Investigation of the Mechanisms of Neuroprotection Mediated by Lobelia Species via Computational Network Pharmacology and Molecular Modeling

Several species of the medicinally valuable genus Lobelia (Campanulaceae) exhibit neuroprotection. While the neuroprotective mechanisms of some components (e.g. lobeline, lobelanine, and lobelanidine) belonging to the L. nicotianaefolia or L. inata are extensively characterized, there remains the need to study and elucidate the mechanism of action of other species and their active components. In this work, we have studied the neuroprotective mechanism of the pharmacokinetically favorable active compounds of 17 Lobelia species.


Background
Neuronal injury is a pathological hallmark of some of the most commonly-known neurodegenerative diseases. In general, the severity of the neuronal damage determines the consequences such as neuronal degeneration or death (1). The etiology and progression of the common neurodegenerative and neuronal disorders such as Alzheimer's disease (AD), Parkinson's disease (PD), Huntington's disease (HD), epilepsy, and Amyotrophic lateral sclerosis (ALS) are some well-known examples of the severe consequences of the neuronal damages. The oxidative stress, excitotoxicity, neuroin ammation, mitochondrial dysfunction, and protein aggregations in the brain have been reported to play the central role in the neuronal damages (2).
Neuroprotective strategies and mechanisms that aim to limit the neuronal loss and/or rescue the neuronal damage progression and/or regenerate the neuronal structural and functional integrity are commonly used in several neurodegenerative diseases. Several investigations are also currently underway to nd novel ways to protect the nervous system from injury and damage (3). The concept of intervening the neurotransmission receptors by the agonist or antagonist of the natural neurochemical modulators to exhibit the neuroprotection effects has been long-established (4). For example, caffeine, an A2 receptor antagonist was found to protect dopaminergic neurons against the 1-methyl-4-phenyl-1,2,3,6tetrahydropyridine (MPTP)-induced neuronal toxicity in the experimental model of PD (5). Other neurotransmitters such as serotonin, gamma-aminobutyric acid (GABA), and glutamate are also associated with regulating the inhibition and excitation of motor neurons. For example, drugs like umazenil interact with the GABAergic system and alter motor behavior (6). Other pharmacological interventions are also known to provide neuroprotection as well as disease-modifying activities.
The neurotransmitter receptors modulation, anti-in ammatory responses, and anti-oxidative stress are well-studied pharmacological intervention approaches to protect the neurons. Due to the complex pathways associated with the pathogenesis of neurodegenerative diseases, the paradigm of multi-targetdirected ligands (MTDLs) has been long warranted and has been well-researched in recent years (7)(8)(9).
Several novel MTDLs with synthetic optimizations have been published in recent years (10,11). In this respect, the herbal medicine or natural products provide an unprecedented advantage of exerting multiple effects on different biological targets (12)(13)(14). For example, Crocus sativus, Nigella sativa, Coriandrum sativum, and Ferula assafoetida, Curcuma longa, to name a few, showed neuroprotective effects by regulating multiple disease-associated targets and signaling pathways (15). Curcumin, the major constituent of Curcuma longa, exerts a neuroprotective role in PD via multiple mechanisms including the restoration of the GSH decreased levels, mediation of the overexpression of BCl-2 (inducible nitric oxide synthase (iNOS) antagonist) (16), reduction of pro-in ammatory cytokines such as IL-1β, IL-6, TNF-α, total nitrite generation, and decreased activation of NF-κB (17,18). Lobelia (Campanulaceae) is a genus of owering plant natives to the temperate and warmer regions with 450 species currently known. Several Lobelia species were traditionally used for medicinal purposes and several species have been continuously evaluated for their distinct pharmacological activities (19). Lobelia chinensis is reported to have anti-oxidative (20), anti-in ammatory (21), anti-viral (21), anti-cancer (22,23) and anti-diabetic properties (24). Lobelia in ata, also known as Indian tobacco, has a long history of use in the treatment of severe breathing problems including asthma, whooping cough, and bronchitis. Amongst the chemical constituents of the Lobelia species, a majority of the research has been performed on pyridine alkaloids such as lobeline, lobelanine, and lobelanidine (25)(26)(27). For example, Li et al., (28) reported the dopaminergic neuroprotective effects of lobeline against the MPTP-induced dopaminergic neuron death. The lobeline extracted from the leaf of Lobelia nicotianaefolia was shown to have antiepileptic activity by modulating the GABAergic mechanism (29).
Several other chemical classes such as glycosides, lignans, avonoids, avonoid, and amino acids endowed with diverse pharmacological effects were revealed (30)(31)(32)(33)(34). In a recent study, Ge et al., (24) studied the anti-diabetic mechanism of the metabolites extracted from Lobelia chinensis using network pharmacology approaches. Their study revealed 5-hydroxymethylfurfural and acacetin as two major active ingredients modulating the insulin resistance signaling pathway and diabetic pathway. Moreover, their study also provided a basis for the further study of the active constituents of the Lobelia chinensis in other diseases and their pharmacological mechanisms.
In this study, we performed network pharmacology and molecular modeling analysis of the active ingredients and corresponding targets of 17 herbs of the genus Lobelia. The aim was to decipher their neuroprotective mechanism of action and provide a rationale for future pharmacological studies. The relationships between the active ingredients and potential targets/pathways were established and presented systematically.

Active Ingredients and Biological Targets
A total of 17 herbs of the genus Lobelia were studied. A total of 233 known active ingredients were retrieved from the Natural Product Activity and Species Source Database (NPASS) (35). The chemical structures and the corresponding targets were mapped to the ChEMBL database (36). Additionally, TCM systems pharmacology database and analysis platform (TCMSP) (37) was also used for the mapping. Three ADME-related properties including OB (oral bioavailability) ≥ 30%, drug-likeness (DL) ≥ 0.18, and half-life (HL) ≥ 4 were used to screen the active constituents. The open-source cheminformatics package RDKit (http://www.rdkit.org) was used to standardize the chemical structures and determine the pharmacokinetic parameters (whenever necessary). After that, 49 distinct chemical ingredients and their 411 corresponding targets were selected. Finally, 12 targets that belonged to the cytochrome 450 families were discarded.

Neuroprotection and Neurodegenerative Disease-Related Gene
The associated-genes of ve neurodegenerative diseases viz. AD, ALS, epilepsy, HD, and PD along with the genes associated with the common neuroprotection mechanism were collected. The gene databases viz. GeneCards (38), The Comparative Toxicogenomics Database (CTD) (39), Human Genome Epidemiology (HuGE) Navigator (40) and The Online Mendelian Inheritance in Man (OMIM) (41) were used. Many custom ltering criteria were applied. For example, only the genes annotated to have ≥ 20 publications associated with them were retrieved from the HuGE navigator database. Only the genes annotated with the "protein-coding" function were retrieved from the GeneCards database. Only the genes annotated with the label "M" and/or "T" were retrieved from the CTD database. Finally, the total genes collected from the various databases were ltered using the NCBI Gene database and only the "proteincoding" genes were selected.

Enrichment Analysis of Genus Lobelia Targets
The overlapping targets of genus Lobelia related to the neuroprotection and neurodegenerative diseases were mapped into Gene Ontology (GO), Kyoto Encyclopedia of Genes and Genomes (KEGGs) and REACTOME Pathway (42). The GO functional annotations were carried for the biological process (BP), molecular function (MF), and cellular components (CC) terms (43).

Compound-Target Network Analysis
The Cytoscape v3.7.2 (44,45) was used to build and analyze the compound-target network. The ECFP4 ngerprints (46) was used to evaluate the chemical diversity of the compounds.

Pharmacology Network Analysis
The STRING v11 (47) was used to construct the protein-protein interaction (PPI) networks of the overlapping targets related to the genus Lobelia, common neuroprotection, and neurodegenerative diseases. The analysis and modularization were performed using Cytoscape v3.7.2. The MCODE algorithm (48) was used to determine highly interconnected regions in the PPI network. The degree cutoff, node density cutoff, and node score cutoff were kept to 2, 0.1, and 0.2, respectively.

Molecular Docking
The docking studies of the representative compounds and targets were performed with AutoDock 4.2 (49) using the UNIX automation script. The BIOVIA Discovery Studio (San Diego: Dassault Systèmes) was used for pre-processing the chemical structures and biomolecules. The metal ions and/or substrate molecules (if any) were kept in the binding pocket of the targets. The Lamarckian genetic algorithm search was employed for the docking. The key residues of the binding pocket were kept exible. The center of the binding pockets of the individual targets was selected for the grid placement. A total of 60 runs along with 2.5 million energy evaluation steps were employed. The representative pose selection was done based on the cluster analysis of the docked poses. The PyMOL Molecular Graphics System (Version 1.8.4.0, Schrödinger, LLC) was used for the visualizations and graphics generations.

Lobelia herbs, Active Ingredients, and Known Targets
In this study, a total of 17 herbs belonging to the genus Lobelia were selected. The criteria of their inclusion in this study were based on available information related to the metabolites and their biological targets. For the herbs are known to have at least one experimentally characterized metabolite and one corresponding target was included. The collected herbs are listed in Supplementary Table S1. Initially, the chemical constituents collected from the NPASS database were mapped to the ChEMBL database and the corresponding bioactivity data were retrieved. A total of 194 compounds were successfully mapped in this manner. Additionally, the chemical constituents and corresponding biological targets of Lobelia chinensis were also retrieved from the TCMSP database. A total of 71 chemical constituents were retrieved from the TCMSP database. The redundant chemical structures between the Lobelia chinensis collected from both databases were merged into one and the nal selection was based on the TCMSP database. Finally, a total of 233 molecules were subjected to the ADME-ltering. In this study, we used three ADME-ltering criteria; OB ≥ 30%, DL ≥ 0.18, and HL ≥ 4. The aim of this ltering was to select the molecules with good absorption, slow metabolism after oral administration, and suitability for the drugdevelopment. A similar approach has been used in other studies (50). In this work, we did not consider the blood-brain barrier (BBB) permeability as ltering criteria. The reason was that certain natural products (e.g. quercetin) with the theoretical prediction of poor BBB permeability when tested experimentally exhibited satisfactory permeability (51). Following the ADME-ltering, 49 unique chemical compounds associated with 411 corresponding targets were selected. Out of that, 12 targets related to the cytochrome 450 families were discarded from further study. Finally, a total of 153 non-reductant targets were studied.

Neuroprotection and Neurodegenerative Disease-Related Gene
The candidate genes of the ve most common neurodegenerative diseases viz. AD (180 genes), ALS (121 genes), epilepsy (2667 genes), HD (65 genes), and PD (127 genes) along with the genes commonly associated with the neuroprotection (NP, 101 genes) mechanism were collected. A series of gene databases (see Material and Methods) were used. Initially, the genes related to each disease and NP mechanism were compared and analyzed individually with the Lobelia genes (153 genes). For the sake of simplicity, this set of genes which includes individual diseases will be called Set I. The overlapped genes of the Set I are shown in Supplementary Figures S1. In the nal selection, all the genes associated with ve neurodegenerative diseases mentioned above and those of NP mechanisms were overlapped with the Lobelia genes. For the sake of simplicity, this set of genes will be called Set II. To be noted that all the subsequent studies reported in the main paper were performed on Set II. The reason for the selection of Set II for the detailed analysis was because of the fact that different neurodegenerative diseases including those mentioned in this study broadly share the common pathogenesis and signaling pathways related to neuronal damages and protection. Therefore, the genes of the Set II best represented the neuroprotective mechanism associated with different neurodegenerative diseases. Moreover, the analysis related to Set I was also performed (see Supplementary Information).
A total of 31 overlapping genes in Set II were retrieved and the IUPHAR classi cation (52) was performed on them. The categorical distribution of the overlapped gene is shown in Fig. 1A. As expected, a majority of the genes belonged to the enzyme class (56.3%) followed by ion transporter (12.5%), nuclear hormone receptors (12.5%), other protein (12.5%), and catalytic receptor (6.3%).
We have also performed a clustering analysis to determine the chemical diversity of the collected chemical constituents. The 49 unique compounds could be represented into 10 well-de ned clusters as shown in Fig. 1B. The representative chemical structure of each cluster is shown in Fig. 1C. The aglycone part of the avonoid such as diosmetin and 18 other aglycones constituted the most populated cluster (Cluster 1).

Enrichment Analysis of the Candidate Genes
All the overlapped genes belonging to Set I and Set II were tested for functional enrichment with three GO terms (BP, CC, and MF) and KEGG/REACTOME pathways. The result of the GO terms and KEGG/REACTOME pathways enrichment of genes in Set I are shown in Supplementary Figures S2-S7. The result of the GO enrichment of the genes in Set II is shown in Fig. 2A. The description is provided in the following section.
The "positive regulation of transcription from RNA polymerase II promoter", "positive regulation of transcription, DNA-templated", "response to drug", and "negative regulation of apoptotic process" were the most signi cantly enriched terms in the BP. Interestingly, the "response to gamma radiation (GO:0010332)" was also found to be within the top ten most enriched BP terms. The ionizing radiation has been shown to have a debilitating impact on neurodegenerative diseases. Several studies reported that radiation inhibits neurogenesis through different mechanisms such as neuroin ammation, elevate reactive oxygen and nitrogen species, oxidative stress, protein degradation, and mitochondrial dysfunctions, among others (53)(54)(55)(56). The major cellular components such as nucleus, cytosol, cytoplasm, mitochondria along with synapse were indicated as the location of the overlapped genes in Set II.
The KEGG/REACTOME pathways analysis of overlapped genes were analyzed with BH-corrected Pvalues < 0.05 (Fig. 2B). The enrichment of the overlapped genes was mostly found in the pathways in cancer, hepatitis B, Akt signaling pathway, proteoglycans in cancer, MAPK signaling pathway, HTLV-1 infection, prostate cancer, colorectal cancer, in uenza A, thyroid hormone signaling pathway, tuberculosis, Chagas disease, endometrial cancer, Amyotrophic lateral sclerosis, and bladder cancer, among other. Indeed, a number of studies supported the intriguing cross-talks between cancer and neurodegeneration (57)(58)(59). The key candidate pathways-targets interaction network is shown in Fig. 2C.

Compound-Target Networks
The compound-target network was constructed to establish the role of the active ingredients of the genus Lobelia and the overlapped targets found in Set I and Set II. The compound-network diagram diagrams of the overlapped genes in AD, ALS, epilepsy, HD, NP, and PD are shown in Supplementary Figures S8-S13 and for Set II is shown in Fig. 3.

Construction and Analysis of Target Proteins PPI Network
The PPI network was constructed to understand the interrelation between the neuroprotection associated candidate genes of the genus Lobelia. The constructed PPI network is shown in Fig. 4. The network edges were rst created based on the molecular interaction by keeping the interaction score to ≥ 0.400. A total of 31 nodes and 249 edges were found in the network with an average node degree of 16.1. As expected, AKT1, TP53, MYC, TNF, EGF, EGFR were amongst the central targets in the PPI network. In the next step, the highly interconnected regions in the PPI network were analyzed using the MCODE algorithm. A wellorganized and highly interconnected hub region with 20 nodes were retrieved. The targets ESR1, MYC, IL1B, IFNG, CXCL8, CASP9, IL2, CCL2, EGF, EGFR, FOS, MMP2, HSPB1, AKT1, TP53, BCL2L1, AR, HIF1A, TNF, and CCND1 constituted the cluster. Interestingly, all the targets associated in this cluster had an association score ≥ 0.9, which in turn suggests the high con dence in their interactions. The topological parameters of the PPI network are shown in Table 1.    Quercetin was reported to stimulate cancer cell proliferation via the estrogen receptor (74).

Discussion
Neurodegenerative diseases such as Alzheimer's disease, Parkinson's disease, Huntington's disease, epilepsy, and Amyotrophic lateral sclerosis are some of the most pressing burdens on the global healthcare system. The high rate of the mortality and morbidity associated with these diseases, particularly in the elderly population, demands novel therapies. Unfortunately, to date, not much success has been achieved in developing effective therapeutics. However, tremendous progress has been made in unraveling the causative mechanism and pathogenesis of these diseases.
One very critical causative mechanism of the most neurodegenerative diseases is the neuronal damage (1). It is now well-established that acute as well as chronic neuronal damage initiates and helps in the progression of neurodegenerative diseases. Thus newer therapeutic strategies to protect the neurons from the damages are needed to be developed (3). In this direction, the use of multi-component herbal products offers a great alternative. Several herbal medicines or natural products are known to exhibit neuroprotection and provide bene cial effects in the treatment of neurodegenerative diseases (12)(13)(14).
However, there have been several clinical concerns that hinder a wide-adoption of herbal/natural products in the prevention and therapy of neurodegenerative diseases. The lack of scienti c evidence or support for patient safety and their e cacy has been often touted as the main reason. Thus, studying the rationale for the explanation of the molecular mechanism of actions of the pharmacokinetically suitable herbal components are of paramount importance.
In this study, through integrated network pharmacology and molecular modeling approach, we aimed to provide an explanation for the neuroprotective mechanism of the chemical constituents of a medically valuable genus called Lobelia. We studied a total of 17 herbs belonging to this genus and their impact on the ve neurodegenerative diseases viz. Alzheimer's disease, Parkinson's disease, Huntington's disease, epilepsy, and Amyotrophic lateral sclerosis. Since the neuronal damages associated with the initiation and progression of these diseases share the common mechanism, the overlapped genes associated with these diseases and those related to the genus Lobelia were studied and presented.
KEGG/REACTOME pathways analysis results showed that the candidate genes involved in enriched pathways were related to the relevant pathways in cancer (Fig. 2). Several studies established the crosstalks between the etiology of cancer and neurodegenerative diseases. For example, the role of the cancerassociated enriched targets in this analysis such as the AKT serine/threonine kinase (AKT1) (

Conclusions
Neurodegenerative diseases have complex mechanisms with several intertwined signaling pathways and multiple targets associated with it. Therefore, targeting based on multi-ingredient-multi-target-multipathway is highly desirable. In this work, we provide the neuroprotective mechanisms of the 17 herbs belonging to the genus Lobelia. We establish the relationship between their active ingredients and target proteins and signaling pathways. We further provided theoretical validation of the binding mechanism of the compounds with the target proteins. Although our theoretical ndings are consistent with other biological reports, further experimental validation of the pharmacological effects would be highly valuable. Moreover, this study provides adequate background and con dence to do so. Figure 1 Property analysis of the compounds related to genus Lobelia and targets associated with neuroprotection mechanism. A) IUPHAR classi cation of the overlapped genes (Set II) shown in terms of percentage; B) chemical diversity of the collected compounds belonging to the genus Lobelia. ECFP4 ngerprint was used. The numbering of the cluster is shown in descending order of the cluster size; C) chemical structure of the representative compound in each cluster.

Figure 3
The compound-target network of the active ingredients of the genus Lobelia and the targets related to the neuroprotective mechanism.