Effect of the marine β-carboline alkaloid Manzamine A on RSK1 vs RSK2 inhibition: a biochemical and computational study

Manzamines are complex polycyclic marine-derived β-carboline alkaloids with reported anticancer, immunostimulatory, anti-inflammatory, antibacterial, antiviral, antimalarial, neuritogenic, hyperlipidemia and atherosclerosis suppression bioactivities, putatively associated with inhibition of glycogen synthase kinase-3, cyclin-dependent kinase 5, and vacuolar ATPases. We hypothesized that additional and yet undiscovered molecular targets might be associated with Manzamine A (MZA) reported pharmacological properties. We report herein for the first time to our knowledge that MZA inhibited a 90kDa ribosomal protein kinase S6 (RSK1) when screened against a panel of 30 protein kinases. Furthermore in vitro RSK kinase assays demonstrated a 10-fold selectivity in potency of MZA against RSK1 versus RSK2. MZA’s differential binding and selectivity toward the two isoforms is also supported by computational docking experiments. Specifically, the RSK1-MZA (Nand Ctermini) complexes appear to have stronger interactions and preferable energetics contrary to the RSK2-MZA ones. In addition, our computational strategy suggests that MZA binds to the N-terminal kinase domain of RSK1 rather than the C-terminal Preprints (www.preprints.org) | NOT PEER-REVIEWED | Posted: 4 November 2020 doi:10.20944/preprints202011.0174.v1 © 2020 by the author(s). Distributed under a Creative Commons CC BY license. domain. RSK is a vertebrate family of cytosolic serine-threonine kinases that act downstream of the ras-ERK1/2 (extracellular-signal-regulated kinase 1/2) pathway, which phosphorylates substrates shown to regulate several cellular processes including growth, survival and proliferation. Consequently, our findings have lead us to hypothesize that MZA and the 80 currently known manzamine-type alkaloids isolated from several sponge genera, may have novel pharmacological properties.


Introduction
Marine biological and chemical diversity continues to demonstrate great potential to contribute novel pharmacology for multiple therapeutic categories [1]. Although the current marine pharmaceutical clinical pipeline consists mainly of compounds developed for cancer chemotherapy [2], several marine natural products have been shown to target a range of protein kinases as well [3].
The 90 kDa ribosomal S6 Kinase or RSK is a vertebrate family of cytosolic serine-threonine kinases that contains four homologous isoforms namely RSK1-4, which act downstream of the ras-ERK1/2 (extracellular-signal-regulated kinase 1/2) pathway [16]. RSK1 and RSK2 adult and embryonic tissue expression has been investigated, with RSK1 found in lung, kidney, pancreas and brain (cerebellum and microglia) [16][17][18], while RSK2 is more abundant in skeletal muscle, heart, pancreas and brain (neocortex, hippocampus, and cerebellum) [16]. RSK kinases are composed of two functional kinase catalytic domains: the C-terminal kinase domain (CTKD) that belongs to the calcium and calmodulin-regulated kinases CamK family, is phosphorylated by ERK1/2, and activates the N-terminal kinase domain (NTKD) that belongs to the protein kinase A, G, and C (AGC) family [16]. NTKD has been shown to phosphorylate several substrates, and in turn regulates several cellular processes including growth, survival, and proliferation [16].
Interestingly, loss of RSK2 function causes a rare form of mental retardation known as Coffin-Lowry syndrome [19], while sustained activation of RSKs appears to be linked to cancer [20,21]. RSK inhibitors used in preclinical studies include two NTKD-interacting inhibitors at the ATPbinding site, namely the reversible dihydropteridinone BI-D1870 [22], and the kaempferol glycoside SL0101 [23], and the pyrrolopyrimidine FMK, a CTKD-interacting irreversible inhibitor that binds to the ATP-binding pocket of RSK [24].
Herein we report that MZA inhibited the 90kDa RSK1 when screened against a panel of 30 protein kinases. In vitro kinase assays demonstrated a 10-fold selectivity in potency between RSK1 and RSK2 (IC50 values of 15.01 μM and 108.4 μM, respectively). Furthermore, MZA is predicted to bind to the ATP-binding pocket of NTKD of RSK1 and its selectivity towards RSK1 versus RSK2 are supported by our computational studies reported herein.
As shown in Table 1, MZA (1 M) reduced the activity of rat p90 ribosomal S6 kinase 1 or RSK1 by 68%. Furthermore, MZA showed no significant inhibitory effect on the other 29 kinases which were part of the University of Dundee protein kinase enzyme panel.

Computational Studies
Superimposition of respective NTKDs and CTKDs of crystal RSK1 and RSK2 structures (see Methods) to identify 'a' representative structure for subsequent docking experiments was undertaken. Selected structures are presented in Table 2.
The active sites of NTKD and CTKD RSK1 (PDB entries 2Z7Q and 3RNY, respectively) with amino acids reportedly involved in binding interactions are depicted in Figure 3 A Table 2

B
Comparison of the binding pockets in either terminus of RSK1 and RSK2 was performed by overlaying the selected CTKD of RSK1 (PDB 3RNY) onto the CTKD of RSK2 (PDB 4D9T).   Because we had no prior knowledge as to whether MZA preferentially binds to the one terminus over the other in RSK1, we superimposed the two termini of RSK1. Figure 6A shows the overall NTKD and CTKD RSK1 crystal structures overlayed with the amino acids in the ATP binding pocket of CTKD RSK1 labelled. It should be pointed out there is notable similarity. Figure 6B shows a comparison of the active sites.

Discussion
The RSK family of proteins consists of four isoforms that are highly homologous, with the exception of the N-and C-terminal sequences where divergence is observed. RSKs can receive signals to their CTKD and in turn will transmit an activating signal to their NTKD. The two catalytic domains are connected via a conserved linker of approximately 100 amino acids. As noted earlier, NTKD belongs to the AGC family and is responsible for substrate phosphorylation. CTKD is homologous with the Ca 2+ /calmodulin-dependent kinase family and responsible for NTKD's activation via authophosphorylation.
Although the search for novel RSK inhibitors is ongoing, to our knowledge there is no RSK isozyme specific inhibitor. The RSK reversible dihydropteridinone BI-D1870 [22], and kaempferol glycoside SL0101 [23] inhibit the NTKD ATP-binding site with comparable potency.
Similarly, the pyrrolopyrimidine FMK is a CTKD-irreversible inhibitor that binds to the ATPbinding pocket of RSK [24].
Our current study suggests that MZA is a moderate yet selective inhibitor of the RSK1 isozyme based on the biochemical evidence. Thus, we set out to explore computationally MZA's potential binding in order to shed light on the observed selectivity towards RSK1 versus RSK2. Because multiple crystal structures of CTKD and NTKD of both RSK1 and RSK2 have been reported in the literature, we were not certain whether MZA binds to the N-or C-terminal domains.
Consequently, we opted to perform docking experiments toward both NTKD and CTKD of the two isoforms in order to: 1) Rationalize the observed selectivity between the two isozymes and, 2) Identify the plausible preferential binding for one domain over the other.
At first, we superimposed the NTKDs of all crystal RSK1, CTKDs of RSK1, and repeated with an overlay of the NTKDs and CTKDs of RSK2 in order to identify representative structures. Table   2 shows the structures selected for all subsequent work. It should be noted that we favored the active conformation of NTKD RSK2 (PDB 3G51) [27] because we felt it would be more informative regarding the conformational transitions of the domain. Furthermore, even though the CTKD of RSK1 is an apo structure, its ATP-binding cleft has been described [28]. with Glu 500 of 4D9T being the corresponding residue (see Figure 4 and Tables 3b) [28]. A similar overlay was undertaken for the NTKDs of RSK1 and RSK2 ( Figure 5 and Table 3a). It can be seen that the ATP-binding cleft of RSK1 consists of Asp 142, Asp 205, Leu 144, and Asn 192, as reported by Ikuta et al. [29]. The active conformation of RSK2 is lined by polar charged residues Because MZA is shown herein to be selective towards RSK1 over RSK2, and since we have no prior knowledge as to whether it preferentially binds to the one terminus over the other, we superimposed the two termini of RSK1. Figure 6A shows the overall NTKD and CTKD RSK1 crystal structures overlayed with the amino acids in the ATP binding pocket of CTKD RSK1 labelled. In comparing the two topologies, there is a notable similarity, in that with the exception of the N-lobe consisting of a five-stranded antiparallel β sheet and a missing helix corresponding to αE of CTKD, the remaining structures superimpose rather well. Figure 6B      Consequently, it seems that MZA forms a tighter complex with RSK1 and is positioned deeper in the pocket contrary to RSK2. Consequently, using docking experiments coupled with visual comparisons of reported crystal structures we have been able to explain observed binding affinities and selectivity of MZA towards RSK1. Another point of concern was whether we could predict the ligand's binding to NTKD or CTKD. Because the MZA-NTKD RSK1 complex appears to be more stable due to stronger interactions with the surrounding amino acid residues in the ATPbinding pocket, in contrast to the MZA-CTKD RSK1 complex, we suggest that MZA preferentially binds to NTKD. Furthermore, an examination of the predicted binding energies of the NTKD and CTKD complexes with MZA (emodels of -62.132 and -55.497, respectively) support the notion that MZA shows a preference towards the NTKD of RSK1.
In conclusion, we report herein an integrated biochemical and computational study investigating the marine-derived alkaloid MZA. We show that MZA selectively inhibits RSK1 versus RSK2 and predict that it binds to its N-terminal kinase domain. To our knowledge, this is the first report of RSK1 as a new macromolecular target of MZA, a finding that extends the pharmacology of the manzamine-type alkaloids.

Materials
Manzamine A (MZA) (Figure 1) was isolated from a marine sponge species of the genus

Haliclona
collected off Manzamo, Okinawa, Japan as described [8]. The reversible dihydropteridinone BI-D1870, an NTKD-interacting ATP-binding site specific RSK inhibitor was provided by P. Cohen, University of Dundee, Scotland, U.K. [22]. A 10 mM stock of BI-D1870 and MZA were prepared in DMSO and stored at -80 C prior to use in the experiments.

Protein kinase activity assays
Thirty protein kinase activity assays were performed by the Kinase Profiling Screening Service, University of Dundee, as described elsewhere [25,26]. MZA (1 μM) was used in all protein kinase assays. Results are presented as a percentage of kinase activity of control incubations (average of duplicate determinations) and are tabulated in Table 1.
Structural comparisons were carried out within the Maestro interface (Schrodinger, LLC, New York, NY) in order to 1) select the structure that would be similar to the majority of representatives of each terminus (Table 2), and 2) identify critical residues that would later serve to define respective binding pockets. Amino acids lining the binding pockets of the crystal RSK1 and RSK2 structures reveal the high degree of similarity (Tables 3a and 3b).  [31]. b Residues in RSK2 that are equivalent to residues reported in reference [31] for RSK1. c Equivalent residues reported in [31,32]. d Residue in RSK1 corresponding to the one reported for RSK2 [32]. e Reference in RSK1 that is equivalent to the one reported for RSK2 [27].

Protein preparation
The receptors were prepared by assigning bond orders, adding hydrogens and finding overlaps, followed by hydrogen bond optimization and minimization using the Protein Preparation Wizard.
Minimization employed the 'Impref' utility, which runs a series of constrained impact minimizations with gradually decreasing strength of the heavy-atom restraining potential. Two minimizations were initially performed with the heavy-atom restraint potential force constant at 10. In the first minimization, the torsional potential was turned off to improve hydrogen optimization, whereas the second minimization restored the torsional potential. The restraining potential force constant was subsequently reduced to 3, 1, 0.3 and 0.1. If the output structure from a minimization exceeds the specified RMSD threshold, relative to the starting structure, the program stops and returns the structure from the previous minimization. Thus, the RMSD is checked at the end of each round of minimization. Receptor grid generation was subsequently employed with the van der Waals radius scaling factor set to 1.0, and partial charge cut-off at 0.25.

Docking Glide 5.7
Glide version 5.7 [30,35] was used in this study. Glide grids were generated with ligand scaling of 0.8 for the van der Waals radii. Ligands were docked and ranked using standard precision, followed

Conflicts of Interest
The authors declare no conflict of interest.