipidemia	Insulin, Metformin, Saxagliptin, Amoxicillin	Cataract
N-42	57	Male	Right	Polymyxin B, Cyclopentolate, Flurbiprofin, Phenylephrine, Lidocaine			Cataract
N-43	74	Male	Left	Polymyxin B, Cyclopentolate, Flurbiprofin, Phenylephrine, Lidocaine	Presbyacusis, Hyperlipemia	Albuterol, Atorvastatin, Budenoside	Cataract
N-44	65	Male	Left	Cyclopentolate, Flurbiprofen, Phenylephrine, Polymyxin B, Tetracaine, Lidocaine		Famotidine	Cataract
N-45	69	Male	Left	Cyclopentolate, Flurbiprofen, Phenylephrine, Polymyxin B, Tetracaine, Lidocaine			Cataract
N-46	75	Male	Left	Polymyxin B, Cyclopentolate, Flurbiprofin, Phenylephrine, Lidocaine			Cataract
N-47	72	Male	Right	Cyclopentolate, Flurbiprofen, Phenylephrine, Polymyxin B, Tetracaine, Lidocaine			Cataract
N-48	75	Male	Left	Polymyxin B, Cyclopentolate, Flurbiprofin, Phenylephrine, Lidocaine	Hyperlipidemia, Diabetes Mellitus, Hypertension		Cataract
N-49	91	Male	Right	Flurbiprofen, Phenylephrine, Polymyxin B, Tetracaine, Lidocaine	Hypothyroidism	Lisinopril, Metoprolol, Simvastin	Cataract
N-50	71	Male	Right	Cyclopentolate, Flurbiprofen, Phenylephrine, Polymyxin B, Tetracaine, Lidocaine	Diabetes Mellitus, Hypertension, Kidney Cancer, Macular Degeneration	Atorvastatin, Terazopam, Albuterol	Cataract
N-51	82	Male	Left	Polymyxin B, Cyclopentolate, Flurbiprofin, Phenylephrine, Lidocaine			Cataract
N-55	61	Male	Left	Flurbiprofen, Phenylephrine, Polymyxin B, Tetracaine, Lidocaine			Cataract
N-57	72	Male	Right	Polymyxin B, Cyclopentolate, Flurbiprofin, Phenylephrine, Lidocaine			Cataract
N-58	77	Male	Right	Flurbiprofen, Phenylephrine, Polymyxin B, Tetracaine, Lidocaine			Cataract
N-59	59	Male	Left	Polymyxin B, Cyclopentolate, Flurbiprofin, Phenylephrine, Lidocaine			Cataract
N-60	64	Male	Left	Cyclopentolate, Flurbiprofen, Phenylephrine, Polymyxin B, Tetracaine, Lidocaine			Cataract
N-61	64	Male	Right	Polymyxin B, Cyclopentolate, Phenylephrine, Flurbiprofen Lidocaine			Cataract
N-62	73	Male		Polymyxin B, Cyclopentolate, Flurbiprofin, Phenylephrine, Lidocaine			Cataract
N-64	75	Male		Polymyxin B, Cyclopentolate, Flurbiprofin, Phenylephrine, Lidocaine, Tetracaine			Cataract
N-65	68	Male		Polymyxin B, Flurbiprofin, Cyclopentolate, Phenylephrine, Lidocaine			Cataract
N-66	85	Male	Right	Polymyxin B, Cyclopentolate, Flurbiprofin, Phenylephrine, Lidocaine			Cataract
N-67	71	Male	Left	Flurbiprofen, Polymyxin B, Phenylephrine, Tetracaine, Lidocaine			Cataract
N-68	64	Male	Right	Polymyxin B, Cyclopentolate, Flurbiprofin, Phenylephrine, Lidocaine			Cataract
N-69	68	Male	Left	Cyclopentolate, Flurbiprofen		Cyanocobalamin, Diclofenac Na, Pravastatin	Cataract
N-70	72	Male	Right	Flurbiprofen, Phenylephrine, Polymyxin B, Tetracaine, Lidocaine		Cyanocobalamin, Insulin, Losartan, Meloxicam, Vanlafaxine, Pregabaline, Bisacodyl, Polyethylene, Hydrocortisone, Morphine Sulfate, Naproxen, Hydrocodone	Cataract
N-72	82	Male	Left	Ketorolac Tromethamine, Polmyxin B/Trimethoprim, Prednisolone, Cyclopentolate, Flurbiprofin, Phenylephrine, Lidocaine		Asprin, Clopidogrel, Tamsulosin, Finasteride, Vitamin D, Metoprolol, Hydrocodone	Cataract
N-73	77		Right	Cyclopentolate, Flurbiprofin, Phenylephrine, Polymyxin B, Lidocaine	Diabetes Mellitus	Glipizide, Apixaban, Ergocalciferol, Atorvastatin, Metoprolol Tartrate, Terazosin, Amoxicillin,	Cataract
N-75	72	Male	Left	Polymyxin B, Cyclopentolate, Flurbiprofin, Phenylephrine, Lidocaine		Ammonium Lactate, Citalopram Hydrobromide, Mupirocin Ointment, Urea Cream, Tramadol	Cataract
POAG
G-10	86	Male	Right	Lidocaine, Tetracaine, Polymyxin B, Flurbiprofen, Cyclopentolate, Phenylephrine, Latanoprost	Benign essential hypertension, Hyperlipidema		No Cataract
G-12	84	Male	Left	Mitomycin, Polymyxin B, Tetracaine, Lidocaine, Dorzolamide, Brimonidine	Dyslipidemia, Vitamin D deficiency		No Cataract
G-16	64	Male		Brimonidine, Polymyxin B, Dorzolamide, Tetracaine, Latanoprost, Lidocaine	Mixed Hyperlipidemia		No Cataract
G-18	54	Female		Latanoprost, Polymyxin B, Dorzolamide, Tetracaine, Lidocaine	Prediabetes		No Cataract
G-19	85	Male	Left	Dorzolamide, Latanoprost			No Cataract
G-20	73	Male	Right	Latanoprost, Brimonidine, Cyclopentolate, Flurbiprofen, Phenylophrine, Polymyxin B, Tetracaine, Lidocaine	Diabetic neuropathy, Type 2 diabetes mellitus		No Cataract
G-26	71	Male	Right	Dorzolamide, Brimodinine, Latanoprost	Ocular Hypertension, Hyperlipemia, Diabetes		No Cataract
G-27	61	Male	Right	Latanoprost, Brimonidine, Polymyxin B, Lidocaine	Mellitus		No Cataract
G-29	73	Male	Left	Brimonidine, Latanoprost, Polymyxin B, Tetracaine, Lidocaine, Cyclopentolate			No Cataract
G-30	79	Male	Left	Latanoprost, Brimodinine, Fluriprofen, Phenylephrine, Lidocaine			No Cataract
G-31	76	Male		Dorzolamide, Latanoprost, Polymyxin B, Flurbiprofen, Tetracaine, Lidocaine			No Cataract
G-33	73	Male	Left	Latanoprost, Brimonidine, Tetracaine, Polymyxin B/Trimethamine, Lidocaine,			Cataract
G-34	69	Male	Right	Dorzolamide, Latanoprost, Polymyxin B/Trimethoprim, Predonisolone, Cyclopentolate, Flurbiprofen, Phenylephrine, Tetracaine, Lidocaine			Cataract
G-35	77	Male	Right	Ketorolac Tromethamine, Polymyxin B/Trimethoprin, Prednisolone, Latanoprost, Cyclopentolate, Flurbiprofin, Phenylephrine, Lidocaine		Ketoconazole, Urea, Allopurinal, Carvedilol, Pantoprazole	Cataract
G-36	82	Male	Left	Polymyin B/Trimethoprim, Cyclopentolate, Flurbiprofin, Phenylephrine, Lidocaine	Alzheimers disease	Edoxaban, Cyanocobalamin, Metoprolol Succinate, Vitamin B, Donepezil, Atropine Sulfate	Cataract
G-38	78	Male	Right	Polymyin B/Trimethoprim, Cyclopentolate, Flurbiprofin, Phenylephrine, Lidocaine	Diabetes Mellitus, Systemic Hypertension, Hypercholesterolemia	Levothyroxine, Allopurinol, Atrovastatin, Metfromin, Cyanacobalamin, Vitamin D, Cephalexin, Clohidine, Hydrochlorothyazide, Losartan	No Cataract
G-39	66	Male	Left	Brimonidine, Dorzolamide, Latanoprost, Cyclopentolate, Flurbiprofin, Phenylephrine, Polymyxin B/Trimethoprim, Lidocaine		Terazosin, Sildenafil, Lisinopril	No Cataract
G-40	70	Male	Left	Ketorolac tromethamine, Polymyxin B/Trimethoprim, Dorzolamide HCl, Latanoprost, Olopatadine HCl, Brimonidine	Diabetes Mellitus, Carpal Tunnel, Neurological Disorder associated with Diabetes Mellitus Type 2	Insulin, Amlodipine, Gabapentin, Metformin, Fluorouracil 5% Cream, Silvadene Cream, Doxycycline monohydrate, Losartan, Rosuvastatin CA	No Cataract
G-41	71	Female	Right	Dorzolamide HCl, Timolol Maleate, Latanoprost, Cyclopentolate, Flurbiprofin, Phenylephrine, Polymyxin B/Trimethoprim, Lidocaine		Potassium Chloride, Aspirin, Fumotidine, Furosemide, Lisinopril, Loratadine, Nifedipine, Allopurinol, Atorvastatin Calcium, Gabapentin	No Cataract
G-42	95	Female	Right	Ketorolac Tromethamine, Polymyxin B/Trimethoprim, Latanoprost, Cyclopentolate, Flurbiprofin, Phenylephrine, Lidocaine	Systemic Hypertension	Amlodipine, Furosemide, Levothyroxine, Metoprolol, Potassium Chloride	No Cataract
G-43	65	Male	Right	Brimonidine, Dorzolamide HCl, Latanoprost, Prednisolone, Cyclopentolate, Flurbiprofin, Phenylephrine, Polymyxin B/Trimethoprim, Lidocaine			Cataract
G-44	73	Male	Right	Prednisolone, Dorzolamide, Polymyxin B/Trimethoprim, Atropine, Cyclopentolate, Flurbiprofin, Phenylephrine, Lidocaine	Diabetes Mellitus Type 2	Sildenofil, Aspirin, Atorvastatin, Calcium, Glipizide, HCTZ, Insulin, Amlodipine	No Cataract
G-45	71	Male	Left	Ketorolac Tromethamine, Polymyxin B/Trimethoprim, Latanoprost. Surgery-Cyclopentolate, Flurbiprofin, phenylephrine, Lidocaine.	Prediabetes	Amlodipine	No Cataract

2.2. Nuclear magnetic resonance (NMR) Spectroscopy

The AH samples were subjected to ¹H NMR spectroscopy (Ravanbakhsh et al., 2015) at the University of Florida (Gainesville, FL) for the identification of metabolites using One dimensional (1D) ¹H NMR. No extraction was performed on these aqueous humor samples prior to NMR analysis. Each sample was centrifuged for 30 minutes at 13,200 rpm at 4°C. Supernatant was transferred to a new microcentrifuge tube in an ice bath. Then 50 µL of NMR sample was prepared by using 45 µL of the supernatant of AH sample with 5 µL of D₂O containing 1.11 mM DSS, resulting in a final concentration of DSS to 0.11 mM. The mixture was further centrifuged at 13,200 rpm at 4°C for 15 minutes before acquiring NMR on the resulting supernatant portion.

NMR spectra were acquired using a Bruker NMR 14.1 T magnet equipped with the Avance Neo console and 1.7 mm TCl CryoProbe (Bruker BioSpin Corporation, Billerica, MA). All 1D ¹H spectra were acquired with zgpurge gprefocus4 pulse sequence (Le Guennec et al., 2017), at 25 °C. One for each sample, 256 scans (nt) were accumulated on a 9615.4Hz spectral width (sw) with 65536 data points (Td). A 90-degree pulse width (pw) was used with 2 s of relaxation delay (d1) and 1.42 s of acquisition time (at). The pre-saturation of the water signal was achieved by irradiating at low power radiation on resonance (during d1).

All NMR spectra were processed using MestReNova 14.0.1–23559 (Mestrelab Research, S.L., Santiago de Compostela, Spain). For each of the ¹H NMR spectra, an exponential line-broadening function of 0.5 Hz was applied followed by Fourier transformation, phasing, spline baseline correction and calibration to DSS resonance at 0.0 ppm. All 1D ¹H spectra were normalized with respect to the DSS signal before integrated areas were extracted for selected chemical shifts. These integrated areas were utilized as raw data to determine concentrations of selected metabolites with respect to the DSS resonance signal. All NMR metabolite identification data was uploaded to metabolomics workbench [Project ID ST001284]. Verification of the metabolite assignments were done by collecting Heteronuclear Single Quantum Coherence (HSQC) and Total Correlated Spectroscopy (TOCSY) spectra for a subset of the AH samples (control and POAG).

2.3. Isotopic Ratio Outlier Analysis (IROA)

Following NMR spectroscopy, the samples were prepared for isotopic ratio outlier analysis (IROA) (Beecher and de Jong, 2019; (Stupp et al., 2013). First, 800 µL of precipitate solution (8:1:1 Acetonitrile: Methanol: Acetone) was added to 50 µL of AH. The metabolites were vortexed and incubated at 4 °C for 30 minutes. This was followed by another incubation period at −20 °C for 1 hour. Each sample was centrifuged at 20,000 x g for 10 minutes at 4 °C (Beckman Microfuge 18) to form a pellet and 375 µL of supernatant was collected. The supernatant was dried in a speed vacuum for approximately 20 minutes or until the sample was fully dry and stored at −20 °C. The IROA internal standard (IROA-IS, IROA technologies) (U-95% 13C) was reconstituted in 1.2 mL of LC-MS grade water with 0.1 % formic acid (FA). The samples were reconstituted in 25 µL of LC-MS grade water with 0.1% FA and 10 µL was combined with 20 µL of IROA-IS and subjected to LC-MS/MS analysis. A long-term reference standard (LTRS) was reconstituted in 50 µL of LC-MS grade water with 0.1 % FA and subjected to LC-MS/MS as the first, last, and every 10th sample.

Metabolite identification and relative quantification were performed using Clusterfinder Build 3.1.10 (IROA Technologies). Thermo raw files were converted into mzxml files before importation into the Clusterfinder software. The manufacturer protocols were followed to recognize IROA peak pairs and determine molecular formulas. Metabolites were identified by comparing retention time, molecular formula and molecular ion m/z with the Mass Spectrometry Metabolite Library of Standards (MSMLS, IROA technologies).

2.4. High Performance Liquid Chromatography (HPLC)

Reversed-phase chromatographic separation utilized an Accela autosampler and an Accela 600 pump (Thermo Fisher Scientific, Waltham, MA) and an ACE Excel 2 C18-PFP (100 × 2.1 mm, 2 μm) column. The flow rate was 350 μL/min and the injection volume was 5 μL. The column temperature and tray temperature were 40°C and 4°C respectively. Solvent A was composed of water with 0.1% formic acid (FA) and solvent B was composed of acetonitrile with 0.1 % FA. All solvents were of LC-MS grade. The gradient was 0–40% solvent B over 10 minutes, 40% solvent B over 10–12 minutes, 60% solvent B over 12–14 minutes, 95% solvent B over 14–20 minutes, 0% solvent B over 20–27 minutes, 0% solvent B was held for 27–30 minutes.

2.5. Mass Spectrometry

Metabolites were analyzed by a Q Exactive Orbitrap Mass Spectrometer (Thermo) equipped with a heated electrospray ionization source (HESI) in both positive and negative ionization modes. The HESI source operated at a spray voltage of 4 kV, HESI vaporization temperature of 300 °C, and heated capillary temperature of 325 °C and 310°C for positive and negative ionization modes respectively. The sheath gas flow rate was 45 (negative mode) and 40 (positive mode). The auxiliary gas flow rate was 10 (negative mode) and 5 (positive mode). The S-lens radio frequency (RF) level was set to 30. Full scan used a scan range of 70–800 m/z with a resolution of 70,000, automatic gain control (AGC) target of 1 × 10⁶, and maximum injection time (IT) of 100 ms. data-dependent MS/MS used a loop count of 10, resolution of 17,500, AGC target of 5 × 10², maximum IT of 50 ms, isolation window of 2 m/z, and a collision energy of 30. All mass spectrometric metabolite identification data is available at metabolomics workbench [Project ID ST001278].

¹H NMR and IROA thus served as independent validations of the identification of metabolites and estimates of their concentration. A subset of AH samples was also subjected to parent ion or neutral loss scan of selected metabolites using triple quadrupole TSQ Quantum Max as yet another independent approach for validation (Bhattacharya, 2013).

2.6. Statistical Analysis

The array of POAG metabolite concentrations found through both ¹H NMR and IROA was normalized against the metabolite concentrations of the control samples to find the fold change. The mean fold change, along with the standard deviation were then calculated. Fold changes were calculated using only common metabolites. Common metabolites are those, which made it into at least one sample in each group. For unique metabolites, they should have all in one group (at least two occurrences) and none in the other. Significant fold changes were determined by a two-tailed t-test with equal variance in Excel. We used an F-test to confirm the use of a t-test with homogenous variances. The significant metabolites fold changes were plotted adjacent to the control sample metabolites in GraphPad Prism 8.2.1.

The normalized results were fed into Metaboanalyst 4.0 (Xia and Wishart, 2011) and log₂ transformed to ensure a normal distribution of metabolites. Principal Component Analysis (PCA) was employed to explain the percentage of variance between the two groups. A 2D plot showing the grouping and separation of the two conditions as a function of the first and second principal components was generated. Metaboanalyst was also used to generate heat maps showing the over and under expression of metabolites as a function of the sample.

Partial least squares-discriminant analysis (PLS-DA) was used to classify the samples based on their attributes (metabolite concentration fold change profile). It utilized dimensionality reduction, along with discriminant analysis to group the samples into their respective classes. PLS-DA also generated a list of the attributes that scored highest in terms of variable importance in projection (VIP). These metabolites provided the most weight to the classification model.

2.7. Metabolite-proteome interaction maps

A literature search was conducted to ascertain the significance of differentially expressed proteins and genes found in the TM and the AH (Liton et al., 2006; (Youngblood et al., 2019). The gene and protein names were converted to Entrez IDs. These Entrez IDs were then fed into the OmicsNet interaction network generator to construct networks showing the protein-protein and protein-metabolite interactions (Zhou and Xia, 2018, 2019). OmicsNet helps map input genes and proteins to their known Kyoto Encyclopedia of Genes and Genome (KEGG) identities, and then to the known KEGG pathways. It then reconstructed a 3D visualization of enriched pathways. The minimum network was constructed to ensure clarity and conciseness. The network was translated from its 3D network on OmicsNet into a 2D representation in Cytoscape Version 3.7.2. Filtered networks of the two most over-expressed and under-expressed metabolite networks were further generated to highlight the interaction of these metabolites with neighboring protein nodes.

3. Results

3.1. Metabolomic Profiling of POAG Aqueous Humor

We performed high performance liquid chromatography (HPLC) tandem mass spectrometry (LC/MS/MS) and nuclear magnetic resonance (NMR) spectroscopy analysis on aqueous humor from 23 POAG and 35 non-POAG control patients. Table 1 shows the characteristics of the patients.

A total of 77 metabolites including unknowns were identified in NMR based on the characteristics of the chemical shifts. These metabolites were common to both control and POAG AH (Fig. 1). There were 206 metabolites identified in IROA LC-MS/MS, 191 of which were common to both control and POAG. The unpaired two-tailed T-test revealed 28 total significant metabolites between IROA LC-MS/MS and NMR combined (Fig. 1). There were 6 significant metabolites identified by isotopic ratio outlier analysis (IROA) and 22 significant metabolites identified by NMR. Of the total metabolites, only 5 metabolites were found unique in POAG and 10 in the control (Fig. 1), suggesting that only a small number of metabolites undergo a dramatic all or none change, the rest of the metabolites are common but with altered concentrations between the two groups (Fig. 2). The measurement of metabolites, when common between control and POAG by NMR and IROA mostly showed similar trends by both methods. For example, L-Lysine, was higher in POAG than control by both methods (Fig. 2 A, B).

Fig. 1.

Comparison of the total number of metabolites identified in aqueous humor (AH) from primary open angle glaucoma (POAG) and controls. ¹H-NMR spectroscopy, IROA, or all methods that were unique or common in POAG and control AH. Comparison of the number of unknowns, insignificant metabolites, and significant metabolites identified using ¹H-NMR spectroscopy or IROA that were unique to the control, POAG, or common in both AH groups.

Fig. 2.

Fold changes showing significant aqueous humor (AH) metabolite changes in POAG samples and controls. (A) Fold changes of metabolites identified using ¹H-NMR spectroscopy, (B) or IROA. Unknown metabolites expressing significance were excluded.

The Principle Component Analysis (PCA) showed no distinct grouping between POAG and control samples (Fig. 3). The PLS-DA, however, showed distinct grouping between POAG and control. The first and second principal components explained 16.5% of the variance between samples of both conditions (Fig. 3). The Q2 value is an estimate of the predictive ability of the model and is calculated via cross-validation. For 1 principal component, the Q2 value was 0.15. For two principal components the Q2 value was 0.05. There were 22 metabolites that had variable importance in projection value (VIP) score greater than 1 excluding unknowns (Fig. 3).

Fig. 3.

Statistical analysis of metabolites in aqueous humor (AH) from primary open angle glaucoma (POAG) and controls. (A) Principal component analysis (PCA) scores plot for discrimination of metabolites. (B) Partial least squares discriminant analysis (PLS-DA) for discrimination of metabolites. (C) Variable importance in projection (VIP) scores plot in the PLS-DA model of 22 metabolites with VIP values ≥ 1. Metabolites are those identified using ¹H-NMR spectroscopy and IROA, excluding unknowns. Metabolites identified using IROA are fully capitalized. Red and green ellipses are included to note the 95% confidence intervals of each group. N= control and G= POAG.

3.2. Metabolite Interpretation

Compared to controls, the metabolomics profile of POAG patients had a significant increase in abundance of 4 amino acids (lysine, arginine, cysteine, and glycine), anthranilate, ascorbate, 4-hydroxybenzoate, myo-inositol, acetate, propylene glycol, 2-hydroxy-butyrate, creatine and choline (Fig. 2). Lysine, acetate, arginine, glycine, and ascorbate concentrations are increased in POAG as has been previously reported (Buisset et al., 2019; (Hannappel et al., 1985; (Hysi et al., 2019; (Mayordomo-Febrer et al., 2015). Threonine was found to increase in POAG using NMR spectroscopy but decreased in POAG using IROA (Fig. 2 A, B). However, the differences in the level of L-Threonine is statistically more significant in IROA than NMR measurements. Cysteine, anthranilate, 4-hydroxybenzoate, myo-inositol, propylene glycol, 2-hydroxy-butyrate, creatine and choline have never been specifically associated with POAG AH. The metabolites that decreased in POAG patients were amino acids: phenylalanine and glutamine, 4-aminobutanoate and isopropanol (Fig. 2).

3.3. Networks

The metabolite-protein interactions are important for overall function of the cells and tissue. We used our experimentally determined metabolites and literature derived curated (reviewed) TM and AH proteomes to depict where the metabolite potential interactors reside in the protein-pathway. The results of the superimposition of the differentially expressed POAG AH metabolites and TM genes and proteins yielded a map that showed protein-protein and protein-metabolite interactions (Fig. 4). Using the log₂ fold changes of the metabolites, genes, and proteins, we were able to show which proteins and metabolites were over and under expressed graphically. This has been performed to integrate existing peer-reviewed data on genes and proteins with our experimentally identified metabolites. From this network, we then determined the top 5 over expressed genes: GARS, NOS2, CARS, TARS and HNF4A. The top 5 under expressed genes were: NOS3, SHMT1, SHMT2, DLD, and FARSB.

Fig. 4.

Interactome network of aqueous humor (AH) metabolites, trabecular meshwork (TM) proteins, and TM genes that are significant and differentially present in primary open angle glaucoma (POAG) compared to controls. (A) An expanded network of metabolite interactors. (B) A snapshot of the upregulated acetate network, and, (C) downregulated L-phenylalanine network. The B, C are examples of upregulated and downregulated network parts within total network. Triangles indicate metabolites, circles indicate genes and proteins. Red indicates upregulation while green indicates downregulation. An interactive version of this network can be viewed by uploading the supplemental GRAPHML file into either OmicsNet.ca or Cytoscape.

4. Discussion

We present the analysis of aqueous humor (AH) metabolites from non-glaucoma patients undergoing cataract surgery with otherwise normal healthy eyes (control) and from POAG patient eyes here. POAG patients also underwent cataract surgery, thus AH samples are not different with respect to cataract surgeries. Aqueous humor composition is dependent on metabolites produced during its generation plus the metabolic interchanges with other intraocular tissues during its passage (De Berardinis et al., 1966). AH is mainly composed of carbohydrates, amino acids, proteins, organic ions, inorganic ions, glutathione and water.

There are several factors that may contribute to changes in the metabolite composition in AH such as the time of collection, the severity of disease and the body temperature. In addition, there are diurnal fluctuations of IOP and AH outflow that could contribute to changes in the metabolites. It should be noted in this context that large spikes of diurnal fluctuations have been noted in POAG patients (Asrani et al., 2000). As depicted schematically in Figure 5A, B, a contributing factor to the variation in metabolites in the AH could be the location of the AH collection. Metabolites are secreted from different anterior segment regions (for example, lens, cornea, CB) (Margolis et al., 2018). If it is possible to compartmentalize these metabolites, one could potentially observe changes in the composition of the fluid. However, this is unlikely due to significant mixing by diffusion and eddy currents. In this study, the AH samples were collected within a timeframe of 8 AM to 3 PM. Although most of these variables could not be controlled for human subjects, we still found significant differences in the metabolite composition between POAG and control subjects (Fig. 1, 2). Since it is not possible to control several variables in human subjects, model organisms may be better suited. Are metabolite changes concomitant with aberrant AH dynamics and IOP fluctuations? Research along these lines in model organisms will provide answers to this question. The study of AH metabolites and multi-omics in model systems, however, will be helpful if changes in control AH and ocular hypotensive AH animals capture the changes that occur in humans. Given that all glaucoma patients/subjects are under IOP lowering medications, it is unclear if the changes in metabolites would be a function of the disease itself or treatments. However, this would be an issue for any study with human samples and is a limitation of such studies including this study. Our analyses included patients who are in diverse treatments and on monotherapies. If a metabolite or certain metabolites are commonly captured in POAG subjected to several different monotherapies, then logically they cannot be ascribed to one singular therapy as an effect of treatment by that singular therapeutic. However, one cannot rule out that different drugs targeting IOP reduction might converge on a few common pathways causing the same molecules to show up on metabolomics assays of multiple patients. We have presented in Supplemental Figure S2 and Figure S3 how these metabolites vary across patients.

Fig. 5.

Schematic illustration of the multifactorial influence on aqueous humor (AH) metabolite levels and potential confounding effects of asynchronous collection. (A) The metabolites secreted from different anterior segment regions (indicated by white arrows) may contribute to total metabolite repertoire of AH. As indicated by dashed lines specific ports may enable collection of AH with metabolites predominantly from a specific tissue region. In absence of significant mixing, metabolites secreted from corneal tissue may be predominantly sampled via port 1, similarly port 2, port 3 and port 4 may collect lens, ciliary body and trabecular meshwork metabolites, respectively. However, the flow of AH is likely non-ideal with significant mixing due to diffusion and eddy currents (indicated by semi-circular arrow). The grey shade has been used to depict 12 noon (12 PM) and the black to depict 12 midnight (12 am) as indicated. (B) The sample collected via same ports at midnight (12 AM) is expected to have differences in metabolites from that collected via same ports in (A) due to diurnal fluctuations. (C) Asynchronous collection and variation in metabolite levels. Representative factors: severity of disease state (0 and 1 indicate none/early and highly severe glaucoma), time of collection (12 AM-12 PM) and body temperature (between 36.5–37°C) that may contribute to variation in metabolite levels. Theoretically if the variation were for each of these factors individually then red blue and yellow dots will depict variation in metabolites. A relative spectrum of levels from low (0) to high for each factor has been provided to present the concept. Without synchronization of several factors, AH will demonstrate asynchronous metabolite variations. (D) Diurnal variations in IOP levels and outflow rate are known to exist, they show higher amplitude in glaucoma and thus should be expected to be coupled with metabolite variations. The dark and light cycles are in the figure as indicated.

The high throughput metabolomics study presented here used two different approaches, ¹H NMR analyses and IROA (Beecher and de Jong, 2019; (Stupp et al., 2013). Our ¹H NMR analysis (Supplemental Fig. S4A, B) was complemented by 2D NMR analysis with standards for a subset of control (Supplemental Fig. S5A, B) and POAG samples (Supplemental Fig. S6 A, B). We therefore used two independent high confidence and high throughput approaches validating the findings of metabolites and their concentrations presented here. For selected metabolites we also used a shotgun mass spectrometry approach utilizing a triple quadrupole TSQ Quantum Access Max and direct infusion (Bhattacharya, 2013) to validate the findings with the high-resolution Orbitrap type instrument. Most metabolites commonly identified by NMR and IROA showed similar trends. For example, L-Lysine consistently showed higher level in POAG compared to controls (Fig. 2 A, B). Only a few metabolites did not show the same trend and differences were not statistically more significant by IROA than NMR measurement. For these minor number of metabolites, the parent ion or neutral loss scans, acquired by triple quadrupole mass spectrometry in a TSQ Quantum Access Max, are consistent with IROA analysis. Our quantitative measurements of several metabolites are consistent with individual measurements of specific metabolites in previous studies of control or POAG AH (Buisset et al., 2019; (Hannappel et al., 1985).

We identified increased levels of L-arginine, L- lysine, cysteine, threonine, glycine anthranilate, ascorbate, 4-hydroxybenzoate, myo-inositol, acetate, propylene glycol, 2-hydroxy-butyrate, creatine and choline, and decreased levels of L-phenylalanine, threonine, glutamine, 4-aminobutanoate and isopropanol in POAG AH (Fig. 2A, B). Our findings are in consonance with several biological processes that have been implicated in the regulation of AH production or outflow. An increase in L-arginine has been reported previously in POAG patients (Hannappel et al., 1985). L-arginine is converted to nitric oxide (NO) by nitric oxide synthase (NOS) in the bloodstream, and NO has been found to decrease in glaucoma patients in comparison to control cataract patients. NO plays an important role in regulation of outflow facility (Becquet et al., 1997; (Chang et al., 2015), however, NO levels vary among different types of glaucoma (Kotikoski et al., 2002). Our analysis of the OmicsNet interaction network showed the downregulation of NOS3 (Fig. 4). NOS (in particular, eNOS) has been found to be a pressure-dependent regulator of IOP (Stamer et al., 2011). An increase in L-arginine and downregulation of NOS3 could possibly support NO’s role in the pathogenesis of POAG. The L-arginine is also involved in the synthesis of other amino acids, such as creatine. Increased concentrations of creatinine in AH have been found in a previous LC-MS/MS study (Buisset et al., 2019). Creatine is a product of creatine kinase as well as ATP. Creatine provides energy to muscles including the ciliary body. An increased creatine level may result in increased AH production and elevated IOP.

The L-arginine is also involved in the formation and degradation of glutamate and GABA. In contrast to our current study, an increased concentration of glutamate was reported in a rat model of glaucoma (Mayordomo-Febrer et al., 2015). Metabolite levels in the model systems may be different than human AH. This suggests that rigorous validation of AH metabolites similarities is necessary before one can derive conclusions from model systems. Despite the importance of glutamate induced damage to retinal ganglion cells (RGCs) and relevance for excitotoxic chronic neurotoxicity and cell death (Almasieh et al., 2012; (Lebrun-Julien and Di Polo, 2008; (Naskar and Dreyer, 2001), the exact role of glutamate in the AH is unclear. Whether AH traverses to the optic nerve has not been extensively investigated and remains unresolved. Acetate is involved in the synthesis of other biomolecules such as cholesterol (Boets et al., 2017). It is a precursor of acetyl-CoA. When cells in the surrounding tissues are under stressful conditions, they will secrete acetate into the AH. Changes in AH outflow dynamics has been conjectured to precede POAG as a consequence of cell loss or dysfunction of subcellular structures (Tan et al., 2006). This could explain the increase in acetate in diseased AH in this study. Diurnal variation in diurnal periodicities in rates of cholesterol and fat synthesis is now well known (Faix et al., 1993). Whether such cycles overlap with IOP fluctuations (Asrani et al., 2000) and diurnal AH turnover rate variations remains to be investigated.

Threonine was found to increase in POAG using NMR spectroscopy (Fig. 2A) but decreased in POAG using IROA (Fig. 2B), though the difference between control and POAG by IROA was more statistically significant than NMR. Threonine was reported in the literature to increase in POAG (Hannappel et al., 1985). Phenylalanine, 4-aminobutanoate, and isopropanol have not been previously reported in the AH in POAG. Glutamine has been previously reported to increase in POAG (Buisset et al., 2019; (Mayordomo-Febrer et al., 2015). Phenylalanine, glutamine, 4-aminobutanoate and isopropanol were found to undergo a decrease in POAG in our analyses (Fig. 2). Similarly, cysteine, anthranilate, 4-hydroxybenzoate, myo-inositol, propylene glycol, 2-hydroxy-butyrate, creatine and choline found in our analyses (Fig. 2) have never been specifically associated with POAG AH in individual or high throughput analyses, attesting to expanded resolution in identification of metabolites due to combinatorial use of IROA and NMR.

From the OmicsNet interaction networks, we determined the top 5 over-expressed genes (GARS, NOS2, CARS, TARS and HNF4A) and the top 5 under-expressed genes (NOS3, SHMT1, SHMT2, DLD, and FARSB) (Fig. 4). GARS directly interacts with glycine, and the concurrent over-expression of GARS recorded in the literature and glycine in our findings is evidence of an overexpressed pathway related to POAG (Fig. 4). CARS was also found to be upregulated in POAG (Liton et al., 2006) and this is further confirmed by our finding that cysteine is significantly over-expressed in POAG. TARS catalyzes the amino acylation of tRNA by its cognate amino acid, threonine. Its upregulation corresponds to the upregulation of threonine in our ¹H-NMR findings. SHMT1 and SHMT2 are responsible for the interconversion of serine to glycine. The high levels of L-serine found in our VIP analysis indicate that the under expression of these genes are linked to higher residual serine levels. FARSB is responsible for attaching L-phenylalanine to the terminal adenosine of the appropriate tRNA. We emphasize that we have used reviewed existing data of genes and proteins and integrated them with our experimental metabolomics data (Fig. 4). This integration (Fig. 4) is based on existing pathway information without actual experimental validation of their interaction.

The analysis of the AH proteome, lipidome and metabolome is expected to provide a multi-omics perspective and complementary insight into pathologic changes associated with the impeded AH outflow. A multi-omics picture of AH as well as all the surrounding tissues such as the cornea, CB and lens will be useful to understand impaired metabolism within the anterior eye segment in POAG. In the tissue and cellular milieu, any knowledge of compartmentalized concentration as well as the location of multi-omics components are lacking. AH is a fluid that bathes the anterior chamber tissues and precise knowledge of its residence time distribution (El Korchi et al., 2019) is lacking. This affects the identification and quantification of its constituent biomolecules including metabolites. Due to a large number of isomeric lipids and metabolites, multiple studies of AH lipidomics and metabolomics are desirable. AH offers non-invasive NMR analysis followed by destructive analysis using mass spectrometry, yielding a better confidence quantification of metabolites and thus possible wider identification, but such studies remain yet to be done. The experimental metabolome can be compared with the proteome or multi-ome of AH, TM, lens and other anterior segment structures, expanding our understanding into metabolism, turnover and their overall impairment in glaucoma. Importantly, such molecular detailed multi-omics studies of tissue and cellular compartments are likely to provide a greater insight into our understanding of IOP homeostasis and diurnal fluctuations.

A comparison of CB secretions (Margolis et al., 2018) with AH has suggested that metabolites of AH may likely be derived from more than one tissue in the anterior eye segment. This is consistent with emerging localized studies of lipids and metabolites using imaging mass spectrometry (Guerra et al., 2015). As noted above, the possible involvement of distal outflow regions and all tissues of the anterior eye segment towards AH level balance and IOP homeostasis is increasingly recognized. There is a lack of compartmental or boundary pathway maps in the literature, and this kind of experiment is not feasible in humans. In addition, there is a lack of data from mammalian systems that show a multi-omics approach to study the pathogenesis of glaucoma. Our analysis may aid in the understanding of the interactions between the various types of compounds present in the eye simultaneously.

5. Conclusion

Current analyses of metabolites identified the presence of L-arginine, L- lysine, cysteine, threonine, glycine, anthranilate, ascorbate, 4-hydroxybenzoate, myo-inositol, acetate, propylene glycol, 2-hydroxy-butyrate, creatine, choline, L-phenylalanine, glutamine, 4-aminobutanoate and isopropanol total from human control and POAG AH respectively. Our results show a significant increase and decrease in metabolites that have been previously implicated in specific biological roles in AH production and outflow in the anterior chamber tissue. Our results are likely to be part of future multi-omics analyses providing greater and integrated insight into these processes.

Highlights.

• We report the metabolites of control versus primary open angle glaucoma (POAG) aqueous humor (AH).

• We have used ¹H nuclear magnetic resonance (NMR) and isotopic ratio outlier analysis (IROA) using a Q Exactive orbitrap mass spectrometer.

• There were significant metabolite changes found in the POAG AH.

Abbreviations

AH

Aqueous humor

POAG

Primary open angle glaucoma

IROA

Isotopic ratio outlier analysis

CB

Ciliary body

TM

Trabecular meshwork