A comprehensive map of disease networks and molecular drug discoveries for glaucoma

Glaucoma is the leading cause of irreversible blindness worldwide. The molecular etiology of glaucoma is complex and unclear. At present, there are few drugs available for glaucoma treatment. The aim of the present study was to perform a systematic analysis of glaucoma candidate drugs/chemicals based on glaucoma genes, including genetic factors and differentially expressed (DE) genes. In total, 401 genes from the genetic databases and 1656 genes from the DE gene analysis were included in further analyses. In terms of glaucoma-related genetic factors, 54 pathways were significantly enriched (FDR < 0.05), and 96 pathways for DE genes were significantly enriched (FDR < 0.05). A search of the PheWAS database for diseases associated with glaucoma-related genes returned 1,289 diseases, and a search for diseases associated with DE glaucoma-related genes returned 1,356 diseases. Cardiovascular diseases, neurodegenerative diseases, cancer, and ophthalmic diseases were highly related to glaucoma genes. A search of the DGIdb, KEGG, and CLUE databases revealed a set of drugs/chemicals targeting glaucoma genes. A subsequent analysis of the electronic medical records (EMRs) of 136,128 patients treated in Sichuan Provincial People’s Hospital for candidate drug usage and the onset of glaucoma revealed nine candidate drugs. Among these drugs, individuals treated with nicardipine had the lowest incidence of glaucoma. Taken together with the information from the drug databases, the 40 most likely candidate drugs for glaucoma treatment were highlighted. Based on these findings, we concluded that the molecular mechanism of glaucoma is complex and may be a reflection of systemic diseases. A set of ready-to-use candidate drugs targeting glaucoma genes may be developed for glaucoma clinical drug treatments. Our results provide a systematic interpretation of glaucoma genes, interactions with other systemic diseases, and candidate drugs/chemicals.

Glaucoma is a set of progressive optic neuropathies 1 and the leading cause of irreversible blindness worldwide 2 . Glaucoma is characterized by a loss of retinal ganglion cells and consequent visual field loss. The two most common forms of glaucoma are primary open-angle glaucoma (POAG) and primary angle-closure glaucoma 3 . The main known risk factors for glaucoma include high intraocular pressure (IOP), older age, African race, high myopia 4 , a high vertical cup/intervertebral disc ratio 5 , and a reduction in the optic disk area and central corneal thickness 6 . Epidemiological studies have shown that the prevalence of glaucoma is expected to reach 76 million by 2020 and 118 million globally by 2040 due to population aging 7 . The mechanism underlying the development of glaucoma is not fully understood.
The treatment of glaucoma includes drug use and surgery 16 . Antiglaucoma drugs reduce IOP mainly by reducing aqueous humor production and promoting aqueous humor discharge 17 . At present, there are four kinds of drugs for glaucoma treatment: β-receptor blockers, prostaglandins, α-2 agonists and carbonic anhydrase inhibitors. Laser peripheral iridectomy (LPI) is used in glaucoma with anterior chamber angle occlusion and occlusion 18 . Selective laser trabeculoplasty (SLT) 19 can be used as primary or auxiliary treatment for primary open-angle glaucoma or early-and late-stage glaucoma after LPI. Microinvasive glaucoma surgery (MIGS) reduces IOP and reduces dependence on glaucoma medications 20 .
It can cost billions of dollars to develop a new drug and often takes several years. Drug reuse is a strategy for the identification of new uses of drugs for approval or research beyond the scope of the original medical indications 21 . This new phase of genomics, which is increasingly referred to as precision medicine, has sparked a new chapter in the relationship between genomics and drug development 22 . Compared with the development of new drugs for specific indications, drug reuse has several advantages 23 , including safety and less investment 24 . For example, thalidomide, developed in 1957, was originally used as a sedative 25 . Later, it was found to be effective in patients with intermediate thalassemia 26 and multiple xanthogranuloma in adults 27 . Given the high failure rate and high costs of new drug development, the reuse of "old" drugs to treat human diseases is becoming an attractive proposition. Some trials for the reuse of "old" drugs in glaucoma are underway. e.g., nicotinamide. Thus far, there has been no systematic analysis of drug reuse for glaucoma.
Given recent progress in genomics, it is now possible to rapidly identify and interpret genetic variations underlying a single disease in a single patient, thereby aiding individualized (tailored) drug therapy 22 . The aim of the present study was to provide new information for candidate drug development for glaucoma. The study design is shown in Fig. 1.

Methods
Gene mining for glaucoma. Information on glaucoma-associated/mutated genes was obtained from the GWAS Catalog 28 , OMIM 29 , Phenolyzer 30 , and published papers 14 . We searched the GEO database for human glaucoma-related tissues, optic papillary astrocytes, and the trabecular meshwork. In all the datasets, a large amount of data were available only for human optic nerve head astrocytes (GSE9963) 31  Glaucoma pathway analysis. The Database for Annotation, Visualization and Integrated Discovery (DAVID) 33 is a bioinformatics data resource composed of an integrated biological knowledge base and analysis tools, which are used to extract meaningful biological information from a large number of gene and protein collections. The database collects and integrates a variety of gene identifiers. We used the DAVID for glaucoma pathway analysis.

Mining gene interactions.
Cytoscape can be used in conjunction with large databases of protein-protein, protein-DNA, and genetic interactions to develop biomolecular interaction networks 34 . ClueGO is a user Cytoscape plug-in that analyzes interrelations of terms and functional groups in networks 35 . We used ClueGO embedded in Cytoscape 3.6 for the gene interaction analysis and selected only pathways with p values ≤ 0.05.

Electronic medical records (EMRs).
To obtain information about the candidate drugs and glaucoma, we searched the EMR data of Sichuan Provincial People's Hospital from August 2015 to August 2018 (N = 136,128) for the usage of the candidate drugs and the onset of glaucoma. For each drug usage, the number of total patients and the proportions of glaucoma patients were calculated. The study was approved by the institutional ethics committee of Sichuan Provincial People's Hospital and was conducted according to the Declaration of Helsinki principles. Informed consent was obtained from the participants.

Results
Mining glaucoma genes. The search included mutation/association genes, as well as IOP genes for glaucoma genetic factors [42][43][44] . The search identified 159 genes in the OMIM database, 144 genes in the GWAS catalog (before 06/20/2017), and 86 genes in the Phenolyzer database. In total, 401 genes were obtained for further analysis as glaucoma genetic factors. In addition, 1,024 DE genes and 495 DE genes were identified in the POAG individuals in the GSE9963 and GSE27276 datasets, respectively. In total, 1,656 genes from the DE gene analysis corresponding to glaucoma were included in subsequent analysis.
Gene pathways for glaucoma. Among the 401 glaucoma genetic factors, 13 pathways were significantly enriched (FDR < 0.05), including pathways in cancer, focal adhesion, amoebiasis, the PI3K-Akt signaling pathway, and the TGF-beta signaling pathway (Table S1A). We used ClueGO 35 , a Cytoscape plug-in, for biological interpretation of the genetically associated genes for glaucoma ( Fig. 2A).
Glaucoma-related diseases. The search of the PheWAS database identified 1,289 diseases related to glaucoma genetic factors (Table S2). Pathways related to cardiovascular disease (13%), endocrine disease (9%), nervous system disease (8%), and eye disease (6%) were enriched in glaucoma-related genes (Fig. 3A). The database search revealed 1,357 diseases related to DE glaucoma genes (Table S3). The disease spectrum for these DE genes was very similar to the disease spectrum of glaucoma genetic factors (Fig. 3B). Cardiovascular disease (11%), endocrine disease (11%), nervous system disease (9%), and eye diseases (6%) were enriched in glaucoma DE genes.
Drug discovery for glaucoma genes. The search of the DGIdb identified 114 drugs with effects on glaucoma-related genes with an FDR < 0.05 (Table S4). The above results indicate that retinoic acid, cyclosporin, collagenase clostridium histolyticum, and taprazole have smaller P values. Of these candidate drugs, dorzolamide is used for antiglaucoma treatment 45 . The top 50 drugs are presented in Table 2.
The search of the KEGG DRUG database returned 24 drugs for glaucoma. Then, we identified 13 target genes for the 24 drugs from the KEGG database. A search of the DGIdb revealed drugs targeting these genes (Table S5). The top 100 candidate drugs with antiglaucoma or therapeutic effects on glaucoma are listed in Table 3. Acetylcholinesterase inhibitors, alpha-adrenergic receptor antagonists, antihypertensive drugs, and antipsychotic drugs had higher frequencies.
The search of the CLUE database for glaucoma genetic factors returned 133 drugs/chemicals with an FDR < 0.05 (Table S6). The top 50 candidate drugs/chemicals for glaucoma are listed in Table 4. Most of the drugs identified in the search were antitumor, antihypertension, and analgesic drugs. In the CLUE database, 3 candidate gene-matched drugs/chemicals with an FDR < 0.05 were returned by searching for glaucoma DE genes (Table 5).
For the 13 KEGG glaucoma drug-targeted genes, 166 drugs/chemicals were enriched in the CLUE database with an FDR < 0.05 (Table S7), including carteolol, betaxolol and latanoprost. Among these, propranolol derivatives were the most enriched chemical. The US FDA approved a new glaucoma drug, netarsudil/latanoprost    Table S7). The top 25 selected candidate drugs/chemicals for glaucoma are listed in Table 6.
The analysis of 136,128 EMR histories revealed nine candidate drugs of all the mined glaucoma drugs mentioned above, which were used in Sichuan Provincial People's Hospital from August 2015 to August 2018 (Table 7). Of 435 patients treated with cytarabine (242 of whom were older than 40 years), none of the patients had glaucoma. The prevalence of glaucoma was 0.11% in theophylline-treated patients (N = 4,594), 0% in nicardipine-treated patients (N = 564), 0.058% in celecoxib-treated patients (N = 1,488), and 0.035% in nicardipine-treated patients (N = 564). The incidence of glaucoma was significantly lower in these drug-use cohorts than in healthy individuals (1% in those aged older than 40 years). Thus, these drugs may have antiglaucoma effects. Among 1,293 hospitalized AD patients, 48 (3.8%) patients had glaucoma. This was significantly higher than the prevalence rate of glaucoma in the healthy population, suggesting that the incidence of glaucoma may be elevated in individuals with AD (P = 7.99E-05 assuming 1000 samples, OR = 3.9). Finally, by P-value and FDR ranking, we selected 40 drugs/chemicals most likely to prevent or treat glaucoma, and we selected 40 most likely drugs/chemicals for the prevention or treatment of glaucoma (Table 8).

Discussion
Glaucoma is a set of disorders that cause damage to the optic nerve and worsen over time. Pathological ocular hypertension, race, a family history of glaucoma, vasospasms, and peripheral vascular disease are common contributing factors 46   www.nature.com/scientificreports www.nature.com/scientificreports/ research reported that various pathways, including focal adhesion, extracellular matrix-receptor interaction, cancer, and the PI3K-Akt pathway, were significantly related to IOP 43 , pointing to its complex etiology. Current glaucoma medications reduce IOP by reducing the production of fluid in the eye or increasing its outflow.   www.nature.com/scientificreports www.nature.com/scientificreports/ In the present study, the pathway analysis identified several nervous system disorders, including AD, that were strongly associated with glaucoma. Glaucoma is part of a set of age-related neurodegenerative diseases 47 . AD is the most common neurodegenerative disease among elderly individuals. Previous studies have demonstrated that AD and glaucoma share several biological characteristics 48,49 . Signaling pathways such as dilated cardiomyopathy, arrhythmogenic right ventricular cardiomyopathy (ARVC), and hypertrophic cardiomyopathy (HCM) are enriched, suggesting that glaucoma may be associated with signaling pathways in cardiovascular disease.
In the present study, we presented an enhanced and updated perspective on glaucoma-related genes, associated diseases, and drugs targeting these diseases (Fig. 1). The analysis of diseases linked to glaucoma genes showed that cardiovascular diseases were most closely associated with glaucoma 50 . Previous research pointed to an abnormal hemorheological pattern in glaucoma patients, with increased plasma viscosity leading to the hypoperfusion of the ophthalmic artery, which can potentially aggravate optic nerve injury 51 . Other research has suggested that peripheral vascular endothelial dysfunction may be related to the progression of glaucoma 52 . High blood viscosity plays a role in the occurrence of glaucoma 53 . A previous study showed that when the shear rate of retinitis decreased and the blood viscosity increased, the low perfusion of retinal blood flow can lead to local ischemia and aggravate vision atrophy 54 . The hemodynamics of the ophthalmic artery and central retinal artery are correlated with POAG 55 .
From a drug discovery standpoint, the identification of glaucoma genes provides valuable information to reveal its causative mechanisms and drugs targeting this disorder. In the CLUE database, carbonic anhydrase inhibitors acetazolamide and dipivefrin had antiglaucoma effects 56 . The results of the present study suggested that anticancer drugs, analgesics, and antihypertension drugs may have potential in the prevention and treatment of glaucoma. Potential candidate drugs identified in the search of the CLUE database included propranolol derivatives, which are nonselective beta-1 and beta-2 adrenergic receptor blockers. Propranolol derivatives have similar pharmacological effects to those of currently used glaucoma drugs and have potential for drug development as   www.nature.com/scientificreports www.nature.com/scientificreports/ antiglaucoma agents. In the DGIdb, celecoxib, paclitaxel, and cyclosporine appeared frequently, suggesting that antineoplastic drugs may represent a new direction for glaucoma drug screening 57 .
Screening of the KEGG database, DGIdb database, and CLUE database identified α1/β adrenergic receptor antagonists. This antagonist can reduce cyclic adenosine phosphate in ciliary epithelial cells. Alpha 1/beta   www.nature.com/scientificreports www.nature.com/scientificreports/ adrenergic receptor antagonists might have potential in reducing not only the generation of aqueous humor but also the outflow of aqueous humor through the trabecular meshwork. Carvedilol, a nonselective beta-adrenergic receptor blocker (beta 1 and beta 2) and alpha-adrenergic receptor blocker (alpha 1), is currently used for the treatment of hypertension 58 .  www.nature.com/scientificreports www.nature.com/scientificreports/ This study has two limitations. First, although we discovered many candidate drugs for glaucoma and provided gene-drug pair information in the supplementary data, it was based on the mixed forms of glaucoma (most genes are involved in POAG). Because different glaucoma types are very different diseases and should have different gene involvement and drug targets, the readers should refer to the gene of a subtype of glaucoma to find the proper candidate drugs in the supplementary data. Second, although we performed statistical analysis of significance, further experiments are still needed for the verification of the treatment of glaucoma.

Medical uses FDR
In summary, we investigated genetic factors and DE genes in glaucoma. We interpreted the pathways of these glaucoma genes and systematically investigated diseases related to glaucoma genes. In this study, we screened the top drug candidates for glaucoma, such as tretinoin, ocriplasmin, collagenase clostridium histolyticum, Talarozole and bevacizumab. Tretinoin is also known as all-trans retinoic acid. Talarozole is a systemic all-trans retinoic acid metabolism blocking agent that increases intracellular levels of endogenous all-trans retinoic acid. Tretinoin is an intermediate product of vitamin A metabolism in the body. Vitamin A is known for its function in the retina with importance for rhodopsin visual phototransduction, and it protects against free radicals, i.e., it acts as an antioxidant 59 . Regarding dietary intake of retinol equivalents, two large studies reported a protective effect on POAG [59][60][61] . Ocriplasmin is a recombinant protease with activity against fibronectin and laminin, components of the vitreoretinal interface, and may lower IOP by degrading vitreous or connective tissue. McClintock et al. reported the case of a glaucoma patient who received a single intravitreal injection of 125 µg ocriplasmin for vitreomacular traction in the right eye. Its final visual acuity was 20/50 + , and IOP was 18 mmHg at 16 weeks after surgery, with IOP reduction and serous choroidal effusion after ocriplasmin injection 62 . Collagenase clostridium histolyticum is an enzyme produced by the bacterium Clostridium histolyticum that dismantles collagen. The collagen matrix is the main structure of the trabecular meshwork, which plays an important role in high-tension glaucoma 63 . Collagenase clostridium histolyticum had a drug effect that may lower IOP by degrading adhesive collagens in the hole of the trabecular meshwork. Bevacizumab is a monoclonal antibody developed against vascular endothelial growth factor (VEGF). It is used for neovascular glaucoma 64 and for reducing glaucoma surgical scars 65 . Of these drug candidates, we still need more mechanistic studies in the future. Subsequently, we mined drugs/chemicals targeting glaucoma genes. In addition, we analyzed the usage of candidate drugs and the onset of glaucoma in clinical EMRs. Finally, we selected the 40 most likely candidate drugs for the prevention and treatment of glaucoma. The results provide a systematic interpretation of glaucoma-related genes, diseases, and candidate drugs. Our research provides comprehensive data that can enrich the understanding of glaucoma and potential glaucoma drugs.