The non-ribosomal assembly and frequent occurrence of the protease inhibitors spumigins in the bloom-forming cyanobacterium Nodularia spumigena

Authors

  • David P. Fewer,

    1. Department of Applied Chemistry and Microbiology, University of Helsinki, PO Box 56, Viikki Biocenter, Viikinkaari 9, FIN-00014, Finland.
    Search for more papers by this author
  • Jouni Jokela,

    1. Department of Applied Chemistry and Microbiology, University of Helsinki, PO Box 56, Viikki Biocenter, Viikinkaari 9, FIN-00014, Finland.
    Search for more papers by this author
  • Leo Rouhiainen,

    1. Department of Applied Chemistry and Microbiology, University of Helsinki, PO Box 56, Viikki Biocenter, Viikinkaari 9, FIN-00014, Finland.
    Search for more papers by this author
  • Matti Wahlsten,

    1. Department of Applied Chemistry and Microbiology, University of Helsinki, PO Box 56, Viikki Biocenter, Viikinkaari 9, FIN-00014, Finland.
    Search for more papers by this author
  • Kerttu Koskenniemi,

    1. Department of Applied Chemistry and Microbiology, University of Helsinki, PO Box 56, Viikki Biocenter, Viikinkaari 9, FIN-00014, Finland.
    Search for more papers by this author
  • Lucas J. Stal,

    1. NIOO-KNAW, Centre for Estuarine and Marine Ecology, PO Box 140 4400, AC Yerseke, the Netherlands.
    Search for more papers by this author
  • Kaarina Sivonen

    Corresponding author
    1. Department of Applied Chemistry and Microbiology, University of Helsinki, PO Box 56, Viikki Biocenter, Viikinkaari 9, FIN-00014, Finland.
    Search for more papers by this author

*E-mail kaarina.sivonen@helsinki.fi; Tel. (+358) 9 19159270; Fax (+358) 9 19159322.

Summary

Nodularia spumigena is a filamentous nitrogen-fixing cyanobacterium that forms toxic blooms in brackish water bodies worldwide. Spumigins are serine protease inhibitors reported from a single strain of N. spumigena isolated from the Baltic Sea. These linear tetrapeptides contain non-proteinogenic amino acids including a C-terminal alcohol derivative of arginine. However, very little is known about these compounds despite the ecological importance of N. spumigena. We show that spumigins are assembled by two non-ribosomal peptide synthetases encoded in a 21 kb biosynthetic gene cluster. The compact non-ribosomal peptide synthetase features a reductive loading and release mechanism. Our analyses demonstrate that the bulk of spumigins produced by N. spumigena are released as peptide aldehydes in contrast to earlier findings. The main spumigin E variant contains an argininal residue and is a potent trypsin inhibitor. Spumigins were present in all of the N. spumigena strains isolated from the Baltic Sea and comprised up to 1% of the dry weight of the cyanobacterium. Our results demonstrate that bloom-forming N. spumigena strains produce a cocktail of enzyme inhibitors, which may explain in part the ecological success of this cyanobacterium in brackish water bodies worldwide.

Introduction

Cyanobacteria are a rich source of secondary metabolites with potential applications as antimicrobial, anticancer agents or useful probes in cell biology studies (Burja et al., 2001; Welker and von Döhren, 2006; Sivonen and Börner, 2008). Many of these secondary metabolites can be classified as peptides or possess peptidic substructures and are generally assumed to be the end-products of a non-ribosomal biosynthetic pathway (Welker and von Döhren, 2006). Non-ribosomal peptides are a very diverse family of natural products with a broad range of biological activities and pharmacological properties (Marahiel et al., 1997; Finking and Marahiel, 2004; Sieber and Marahiel, 2005). Non-ribosomal peptides are synthesized on large multifunctional enzymes complexes that catalyse the assembly of complex natural peptides from amino acid monomers (Marahiel et al., 1997; Finking and Marahiel, 2004; Sieber and Marahiel, 2005).

Nodularia spumigena is a diazotrophic filamentous cyanobacterium found in brackish water bodies. The extensive late summer blooms of N. spumigena in the Baltic Sea are among the largest in the world. These blooms are toxic through the production of cyclic pentapeptide nodularin (Sivonen et al., 1989). Nodularins are eukaryotic protein phosphatase inhibitors and act as hepatotoxins and constitute a health risk to humans as well as domestic and wild animals (Kuiper-Goodman et al., 1999; Sivonen, 2009). The production of nodularin may contribute to the ecological success of this cyanobacterium by acting as a grazing deterrent (Karjalainen et al., 2003; Ibelings and Havens, 2008). Nodularins are assembled on large enzyme complexes comprised of non-ribosomal peptide synthetases and polyketide synthases (Moffitt and Neilan, 2004). N. spumigena produces cyclic nodulapeptins as well as linear spumigins (Fujii et al., 1997a) while Nodularia sphaerocarpa produces the highly modified glycopeptide suomilide (Fujii et al., 1997b). However, these compounds are known only from single strains of the genus Nodularia. The frequency, biosynthetic origins and bioactivities of these compounds remain poorly characterized despite the ecological importance of Nodularia.

Spumigins are linear tetrapeptides reported to contain N-terminal d-hydroxyphenyllactic acid, d-homotyrosine (2S,4S)-4-methylproline (mPro) and C-terminal argininol (Fig. 1). Four variants of spumigin have been described from the cyanobacterium N. spumigena AV1, varying in the derivatization of the C-terminal amino acid and whether or not they contain l-Pro or (2S,4S)-4-methylproline (Fujii et al., 1997a). Spumigins are also protease inhibitors (Fujii et al., 1997a) and considered members of the aeruginosin class of peptides in which the Choi residue is replaced by mPro (Welker and von Döhren, 2006; Ishida et al., 2007). The presence of non-proteinogenic amino acids (Fujii et al., 1997a) coupled with structural similarity to aeruginoside (Ishida et al., 2007) suggested that spumigins might also be the product of a non-ribosomal assembly line. Here we report the discovery, annotation and biochemical analysis of the spumigin biosynthetic gene cluster from N. spumigena CCY9414. In contrast to earlier results (Fujii et al., 1997a) we demonstrate that the bulk of the spumigins are peptide aldehydes, potent protease inhibitors and show that they are common in N. spumigena strains.

Figure 1.

The chemical structure of the linear spumigin tetrapeptides produced by Nodularia spumigena AV1 and CCY9414. The linear peptides, spumigins A–H, differ through the presence of Hty or Hph, Pro or mPro, or the presence of argininal, argininol or arginine. Abbreviations: Hty, homotyrosine; Hph, homophenylalanine; Pro, proline; and mPro, (2S,4S)-4-methylproline.

Results

Identification and structural elucidation of spumigin variants in N. spumigena CCY9414 and AV1

We identified eight variants of the spumigin tetrapeptide from N. spumigena AV1 and seven variants from the CCY9414 strain (Table 1). These variants differ from one another by the presence of Hty or Hph in the second position, Pro or mPro in the third position, and the presence of aldehyde, alcohol and acid forms of Arg at the C-terminus (Table 1, Fig. 1). Liquid chromatography-mass spectrometry (LC-MS) analysis showed that spumigin E was the main variant produced by both N. spumigena strains (Table 1). The presence of an aldehyde functional group was explored by derivatization with 2,4-dinitrophenylhydrazine (DNPH) which specifically reacts with ketones and aldehydes. The protonated ion m/z 791 corresponding to a DNPH derivative of spumigin E was found from the chromatograms of both Nodularia strains. Product ion spectra of both native and derivatized spumigin E (Fig. 2) confirmed the structure and the presence of an aldehyde functional group. Commercially available leupeptin, which also contains a C-terminal argininal, was derivatized with DNPH and the product ion spectrum produced from the protonated molecular ion showed the same fragmentation pattern as spumigin E (Supporting information). Other new argininal containing spumigin variants, spumigin F, G and H differ in the presence of Hph or mPro (Table 1). Hydrated forms of spumigins E–H were also detected together with the free aldehyde forms of these spumigins (Table 1). Our results demonstrate that approximately 80% of the spumigins produced by N. spumigena AV1 and CCY9414 are released from the peptide synthetase enzyme complex as aldehydes (Table 1).

Table 1.  Structure, ion mass, retention time and abundance of spumigins identified from Nodularia spumigena AV1 and CCY9414.
Peptide variantSubunits and their order[M+H]+
(m/z)
Rt
(min)
[M+H2O+H]+
(rel inta)
Relative amount of each variant (%)
1234AV1CCY9414
  1. a. Relative intensity compared with [M+H]+.

  2. Argal, argininal; Argol, argininol; Arg, arginine; Hpla, hydroxyphenyllactic acid; Hty, homotyrosine; Hph, homophenylalanine; Pro, proline; mPro, (2S,4S)-4-methylproline; ND, not detected.

Spumigin AHplaHtymProArgol61315.2ND1517
Spumigin B1HplaHtymProArg62715.4ND< 1< 1
Spumigin B2HplaHtymProArg62714.8ND< 1< 1
Spumigin DHplaHtyProArgol59914.1ND73
Spumigin EHplaHtymProArgal61115.1629 (27)6169
Spumigin FHplaHtyProArgal59714.0615 (23)149
Spumigin GHplaHphmProArgal59519.0613 (20)21
Spumigin HHplaHphProArgal58117.8599 (20)< 1ND
Figure 2.

DNPH derivative of the main spumigin variant, spumigin E, analysed by LC-MS. (A) The m/z 791 ion chromatogram shows the location of DNPH-spumigin E in the total ion chromatogram, (B) product ion spectrum from ion 791. The precursor ion is indicated with a diamond. (C) Assignments of the product ions.

The majority of the remaining spumigins, spumigin A and D, contained the argininol alcohol version of arginine (Table 1). Trace amounts of spumigins containing the acid version of arginine were present (Table 1). Spumigin H was detected only from N. spumigena strain AV1. All structures were verified with product ion spectra from the protonated precursor spumigin ions and are presented in Supporting information. Spumigin concentrations in dried biomass gave a total spumigin concentration of 7 mg g−1 in N. spumigena AV1. The amounts of spumigins produced were three to eight times higher in N. spumigena AV1 than in strain CCY9414.

Spumigin gene cluster

Several features of the spumigin chemical structure suggest that they could be synthesized on a non-ribosomal peptide synthetase complex. We identified a 21 kb gene cluster in the partial genome of N. spumigena CCY9414. The gene cluster was located on a single scaffold and spread over four contigs in the partial genome of N. spumigena CCY9414 (AAVW01000147, AAVW01000160, AAVW01000102, AAVW01000188). We designed oligonucleotide primers and PCR amplified the missing 98, 606 and 39 bp regions from the genome of N. spumigena CCY9414. The 21 kb spu gene cluster contains six genes encoding proteins which we postulated to be responsible for the assembly and transport of the linear tetrapeptide spumigin in N. spumigena CCY9414 (Table 2). The spumigin cluster encodes two multidomain NRPS proteins, SpuA and SpuB, that make up the four modules (Fig. 3). The first protein encoded in the spu gene cluster, SpuA (1401 aa), is an unusual mixed NRPS-PKS protein with a mass of 156 498 Da containing an adenylation (A), a ketoreductase (KR) and a peptidyl carrier protein (PCP) domains (Fig. 3). The second protein of the spu gene cluster, SpuB (4171 aa), encodes three NRPS modules with a mass of 468 044 Da. The first module of SpuB contains a condensation (C), an A, a PCP and an internal epimerase (E) domain (Fig. 3) with a predicted substrate specificity of homotyrosine (Table 3). The second module of SpuB possesses a C, A and PCP domain with a predicted substrate specificity of (2S,4S)-4-methylproline or l-Pro (Table 3). The third module of SpuB contains a C, A and PCP domain with a predicted substrate specificity of l-Arg (Table 3). This module also has a C-terminal reductase (R) domain (Fig. 3). The R domain containing a putative NAD(P)H binding site was identified at the C-terminal end of SpuB in place of the more common thioesterase domain. The genetic architecture and domain organization of SpuA and SpuB NRPS is colinear with the order of the putative catalytic reactions necessary for the assembly of the linear tetrapeptide.

Table 2.  Proposed functions of SpuA–F proteins encoded in the spumigin biosynthetic gene cluster.
ProteinAmino acidsProposed functionSequence similarityOrganismIdentity (aa)Accession number
SpuA1401NRPSNRPSAnabaena variabilis
ATCC29413
70%, 82% (1010)ABA21232
SpuB4171NRPSNRPSBrevibacillus brevis44%, 62% (2595)CAD92852
SpuC265UnknownHypothetical proteinNostoc punctiforme
PCC73102
87%, 93% (246)ZP_00110901
SpuD368DehydrogenaseThreonine dehydrogenase and related Zn-dependent dehydrogenasesNostoc sp. GSV22491%, 96% (336)AAF17283
SpuE274ReductasePyrroline-5-carboxylate reductaseNostoc sp. GSV22481%, 90% (218)AAF17284
SpuF683ABC-transporterABC-type uncharacterized transport system, permease and ATPase
components
Nostoc sp.
ATCC53789
66%, 82% (455)AAO23332
Figure 3.

Map of the spu gene cluster and proposed biosynthetic pathway of spumigins.
A. Structure and genetic organization of the 21 kb spu biosynthetic gene cluster from Nodularia spumigena CCY9414. Oligonucleotide primer pairs P1: gapAF and gapAR; P2: gapBF and gapBR; and P3: gapCF and gapCR; were used to close gaps between contigs. (Black) genes encoding peptide synthetases; (white) genes encoding proteins involved in biosynthesis of (2S,4S)-4-methylproline and an ABC transporter.
B. Proposed catalytic scheme with activation of prephenate and reductive release of the final spumigin. Abbreviations: A, adenylation; ACP, acyl carrier protein; C, condensation; PCP, peptidyl carrier protein; KR, keto reductase domain; E, epimerase domain; R, reductase domain.

Table 3.  Proposed and activated substrates as well as the 10 amino acid residues predicted to line the substrate binding pocket of the SpuA1 and SpuB1−3 adenylation domains (Stachelhaus et al., 1999).
NRPS moduleProposed substrateActivated substrateResidue
235236239278299301322330331517
  1. Hpla, 4-hydroxyphenyllactic acid; Hpp, 4-hydroxyphenylpyruvate; Hty, homotyrosine; mPro, (2S,4S)-4-methylproline.

SpuA1HppHplaVGVWIAASGK
SpuB1HtyHtyDLAFTGCVTK
SpuB2Pro/mProPro/mProDVQFIAHAVK
SpuB3Arg/LysArgDVETTGAVTK

SpuC (265 aa) codes for a hypothetical protein with a mass of 30 744 Da homologous to an open reading frame downstream of the genes coding for proteins assembling (2S,4S)-4-methylproline in both the nostocyclopeptide and nostopeptolide gene clusters (Table 2). While no function can be assigned to this hypothetical protein its conserved position in the nostocyclopeptide, nostopeptolide and spu gene clusters argues for a role in the biosynthesis of (2S,4S)-4-methylproline. SpuD (368 aa) encodes a protein with a mass of 39 898 Da while SpuE (274 aa) codes for a protein with a mass of 28 677 Da. The SpuD and SpuE proteins are homologous to NosE and NosF proteins from the nostopeptolide gene cluster (Table 2). We postulate that three stand-alone proteins SpuC, SpuD and SpuE function in the assembly line through the biosynthesis of the rare non-proteinogenic amino acid (2S,4S)-4-methylproline (Table 2). The SpuF (683 aa) encodes a protein with a mass of 79 085 Da and exhibits high levels of similarity to members of the ATP-binding cassette (ABC) transporter superfamily (Table 2). This protein may be associated with the translocation of spumigins.

ATP–pyrophosphate exchange

Nodularia spumigena is recalcitrant to genetic manipulation and in order to verify that the 21 kb gene cluster encodes the spumigin peptide synthetase we undertook biochemical analysis of A domain substrate specificity. Four A domains encoded in the spumigin gene cluster were overexpressed. An ATP–pyrophosphate (ATP–PPi) exchange reaction was carried out to determine the activity of enzymes forming enzyme-adenylate intermediates in the course of their catalytic action of the A domains from SpuA1 and SpuB2. The results of the ATP–PPi exchange assay are consistent with the proposed catalytic scheme (Fig. 4). The ATP–PPi exchange assay clearly demonstrates a preference of the SpuA1 A domain for the activation of 4-hydroxyphenylpyruvic acid over 4-hydroxyphenyllactic acid and other substrates in vitro. The ATP–PPi exchange assay results suggest that the SpuB2 A domain preferentially activates l-Pro over the other hydrophobic amino acids tested in vitro (Fig. 4). The expression of SpuB1 and SpuB3 A domains resulted in insoluble proteins (data not shown).

Figure 4.

Adenylation domain activities of (A) SpuA1 and (B) SpuB2 as determined by the ATP–pyrophosphate exchange assay. HPL, 4-hydroxyphenyllactic acid; HPP, 4-hydroxyphenylpyruvate; COU, p-coumarate; CIN, cinnamate; Hty, homotyrosine.

Distribution of the spu gene cluster

We used oligonucleotide primers specific for spuA, spuB and spuF to screen 24 strains of N. spumigena, N. sphaerocarpa and N. harveyana by PCR for the presence of the spu gene cluster (Table 4). We identified products from the spumigin biosynthetic gene cluster in all strains of N. spumigena included in the analysis. These include 14 strains isolated in different years from sampling stations throughout the Baltic Sea and one strain originating from Orielton Lagoon in Tasmania, Australia. A strain of N. harveyana (Hübel 1983/300) had spu genes and was found to produce a spumigin (spumigin I, Hpla–Leu–Pro–Argol, spectral data in Supporting information) which differed from the spumigins produced by strains of N. spumigena. PCR products of the expected length were amplified from N. sphaerocarpa strains PCC7804 and PCC73104/1 using primers specific for the spuB gene. These were sequenced to explore the possibility of a non-functional spumigin gene cluster in these strains. However, blast results of these sequences indicated homology to other non-ribosomal peptide synthetase gene clusters from cyanobacteria (data not shown). LC-MS analysis confirmed that the strains containing spuA, spuB and spuF genes produced spumigins (Table 4). We also identified nodularins and nodulapeptins in all of the N. spumigena strains in addition to spumigins (Table 4). Suomilide was detected only in N. sphaerocarpa strains (Table 4).

Table 4.  The strain number, year of isolation and origin of the planktonic Nodularia spumigena and benthic N. sphaerocarpa and N. harveyana strains used in this study.
SpeciesStrainaYearOriginspu gene clusterPeptides
spuAspuBspuFSpumiginsNodularinNodulapeptinsSuomilide
  • a. 

    Strain codes of N. spumigena strains refer to Baltic Sea sampling stations shown in a previous publication (Sivonen et al., 1989).

  • The presence of spumigin biosynthetic genes was determined by PCR (○). The presence of spumigins, nodularin, nodulapeptins and suomilide was determined by LC-MS (●).

N. spumigenaBY11986Baltic Sea
N. spumigenaP381986Baltic Sea
N. spumigenaAV11987Baltic Sea
N. spumigenaAV31987Baltic Sea
N. spumigenaAV631987Baltic Sea
N. spumigenaF811987Baltic Sea
N. spumigenaHEM1987Baltic Sea
N. spumigena55/151987Baltic Sea
N. spumigenaTEILI1987Baltic Sea
N. spumigenaCCY94141988Baltic Sea
N. spumigenaGR8a1992Baltic Sea
N. spumigenaGR8b1992Baltic Sea
N. spumigenaGDR1131992Baltic Sea
N. spumigenaTR1831993Baltic Sea
N. spumigenaNSOR-121993Australia
N. sphaerocarpaHKVV1986Baltic Sea
N. sphaerocarpaUP16a1994Baltic Sea
N. sphaerocarpaUP16f1994Baltic Sea
N. sphaerocarpaPCC73104/11972Canada
N. sphaerocarpaPCC78041966France 
N. harveyanaHübel 1983/3001983Baltic Sea
N. harveyanaBo531992Baltic Sea
N. harveyanaBECID 272001Baltic Sea
N. harveyanaBECID 292001Baltic Sea

Trypsin inhibition

The bulk of the spumigins identified were peptide aldehydes and highly reactive. We purified 200 μg of spumigin E variant from 200 mg of dried N. spumigena AV1 biomass in order to test this compound for protease inhibition. The final compound was 80–95% pure. The spumigin E variant inhibited trypsin at the nanomolar level in a time-dependent manner (Fig. 5). The time-dependent inhibition of trypsin prevented the estimation of an IC50 value for the compound.

Figure 5.

The progressive inhibition of trypsin with increasing concentrations of the spumigin E variant measured at 25°C. There was no inhibition in the control treatment. Standard deviation was measured with four replicates.

Phylogenetic analysis

Spumigins and aeruginosins are both tetrapeptides which similar chemical structures. We reconstructed the evolutionary history of the spumigin and aeruginosin gene clusters using the conserved adenylation and ketoreductase of the loading module and the epimerase domain of second module as a proxy (Fig. 6). There was strong support for the non-monophyly of the adenylation, ketoreductase and epimerase domains from each gene cluster despite the structural similarities between spumigin and aeruginosins (Fig. 6).

Figure 6.

Distance trees showing the non-monophyly of conserved catalytic domains from the spumigin and aeruginosin biosynthetic gene clusters.
A. A neighbor-joining tree based on a concatenation of adenylation and ketoreductase domains from spumigin and aeruginosin biosynthetic gene clusters.
B. A neighbor-joining tree based on the epimerase domains from spumigin and aeruginosin biosynthetic gene clusters.
Bootstrap values above 50% from 1000 bootstrap replicates are given at the nodes. Branch lengths are proportional to sequence change.

Discussion

Here we report the discovery and biochemical analysis of the spumigin biosynthetic gene cluster. The predicted substrate specificities of the SpuA and SpuB adenylation domains, the presence of a C-terminal reductase and genes encoding proteins catalysing the terminal stages in the biosynthesis of the non-proteinogenic amino acid (2S,4S)-4-methylproline, all support the involvement of this cluster in spumigin biosynthesis. In contrast to previous work (Fujii et al., 1997a) our analysis demonstrates that the majority of the spumigins possess a C-terminal aldehyde derivative of arginine.

The SpuA protein contains an A domain at the N-terminus indicating that it is the loading module. The α-amino group-stabilizing Asp, which is found in amino acid-activating adenylation domains, is replaced by Val in the amino acids predicted to line the binding pocket of the SpuA1 A domain (Table 2). This replacement is common in hydroxy amino acid-activating adenylation domains (Fujimori et al., 2007). The SpuA1 A domain bears homology to the A domains of AerA, BarE, HctF, McyG and NdaC all of which are involved in the activation of non-amino acid substrates in the biosynthesis of aeruginoside (Ishida et al., 2007), barbamide (Chang et al., 2002), hectochlorin (Ramaswamy et al., 2007), microcystins (Tillett et al., 2000) and nodularin (Moffitt and Neilan, 2004). The predicted substrate specificity of the SpuA1 A domain is similar to that from AerA which is proposed to activate phenylpyruvate during the biosynthesis of aeruginoside (Ishida et al., 2007). The presence of a KR domain between the A and PCP domain of the SpuA loading module (Fig. 2) is atypical but has been reported from the cereulide (Ehling-Schulz et al., 2005), valinomycin (Cheng, 2006), cryptophycin (Magarvey et al., 2006), heterochlorin (Ramaswamy et al., 2007), aeruginoside (Ishida et al., 2007), and aeruginosin (Ishida et al., 2009) biosynthetic gene clusters. We postulated that activation of 4-hydroxyphenylpyruvic acid followed by the reduction of the C2 carbonyl group to a hydroxyl group would lead to the incorporation of 4-hydroxyphenyllactic acid into spumigin. 4-Hydroxyphenylpyruvic acid is an intermediate in the biosynthesis of l-tyrosine and could be supplied to the peptide synthetase by oxidative decarboxylation of prephenate by the shikimate pathway. In order to test this hypothesis we expressed the SpuA1 A domain in vitro and tested for substrate specificity. The results of the ATP–PPi exchange assay indicated a preferential activation of hydroxyphenylpyruvic acid suggests that the KR domain reduces C2 carbonyl group to a hydroxyl group during the biosynthesis of spumigin (Fig. 3). Interestingly, the SpuA1 A domain also activates 4-hydroxyphenyllactic acid suggesting that 4-hydroxyphenyllactic acid may be incorporated directly into spumigin without the action of the KR domain. The substrate specificity of McyG analysed by Hicks et al. (2006) differs from that determined for SpuA1 despite the sequence similarity between the adenylation domains of McyG and SpuA1 (48/64% identity/similarity in the expressed part of SpuA1).

We postulated that the first module of SpuB incorporates d-Hty as the second residue in the growing chain based on the presence of the E domain in the first module of SpuB and the predicted l-Hty substrate specificity (Table 2). The organization of catalytic domains suggests that l-Hty is activated by the A domain bound to the PCP domain and racemized to the d-Hty by the E domain in SpuB. However, expression of the SpuB1 A domain resulted in insoluble proteins and prevented an explicit test of this hypothesis by the ATP–PPi exchange assay. Spumigins containing homophenylalanine instead of homotyrosine were also produced in trace amounts. Homo amino acids contain a methylene (CH2) group on the α-carbon and appear to be common in cyanobacterial non-ribosomal peptides (Welker and von Döhren, 2006). Homotyrosine is common in a number of cyanobacterial secondary metabolites including members of the spumigins, lyngbyastatin, microcystins, cyanopeptolins and anabaenopeptins (Welker and von Döhren, 2006). However, homotyrosine is not part of typical tyrosine metabolism and the metabolic origin of homotyrosine in cyanobacteria is unclear. Homotyrosine is produced through condensation of tyrosine and acetate with loss of the carboxyl group of tyrosine during the biosynthesis of the l-671 329 antibiotic by the fungus Zalerion arboricola (Adefarati et al., 1991). Homotyrosine is present in both nodulapeptin and the spumigins produced by N. spumigena AV1 (Fujii et al., 1997a) and the genes for the biosynthesis of homotyrosine are likely to be encoded in a different part of the genome.

The second module of SpuB contains a C, A and PCP domain with a predicted substrate specificity of (2S,4S)-4-methylproline or l-proline (Table 3). This unusual non-proteinogenic amino acid is present in cyclic non-ribosomal peptides synthesized by Nostoc spp. GSV224, ATCC53789 and PCC73102 (Golakoti et al., 2000; Hoffmann et al., 2003; Becker et al., 2004; Hunsucker et al., 2004). Approximately 80–90% of the spumigin variants produced by N. spumigena AV1 and CCY9414 contain (2S,4S)-4-methylproline while the remaining spumigins had l-proline (Table 1). The results of the ATP–PPi exchange suggest that the SpuB2 A domain preferentially activates l-proline over the other hydrophobic amino acids tested (Fig. 4). SpuC, SpuD and SpuE code for proteins responsible for synthesizing (2S,4S)-4-methylproline in both the nostocyclopeptide and nostopeptolide gene clusters (Hoffmann et al., 2003; Luesch et al., 2003; Becker et al., 2004). The terminal steps in the biosynthesis of the rare non-proteinogenic amino acid (2S,4S)-4-methylproline have been shown to be catalysed by NosE and NosF in the nostopeptolide gene cluster (Luesch et al., 2003). We postulate that three stand-alone proteins SpuC, SpuD and SpuE also function in the assembly line through the biosynthesis of the rare non-proteinogenic amino acid (2S,4S)-4-methylproline (Table 2).

The third module of SpuB contains a C, A and PCP domain with a predicted substrate specificity of l-Arg (Table 3). This module also contains a C-terminal reductase (R) domain. The R domain containing a putative NAD(P)H binding site was identified at the C-terminal end of SpuB in place of the more common thioesterase domain. Thioesterase domains are typically responsible for the hydrolytic cleavage of the covalently bound peptide from the NRPS (Kohli et al., 2001). The presence of a C-terminal R domain suggests the reductive offloading of the PCP-tethered tetrapeptide as a linear aldehyde (Kopp et al., 2006) as found in spumigins E–H (Table 1). The aldehyde may be further reduced to an alcohol and spumigins A and D contain argininol residue as previously reported (Fujii et al., 1997a). Spumigins containing argininol comprised approximately 20% of the spumigins produced by N. spumigena AV1 and CCY9414. However, the exact mechanism for the second reduction step is not clear. Either it is spontaneous or the R domain itself is responsible for the second reduction step. Trace amounts of spumigins B1 and B2 containing the acid form of arginine were found (Table 1). In all the N. spumigena strains studied acidic arginine were absent or present in trace amounts. This could be the result of release of the PCP-tethered peptide by a thioesterase type II through hydrolytic cleavage prior to reductive release taking place. Type II thioesterases, whose gene is associated with the gene cluster of many NRPSs, improves the efficiency of product formation in these systems and has been proposed to edit modules through hydrolysis of acyl groups preventing stalling in NRPS assembly (Schwarzer et al., 2002). Alternatively, the acid form of arginine could be an artefact of the extraction procedure.

Aeruginosins are a group of low-molecular-weight linear tetrapeptide metabolites isolated from various cyanobacterial genera and from marine sponges that inhibit different types of serine proteases (Ishida et al., 1999). They are characterized by hydroxyphenyllactic acid at the N-terminus, the amino acid 2-carboxy-6-hydroxyoctahydroindole (Choi) and an arginine derivative at the C-terminus (Welker and von Döhren, 2006). The N-terminal hydroxyphenyllactic acid is commonly chlorinated, brominated or sulphated (Welker and von Döhren, 2006). Spumigins are considered to be members of the aeruginosin structural class of peptides (Ishida et al., 1999; Welker and von Döhren, 2006). However, spumigins differ from aeruginosins by the presence of (2S,4S)-4-methylproline in place of the 2-carboxy-6-hydroxyoctahydroindole (choi) residue (Welker and von Döhren, 2006). Despite similarity in chemical structure blastp searches show low similarity between the non-ribosomal peptide synthetase proteins involved in the assembly of aeruginosins and spumigins (Table 2). We reconstructed the evolutionary history of the two gene clusters using the conserved catalytic domains present in both clusters in order to explore this finding further (Fig. 6). We found strong evidence for the non-monophyly of the loading adenylation domain, the ketoreductase domain and the epimerase domains from each gene cluster (Fig. 6). Recombination, gene loss, gene conversion and horizontal gene transfer have been already described from cyanobacterial non-ribosomal gene clusters (Fewer et al., 2007; Cadel-Six et al., 2008; Rounge et al., 2008 Ishida et al., 2009). However, the low sequence homology between spumigin and aeruginosin gene clusters suggests that an alternative mechanism involving the shuffling of non-ribosomal peptide synthetase gene clusters may have led to convergence in the chemical structures of the two tetrapeptides. More detailed phylogenetic analysis are necessary to resolve this issue.

Our results demonstrate that approximately 80% of the spumigins produced by N. spumigena AV1 and CCY9414 are released as aldehydes (Table 1). Leupeptin is biosynthetically produced in arginine and then enzymatically reduced to argininal (Kim and Lee, 1995). LC-MS analysis showed that the commercially obtained leupeptin used in this study contained no arginine variant but about 10% argininol variant relative to argininal variant. This information further supports the claim that spumigins containing argininal are the biochemical products and argininol variants are formed at least partly via spontaneous reduction. The ratio between the amounts of arginine and argininol containing spumigins was 80:20 (calculated from the data of Fujii et al., 1997a) as it was in our study between argininal and argininol containing spumigins. Oxidation of argininal to arginine during the purification of spumigins may explain why argininal containing variants were not previously described from the AV1 strain (Fujii et al., 1997a).

Serine proteases regulate important biological processes and are attractive targets in therapy (Radau, 2005). Serine protease inhibitors are important in the treatment of a wide variety of human diseases (Rubin, 1996; Leung et al., 2000; Groll et al., 2009). Spumigins are reported to be serine protease inhibitors (Fujii et al., 1997a) and could represent candidates for drug development. Spumigin A, which contains a alcohol derivative of arginine, is reported to inhibit thrombin, plasmin and trypsin in the micromolar range (Fujii et al., 1997a). Spumigin B1, which contains arginine, also inhibits trypsin in the micromolar level (Fujii et al., 1997a). The C-terminal arginine derivatives are reported to play a critical role in specific serine protease inhibition by the structurally related aeruginosins (Ishida et al., 1999). In contrast to earlier findings we show that the main spumigin variant produced by N. spumigena also contains an aldehyde group. The spumigin E variant is a peptide aldehyde and inhibited trypsin at the nanomolar level (Fig. 5). We also detected a time-dependent inhibition of trypsin by spumigin E (Fig. 5). Together, these results demonstrate that spumigins are potent serine protease inhibitors. This time dependence suggests that there may be irreversible inhibition of trypsin by the spumigin E variant. Our results suggest that spumigins are even more potent protease inhibitors than previously reported.

Nodularia spumigena is a nitrogen-fixing genus that forms surface blooms in brackish water bodies, estuaries and saline lakes in Europe, Australia, New Zealand, North America and South Africa (Galat et al., 1990; Sivonen and Börner, 2008; Sivonen, 2009). Cyanobacterial blooms in the Baltic Sea are dominated by the filamentous cyanobacterium N. spumigena (Sivonen et al., 1989). These blooms are among the largest in the world and invariably toxic through the production of the pentapeptide eukaryotic protein phosphatase inhibitor nodularin (Sivonen et al., 1989). N. spumigena has a wide vertical distribution of the water column from the top down to 30 m below the surface (Koskenniemi et al., 2007). There is a growing body of evidence that cyanobacteria in the Baltic Sea are important food for grazers (Karjalainen et al., 2007; Ibelings and Havens, 2008). We found spumigins, nodularins and nodulapeptins in all N. spumigena strains examined (Table 4). We found the serine protease inhibitor suomilide (Fujii et al., 1997b) in four out of five N. sphaerocarpa examined (Table 4). The distribution of peptides among the planktonic N. spumigena and benthic Nodularia species (Table 4) goes well together with the identified taxonomic differences between these species (Lyra et al., 2005). Spumigins are commonly produced by planktonic N. spumigena and furthermore are potent serine protease inhibitors. Planktonic bloom-forming cyanobacteria produce a range of protease inhibitors (Welker and von Döhren, 2006; Sivonen and Börner, 2008). The fatal disruption of molting in Daphnia is attributed to microviridin J and it is suggested that other cyanobacterial protease inhibitors should be considered potentially toxic to zooplankton (Rohrlack et al., 2003; 2004). The amounts of spumigins produced by N. spumigena strains examined are quite high. This suggests that the production of spumigin is important to N. spumigena and that they may act as a feeding deterrent in nature. Our results demonstrate that bloom-forming N. spumigena produces a broad spectrum of enzyme inhibiting compounds. The role of these bioactive compounds in brackish water ecosystems remains to be determined.

Experimental procedures

Strain cultivation

Twenty-four strains of N. spumigena, N. harveyana and N. sphaerocarpa (Table 4) were grown at a photon irradiance of 15 μmol m−2 s−1 in modified Z8 medium for 21 days (Lyra et al., 2005). N. spumigena CCY9414 was grown at 20°C in medium free of combined nitrogen, comprising one part ASN3 and two parts BG 11 giving a salinity of 11 (Rippka et al., 1979). Cells were harvested by filtration with 10 μm polycarbonate filters (Poretics, Osmonics).

Liquid chromatography-mass spectrometry

From 7 to 31 mg of freeze-dried cells of Nodularia strains were extracted for 20 s with 1 ml of ACN : DMSO (3:1) or methanol in 2 ml plastic tubes containing approximately 200 μl of 500 μm glass beads (Scientific Industries, New York) using Fast Prep homogenizer (FP120, Bio 101, Savant) at speed value of 6. Extracts were centrifuged for 5 min at 10 000 g prior to LC-MS analysis. DNPH derivatives were prepared from strains AV1 and CCY9414 using 10 μl of reagent [clear solution from mixing of 20 mg of moist 2,4-dinitrophenylhydrazine (Fluka, Switzerland) + 1 ml of methanol + 10 μl of concentrated HCl for 15 s] which was then added to 90 μl of methanol extract. Reaction mixture was incubated at 37°C for 60 min prior to LC-MS analysis. We used spumigins B1, B2 and C (Fujii et al., 1997a) as reference standards. The tetrapeptide leupeptin (Sigma-Aldrich) contains a C-terminal argininal and was used as reference compound in DNPH derivatization.

LC-MS analyses of extracts were performed with an Agilent 1100 Series LC/MSD Ion Trap XCT Plus System (Agilent Technologies, Palo Alto, CA, USA) using a Phenomenex Luna C8 (150 × 2.0 mm, 5 μm, Phenomenex, Torrance, CA, USA) LC-column. The mobile phase was composed of 0.05% trifluoroacetic acid (or 0.1% formic acid) in water (A) and 2-propanol (B). The gradient run was from 5–100% B for underivatized spumigins and 5–50% B for DNPH-derivatized spumigins over 35 min at a flow rate of 0.15 ml min−1 at 30°C using 1 μl injections. Peptides were detected with a diode array detector and with MS using electrospray ionization (ESI) set in positive mode. The nebulizer gas (N2) pressure was 30–35 psi (207–240 kPa), desolvation gas flow rate 8 l min−1 and the desolvation temperature 350°C. The structures of spumigins were confirmed by fragmentation using multiple reaction monitoring (MRM) mode. The capillary voltage (CV) was set to 3200 V, the capillary exit offset was 300 V, the skimmer potential (SP) was 42.0 V and the trap drive (TD) value was 85.0. Spectra were recorded at a scanning range (SR) of 50–1200 m/z at a rate of 26 000 m/z s−1. The MS2 fragmentation amplitude was 0.65 V and MS3 fragmentation amplitude was 0.50 V. Other peptides were identified in auto MS2 mode with the following parameters: CV 5000 V; SP 85 V; TD 144; SR 100–2200 m/z; and MS2 0.35 V. Identification of nodularins, nodulapeptins, suomilide and spumigins from Nodularia strains other than AV1 and CCY9414 was based on the retention times, identification of the protonated molecular ions and product ion spectra of the protonated molecular ions.

Identification of the spumigin gene cluster

We identified a 21 kb NRPS gene cluster on a single scaffold of the partial genome N. spumigena CCY9414 (AAVW00000000) through reverse-position-specific blast searches. This cluster contained two open reading frames encoding four NRPS modules located downstream of genes similar to genes encoding proteins involved in the biosynthesis of (2S,4S)-4-methylproline in the nostocyclopeptide and nostopeptolide synthetase gene cluster (Hoffmann et al., 2003; Luesch et al., 2003; Becker et al., 2004). The 21 kb gene cluster was located on a single scaffold and spread over four contigs in the partial genome of N. spumigena CCY9414. We undertook biochemical and bioinformatic studies to confirm this initial prediction.

Spumigin gene cluster assembly

We extracted high-molecular-weight genomic DNA from N. spumigena CCY9414 using enzymatic lysis of cyanobacterial cells with Proteinase K and Lysozyme followed by phenol:chloroform separation (Golden et al., 1988). We designed oligonucleotide primers in order to confirm the assembly and close the gaps between the three contigs (Supporting information). We amplified the three missing regions of the putative spu gene cluster from genomic DNA of N. spumigena CCY9414 by PCR. The PCR reactions were performed in a 40 μl final volume containing 1 μl of DNA, 1× DyNAzyme EXT PCR buffer, 250 μm of each deoxynucleotide, 0.5 μm of each oligonucleotide primers, and 0.5 unit of DyNAzyme EXT DNA polymerase (Finnzymes, Espoo, Finland). The following protocol was used: 95°C, 3 min; 30 cycles of 94°C, 30 s; 56°C, 30 s; 72°C, 1 min; and 72°C, 10 min. Sequence data were obtained for both strands using the BigDye Terminator Cycle Sequencing kit (Applied Biosystems) and analysed on the ABI 310 Genetic Analyser.

Amplification and cloning of NRPS fragments for overexpression

Nodularia spumigena is recalcitrant to genetic manipulation. In order to prove that the spu gene cluster was responsible for spumigin biosynthesis we cloned and expressed adenylation domains and tested for substrate specificity. The four adenylation domains of spuA and spuB were PCR amplified from the putative spumigin synthetase of N. spumigena CCY9414 using oligonucleotide primers designed for the substrate conferring portion of each adenylation domain (Supporting information). Priming sites were chosen to allow expression of a protein with a N-terminus approximately 280 aa in front of motif 1 (LKAGGA) and the C-terminus approximately 300 aa behind motif 5 (RIELGEIE). These two core motifs are highly conserved among peptide synthetases (Mootz and Marahiel, 1997). The PCR reactions were performed in a volume of 50 μl consisting of Pfu buffer [2 mM, pH = 8.8, 10 mM (NH4)2SO4, 10 mM KCl, 0.1% Triton X-100, 0.1 mg ml−1 BSA and 2 mM MgSO4, Fermentas], 0.5 mM primers, 0.2 mM dNTP, 1.25 U of Pfu DNA polymerase (Fermentas) and 20–50 ng of genomic DNA. The following protocol was used: 1 × (95°C, 2 min; 57°C, 40 s; 70°C, 2 min); followed by 25 × (94°C, 30 s; 57°C, 30 s; 70°C, 2 min); and 72°C, 5 min. PCR products were gel excised and purified with the Quick-Step 2 Purification kit (Edge BioSystem, Gaithersburg, MD, USA) and cloned in the pET101/D-TOPO expression vector (Invitrogen) according to the manufacturer's instructions. Plasmids were amplified in the Escherichia coli TOP10 cloning strain and purified for cycle sequencing. The resulting clones were end-sequenced using the BigDye Terminator Cycle Sequencing kit (Applied Biosystems) and analysed on the ABI 310 Genetic Analyser in order to verify the end sequences and correct orientation of the inserts.

Overexpression and purification of HIS6-tagged proteins

The plasmid clones were used to transform the Ecoli BL21 Star (DE3) expression host (Invitrogen). Transformants were grown in LB medium supplemented with 100 μg ml−1 ampicillin and grown at 37°C overnight with shaking at 160 r.p.m. and 400 μl was used to inoculate 20 ml of 2× YT medium containing 50 μg ml−1 carbenicillin. Cultures were then grown at 37°C by shaking at 160 r.p.m. until they reached an OD600 of 0.7–1.0 and then induced by the addition of 1 mM IPTG. Shaking was continued at 23–24°C overnight for 18–20 h at 100 r.p.m. Cells were collected by centrifugation at 3000 g for 5 min, re-suspended in 1.5 ml of lysis buffer containing 50 mM sodium phosphate, 300 mM NaCl, 100 mM KCl, 10 mM imidazole, 0.5% Triton X-100, 1 mM β-mercaptoethanol, 1 mM PMSF and Complete Mini protease inhibitor cocktail (Roche) and placed on ice. The cells were sonicated in ice (Braun Labsonic U sonicator), power level 50%, at 0.7 repeating duty cycles 4 × 0.5 and 3 × 10 cycles, respectively, and the soluble proteins were recovered by centrifugation at 12 000 g for 20 min. Expression level and protein purity were tested with the SDS-PAGE runs. The soluble adenylation domains were purified with the His affinity tag method using Ni-NTA agarose (Qiagen). Protein concentration of the preparations was measured with the BCA protein assay kit (Pierce). The SpuA1 A domain and the three A domains of SpuB (B1, B2 and B3) were overexpressed in E. coli, but only SpuA1 and SpuB2 adenylation domains were obtained in soluble form and could be used in the substrate specificity assay.

ATP–pyrophosphate exchange assay

The substrate specificity of the expressed adenylation domains was determined using the ATP–PPi exchange reaction as described previously (Gruenewald et al., 2004). The assay buffer was modified to 75 mM Tris–HCl, pH 7.3, 10 mM MgCl2, 10 mM ATP, 5 mM DTT, 1 mM substrate amino acid and 80–90 μg of enzyme. The reaction was initiated by addition of 2 mM ATP, 0.2 mM tetrasodium pyrophosphate, and 0.15 μCi of tetrasodium [32P]pyrophosphate (Perkin-Elmer Life Science, Boston, USA). The reaction was allowed to proceed for 15 min at 37°C and quenched as previously described (Gruenewald et al., 2004). All reactions were performed in duplicates and the charcoal-bound radioactivity was measured by liquid scintillation counting in a Wallac 1411 Liquid Scintillation Counter after addition of 3 ml of liquid scintillation fluid (Optiphase Hisafe3, Perkin Elmer).

Distribution of the spumigin gene cluster

Genomic DNA was extracted from the cultivated Nodularia cells as previously described (Koskenniemi et al., 2007). We PCR amplified three genes from the spu gene cluster using oligonucleotide primers specific for the KR domain of spuA, the third A domain of spuB and the spuF ABC transporter (Supporting information). The PCR reactions were performed in a 20 μl final volume containing 1 μl of DNA, 1× DyNAzyme II PCR buffer, 100 μm of each deoxynucleotide, 0.4 μm of both PCR primers, and 0.4 units of DyNAzyme II DNA polymerase (Finnzymes, Espoo, Finland). The following protocol was used: 95°C, 5 min; 35 cycles of 94°C, 30 s; 57–60°C, 45 s; 72°C, 1 min; and 72°C, 10 min. PCR products were visualized on 1.5% agarose gels containing 0.5× TAE run at 120 V for 20–25 min and scored for the presence or absence of PCR products of the expected length. Cultures were assayed for the production of spumigins as earlier described.

Purification of spumigin E

Two batches of freeze-dried N. spumigena AV1 cells (100 mg) were extracted for 20 s with a speed of 6.0 m s−1 in 1 ml of methanol/dichloromethane (1:1 v/v) and 0.6 g of glass beads in FastPrep homogenizer (FP120; Bio101 Savant, Thermo Electron Corporation, Mitford, MA). After centrifugation (10 000 g, 5 min) 300 μl of water was added to 600 μl of supernatant and the solution was mixed in FastPrep homogenizer with previous settings. The upper phase was collected after centrifugation (10 000 g, 5 min) and dried with a vacuum centrifuge (Heto vacuum centrifuge; Heto-Holten A/S, Allerød, Denmark). Dried residues were dissolved in 1 ml of methanol/dichloromethane (1:1 v/v) and after pooling, solutions were passed through Sep-Pak® Vac 12cc (2 g) Silica cartridge first preconditioned and then eluted with the same solvent. The first 10 ml of effluent was collected and vacuum dried.

Spumigin E was further purified with HPLC (Agilent 1100 Series LC/MSD Ion Trap XCT Plus System, Agilent Technologies, Palo Alto, CA, USA). Astec CHIROBIOTIC™ T (15 cm × 2.1 mm, 5 μm, Supelco) column was eluted isocratically 0.15 ml min−1 with 30% of 0.1% ammonium formate (A) and 70% of acetonitrile (B). The dried residue was dissolved in 0.3 ml of eluent and was then injected in 100 μl batches into the column. Pooled spumigin E fractions were vacuum dried. Then TSKgel Amide-80 (25 cm × 2.0 mm, 5 μm, Tosoh Bioscience) column was eluted isocratically 0.2 ml min−1 with 12% of solution A and 88% of solution B. Partially purified spumigin E was dissolved in 0.3 ml of eluent and was then injected in 100 μl batches into the TSKgel column. The pooled fractions were vacuum dried, the residue mass was measured with micro balance and purified spumigin E was dissolved in water (1 mg ml−1). Spumigin E purity was 80–95%, based on spumigin-specific mass signals versus total ion signal or UV210 signals respectively.

Protease inhibition assay

Trypsin (Sigma-Aldrich, Type IX-S, Porcine Pancreas) activity was measured on a well plate at 25°C in a reaction mixture containing 50 mM Tris/HCl buffer, pH 8.0, 0.15 M NaCl, 1 mM CaCl2, 0.1 mg ml−1 of bovine serum albumin and 0–5.5 μm inhibitor (spumigin E). Two microlitres of substrate Boc–Gln–Ala–Arg–MCA (Peptide Institute, Japan) in DMSO (100 μm in assay) was added to 193 μl of reaction mixture. The reaction was initiated with 5 μl of trypsin solution and hydrolysis was followed by measuring the fluorescence of the product 7-amino-4-methylcoumarin (MCA) with Infinite M200 spectrofluorometer (Tecan Austria GmbH) with excitation at 380 nm and emission at 460 nm.

Phylogenetic analysis

In order to investigate the evolutionary relationships between the spumigin and aeruginosin gene we constructed an amino acid sequence alignment of the conserved epimerase domain from both gene clusters in bioedit. We obtained a selection of adenylation, keotreductase and epimerase domains from the ceruelide, cryptophycin, cyanopeptolin, hectochlorin, microcystin and nostocyclopeptide gene clusters as well as uncharacterized NRPS gene clusters from the complete genomes obtained from NCBI and aligned them against the corresponding domains from the spumigin and aeruginosin gene clusters. The adenylation and ketoreductase domains were concatenated and ambiguous regions excluded and 694 aa were considered for phylogenetic analysis. Regions of ambiguous alignment were excluded and we considered 429 aa of the epimerase domain for phylogenetic analyses. Protein neighbour joining phylogenies were inferred using prodist implemented in the phylip 3.6 package (Felsenstein, 1993) with a Dayhoff-pam substitution model and a gamma distribution of rates. Ten random additions with global rearrangements were used to find the optimal tree. We performed 1000 distance bootstrap replicates using the seqboot, protdist, neighbor and consense programs of the phylip 3.6 package (Felsenstein, 1993).

Acknowledgements

We thank Lyudmila Saari for her valuable help in handling the cultures. This work was supported by grants from the European Union PEPCY (QLK4-CT-2002-02634) and grants from the Academy of Finland to D.P.F. (1212943) and K.S. (53305, 118637 and 214457). This is publication 4587 of NIOO-KNAW.

Ancillary