Protein aggregation and disease
  Proteins are macromolecules responsible for a wide array of functions in every living organism. Right interactions of proteins either with other proteins or other macromolecules or drugs are crucial for the healthy functioning of any organism. When there is modification or disruption in these interactions, proteins tend to aggregate leading to several diseases (https://bicmku.in/ProADD) like Alzheimer, Parkinson, Huntington, Amyotropic Lateral Sclerosis, Spongiform encephalopathy, Drepanocytosis, Type II Diabetes, Primary Open Angle Glaucoma (POAG), Blepharophimosis Ptosis Epicanthus Inversus Syndrome (BPES)^[1-4]. Protein aggregation occurs either due to mutations, misfolding and/or intrinsic disorder in proteins^[5-7]. Proteins like β 2 microglobulin, transthyretin, prion protein, lysozyme aggregate due to misfolding. But proteins like amylin, alphasynuclein, β amyloid peptide aggregate due to intrinsic disorder^[7]. Mutations in proteins like prion protein, hemoglobin, superoxide dismutase, forkhead domain containing proteins and myocilin, play role in protein aggregation^[1,4,5,8,9]. Molecular mechanisms involved in the transition of folded biologically-functional protein molecules into aggregates remain poorly understood. In certain cases as in amyloid and prion diseases, protein aggregation follows a pattern, but the exact mechanism is not clear yet^[10]. Studying the protein-protein interactions prevailing in these diseases might help in understanding the molecular mechanism. A tool InterCluster (https://bicmku.in/InterCluster) has been developed to cluster the protein-protein interactions in one of the protein aggregation diseases - POAG and analyzed.

Materials and Methods

Proteins involved and primary interactors
   The list of proteins involved in POAG was collected from literature and UniProt database (www.uniprot.org). Ten proteins were obtained from UniProt. Literature search was also done to collect details of other proteins involved in POAG. The availability of 3D structure was checked for every protein from PDB (www.rcsb.org). The primary interactors of these proteins were then collected from STRING database (www.string-db.org). The confidence score was set as 0.150 and the maximum number of interactors was limited to 500 by the database. The resultant lists of primary interactors were saved in a text file.

Localization and function
  The localization information of the proteins involved in the disease was collected using CoPub Portal (www.copub.org). Function of the proteins was obtained from UniProt and confirmed using literature search.

Tool development
  A tool was developed to cluster the proteins involved in the disease and their primary interactors. Python was used as the back end and the web interface was designed using Hypertext Preprocessor (PHP) and Hyper Text Markup Language (HTML). The web interface allows the user to upload the list of primary interactors in a compressed format. The input to the tool is a compressed folder which consists of list of primary interactors, stored in a text file named with the corresponding protein. Thus there are text files, with list of primary interactors for every protein. The tool results in clusters at all levels. The five output files generated are explained below.

  Occurrence of primary interactor among the proteins: The occurrence of primary interactor among the proteins is given in the file “Name.out”. This includes the list of primary interactors along with the protein with which they interact.

  Interaction number of the primary interactors: The number of interactions for each primary interactor is given in the file “Num.out”.

  Clusters within interaction group: The cluster information at all levels is in the file “Cluster.out”. Under each “Interaction number” group, there are various sub-groups and clusters in each of them are shown. X > Y implies that, there are X primary interactors common to Y proteins involved in the disease. Consolidated list of all clusters: The consolidated list of all clusters is presented in “List.out”. The clusters are obtained in the following format:
(IN : SN) I1, I2, I3 --> P1, P2, P3 [X>Y]

  The numbers IN and SN corresponds to interaction number and sub-group number (Interaction number: Sub-group number). The list of primary interactors (I1,I2,I3) is on the left hand side and the proteins involved in the disease (P1,P2,P3) are on the right hand side. The numbers at the end corresponds to the number of primary interactor and the number of proteins involved in the disease. [Number of primary interactor (X)> Number of proteins involved in the disease (Y)].

  Detailed information about the clusters: The detailed information about the clusters is given in “Output.out”. This file gives elaborate details about the cluster and how the cluster is formed.

Validation of the tool
  To validate the tool, a dataset was created, based on T-cell receptor pathway. The proteins on the membrane were considered equivalent to the proteins involved in the disease. The proteins directly interacting with these membrane proteins were considered as the primary interactors. The files were then given as input to the tool.

Results

Proteins and their primary interactors
   POAG is a major cause of blindness, characterized by progressive degeneration of the optic nerve and is usually associated with elevated intraocular pressure (IOP). This results in loss of retinal ganglion cell axons, along with supporting glia and vasculature. Reducing the intraocular pressure prevents progression of the disease in all stages^[11,12].

  Thirty one proteins are involved in POAG, fifteen of them have three dimensional structure (Table 1). The number of primary interactors varied from tens to five hundreds. The total number of primary interactors for all the proteins together was around 10500.

Table 1: Proteins involved in primary open angle glaucoma.

* proteins with pdb entry

InterCluster
   InterCluster is a web-based server for clustering proteins and primary interactors. InterCluster is freely accessible at https://bicmku.in/InterCluster (Figure 1). The total number of proteins involved in POAG and their primary interactors from STRING database is around 10500. InterCluster clusters these proteins and results in 16 groups and 110 clusters. The cluster with the maximum interaction number [7 > 11] is selected as Cluster 1. The [9 > 3] sub-cluster in the [11:9] cluster is considered as Cluster 2 as it has maximum experimental evidence.

Figure 1: Home page of InterCluster

Validation of the tool
  The tool was validated using primary interactions in T-cell receptor pathway. The tool was able to group the primary interactors and the proteins to form clusters as involved in the pathway (Figure 2). For example, 2 membrane proteins in the pathway PD-1 and CTLA4 interact with the protein SHP1. Here, SHP1 is considered as a primary interactor. Hence the cluster obtained is: SHP1 > PD-1, CTLA4

Figure 2: Validation of the tool. a) Part of T-cell receptor pathway used for validation. (KEGG : www.genome.jp/kegg) b) Clusters obtained from InterCluster

Clusters from InterCluster

Cluster 1: (17:7) POLA1, TNF, ALB, GAPDH, POLD1, TSPO, HRAS --> APOE, CYP1B1, TBK1, GSTM1, ELN, OPA1, MYOC, IL1A, FN1, IL1B, NPPA [7>11]

  This cluster has the maximum number of interaction. Eleven proteins involved in POAG have seven primary interactors (Table 2) in common. The functions of the proteins involved in this cluster were quite diverse. From the localization data acquired, some of the POAG proteins and the primary interactors were found to be co-localized. For example, MYOC (Uniprot ID:Q99972) and HRAS (UniProt ID:P01112) are co-localised in mitochondria^[13,14]. Studies also show that both MYOC and HRAS play a role in cell proliferation and survival^[15,16]. Further analysis was not carried out due to lack of experimental data in literature.

Table 2: Primary interactors in Cluster 1

* proteins with pdb entry

Cluster 2: (11:9) ELANE, STAT3, TIMP1, EDN1, KNG1, MYOC, LDLR, WDFY2, TF --> FN1, APOE, CYP1B1[9 > 3]

  This cluster was chosen for analysis as this had maximum number of experimental analysis. Three proteins involved in POAG has nine primary interactors (Table 3) in common. In this cluster, four proteins, MYOC (Uniprot ID: Q99972), APOE (UniProt ID: P02649), FN1 (UniProt ID: P02751) and CYP1B1 (UniProt ID: Q16678) are proteins involved in POAG. Though MYOC is a protein involved in POAG, in this cluster it occurs only as a primary interactor. It is known that mutations in MYOC and APOE cause POAG^[17]. Accumulation of FN1 in the trabecular meshwork is associated to POAG^[18]. Mutant CYB1B1 causes MYOC upregulation, which in turn causes POAG pathogenesis^[19].

Table 3: Primary interactors in Cluster 2

* proteins with pdb entry

  Clinical studies show that POAG is significantly more prevalent in a group of people with optic cup - retinal venous occlusion (OC-RVO). RVO is the blockage of the small veins that carry blood away from the retina. The mean IOP is significantly higher in OC-RVO than in other types of RVO^[20]. POAG is also associated with increase in IOP and loss of vasculature^[11]. From this analysis the 4 proteins APOE, MYOC, ELANE (UniProt ID: P08246) and KNG1 (UniProt ID: P01042) in this cluster are found to be involved in heparin binding.^[21-24]

  Further, based on the localization of the proteins^[25-31] in Cluster 2 (Figure 3) it is most probable that STAT3 (UniProt ID: P40763) interacts with POAG proteins APOE and CYP1B1 in the cytoplasm. Also MYOC might interact with POAG proteins FN1 and CYP1B1 in the endoplasmic reticulum. The localization information of these proteins gathered from literature is based on experimental analysis.

Figure 3: Localisation and interactions of the proteins in cluster 2. The proteins in the cluster 2 are placed in their subcellular locations and the arrows indicate the interactions between them based on literature.

Discussion

  POAG is associated with IOP, degeneration of the optic nerve and loss of vasculature^[11]. InterCluster, a tool developed to cluster protein-protein interactions, has helped in datamining the proteins involved in POAG and their primary interactions. Among 10500 proteins that were primary interactors of the 31 proteins involved in POAG we could datamine four proteins APOE, MYOC, ELANE and KNG1 in a cluster (Cluster 2) experimentally proved to be involved in heparin binding^[21-24]. This would have been impossible without the aid of the tool. These proteins may probably be involved in the pathogenesis of POAG by playing a major role in RVO leading to increase in IOP and loss of vasculature which is one of the reasons for neuronal death in POAG^[11]. Understanding the protein-protein interactions involved in POAG and other protein aggregation diseases would also help in elucidating protein-drug interactions as in the case of clozapine induced agranulocytosis^[32] which would help in finding appropriate drugs to find cure for these incurable diseases. Pathogenesis on POAG has been a mystery though decades of research has been carried out as only 2-3% of the disease is associated with mutation. With increasing research more proteins are known to be involved in the disease based on mutation analysis. But the real reason has not been yet understood. Our study has opened a new way to look at the pathogenesis of POAG and further elaborate research in the new direction.

Acknowledgement

Department of Biotechnology, New Delhi for facilities at the Center of Excellence in Bioinformatics and fellowship to SD. University Grants Commission, New Delhi is acknowledged for Dr. D. S. Kothari Post-Doctoral fellowship to EPJP.

References