Mechanism of Action of Thalassospiramides, A New Class of Calpain Inhibitors

Thalassospiramides comprise a large family of lipopeptide natural products produced by Thalassospira and Tistrella marine bacteria. Here we provide further evidence of their nanomolar inhibitory activity against the human calpain 1 protease. Analysis of structure-activity relationship data supported our hypothesis that the rigid 12-membered ring containing an α,β-unsaturated carbonyl moiety is the pharmacologically active functional group, in contrast to classic electrophilic “warheads” in known calpain inhibitors. Using a combination of chemical modifications, mass spectrometric techniques, site-directed mutagenesis, and molecular modeling, we show the covalent binding of thalassospiramide's α,β-unsaturated carbonyl moiety to the thiol group of calpain's catalytic Cys115 residue by a Michael 1,4-addition reaction. As nanomolar calpain inhibitors with promising selectivity and low toxicity from natural sources are rare, we consider thalassospiramides as promising drug leads.

less active in the calpain inhibitory assay, strongly indicating that the intact 12-membered ring system is a critical element for the inhibitory activity. Reduction of 1 to 3 also resulted in the saturation of the acyl side chain, which, based on natural thalassospiramide analogues in the series, does not significantly influence the overall calpain bioactivity (see Table 1). These results supported our hypothesis that the a,b-unsaturated carbonyl moiety in the 12membered ring system is essential for the inhibitory activity of calpain. We therefore predicted that Cys115 of calpain attacks the double bond of the unsaturated amide via a Michael-type 1,4addition to form a covalent linkage between the inhibitor and protein. A similar binding mechanism was reported between the active site Thr1 residue of the 20S proteasome and the bacterial natural product syringolin A, a potent proteasome inhibitor that also contains an a,b-unsaturated amide in a 12-membered ring system 18 .
Human Calpain 1 Sample Analysis by MALDI-TOF MS. Topdown and bottom-up mass spectrometry analyses were next used to explore this postulated mode of action [19][20][21][22] . Excess 1 and 3 were added to HCAN1 and incubated separately before being subjected to MALDI-TOF analysis. The results revealed that 1 (957.5 Da) formed a covalent adduct with calpain, as the m/z value was shifted by approximately 974 Da in comparison to the control sample of ''free'' HCAN1 (see Fig. 3A). We measured only a 151 (HCAN1 to 1) complex despite using excessive amounts of 1, suggesting a specific interaction between HCAN1 and 1. Conversely, the HCAN1 1 3 complex did not yield a significant mass shift (see Fig. 3A), as anticipated, which is consistent with the loss of the electrophilic olefin in the 12-membered ring of 1. These findings support the specific binding of 1 to just a single calpain amino acid residue.
To explore the nature of the covalent linkage between thalassospiramide and calpain, we digested three samples (free HCAN1, HCAN1 1 1, and HCAN1 1 3) with trypsin and analyzed the products by MALDI-TOF/TOF MS. The results revealed that the Cys115-containing peptide fragment TDICQGALGDC 115 WLLAA-IASLTLNDTLLHR (cal. m/z 5 3097.6) could be detected in the free HCAN1 and HCAN1 1 3 samples but not in the HCAN1 1 1 sample (see Fig. 3B and Supplementary Fig. 1) in support of the proposed mechanism. Unfortunately, using MALDI-TOF we were unable to detect the anticipated tryptic peptide fragment with 1 (m/z 5 4055.1) 23,24 .
Protein Mutation, Expression, and Identification. Our previous MALDI-TOF MS result showed that only one molecule of 1 binds per calpain (see Fig. 3A), which indicates that the interaction between calpain and thalassospiramide is highly specific. If Cys115 is the specific attachment site, then a Cys115-mutant calpain should not form a covalent bond with 1 and therefore no mass shift should be observed. To test this hypothesis, the protease core of human calpain 1 (domains I-II of m-calpain, ''mI-II'' as abbreviation), which retains the functional and structural elements of intact calpain 25 , was mutated to replace the Cys115 active site residue with Ala (mI-IIC115A). Native and mutated mI-II proteins were expressed, purified and their sequences verified (see Supplementary Fig. 3, 4 and 5). However, the calculated and observed masses of the two proteins, as measured by FT-ICR-MS, were in discordance (mI-II, cal. exact mass: 42239.9 Da, obs. exact mass: 42467.9 Da; mI-IIC115A, cal. Exact mass: 42207.9 Da, obs. exact mass: 42360.1 Da). Considering that the observed mass deviations of mI-II and mI-IIC115A were 228 and 152 amu, respectively, and that both are multiples of 76, we reasoned that these differences were most likely due to the addition of bmercaptoethanol (b-ME, 76.0 Da) from the storage buffer 26 . In order to verify that the observed mass deviations of the intact protein were actually caused by b-ME adducts, we performed a brief (30 minute) trypsin digestion of mI-II, to be followed by MS analysis, without the use of additional reducing agents such as TCEP 27 . HPLC fractionation and MS/MS analysis of the resulting peptides showed that three of the five cysteine residues were modified by b-ME (cys49, cys108 and Cys115; see Supplementary  Fig. 6). This is consistent with the crystal structure of mI-II 28 (PDB file: 1ZCM) in which only these three cysteine residues reside on the solvent-accessible surface of the protein. Verification of the formation of these b-ME adducts clearly explained the observed  masses for both mI-II (1228 Da, 3x b-ME) and mI-IIC115A (1152 Da, 2x b-ME).
Expressed Protein Sample Analysis by FT-ICR and Q-TOF MS. We next incubated excess 1 with mI-II and mI-IIC115A, purified the samples by C4 RP-HPLC, and analyzed them by FT-ICR-MS. HPLC chromatograms showed that the retention time of mI-II was changed by 0.3 min following the addition of 1 and a shift of 1881.5 amu was clearly measured by MS (see Fig. 4 and Fig. 5). This mass shift represents the displacement of b-ME (loss of 76 Da) from Cys115 and the subsequent addition of 1 (addition of 957.5 Da) in its place. On the contrary, neither a change of HPLC retention time nor a mass shift in the MS signal was found for the mI-IIC115A sample after the addition of 1 (see Fig. 4 and Fig. 5). These results clearly signify that 1 interacts specifically with the Cys115 residue. As additional confirmation of these results, both mI-II 1 1 and mI-IIC115A 1 1 samples were digested overnight with trypsin and analyzed by LC-MS/ MS (qTOF), which allowed the detection of the 1-modified fragment in the mI-II 1 1 sample (see Fig. 6 and Supplementary Fig. 8) and not the mI-IIC115A 1 1 sample. Based on these complimentary results, we conclude that thalassospiramide is exclusively associated with the Cys115 residue of calpain by a 1, 4-addition reaction.

Molecular Modeling Study.
To gain further insight into the binding properties between thalassospiramide analogues and calpain, we performed large-scale docking between various thalassospiramides and 30 representative calpain conformations (PDB file: 1ZCM) generated via molecular dynamics simulations. Our docking results showed that thalassospiramide C (4), which experimentally demonstrated the greatest potency, reached Cys115 over the shortest distance, had the lowest docking energy of 28.1 kcal/mol, and experienced the largest number of binding conformations with a distance of less than 5 Å (see Supplementary Table 1 and Supplementary Fig. 9). Moreover, 4 had the shortest lipopeptide side chain that enables tight interaction with the three important regions of calpain (S1, S2, and S3) 14 , which is consistent with our previous observations of markedly different bioactivities associated with the variable hydrophobic tails in the thalassospiramide series. In the case of syringolin A, which shares a similar ring structure to 4, Clerc et al.
found that its proteasome inhibitory activity was enhanced more than 100-fold by modifications to the side chain 29 . Furthermore, the nucleophilic attack of Cys115 leads to the S-configuration of the resulting chiral center, the same as the study of Cysteine Cathepsins 30 .
Preliminary Selectivity and Toxicity Studies. Thalassospiramides 1 and 4 were inactive against the human 20S proteasome at 10 mM, and have been reported to have weak activity against papain and none against trypsin 31,32 . In a preliminary toxicity study, we did not observe any obvious growth inhibition against the bacterial strains Staphylococcus aureus, the fungal strain Candida albicans, and HeLa cells with 10 mM 1 and 4-8. In addition, Oh et al. reported that at a concentration of 10 mM of either 1 or 9 or 20 mM of 10, no cytotoxicity was observed in a mouse model 33 . These preliminary findings suggest that thalassospiramide may possess both low toxicity and good selectivity.

Discussion
We report that thalassospiramide natural products are potent inhibitors of calpain and function differently from standard calpain inhibitors. Most inhibitors possess highly reactive aldehyde groups that bind covalently to the active site of calpain, causing disadvantages such as non-selectivity, instability, and excessive metabolism [34][35][36][37] .
Although other examples of a,b-unsaturated amides have been reported as electrophilic inhibitors of cysteine proteases [38][39][40] , this is the first report that an a,b-unsaturated carbonyl moiety in a rigid ring system functions as an electrophilic warhead at nanomolar scale inhibitory activity against calpain. Interestingly, our previous biosynthetic study revealed that despite the great variety of side chains amongst the 16 known thalassospiramide analogues, the 12-membered macrolide ring is perfectly conserved 17 . These observations    Trypsin Digestion for Free HCAN1, HCAN1 1 1 and HCAN1 1 3 Samples. Trypsin (mass spectrometry grade, Promega) was added to the 3 samples for digestion to a final proteinase: protein ratio of 1550 (w/w) and incubated at 37uC for 24 hours. Digested samples were then desalted and purified using the C18 Zip Tip (Millipore), dissolved in saturated SA matrix solution, and analyzed in reflectron mode and linear mode by MALDI-TOF/TOF-MS.
Expression, Purification and Identification of mI-II and mI-IIC115A. The cDNAs were obtained from HeLa cells by reverse transcription (SuperScriptH III Reverse Transcriptase, Invitrogen TM ), and human mI-II was amplified by PCR using pfu DNA polymerase (Tiangen, China). The active site Cys115 was mutated to Ala by overlap PCR and confirmed by sequencing. The two DNA sequences were inserted into plasmid pet28a using BamHI and SalI, and the recombinant plasmids were sequenced and BLAST against human calpain 1 (see Supplementary Fig. 2). mI-II and mI-IIC115A were expressed in E. coli BL21, and grown on LB medium containing 30 mg/mL kanamycin at 37uC with shaking (225 rpms). Protein expression was induced by the addition of 0.5 mM IPTG at a bacterial concentration of OD 600 0.6, followed by bacterial cell growth for an additional 18 h at 16uC before harvesting. Protein purification was conducted using Ni 21 -affinity columns according to the manufacturer's instructions. Eluted protein was concentrated using a 30 K molecular weight filter and exchanged into storage buffer (20 mM imidazole-HCl, 5 mM b-mercaptoethanol, 1 mM EDTA, 1 mM EGTA, 30% glycol, pH 7.6).
The two expressed proteins were identified by SDS-PAGE and MS/MS annotation of trypsin digests. The map of SDS-PAGE showed that the molecular weight of two expressed proteins were 35-48 kDa (see Supplementary Fig. 3), consistent with the predicted molecular weight. Tryptic peptides of mI-II and mI-IIC115A were analyzed by LC-HR-MS/MS using a Bruker Daltonics micrOTOF-QII. The tryptic peptide coverage of each protein, as determined by MS 1 and verified by MS 2 , and the annotation of active site containing peptides are shown in Supplementary Fig. 4 and Fig. 5.

Intact Protein Sample Preparation and Analysis by FT-ICR-MS.
In preparation for FT-ICR-MS analysis, the 4 proteins samples (mI-II, mI-II 1 1, mI-IIC115A, and mI-IIC115A 1 1) were purified by off-line HPLC (Agilent Infinity 1200 equipped with a multiple wavelength detector). Stock protein samples were thawed and then mixed with a 1 mM CaCl 2 solution. A portion of mI-II and a portion of mI-IIC115A were each mixed with a molar excess of 1 and these were allowed to incubate at room temperature for an additional 10 minutes prior to HPLC purification. Samples, each containing 10 mg of protein, were injected onto a C 4 RP-HPLC column (Jupiter 5 m C 4 , 150 mm 3 4.60 mm, 300 Å , Phenomenex). These samples were loaded at 1 ml/min of 90% mobile phase A (H 2 O 1 0.1% TFA) and 10% mobile phase B (ACN 1 0.1% TFA), which was held for 5 minutes, then eluted by increasing to 90% mobile phase B over 20 minutes. Absorbance at 220 nm was monitored and fractions of the HPLC eluent containing the major peaks were collected, flash frozen, and lyophilized to dryness. Immediately prior to FT-ICR-MS analysis, lyophilized fractions were dissolved in 100 ml of a solution containing a mixture of 49.5% H 2 O, 49.5% methanol, and 1% formic acid (all LC/MS grade). Samples were introduced into an LTQ-FT hybrid mass spectrometer with a 6.4 T magnet (Thermo Electron, North America) using a TriVersa NanoMate (Advion BioSciences, Inc.). The NanoMate was utilized in direct-infusion mode with a spray pressure of 0.3 psig and an ESI voltage of 1.4 kV.  The LTQ-FT ion optics were first tuned to m/z 816 using cytochrome C. The LTQ capillary temperature was maintained at 200uC. For the acquisition of high-resolution FT-ICR-MS spectra of the intact protein samples the following settings were used: 200000 resolution, m/z range of 700-1400, 3 micro-scans per scan, max injection time of 8000 ms, and an FT automatic gain control (AGC) of 8 e5. In order to improve the clarity of the FT-ICR measurements of the intact proteins, an isolation window of 100 m/z units was centered upon m/z 1015 in order to capture multiple charge states of the protein within each of the four samples, and a minimum of 1000 scans were acquired and averaged (see Supplementary Fig. 7). Deconvolution of intact protein FT-ICR mass spectra was performed by using Xtract (Thermo Electron, Bremen, Germany).
Limited Trypsin Digestion and FT-ICR-MS Analysis of mI-II. A sample of protein mI-II was briefly digested using trypsin (Trypsin Singles, proteomics grade, Sigma-Aldrich). A final trypsin:protein ratio of 1520 (w/w) was mixed and incubated at 37uC for 30 minutes in 40 mM ammonium bicarbonate buffer. The reducing agents DTT and TCEP were intentionally excluding as these reducing agents have been shown to remove b-ME from the cysteines 27 . The limited trypsin digest of mI-II was fractionated by HPLC using the same column and gradient that was utilized for the intact protein analysis. The eluent was collected in 1 ml fractions (1 min/fraction), flash frozen, and lyophilized. The collected fractions were dissolved in 30-60 mL of the 49.5% H 2 O, 49.5% methanol, and 1% formic acid mixture and analyzed by FT-MS using similar instrument parameters as previously outlined, with the exceptions being that the FT-ICR resolution was lowered to 50000 and the range of m/z 200-2000 was recorded for all fractions. Peptides were identified on the basis of their intact mass and b-ME modified peptides were verified by MS/MS fragmentation (see Supplementary Fig. 6).
Bottom-up Analysis of Protein Samples by LC-HR-MS/MS. In preparation for bottom-up analysis of the 4 proteins samples (mI-II, mI-II 1 1, mI-IIC115A, and mI-IIC115A 1 1) each one was reduced with DTT, alkylated, and incubated with a trypsin: protein ration of 1520 (w/w) at 37uC for 18 hours. Samples were subsequently quenched with an equal volume of 10% formic acid. For LC-HR-MS/ MS analysis and Agilent Infinity 1290 UPLC connected to a Bruker Daltonics micrOTOF-QII qTOF mass spectrometer. The trypsin digested samples were separated using an RP-C 18 column (Luna 3 m C 18 (2), 100 mm 3 2.0 mm, 100 Å , Phenomenex). Samples were loaded at 0.25 ml/min of 95% mobile phase A (100% H 2 O 1 0.1% formic acid) and 5% mobile phase B (ACN 1 0.1% formic acid), which was held for 3 minutes, then mobile phase B was increased to 65% over 30 minutes, and then B was increased to 100% over 2 minutes. The micrOTOF-QII was configured to have a nebulizer pressure of 1.6 bar, dry gas temperature of 200uC, dry gas flow-rate of 7 L/min. Mass spectra were acquired with an MS 1 spectral rate of 2 Hz, an MS 2 rate of 3 Hz, and the collision RF was stepped from 200 Vpp to 600 Vpp (increments of 100 V) during each MS 2 event in order to maximize observed fragment ions. A "lock-mass" internal calibrant (m/z 922.01; CAS 58943-98-9) was continuously introduced into the mass spectrometer in order to maintain proper calibration throughout the run. The MS/MS annotation of thaA-modified peptide is detailed in Supplementary Fig. 8.

Molecular Dynamics Simulations and Docking Study of Calpain and
Thalassospiramides. The currently available structures for calpain-1 are 1ZCM and 2ARY, from which we selected the former because it is a bound structure in which the ligand is covalently linked to Cys115 28 . To prepare the MD simulations, we removed the ligand and determined the protonation states of all histidines using the PDB2PQR server 41 . The protein was then solvated in a water box with 13309 SPC waters 42 and 9 sodium ions to neutralize the system. The GROMACS 4.5.5 simulation package 43 with the Amber99sb force field 44 were used in the MD simulations due to their speed. The system was minimized with a steepest descent algorithm followed by 200 ps MD simulations with a position restraint for heavy protein atoms, under NPT conditions www.nature.com/scientificreports SCIENTIFIC REPORTS | 5 : 8783 | DOI: 10.1038/srep08783 with 1 bar of pressure and a temperature of 310 K using a V-rescale thermostat 45 . The cut-offs for both VDW and short-range electrostatic interactions were set at 10 Å , and long-range electrostatic interactions were treated according to the Particle-Mesh Ewald method 46 . Water molecules were constrained using the SETTLE algorithm 47 , and all protein bonds were constrained by the LINCS algorithm 48 . Finally, we performed three independent 50-ns NPT simulations with the same initial conformation (i.e., the last frame of the position restraint simulation) and different initial velocities. The conformations were saved every 5 ps.
All of the 30,000 MD simulation frames were collected and divided into 30 clusters according to the C-alpha of two flexible loops (residue 69-82 and 251-261) using the k-center algorithm 49 . The average RMSD within each cluster was approximately 1.5 Å . We chose one representative conformation from each cluster for flexible docking using AutoDock Vina.
We used AutoDock Tools (ADT) to prepare the proteins and ligands prior to docking. The ADT merged nonpolar hydrogens of proteins and automatically detected each bond type of the ligands. Proteins and ligands were added with Gasteiger partial charges 50 . AutoDock Vina, which significantly improved the average accuracy of the binding mode prediction compared to AutoDock 4, was applied for the docking study. The docking grid box was centered near Cys115 with a volume of 24*26*34 (A 3 ), and each docking had 48 (parameter: exhaustive) parallel runs. The sequence index of the selected flexible residues was: 72, 79, 109, 115, 254, 260, 261, 272 and 298 51 .
Fluorescence-based 20S Proteasome Inhibition Assay. Thalassospiramide 1 and 4 were serially diluted in methanol to obtain various concentrations, and then incubated with 1 nM 20S proteasome in 96-well plates (Tris 25 mM, pH 7.5; SDS 0.03%; EDTA 0.5 mM, final reaction volume 40 mL) for 15 min at 37uC. The specific fluorogenic proteasome substrate Suc-LLVY-AMC (10 mL, 200 mM) was added and incubated for another 15 min at 37uC, and then the fluorescence in each well was measured. The assay wells without inhibitor were conducted as positive control (related to 100% of enzyme activity) 52 .
Preliminary Toxicity Assays. The preliminary toxicity assays included antimicrobial and cytotoxic assays. In antimicrobial assay, LB (10 g of tryptone, 5 g of yeast extract, 10 g of NaCl, 1 L of dd H 2 O) and potato dextrose medium (200 g of potato, 20 g of dextrose, 1 L of dd H 2 O) were used to inoculate strain S. aureus ATCC 43300 and C. albicans ATCC 76615 (28uC, for 12 h), respectively. The test samples were prepared at 1 mM in DMSO and added to the broth (96-well plate) at 1% (V/V, final concentration 10 mM). Then, the microbes were incubated at 28uC overnight. Cell growth was recorded by measuring the optical density at 600 nm, and Penicillin (for S. aureus) and Natamycin (for C. albicans) were used as positive controls. Human HeLa cells were used in cytotoxic assay. Cells (80 mL, 1 3 10 5 /mL) were planted into 96-well plates for 12 h, then the test samples were dissolved in DMSO and added to the assay medium to inoculate for another 48 h. The cell viability was assayed by the MTT method.