Lyngbyastatins 8–10, Elastase Inhibitors with Cyclic Depsipeptide Scaffolds Isolated from the Marine Cyanobacterium Lyngbya semiplena

Investigation of an extract from the marine cyanobacterium Lyngbya semiplena, collected in Tumon Bay, Guam, led to the identification of three new cyclodepsipeptides, lyngbyastatins 8–10 (1–3). The structures of 1–3 were determined by NMR, MS, ESIMS fragmentation and chemical degradation. Compounds 1–3 are closely related to lyngbyastatins 4–7. Like the latter compounds, we found 1–3 to inhibit porcine pancreatic elastase, with IC50 values of 123 nM, 210 nM and 120 nM, respectively.


Introduction
Marine cyanobacteria have yielded many novel and bioactive secondary metabolites [1], including many cytotoxins [2][3][4]. Our efforts have been focused on discovering novel cyanobacterial metabolites with cytotoxic and other activities. Indeed, we have recently identified several cyanobacterial compounds that are potent protease inhibitors, such as grassystatins AC that selectively inhibit cathepsin E [5] and lyngbyastatins 47 [6,7] (see Figure 1) which inhibit porcine pancreatic elastase. The latter group are part of a class prolifically produced by marine and aquatic cyanobacteria [8][9][10],

OPEN ACCESS
containing 3-amino-6-hydroxy-2-piperidone (Ahp) as part of a six-unit cyclic core with a pendant side chain. These generally inhibit certain serine proteases, due to extensive complementarity to the enzymes' active sites. Additionally, the cyclic core is very rigid, due to hydrogen bonds from the Ahp unit to the opposite Val, making hydrolysis of the inhibitor difficult for the protease enzymes [11][12][13]. Elastase is particularly susceptible to the lyngbyastatin series of this inhibitor prototype, which contain 2-amino-2-butenoic acid (Abu) adjacent to the Ahp unit [6,7,14]. In humans, neutrophil elastase is known to contribute to immunogenic tissue damage in diseases such as cystic fibrosis and asthma [15]. We now report three new members of this class, lyngbyastatins 810 (13, see Figure 1), which are close relatives of lyngbyastatins 47.

Results and Discussion
A sample of Lyngbya semiplena was collected in Tumon Bay, Guam. Samples were freeze-dried and stored frozen for ~8 years before extraction with EtOAc-MeOH (1:1). The resulting non-polar extract was partitioned between hexanes and MeOH-H 2 O (80: 20). The MeOH-H 2 O fraction was further partitioned between n-BuOH and H 2 O. The n-BuOH fraction was subjected to silica and reversed-phase chromatography and finally HPLC to furnish 1 (1.8 mg), 2 (1.1 mg) and 3 (~100 g).  . The 1 H-NMR spectrum suggested that 1 was a depsipeptide, with several exchangeable amide proton signals (~ H 68) and -protons (~ H 45.5). The most downfield -proton also had a corresponding downfield 13 C shift ( H / C 5.53/71.4), as shown in the edited HSQC spectrum (see Table 1), indicating a methine adjacent to an ester linkage. In addition, one N-methyl was present ( H 2.74). A methyl singlet at  H 1.82 suggested the presence of an acetyl group, and there was a downfield singlet ( H 9.38) characteristic of a phenolic OH in DMSO. The 1 H-NMR spectrum of 1 was strikingly similar to the spectra of lyngbyastatins 47 [6,7], and indeed examination of the COSY, edited HSQC, HMBC, ROESY and TOCSY spectra of 1 revealed the same units present in the cyclic core as some other members of the series: Thr, Val, N-Me-Tyr, Phe, Ahp and Abu. Like previously described lyngbyastatins, there was a general paucity of HMBC signals within the cyclic core, but examination of ROESY correlations made it apparent that 1 had the same cyclic core as lyngbyastatins 47. Unlike the latter, 1 contains only hydrophobic residues in the pendant side chain, with Val replacing Htyr and an acetyl group replacing Gas/Ga/GasNa (Table 1 and Table S1, Supporting Information).  Table 1 and Table S2, Supporting Information) revealed the presence of the same units as for 1. However, in place of acetyl, a butanoic acid unit (Ba) was deduced. The HMBC and ROESY correlations were very similar to those observed for 1. Along with the close similarity in chemical shifts, NMR data allowed the assembly of the structure shown ( Figure 1).
Compound 3 was obtained in much lower yield (~100 g). The HRESI/APCIMS spectrum contained ion clusters characteristic of species containing one bromine atom, with two major peaks 2 amu apart in a ratio close to 1:1, and the molecular formula was calculated to be C 49 H 67 N 8  . This is the same molecular formula as 2, except one Br atom was present in place of a hydrogen atom. Additionally, proton and carbon chemical shifts were very similar to 2. Examination of the 1 H-NMR, COSY, edited HSQC and ROESY spectra of 3 in DMSO-d 6 (see Table 1 and Table S3, Supporting Information) suggested the presence of the same units as in 2, except that there were only signals for three aromatic protons attributable to the N-Me-Tyr unit, including a singlet ( H 7.25), indicating that either N-methylated 2'-or 3'-Br-Tyr was present. Chemical shift data and ROESY correlations between the Br-Tyr singlet to both -methylene protons (see Table 1 and Table S3, Supporting Information), supported the proposal that the bromine atom is present at the 3' position. This is consistent with previously reported compounds in this series such as kempopeptin B [14] and cyanopeptolin 954 [16], which possess bromine and chlorine at the 3'-position of Tyr, respectively.
Further evidence for the assigned 2D structures of 13 was obtained by ESIMS fragmentation (Figure 2). Notably, this data unambiguously placed the Br atom in 3 with the N-Me-Tyr unit. Fragmentation data for a related series of cyclic peptides, the aeruginopeptins, has been reported [17]. We found a similar fragmentation pattern using an authentic sample of lyngbyastatin 7 [7], plus another series that was consistent with loss of the pendant side chain followed by ring cleavage between Ahp and Abu ( Figure S1, Supporting Information).
The configuration of the Abu unit was established for 1 by a ROESY correlation between H 3 -33 and the Abu NH (see Table S1, Supporting Information), indicating Z configuration. In 2, this correlation was not observed. However, an unusual 4-bond HMBC correlation was seen between C-30 and H 3 -33 (see Table S2, Supporting Information). Such 4-bond correlations are typically observed in substructures where the bonds between the H and C atoms can form a "w" configuration [18][19][20]. This correlation therefore supports Z configuration for Abu in 2. Configuration of Abu in 3 could not be determined by ROESY, but it is presumed to be Z on the basis of proton chemical shifts in this unit, which are very close to those in 1 and 2, and on biogenetic grounds. If the configuration were E, it is likely that these shifts would be different due to the anisotropic influence of the nearby carbonyl group (C-30). The absolute configurations of other units in 13 were determined by FDLA-based Marfey's analysis [21], using both UV and MS detection, to all be the L-form. The configuration of C-3 of the Ahp unit was determined by CrO 3 oxidation followed by hydrolysis to liberate L-Glu, as described previously [6,7]. The configuration of C-6 of Ahp was determined by examination of ROESY correlations within the unit and by considering coupling constants, such that (3S,6R) configuration was assigned for this unit. CrO 3 oxidation was not carried out for 3 due to lack of material, but the Ahp unit is most likely the same configuration as in 2 and 3, based on overall similarity of NMR data and the fact that the result of Marfey's analysis was the same for other units. ESIMS fragmentation pattern for lyngbyastatins 810 (13). Two series of b ions were observed. In one (top), ring opening at the ester bond occurred, followed by sequential loss of units. In the other (bottom), the side chain was lost, before ring opening at the amide bond between Ahp and Abu. For comparison, see ESIMS fragmentation pattern of lyngbyastatin 7, Figure S1. Note: All ions retained Na + .
Compounds 13 were assessed for inhibition of porcine pancreatic elastase. Lyngbyastatin 7, tested concurrently, exhibited an IC 50 of 47.3  7.6 nM in this assay. By comparison, 13 were all less potent, with IC 50 values of 123  2 nM, 210  10 nM, and 120  16 nM, respectively. Compounds 13 possess the same depsipeptide core as lyngbyastatin 7. And therefore, differences in the side chain residue may play some part in their reduced potency. The presence of exclusively hydrophobic residues in the pendant chain may allow for fewer favorable electrostatic interactions and hydrogen bonding with the enzyme. Conversely, the presence of Br does not seem to have much of an influence of the potency of 3 compared with 1 and 2.

General Experimental Procedures
Optical rotation was measured on a Perkin-Elmer 341 polarimeter. UV spectra were obtained on a SpectraMax M5 (Molecular Devices) and IR data were obtained on a Bruker Vector 22 instrument. 1 H and 2D NMR spectra for 13 in DMSO-d 6 were recorded on a Bruker 600 MHz Avance II Spectrometer, using a 1-mm triple-resonance high-temperature superconducting cryogenic probe [22] for 1 and 2, and a 5-mm cryogenic probe for indirect detection (Bruker CryoProbe TXI) for 3. Spectra were referenced to residual solvent signals [ H/C 2.49/39.5]. HSQC experiments were optimized for 1 J CH = 145 Hz, and HMBC experiments were optimized for n J CH = 7 Hz. HRESI/APCIMS data were recorded on an Agilent LC-TOF mass spectrometer equipped with an APCI/ESI multimode ion source detector in positive ion mode. LC-MS data were obtained using an API 3200 triple quadrupole MS (Applied Biosystems) equipped with a Shimadzu LC system. ESIMS fragmentation data were obtained on an API 3200 by direct injection with a syringe driver. Enzymatic assays were read using a SpectraMax M5. The standard for N-Me-3'-Br-Tyr was prepared as previously described [23].

Extraction and Isolation
A sample of Lyngbya semiplena was collected from Tumon Bay, Guam on December 17, 1998. The freeze-dried organism (dry weight 1.85 kg) was extracted with EtOAc-MeOH (1:1). The extract was concentrated to dryness and partitioned between hexanes and MeOH-H 2 O (80:20). After removal of the solvents, the latter fraction was further partitioned between n-BuOH and H 2 O. The n-BuOH soluble fraction was subjected to silica gel chromatography, using a gradient system of increasing i-PrOH in CH 2 Cl 2 . The fraction eluting with 50% i-PrOH was further purified by reversed-phase chromatography, using a gradient system of increasing MeOH in H 2 O. The fraction eluting with 80% MeOH was purified by reversed-phase HPLC (YMC-Pack ODS-AQ, 250  10 mm, 2.0 mL/min; UV detection at 220 and 254 nm) using a MeOH-H 2 O linear gradient (20100% over 60 min, then 100% MeOH for 10 min), to furnish compounds 1 (t R 42.8 min; 1.8 mg), 2 (t R 47.0 min; 1.1 mg) and impure 3 (t R 49.7 min). Compound 3 was repurified using a different column (Phenomenex Synergi Hydro-RP, 250  10 mm, 2.0 mL/min, PDA detection, 200800 nm), with the same linear gradient used before to yield pure 3 (t R 47.7 min; ~100 g).  Table 1 and Table S1 Table 1 and Table S3

Marfey's Analysis
Samples (~100 g) of compounds 1 and 2 were treated with 6 N HCl at 110 C for 24 h. The hydrolysates were evaporated to dryness and dissolved in H 2 O (100 L). To this 1 M NaHCO 3 (50 L) and a 1% w / v solution of 1-fluoro-2,4-dinitro-5-L-leucinamide (L-FDLA) in acetone was added, and the mixture was heated at 80 C for 3 min. The reaction mixture was then cooled, acidified with 2 N HCl CrO 3 oxidations of 1 and 2 followed by acid hydrolysis were carried out as previously described [6]. The resulting hydrolysates were derivatized with L-FDLA and aliquots subjected to reversed-phase HPLC as above. When compared to the above results, HPLC profiles for derivatives resulting from compounds 1 and 2 showed one new peak corresponding to L-Glu (t R 16.5 min), and the peak corresponding to L-Phe (t R 28.8 min) showed increased intensity. The retention times of authentic standards for L-Glu and D-Glu were 16.5 and 17.7 min, respectively.
A sample (50 g) of 3 was hydrolyzed then derivatized with L-FDLA in the same manner as 1 and 2. Aliquots were subjected to reversed-phase HPLC (Alltech Alltima HP C18 HL 5 m, 250  4.

Elastase Assays
A stock solution of porcine pancreatic elastase (Elastase-high purity; EPC, EC134) was prepared at a concentration of 75 g/mL in Tris-HCl (pH 8.0). Aliquots of test compounds (1 L, DMSO) were preincubated with 79 L Tris-HCl (pH 8.0) and 5 L elastase stock for 15 min at room temperature in a microtiter plate. After this time, 15 L substrate solution was added (2 mM N-succinyl-Ala-Ala-Alap-nitroanilide in Tris-HCl, pH 8.0) to each well, and the reaction was followed by measuring the absorbance at 405 nm every 30 s. Enzyme activity was determined by calculating the initial slope of each progress curve, expressed as a percentage of the slope of the uninhibited reaction. Lyngbyastatin 7 [7] was used as a positive control for inhibition, and the assay was carried out in triplicate.
resonance HTS cryogenic probe used for NMR of compounds 1 and 2 in DMSO-d 6 was developed through collaboration between UF, NHMFL, and Bruker Biospin. We thank J. R. Rocca (UF) for technical assistance, R. Thacker and J. Manglona for help collecting the cyanobacterium, and S. Matthew for preparation of the standard for N-Me-3'-Br-Tyr. This is contribution #796 from the Smithsonian Marine Station at Fort Pierce.