Pannorin isolated from marine Penicillium sp. SG-W3: a selective monoamine oxidase A inhibitor

Six compounds were isolated from Penicillium sp. SG‑W3, a marine‑derived fungus, and their inhibitory activities against target enzymes relating to neurological diseases were evaluated. Compound 1 (pannorin) was a potent and selective monoamine oxidase (MAO)‑A inhibitor with a 50% inhibitory concentration (IC 50 ) of 1.734 μM and a selectivity index (SI) of > 23.07 versus MAO‑B, and it showed an efficient antioxidant activity. All compounds showed weak inhibitory activities against acetylcholinesterase, butyrylcholinesterase, and β‑secretase. The inhibi‑ tion constant (K i ) of 1 for MAO‑A was 1.049 ± 0.030 μM with competitive inhibition. Molecular docking simulation predicted that compound 1 forms hydrogen bonds with MAO‑A, and binds more tightly to MAO‑A than to MAO‑B (− 25.02 and − 24.06 kcal/mol, respectively). These results suggest that compound 1 is a selective, reversible, and com‑ petitive MAO‑A inhibitor that can be a therapeutic candidate for treating neurological diseases


Introduction
Alzheimer's disease (AD) is a major concerned-neurodegenerative disease, leading cause of dementia [1].Recently, more than 50 million people have been affected by AD, and it is expected to double every 20 years owing to industrialization and aging of the population [1].The most typical symptoms of AD are memory, intellectual, speech, movement, and space-time disorders [2,3].Despite the increasing number of patients with AD, no effective drugs are available to treat AD [3].However, its pathogenesis remains unclear.Several studies have shown that monoamine oxidase (MAO) level increase, acetylcholine (ACh) level reduction, neuro-inflammation, β-amyloid (Aβ) accumulation, and tau phosphorylation play important roles in AD etiology [1,4].Oxidative deamination of various amines catalyzed by MAO (EC 1.4.3.4) has been reported to be a source of reactive oxygen species (ROS) and a cause of various diseases [4].Moreover, Aβ accumulation in the brain also causes AD, due to abnormal amyloid precursor protein (APP) lysis by β-secretase (BACE1) and neurofibromins tangling by tau hyperphosphorylation [5].
Parkinson's disease (PD), along with AD, are major neurological diseases.Many studies have suggested that the degeneration of dopaminergic neurons, specially, in the nigrostriatal pathway causes PD, due to oxidative stress [6,7].Dopamine metabolism involves in ROS formation in the brain as a major process.In general, dopamine levels are mostly regulated by MAO-A, locating in catecholaminergic neurons.Furthermore, increased levels of MAO-B, a major enzyme involved in dopamine catabolism inside glial cells, lead to neuronal degeneration [8,9].
Therefore, dopamine degradation by MAO-B produces ROS, which causes oxidative stress and PD.Additionally, increased MAO-B activity correlates with cognitive impairment in patients with AD [10,11].
MAOs are located in the mitochondrial outer membrane in two isoforms, i.e., MAO-A and MAO-B, catalyzing the oxidative deamination of neurotransmitters, such as monoamines [12,13].In addition, MAOs are also associated with Aβ accumulation in the brain of AD patients [14].MAO inhibitors, moclobemide and clorgyline inhibit MAO-A, whereas lazabemide, pargyline, selegiline, and rasagiline inhibit MAO-B [15].Among them, MAO-A inhibitors have been used to treat anxiety and depression, and MAO-B inhibitors have been used to treat PD; study to find new MAO inhibitors as neurological disorder treatments is ongoing [16,17].
ACh, one of neurotransmitters in the brain, has important functions in both central and peripheral nervous systems.Cholinesterases (ChEs) include acetylcholinesterase (AChE) and butyrylcholinesterase (BChE), which commonly decompose ACh into choline and acetyl coenzyme A [18,22].AChE catalyzes ACh, whereas BChE catalyzes both ACh and butyrylcholine (BCh), but has a higher preference for BCh than ACh.These ChEs are involved in serine hydrolysis in the body and regulate ACh levels in the glial cells, the hippocampus, and the temporal nerve cortex [17,19,20].Several studies have reported that AChE and BChE inhibitors that increase ACh levels in the brain as treatment candidates for AD [21,22].FDA-approved ChE inhibitors such as donepezil, galantamine, and rivastigmine elevate ACh levels and alleviate AD symptoms [20,23,24].
In the healthy brain, neuronal APPs are typically digested by α-and γ-secretase enzymes with three domains such as cell, cell membrane, and extracellular space.The normal Aβ fragments are soluble polypeptides produced through this digestive reaction and can be recycled in cells.However, Aβ fragments produced by BACE1 and γ-secretase are not recycled and can cause AD through accumulation in brain neurons.Until recently, many studies have focused on BACE1 inhibitors, however, there has been no success in drug approval, while several monoclonal antibodies as Aβ aggregation inhibitors like aducanumab [25] and lecanemab [26] have been developed.Nevertheless, BACE1 inhibitor development remains worthwhile.In addition, multi-target inhibitors that improve cognitive function by simultaneously inhibiting MAOs, ChEs, and BACE1 to increase monoamine and acetylcholine levels and decrease Aβ plaque formation have been developed to treat AD [17,27].
Marine and freshwater fungi have distinct metabolite profiles [28].Marine fungi have garnered interest owing to the secondary metabolites produced, which often have distinct structural features and interesting biological and pharmacological characteristics [29].Marine Penicillium species produce various secondary metabolites with antiviral, anti-bacterial, anti-tumor, and anti-inflammatory activities [30,31].Though various MAO inhibitors of marine origin were described [32], only few MAO inhibitors have been isolated from Penicillium spp.till date.
This study describes the inhibitory activities of six compounds isolated from the marine Penicillium sp.SG-W3 against the enzymes MAO, AChE, BChE, and BACE1.

Isolation and identification of a marine fungal strain SG-W3
The fungal strain SG-W3 was isolated as previously described [33] with a few modifications such as growing the fungus on PDA containing 1% (w/v) NaCl at 28 °C for 7 to 14 days.Seawater was collected from Dangjin, Chungcheongnam-do, Republic of Korea (36º53ʹ19.1ʺN,126º49ʹ36.6ʺE)on March 25, 2020, transported to the laboratory, and filtered using a membrane filter (0.45μm, Hyundai Micro Co., Korea).The filters were then incubated on potato dextrose agar (PDA; BD, USA) and yeast mold agar (YM; BD) supplemented with 3% (w/v) NaCl, 0.1% (w/v) ampicillin, and 0.1% (w/v) streptomycin at 20 ºC for 7-14 days and the SG-W3 fungal colony was isolated and transferred on fresh PDA media.SG-W3 spores were stored in 20% glycerol at -80 ºC and deposited at the Microbial Marine Bio Bank (MMBB) of the National Marine Biodiversity Institute of Korea (MABIK).
Fungal DNA extraction, polymerase chain reaction (PCR), and PCR product purification were performed as previously described [33].Briefly, DNA was isolated from SG-W3 mycelia, and the partial beta-tubulin gene sequence was amplified, as a molecular marker for fungal identification [34].The beta-tubulin segment was sequenced by Macrogen Inc. (Seoul, Korea), and used as a query sequence to search for close sequence matches using BLASTN in GenBank.Phylogenetic tree was constructed using MEGA version 6 [35].

MAO, ChE, and BACE1 inhibition studies
In the primary screening, the inhibitory activity of 10 µM test compounds against MAOs, ChEs, and BACE1 were evaluated.The 50% inhibitory concentration (IC 50 ) of the compounds were calculated up to 40 µM by using Graph-Pad Prism software 5 (San Diego, CA, USA) [13].The selectivity index (SI) was calculated using IC 50 MAO-B/ IC 50 MAO-A.The reference inhibitors were included: toloxatone and clorgyline for MAO-A, safinamide and pargyline for MAO-B, and donepezil for AChE or BChE [49,50].For the inhibition kinetics, compound 1 for MAO-A was used at approximately 0.5-, 1.0-, and 1.5times the IC 50 [13,36], and its inhibition constant (K i ) was determined using the secondary plot of their slopes in the LB plots.

Reversibility studies
The reversibility patterns of compound 1 for MAO-A inhibition were analyzed by measuring the undialyzed (A U ) and dialyzed (A D ) residual activities after pre-incubation for 30 min, as previously described.The inhibition type was determined by comparing to the reference compounds [13,36].

Molecular docking and dynamics simulation
MAO-A and MAO-B structures (PDB ID: 2Z5X and 2V5Z, respectively) were obtained from the Protein Data Bank [40].The three-dimensional structures of compounds 1 (PubChem ID: 54692973), 2, and 4 (PubChem ID: 23902332) were generated from the SMILES string in PubChem [41] and in-house program using OpenBabel [42].AutoDock Vina [43] was used to predict the binding positions of molecules.A docking box for each protein was defined as a cube with a length of 22.5 Å along each axis.The distance between voxel points was set to 0.375 Å. AutoDock4 [44] affinity maps were calculated.The receptors and ligands were prepared using Auto-DockTools [44].
To explain the selectivity of compound 1 to MAO-A and MAO-B, predicted complex structures were further prepared for molecular dynamics (MD) simulations using CHARMM-GUI [45].Cubic boxes with a periodic boundary condition were created by extending 10 Å from the docked complexes along each axis.TIP3P water molecules were used to solvate the boxes, and K + and Cl − ions were then added to neutralize them.Proteins and ligands were parameterized using AMBER FF19SB [46] and GAFF [47], respectively.Both systems were prepared in AMBER input format using CHARMM-GUI [48].
AMBER20 [49] was used to simulate the system.The system was initialized with a minimization of 2500 steepest descent steps, followed by 2500 conjugate gradient steps.Subsequently, the NVT equilibration was performed for 125 ps using a Langevin thermostat.FAD and compound 1 were subjected to position restraints of 1.0 kcal/mol/Å 2 during both steps.Three independent 100 ns MD simulations were conducted for both systems using the SHAKE algorithm [50].The binding affinities of compound 1 and MAOs were determined using MMG-BSA [51] for all MD production trajectories.MMGBSA is one of the widely binding affinity prediction methods using MD simulation.Since it was first proposed in 1998 by P. A. Kollman [PMID: 10052623], a couple of thousand papers have been published [PMID: 25835573] to study the binding affinity of biomolecular complexes.The method calculates the interaction between protein and ligand using molecular mechanics (MM), and solvation of each molecule using generalized-Born (GB) and solvent accessible area (SA) for polar and nonpolar atoms, respectively.Compared to the docking score, it shows higher correlation with the experimentally observed binding affinities [PMID: 23268595].In this study, the igb option was set to 2, and the interaction between the compound and binding-site residues was analyzed by decomposing MMGBSA to the residue level.

Identification of the fungal strain SG-W3
The beta-tubulin segment of a seawater-derived strain SG-W3 was analyzed for molecular identification, and the resulting 402-bp sequence was deposited in GenBank (accession number OR639834).Beta-tubulin sequencebased neighbor-joining phylogenetic analysis showed that SG-W3 was closely related to Penicillium cyclopium CBS 144.45 (MN969380), Penicillium polonicum CBS 222.28 (MN969392), and Penicillium melanoconidium CBS 115506 (MN969387) with 97.77%, 97.52%, and 97.52% sequence identity, respectively (Additional file 1: Fig. S1).In the phylogenetic tree, SG-W3 was clustered separately and thus not identified at the species level.Based on these data, SG-W3 was assigned as Penicillium.
Based on the structural features of compounds 1-6, the plausible biosynthetic pathway of them was proposed to begin with the non-reducing polyketide synthase for the generating C 16 -octaketide.Then, this undergoes cyclization to form atrochryone carboxylic acid, which can be autoxidized to produce endocrocin (5).The emodin derived from this process undergoes enzymatic transformation, involving oxidative ringopening between C-4 and C-5.Subsequent dehydration leads to the formation of intermediate calyxanthone (4) [61].

Antioxidant activity
Compounds 1 and 2 showed antioxidant activity, with 37.8% and 33.1% inhibition in the DPPH assay, respectively (Fig. 2).The activity of compound 1 was similar to that of the reference Trolox (40.7%), indicating that compound 1 is an efficient antioxidant.

Inhibition kinetics
MAO-A inhibition by compound 1 was analyzed using five substrate concentrations (kynuramine) and three  inhibitor concentrations.In LB plots, the lines of compound 1 intersected at a point on y-axis and compound 1 was determined to be a competitive MAO-A inhibitor (Fig. 3A), and its secondary plot showed that its K i was 1.049 ± 0.030 μM (Fig. 3B).These results suggested that compound 1 is a potent competitive MAO-A inhibitor.

Reversibility studies
Inhibition reversibility was evaluated by the dialysis method after pre-incubating the enzyme (two times the IC 50 ) and inhibitor for 30 min.Recovery types were analyzed by comparing the undialyzed (A U ) and dialyzed (A D ) relative residual activities.MAO-A inhibition by compound 1 recovered from 32.70% (A U ) to 75.00% (A D ) (Fig. 4).The relative residual activities of compound 1 was similar to those of toloxatone, a reversible inhibitor (A U 29.53%; A D 80.00%), and different from those of clorgyline, an irreversible inhibitor (A U 36.10%;A D 33.85%).
These results indicate that compound 1 is a reversible MAO-A inhibitor.

Molecular dynamics simulations to explain selectivity
The docking poses of compound 1 to MAO proteins, as the starting structure for MD simulations, are displayed in Fig. 5.The hydrophobic scaffolds of compound 1 in both binding poses substantially overlapped since both protein-binding pockets comprised conserved hydrophobic residues.G67 (58), Y69 (61), F177 (168), V182 During the 100 ns MD simulation, the RMSDs of MAO-A and MAO-B in complex with compound 1 fluctuated around 3-4 and 2-3 Å after 30 ns, respectively (Fig. 6).These shifts may be owing to the C-terminal helix of the proteins.Specifically, the long helix in MAO-A, with approximately 14 additional residues or four helix turns, resulted in significant fluctuations during the simulation.The RMSD for the C-terminal helix of MAO-A (V498-L524) and MAO-B (V489-I501) were 1.1-6.2 and 0.8-3.4Å, respectively.The RMSD of compound 1 was < 0.5 Å in all trajectories, with approximately 0.2 Å fluctuations.Guterres and Im [63] have reported that the RMSD threshold for active molecules was < 3 Å.Therefore, compound 1 may function as an active molecule for both MAO proteins.
The binding affinities calculated by MMGBSA showed that MAO-A (− 25.02 kcal/mol) was a stronger binding partner than MAO-B (− 24.06 kcal/mol).Converting the predicted affinities to K i gives 1.03 × 10 -18 M and 5.03 × 10 -18 M for MAO-A and MAO-B, respectively.The predicted K i for MAO-A and compound 1 complex is much smaller than the experimentally determined one.Due to the enthalpy nature of MMGBSA method, the binding affinity might be overestimated.However, the method is good at relative ranking the compounds [PMID: 25835573].Since the IC 50 values and predicted binding affinities are correlated as discussed, the predicted values are consistent with the experimentally observed IC 50 .
Residue-wise pair decomposition analysis using MMG-BSA identified hotspots for compound 1.Table 2 lists the top 10 residues interacting with the compound in both complexes.Hydrophobic amino acids constituted most residues in both proteins.For MAO-A, Q215 (− 4.12 kcal/mol) and N181 (− 3.95 kcal/mol) were the top two residues driven by electrostatic potential (− 2.16 and − 4.29 kcal/mol, respectively).C172 of MAO-B occupied the same position as N181 of MAO-A and was ranked within the top 10 with a score of − 2.81 kcal/mol.However, the van der Waals interaction between C172 and compound 1 was dominant (− 1.39 kcal/mol).The interaction between MAO-B Y326 and compound 1 was ranked third with − 3.53 kcal/mol.Thus, this interaction placed the compound closer to FAD in MAO-B than that in MAO-A and the interaction between FAD and compound 1 was ranked fourth with − 3.27 kcal/mol.From the decomposition analysis, the selectivity of compound 1 for MAO-A over MAO-B was mainly caused by MAO-A N181/MAO-B C172 and MAO-A I335/MAO-B Y326.
In conclusion, marine fungi-produced secondary metabolites often exhibit distinct structural features and various biological and pharmacological characteristics.In this study, six compounds were isolated from marine Penicillium sp.SG-W3 and their inhibitory activities against MAOs, ChEs, and BACE1 related to neurological diseases were evaluated.Among them, compound 1 (pannorin) showed the highest inhibitory activity against MAO-A (IC 50 = 1.734 μM) and SI > 23.07.The kinetic study showed compound 1 was a competitive reversible MAO-A inhibitor with 1.049 ± 0.030 μM K i .In addition, compound 1 exhibited antioxidant activity in the DPPH assay.However, all the compounds weakly inhibited MAO-B, AChE, BChE, and BACE1, except compound 6 against MAO-B (IC 50 : 11.82 μM).Molecular docking simulation predicted that compound 1 formed hydrogen bonds with MAO-A at Q215 and N181, and have stronger binding energy to MAO-A (− 25.02 kcal/mol) than that to MAO-B (− 24.06 kcal/mol) using MMGBSA.These results suggested that compound 1 is a potent, reversible, competitive, and selective MAO-A inhibitor that could be used as a therapeutic candidate for treating neurological diseases.

Fig. 2
Fig. 2 Antioxidant activity of compounds 1-6 using for 2,2-dipheny-1-picrylhydrazyl (DPPH) assay.The results are the means ± standard errors of triplicate experiments.Trolox was used as a reference inhibitor.The % inhibition was calculated as (absorbance of control − absorbance of reaction mixture)/absorbance of control × 100

Docking simulation was performed to investigate the activity trend of compounds 1 , 2 ,
and 4. The predicted binding affinities are shown in Additional file 1: Table S1.Compound 1 exhibits the lowest binding affinity to MAO-A, and compound 2 shows the lowest binding affinity to MAO-B, as observed in IC 50 values.According to the Cheng-Prusoff equation [PMID: 4202581], lower IC 50 corresponds to the lower binding affinity, although they are not directly related.The predicted K i values of compound 1 to MAO-A and MAO-B from the docking scores are 501 nM and 989 nM, respectively.Su et al. [PMID: 30481020] benchmarked the correlation between the experimentally observed binding affinity and the docking score.AutoDock, which is employed in this study, shows the Pearson's corelation coefficient as 0.604.Based on the benchmark, compound 1 might have a larger K i to MAO-B than to MAO-A.

Fig. 3 AFig. 4
Fig.3A Lineweaver-Burk (LB) plots and B its secondary plot of the slope vs inhibitor concentrations for MAO-A inhibition by compound 1.The experiments were conducted using five kynuramine concentrations and three inhibitor concentrations

Fig. 5
Fig. 5 Predicted binding positions of compound 1 to A MAO-A and B MAO-B.The proteins, compound 1, and FAD are indicated in gold, blue, and pink, respectively

Fig. 6
Fig. 6 RMSD plots of MD simulations for A MAO-A, MAO-B (B), and pannorin-docked C MAO-A and D MAO-B.Complex and ligand RMSDs calculated from the same trajectory are presented as the same color codes

Table 1
Monoamine oxidase (MAO), cholinesterase (ChE), and β-secretase (BACE1) inhibition by the compounds a a Results are the means ± standard errors of duplicate or triplicate experiments b Selectivity index (SI) expressed for MAO-A compared to MAO-B.Except for *, the SI for MAO-B c The 50% inhibitory concentration (IC 50 ) against AChE was 20.614 ± 0.556 μM d The IC 50 of donepezil, a reference AChE and BChE inhibitor

Table 2
Top 10 interacting residues (including FAD) of both MAO proteins with compound 1 identified by MMGBSA