Echinocandins are fungal non-ribosomal cyclic hexapeptides with a fatty acid side chain attached to a dihydroxyornithine residue. As they are specific noncompetitive inhibitors of the β-1,3-glucan synthase involved in fungal cell wall biosynthesis, they have a pronounced antifungal bioactivity. Although natural echinocandins are not of clinical use due to their toxicity and low solubility, chemical derivatives such as caspofungin, anidulafungin, and micafungin are most important drugs for the treatment of invasive mycoses. The pharmacological use of echinocandins has been reviewed in several overview articles , , , . In Table 1, some properties of echinocandins applied in therapy are summarized.
Besides the pharmacological properties of the echinocandins, their extraordinary structure is very interesting for biosynthetic studies (Figure 1). Thus, most amino acids in echinocandins are non-proteinogenic, the ring is closed via an unusual N-acyl-hemiacetal, and in some biosynthetic clusters, a polyketide synthase (PKS) for the synthesis of a branched chain fatty acid side chain is included.
The history of exploring echinocandin biosynthesis has been featured in previous review articles , , , , . In brief, the biosynthetic origin of the non-proteinogenic amino acids and the dimethylmyristate side chain was elucidated in the early 1990s by 13C-labeling experiments with the pneumocandin producer G. lozoyensis , . In 2003, an α-ketoglutarate (αKG)-dependent l-proline hydroxylase (PH) activity, which was thought to be involved in pneumocandin biosynthesis, was found in crude protein extracts of G. lozoyensis . However, nothing was known about the genes encoding the biosynthesis of the pharmaceutically important compounds until 2012 when the groups of Tang and Walsh disclosed the genes encoding echinocandin B biosynthesis in the genome of Aspergillus pachycristatus NRRL 11440 (formerly Emericella rugulosa) . According to this study, the genes are arranged in two separate partial clusters (Ecd and Hty) located on different contigs of the assembly. However, we showed recently that Ecd and Hty can be aligned to a single cluster (Ecd/Hty) and the disjunct genomic location of the cluster fragments was likely an artifact of genome misassembly in the absence of a reference genome . It was also found that Ecd/Hty was virtually identical with a putative echinocandin B biosynthetic cluster AE deposited at the NCBI as a sequence of the producer strain Aspergillus delacroxii NRRL 3860 (formerly Emericella nidulans var. echinulatus) (cf. Table 5). Sequence analysis of PCR samples from the genomes of both strains, showed that the sequence of AE cluster is identical with that of A. pachycristatus, but not with A. delacroxii . Moreover, a comparison of the calmodulin genes, a common taxonomic marker for fungi, suggested that strain NRRL 3860 belongs to the Aspergillus pachychrystatus/Aspergillus rugulosa group rather than to A. delacroxii. However, a more detailed investigation will be necessary to clarify the species affiliation , .
The central gene of the Ecd/Hty cluster codes for a non-ribosomal peptide synthetase (NRPS) with a (T0 CAT CAT CAT CAT CAT CAT CT)-domain structure (T=thiolation, C=condensation, A=adenylation), which allows the assembly of six amino acids. It was shown that the initial thioesterase domain T0 is loaded with linoleic acid activated by the fatty acyl-AMP ligase. The first CAT module is selective for ornithine, which is initially N-acylated with the linoleic acid side chain. After assembly of the six amino acids, the peptide is cyclized by the terminal condensation domain (CT) .
Another group of enzyme genes in the Hty section of the cluster encodes the biosynthesis of homotyrosine according to a mechanism already postulated by Adefarati et al. in 1991 based on labeling experiments , . First, acetate is added to the keto group of hydroxyphenylpyruvate by an isopropylmalate synthase (IPMS) (cf. Figure 3). Then, the hydroxyl group of the product is shifted from C-3 to C-2 by an aconitase (ACN). After oxidative decarboxylation, an α-keto acid is obtained, which is finally transaminated to provide homotyrosine. In a subsequent 2013 study, Tang, Walsh, and colleagues revealed the activity of the three oxygenases in the Ecd section of Ecd/Hty, which will be discussed in the next section .
In the same year, the genome of G. lozoyensis ATCC 20868 was reported, including the biosynthesis of pneumocandin encoded in a single cluster (GL) . The most obvious difference to the echinocandin biosynthetic cluster was the presence of a PKS, which is thought to be responsible for the synthesis of the dimethylmyristate side chain. Four non-heme dioxygenases were identified. For one of them, thought to be a glutamine hydroxylase (GH), no homologue has been found in Ecd/Hty. Another dioxygenase was characterized as an αKG-dependent PH . More recently, this and three other oxygenases were deleted in G. lozoyensis and the production of pneumocandins was analysed , . As the oxidation steps are very complex in echinocandin biosynthesis and play a key role in the formation of structural diversity, they are discussed in more detail in the following section.
2 Oxidation reactions in echinocandin biosynthesis
By a combination of predicted activities for the putative proteins with experimental results, a fairly clear picture of the origin and assembly of the building blocks in echinocandin biosynthesis can be obtained (cf. Figure 3) . Nevertheless, the order of biosynthetic steps, especially of the oxidations catalyzed by up to six dioxygenases, remains enigmatic. For most of these enzymes, it is not fully clear whether they accept the free amino acid, an amino acyl or peptidyl-S-NRPS precursor, or a dehydroxyechinocandin framework as substrate.
2.1 Oxygenases in A. pachycristatus
After the discovery of the echinocandin biosynthetic cluster in A. pachycristatus, the groups of Tang and Walsh revealed the activity of the three oxygenases in the Ecd section of the Ecd/Hty cluster, a l-homotyrosine hydroxylase (hT3H), an ornithine hydroxylase (OrnH), and a l-leucine dioxygenase (LDO)-producing (S)-1-pyrroline-5-carboxylate, a precursor of trans-4-methyl-l-prolin (cf. Figure 3 and Table 4) . Deletion of the αKG-dependent LDO EcdK resulted in a breakdown of echinocandin B biosynthesis; however, a minimal activity was detected in vitro with the heterologously expressed enzyme: the pro-(S)-methyl group of leucine was first hydroxylated and then further oxidized to the aldehyde, which undergoes spontaneous condensation to the cyclic imine. It is assumed that, in vivo, the imine is reduced to trans-4-methylproline by a dehydrogenase whose gene is not included in the cluster.
Genomic deletion and in vitro experiments have consistently shown that the αKG-dependent non-heme dioxygenase EcdG is a homotyrosine 3-hydroxylase . It was found that EcdG acts upon the free amino acid, but not upon the homotyrosine residue in echinocandin D (cf. Table 6). Subsequent hydroxylation to 3,4-dihydroxyhomotyrosine was not observed. Cytochrome P450 monooxygenase EcdH (OrnH) is responsible for both hydroxyl groups at the ornithine residue in echinocandins, which was discovered by the analysis of echinocandin products produced by the corresponding deletion mutant. It should be noted that hydroxylation at C-5 of the free ornithine would result in a terminal hemiaminal, which undergoes rapid hydrolysis to the aldehyde. Hence, a cyclization of the linear echinocandin precursor via formation of an N-acyl-hemiacetal is no longer possible. Thus, it has been concluded that ornithine hydroxylation at C-5 must occur after cyclization of an echinocandin precursor and is therefore a very late step in biosynthesis.
Further information about the order of oxidation steps was deduced from deletion of ecdG and ecdH . Echinocandins produced by these strains not only lack the hydroxyl groups introduced by the corresponding gene, the hydroxylation at other positions was also incomplete. This effect was explained by a reduced activity of hydroxylases catalyzing subsequent steps because a hydroxyl group is missing in the substrate. In this way, the order of hydroxylation steps could be deduced. Proline or methylproline hydroxylation was not affected by any of the deletions, supporting the hypothesis that this hydroxylation occurs at an earlier stage of echinocandin B biosynthesis, most likely of the free amino acid. In the case of mutant ΔecdG, the missing hydroxyl group at homotyrosine C-3 strongly influences the insertion of the hydroxyl group at homotyrosine C-4 and ornithine hydroxylation at C-4 and C-5 by EcdH (Figure 2) is also affected.
Consequently, both hydroxylation events are thought to occur after homotyrosine C-3-hydroxylation. The ΔEcdG mutant produced echinocandin B derivatives with non-hydroxylated, C-5-monohydroxylated, and C-4,C-5-dihydroxylated ornithine, but there was no hydroxylation solely at C-4. From that, it was deduced that EcdH hydroxylates ornithine preferentially first at C-5, so that the hemiaminal is formed and then at C-4. In both mutants, the hydroxylation of homotyrosine at C-4 was significantly reduced indicating that this is presumably the last oxidation step of echinocandin B biosynthesis. From these results, a general scheme for echinocandin biosynthesis can be derived (Figure 3). For two biocatalytic steps, the reduction 4-methylpyrroline-5-carboxylate to 4-methylproline and the ortho-sulfation of the homotyrosine, no corresponding genes are found in the clusters. Imine reduction is most likely performed by a 1-pyrroline-5-carboxylate reductase (EC 188.8.131.52) or a proline dehydrogenase (EC 184.108.40.206). Aromatic hydroxylations and sulfations are common processes in catabolism. Clustering of the genes is apparently not necessary.
2.2 Proline hydroxylation in G. lozoyensis
As early as 2003, Petersen et al. detected an αKG-dependent PH activity in the crude protein extract of G. lozoyensis producing trans-4-hydroxyproline and also the trans-3-hydroxy isomer in substantially smaller amounts . About a decade later, our group characterized the corresponding PH (GloF)  in the genome of G. lozoyensis mutant strain ATCC 74030  (GloF is identical with GLOXY2 in the genome of wild-type G.lozoyensis ATCC 20868 , ). It is distantly related to a group of fungal pipecolic acid hydroxylases discovered only recently . In vitro experiments with the heterologously expressed and purified enzyme showed that proline is converted into trans-4-hydroxyproline and a minor amount of the trans-3-hydroxy isomer. Additionally, GloF hydroxylates trans-4-methylproline to provide (S,S)-4-methyl-3-hydroxyproline, so that all three hydroxyproline building blocks for the biosynthesis of pneumocandins A and B are provided by this enzyme . The ratio of trans-4/trans-3 hydroxyproline was about 8:1. This corresponds well with the approximately 7:1 demand for hydroxyprolines for the production of pneumocandins A0 and B0, as found for wild-type G. lozoyensis. For the selectivity of GloF to actually be correlated with the ratio of pneumocandins A0 and B0 products, two prerequisites must be met: (i) the trans-3/trans-4 selectivity of GloF should be similar or identical in vitro and in vivo; (ii) the adenylation (A)-domain in module 6 of the NRPS must accept 3-hydroxyproline very well (besides 4-methyl-3-hydroxyproline), so that it is quantitatively consumed. Although some indirect evidence for this concept can be derived from other biosynthesis studies (see Section 2.3), further experimental confirmation is required. Proline residues in small peptides were not accepted as substrates by GloF. An activity with acyl carrier protein-bound proline could not be excluded; however, the relatively high activity with the free amino acids suggests that these are the native substrates.
Whereas GloF obviously meets the demand for pneumocandin production in wild-type G. lozoyensis, this does not apply to the mutant strain. For example, for mutant strain ATCC 74030, which produces pneumocandin B0 almost exclusively, a trans-4/trans-3-hydroxyproline ratio of 1:1 would be ideal. With the PH activity of GloF, a shortage of trans-3-hydroxyproline and an overflow of the trans-4-hydroxy isomer would be expected. Feeding experiments with hydroxyprolines suggest that this is indeed the case , . Addition of trans-3-hydroxyproline (0.13 M) to a culture of G. lozoyensis ATCC 74030 resulted in a substantial increase in pneumocandin B0 production (+39% relative to the non-supplemented culture, Figure 4). The concentrations of byproducts with proline derivatives other than trans-3-hydroxyproline (pneumocandins C0, D0, and E0) were drastically reduced. In contrast, feeding of trans-4-hydroxyproline provided only a relatively small increase in pneumocandin B0 production (+9%). As expected, the concentrations of pneumocandins C0 and D0 were strongly increased. The feeding of proline itself effected a strongly increased intracellular concentration of this amino acid, as proven by a 349% increase in pneumocandin E0 in the product mixture. Moreover, the production of hydroxyprolines should be enhanced, too. According to the intrinsic selectivity of PH GloF (trans-4/trans-3=8:1), a large quantity of trans-4- and a minor amount of trans-3-hydroxyproline should be produced. However, the metabolic spectrum of the feeding experiment was primarily affected by an increased supply of trans-3-hydroxyproline rather than of the trans-4-hydroxy isomer. The concentrations of pneumocandins C0 and D0 were less than half of those from cultures without proline feeding, even though a substantially increased intracellular concentration of trans-4-hydroxyproline can be assumed. The pronounced effect of trans-3-hydroxyproline on the production of pneumocandins is a strong indicator that this compound is, as predicted, a limiting factor for pneumocandin B0 biosynthesis in G. lozoyensis ATCC 74030.
2.3 Other oxygenases involved in pneumocandin biosynthesis
With the protein sequences of the oxygenases from echinocandin B biosynthesis as templates, it was easy to predict the functions of the other oxygenases in the pneumocandin biosynthetic cluster, as shown in Table 2. An additional gene for an αKG/Fe(II)-dependent dioxygenase was found in pneumocandin biosynthesis, which has no homologue in echinocandin B biosynthesis. As residue 6 in the pneumocandins, but not in echinocandin B, is hydroxyglutamine it can be concluded that the enzyme is a GH .
Currently, two genome sequences of the pneumocandin producer G. lozoyensis are available: one of the wild-type strain ATCC 20868 , which is a producer of pneumocandins A and B, and the other from mutant strain ATCC 74030 which produces pneumocandin B0 predominantly . Sequence comparison of the clusters showed that only one enzyme, leucine dioxygenase GloC (= GLOXY4 in wild type ), is modified at two sites in the mutant strain: T98I and A294T . Although not located in the active site, it is assumed that these variations hamper the activity of the enzyme, which is involved in methylproline biosynthesis and thus essential for pneumocandin A0 production (cf. Figure 2). Recently, the targeted deletion of GLOXY4 in wild-type G. lozoyensis has been reported . As expected, mutant strain ΔGLOXY4 did not produce pneumocandin A. Instead, the production of pneumocandin B0 was increased by a factor of 9.5. In a further study, three other oxygenases were deleted in wild-type G. lozoyensis, and the effects on pneumocandin production were investigated . Deletion of the cytochrome P450 monooxygenase GLP450-1 gave pneumocandins without hydroxylation at homotyrosine C-4 (pneumocandin F and G), while the inactivation of GLP450-2 resulted in products with unmodified ornithine (pneumocandins A2 and B2). In extracts of mutant strain ΔGLOXY1, in which the putative αKG/Fe(II)-dependent homotyrosine C-3 oxygenase was deleted, a complex mixture of nine products was found. It consisted of A- and B-type pneumocandins with various hydroxylation patterns at ornithine C-4 and C-5, as well as at homotyrosine C-4; however, there was no hydroxylation at homotyrosine C-3. These results nicely support the model for echinocandin biosynthesis proposed by the groups of Tang and Walsh , . They also show that the principle of echinocandin biosynthesis is well conserved, even in distantly related fungi.
3 Biosynthetic clusters
To date, the sequences of nine echinocandin biosynthetic clusters are available from the NCBI (Table 3). The corresponding gene maps are depicted in Figure 5 and the genes are explained in Table 4. Basically, there are two groups of clusters corresponding to two different classes of fungi, the Leotiomycetes (GL/Glo, PH, CE_1, CE_2 and CC) and the Eurotiomycetes (Ecd/Hty, Ani, AE and AA). Clusters from Leotiomycetes may be subdivided into CE_1, CE_2 and CC, which are essentially collinear, and GL/Glo and PH. Although the latter appear to be different at first sight, they can formally be interconverted in a single DNA rearrangement if the reverse complement of the last eight genes from the C-terminus of GL/Glo is shifted to the N-terminus.
The genetic composition of the clusters, shown in Figure 5B, allows a relatively precise prediction of the structure of the corresponding products. Only the sulfate groups at the homotyrosine residue in some of the products are not encoded in the clusters. The presence of a PKS indicates a branched chain fatty acid residue, typically dimethylmyristate (GL/Glo, PH). Found in all Leotiomycetes clusters is an additional non-heme dioxygenase, most likely a GH. In the clusters CE_2 and CC, the heme-dependent homotyrosine 4-hydroxylase (hT4H) is missing. Consequently, there is no hydroxyl group at homotyrosine C-4 in metabolites produced by such strains. The aculeacin biosynthetic cluster AA from Aspergillus aculeatus ATCC16872 lacks the gene for the transaminase (TA), which is thought to catalyze the last step in homotyrosine biosynthesis. A BLASTP analysis of the proteome of this strain deposited at the JGI genome portal  with the aminotransferase (AT) from Aspergillus nidulans NRRL 8112 as template resulted to no putative AT with significantly increased sequence identity (>60%). This should be expected for an orthologous enzyme involved echinocandin biosynthesis. Therefore it is doubtful whether this strain is still able to produce aculeacin. In all clusters from Leotiomycetes, a putative gene for an α/β-hydrolase (ABH) was identified downstream from ornithine 4,5-hydroxylase (OrnH), which has not been described previously. Conclusions about its function are highly speculative and lack experimental evidence. The same applies to the relatively large (≈75 kbp) putative protein P1 of unknown function, which is found in all clusters and is always located downstream from the LDO, in reverse orientation.
4 Fungal strains and echinocandin structures
During the search for antifungal echinocandin compounds, a number of producer strains have been identified. Some of these strains have also been cultivated on a technical scale , so that echinocandin byproducts produced in low concentrations could be isolated and structurally characterized. This section provides an overview of the echinocandin producer strains described so far and their diverse products. For discussion, it is necessary to distinguish between the main product and byproducts. The main product is the primarily produced product. As a rule, the activity of all synthetic enzymes of the biosynthetic cluster, and sometimes even external enzymes, is required to produce the main metabolite. The enzymes convert their preferred substrate with the preferred selectivity. As most organisms only produce one main product, structural diversity can be observed by comparing different species. Such diversity is always due to different genetic constitutions of the producers. In contrast, byproducts emerge from enzymes with relaxed substrate specificity or incomplete product selectivity. Some biosynthetic steps yield more than one product or are simply skipped. Furthermore, congeners may arise from biotransformations catalyzed by promiscuous external enzymes from other metabolic pathways.
Finally, as for other secondary metabolites, it should be noted that echinocandin production can be highly dependent on environmental conditions affecting gene expression and activity of enzymes in many ways.
4.1 Echinocandin producer strains
The current collection of echinocandin producers includes 24 different wild-type fungal strains from more than a dozen different species (Table 5). Fifteen strains belong to the class Leotiomycetes, mostly from the order Helotiales, and nine to the Aspergillaceae family (Eurotiomycetes).
The metabolic diversity produced by a single species is best seen for pneumocandin biosynthesis in G.lozoyensis (Helotiales). It is mainly characterized by incomplete hydroxylations of homotyrosine and ornithine C-4 and C-5. A characteristic feature of pneumocandin biosynthesis is the incorporation of diverse hydroxyprolines and proline at residue 6, which is normally reserved for 4-methyl-3-hydroxyproline in echinocandins (Table 6).
In Aspergillaceae, even from the best-characterized producers of echinocandin B or aculeacin, only a portion of the metabolite structures has been established (see footnote d in Table 6). Therefore, it is somewhat speculative to describe metabolic diversity for this group of fungi. In general, the structures of metabolites produced by Aspergillaceae appear to be more consistent. The main products differ, if at all, in the fatty acid side chain or amino acid 5, which is serine instead of threonine in mulundocandins. Based on available data, most byproducts occur through incomplete hydroxylation with hT4H. In echinocandin D, the ornithine residue has not been hydroxylated by OrnH at all. In aculeacin biosynthesis, both palmitic and myristic acids are accepted as fatty acids, resulting in γ- and α-forms of aculeacin, respectively. No variations of threonine 3 and 4-methyl-3-hydroxyproline 6 have been documented for echinocandins from Aspergillaceae. Despite the diversity within echinocandin structures, some structural elements are strictly conserved (Figure 6). It is likely that these are essential for biological activity or required for the conformational stability of the molecule. Most notable among these elements is trans-4-hydroxyproline 3, which is fully conserved in all known echinocandin structures. The uptake of trans-4-hydroxyproline by A-domain 3 of the NRPS must be strictly substrate specific, since similar building blocks, such as 4-methyl-3-hydroxyproline, proline or trans-3-hydroxyproline, are available during biosynthesis. A second element found in all echinocandins described to date is the fatty acid R3 attached to the α-amino group of ornithine 1. This is explained by its essential function as an anchor for membrane binding in target cells. In products of clusters equipped with a PKS for biosynthesis of a branched chain fatty acid like dimethylmyristic acid, this is incorporated exclusively; otherwise, specific fatty acids from primary metabolism are utilized. A further conserved element is found in residues 2 and 5. Apart from one exception (cryptocandin), these amino acids have a hydroxyl group at C-3 (threonine, serine, hydroxyglutamine), which, if chiral, has the (R)-configuration. In pneumocandins and related structures, residue 5 is occupied exclusively by 3-hydroxyglutamine which is generated by hydroxylation of glutamine with a non-heme GH. The complete absence of glutamine 5 derivatives among pneumocandins and related echinocandins suggests that the hydroxylated amino acid is formed before the specific incorporation by A-domain 5 into the peptide chain. Otherwise, incomplete hydroxylation should lead at least to trace amounts of glutamine 5 variants. This hypothesis, however, lacks experimental evidence. As noted, cryptocandin as the main metabolite of Cryptosporiopsis cf. quercina is a unique exception among the echinocandins, not only due to the glutamine residue at residue 5, but also because a 4-hydroxymethyl group (R11) occurs in 3-hydroxyproline 6. It would be interesting to determine whether the additional hydroxyl group is introduced by the PH or if an external enzyme is involved. Finally, it is striking that homotyrosine is an integral element of all echinocandins. Incorporations of tyrosine or phenylalanine have yet to be reported.
No variations in the configuration of the up to 17 stereocenters in echinocandins (cf. Figure 1) have been described. Although it is not clear to what extent the configuration of byproducts has been determined, the stereochemistry appears to be strictly conserved.
Of the 11 variable residues R1–R11 in the general structure depicted in Figure 6, only four contain variations in which no oxidation reactions are involved: the fatty acid side chain (R3), threonine or serine at residue 2 (R4), homotyrosine sulfate (R8), and the different amino acid side chains at residue 5 (R9). The methyl group in proline 6 (R11) is biosynthetically derived from a dioxygenase-catalyzed cyclization of leucine and, thus, the result of an oxidation step. This indicates that metabolic diversity in echinocandin biosynthesis is largely induced by (incomplete) activity of the oxygenases. Structural diversity and especially the generation of byproducts through unselective or incomplete tailoring steps are not limited to echinocandin biosynthesis, but rather are a general feature of secondary metabolism. Nevertheless, the detailed knowledge of echinocandin biosynthesis, at least for some species, provides a basis for a more detailed discussion of the evolutionary background of this phenomenon. Based on current models for secondary metabolite evolution, the significance of metabolic diversity for an organism is discussed in the next section using the example of pneumocandin biosynthesis in G. lozoyensis.
5 Comparison of pneumocandin biosynthesis with models for the evolution of secondary metabolism
It is a common phenomenon in secondary metabolism that biosynthetic pathways produce families of closely related compounds , . Several reasons for this phenomenon have been discussed , , , , , , , . For a profound understanding of metabolic evolution, the concrete functions of metabolites in the organism’s complex natural environment must be understood. However, these functions are notoriously difficult to investigate. Thus, current knowledge in this field is largely based on more general models explaining the characteristics of secondary metabolite biosynthesis. Two more recent and fundamentally different approaches are introduced briefly. In 1989, Williams et al. proposed a model that emphasized the elaborate biosynthesis pathways for secondary metabolites and their sophisticated modes of action with their targets : the complex structures of such metabolites, which allow an optimal binding to the target, and their elaborate biosyntheses could not have arisen by chance; instead, pathways have evolved in a stepwise manner strictly according to the Darwinian principle, driven by the bioactivity of the product. Only mutations with a positive effect for the organism can prevail and, consequently, secondary metabolites only evolve when this is accompanied by a distinct advantage for the organism. Therefore, for each secondary metabolite a specific physiological function can be expected. Metabolic diversity typically occurs in a late stage of biosynthesis. It has developed in a more recent phase of evolution on the basis of the vast number of hazardous species a microorganism has to cope with in its natural environment. Since each metabolite is supposed to be designed by evolution in order to interact with a specific target, Fischbach and Clardy termed this theory the ‘target-based’ model .
Based on the exploding number of secondary metabolites of unknown function, Firn and Jones introduced a ‘diversity-based’ model for plant secondary metabolism in 1991 . The idea was further developed  and about a decade later Firn and Jones introduced the ‘screening’ model for secondary metabolism in general . In brief, the likelihood that an accidentally created new metabolite has an advantageous bioactivity for the organism is generally very low. Therefore, the probability that an organism confronted with a threat (e.g. another hazardous organism) has a compound suitable for defense is much higher, from a statistical point of view, when the organism is equipped with a large library of metabolites. To produce such a library at low costs, enzymes with increased substrate promiscuity and reduced substrate selectivity need to be employed. Thereby, metabolic pathways can branch and combine in multiple ways so that a ‘matrix grid’ of secondary metabolism is formed . Another positive effect of this metabolic network for the organism is an increased metabolic stability. A biosynthetic pathway with strictly substrate-specific enzymes will collapse as soon as one of the enzymes is inactivated or a specific substrate is exhausted. More promiscuous enzymes, however, also accept structurally similar substrates, such as products from other steps in the same biosynthetic route or metabolites from other pathways. Consequently, the production via a pathway permitting diversity does not necessarily cease when a particular catalytic step breaks down. In summary, diversity-inducing enzymes allow an organism to create metabolic libraries for defense, and can stabilize biosynthetic routes and simplify the evolution of novel biosynthesis pathways. A consequence of this model is that numerous metabolites formed by combinatorial biosynthetic routes can be found in organisms, which have no immediate physiological effects. Both the target-based model and the screening model (also referred to as the diversity-based model ) were designed to describe secondary metabolism in its entirety, and not single metabolic pathways such as echinocandin biosynthesis. However, Fischbach and Clardy successfully applied both models in a discussion of the extremely diverse biosynthesis of the gibberellin diterpenoids .
With this background, the example of pneumocandin biosynthesis in G. lozoyensis (Helotiales), which is by far the best-documented system among the echinocandin producers , , , , , ,  is now examined in an effort to ascertain if one of the models can be applied to echinocandin biosynthesis. During pneumocandin fermentation development at Merck & Co. Research Laboratories, over two dozen different pneumocandins were identified , with the chemical structures being reported for 16 of these derivatives. Recently, a total of seven new pneumocandins were isolated, and chemically characterized, from three G. lozoyensis mutants (GL_ΔGLP450-1, GL_ΔGLP450-2, GL_ΔGLOXY1; cf. Tables 5 and 6) , . The biosynthetic routes of the 23 reported pneumocandins are plotted as lines in a schematic representation of the involved enzymes, according to the current biosynthesis model (Figure 7). This graphic clearly shows that in the earlier steps enzymatic conversions are apparently selective and only very few byproducts are formed: more precisely, byproducts which are incorporated into detectable amounts of pneumocandin derivatives. During the last steps, however, the number of derivatives increases dramatically. This increase begins with the exceptional substrate promiscuity of A-domain 6 in the last module of the NRPS, which accepts at least five different proline derivatives. One of them, trans-3-hydroxyproline (trans-3-Hyp), originates from the incomplete regioselectivity of the PH and forms the basis of the pneumocandin B family.
In total, not less than eight pneumocandins are finally released from the NRPS. All of them are accepted by the subsequent two cytochrome P450 oxygenases (OrnH and hT4H), which obviously have a relaxed substrate specificity. However, none of the three hydroxylations catalyzed by these enzymes is performed quantitatively. Due to different hydroxylation patterns, the total number of congeners has almost tripled in the end. The combinatorial nature of these transformations suggests that further pneumocandins with other combinations of functional groups should exist; however, their concentrations might be very low, with resulting difficulties in detection and characterization. Interestingly, a blockade of main pathways by deletion of certain oxygenases unveiled several new combinatorial pathways not found in the wild-type strain of G.lozoyensis , . Considering all data on pneumocandin biosynthesis, elements of both models for secondary metabolism can be identified:
The target-based model: Despite all variations (Figure 7), it can be seen that a sharply defined amino acid backbone is conserved in all echinocandins. (If this is restricted to pneumocandin biosynthesis, this is even more distinct, since there are no variations at R3, R7, R8 and R9.) The biosynthesis of the echinocandin core structure requires a complex interaction of many diverse enzymes and it can be assumed that this is the result of a long evolutionary process. A key precondition of the target-based model is that all metabolites have a distinct physiological function. In fact, virtually all resulting echinocandins have a pronounced antifungal activity. However, most of them are only produced in very low amounts and the manner in which they are produced does not appear to be targeted nor well controlled. Finally, it is striking that in all known echinocandin biosyntheses, the most oxidized compound is also the main product. In other words, despite all possibilities to omit biosynthetic steps and to generate congeners, the preferred product is that produced by the use of all synthetic enzymes.
Since a targeted biosynthesis of the pneumocandins produced in minor concentrations is rather unlikely, a variation of the target-based model is briefly discussed. This variation postulates that only the main products, pneumocandin A0 and possibly pneumocandin B0, have evolved to fit a target; all minor compounds are tolerated as a ‘side effect’ of biosynthesis. Since these byproducts also exhibit bioactivity, and secondary metabolism generally has a relatively low throughput, the loss of substance is only at little cost, possibly even less than the expression of a fully selective enzymatic machinery. Yet, the metabolic diversity of pneumocandin biosynthesis appears to constitute an evolutionary optimum and, in principle, byproducts could be easily decreased or avoided, probably at minimal cost: A-domain 3 of G. lozoyensis NRPS is strictly specific for trans-4-hydroxyproline and A-domain 6 of A. pachycristatus NRPS is specific for 4-methyl-trans-3-hydroxyproline. This shows that hydroxyprolines can be incorporated specifically. In contrast, A-domain 6 of G.lozoyensis NRPS accepts at least five different (hydroxy)prolines. If this was unfavorable, it can be expected that an evolution would have created a substrate-specific domain within a reasonable period. In addition, the amount of partially hydroxylated congeners could be reduced by a simple increase in hydroxylase activity, for instance, via a slight increase in enzyme production.
The diversity-based model: Whereas production of the echinocandin core structure can be readily explained with the target-based model, there are inconsistencies concerning the diversity-generating late steps of the biosynthesis; rather, the seemingly uncontrolled production of diverse pneumocandins, with the branching and converging pathways forming a complex biosynthetic network, corresponds excellently with the screening or diversity-based model. A key notion of this theory is that organisms producing a large number of secondary metabolites have an advantage in defense against hazardous organisms; however, it is debatable whether this assumption can be applied to the limited diversity provided by a single pathway such as pneumocandin biosynthesis. In principle, it can be surmised that a mixture of antibiotic congeners has a broader applicability against diverse hazardous organisms than a single compound. Nevertheless, most pneumocandins are only produced in very low concentrations. Even more important, they are all derived from a highly specialized core structure, so that their mode of action is probably (but not necessarily) always the same. Thus, a fungus with a fundamental echinocandin resistance will not be affected by any of the derivatives.
The second benefit proposed for an organism with diversity-based secondary metabolism is metabolic flexibility, which also supports evolutionary processes. Thus, a second prediction of the diversity-based model is an increase in metabolic stability resulting from enzymes with reduced substrate specificity. Again, experimental data disclosed by Merck & Co. are instructive. Although all experiments focused on an improved pneumocandin B0 production and the fermentation conditions were far removed from the natural environment of G.lozoyensis, the data provide an interesting insight into the range of variations acceptable for pneumocandin production. Besides variations in the fermentation conditions, excellently reviewed by Connors and Pollard , mutants of G. lozoyensis were generated by means of chemical mutagenesis . Fermentation experiments with a descendant of pneumocandin B0 producer strain ATCC 74030 revealed that, in particular, the composition of the medium influenced the metabolite composition rather than temperature or pH. First, the relaxed substrate specificity of A-domains allows a flexible incorporation of alternative amino acids, depending on their availability in the substrate. Specifically, this was shown by feeding experiments with serine, hydroxyproline and proline , . Surprisingly, the feeding of threonine had a slightly inhibitory effect on pneumocandin B0 biosynthesis (−27%). Other modifications of the medium, such as addition of transition-metal ions or osmotic stress induced by fructose, had a strong impact on pneumocandin production. Notably, the effect differed drastically between individual pneumocandins and, in some cases, the reduced concentration of pneumocandin B0 was partly offset by an increased titer of a less hydroxylated congener. For example, the production of pneumocandin B0 was slightly decreased in the presence of nickel and cobalt; however, the production of byproduct pneumocandin B1 increased three- and fivefold, respectively . The metabolic flexibility in pneumocandin biosynthesis is even more pronounced in deletion mutants. By knockout of the LDO in wild-type G. lozoyensis, a mutant was created which was no longer able to produce the main product, pneumocandin A0 ; however, this loss was compensated by a ninefold increase in pneumocandin B0 production. In cultures with fermentation medium H, the pneumocandin B0 production strain ATCC 74030, in which the same enzyme is impaired by two point mutations , produced even more pneumocandin B0 than the wild-type production of pneumocandin A0 . Another mutant derived by classical mutagenesis (ATCC 20958) produced high amounts of the previously unknown pneumocandin A2 (cf. Table 6) instead of pneumocandin A0 , which can now be explained by an inactivation of OrnH. A derivative strain, mutant ATCC 20988, additionally synthesized pneumocandin A4 in which both aliphatic hydroxyl groups at the dihydroxyhomotyrosine side chain are missing. More recently, Li et al. reported the knockouts of three oxygenases in wild-type G. lozoyensis (ΔhT3H, ΔhT4H, ΔOrnH) . The mutants readily produced the dehydroxypneumocandins (A1, B1), (AG, BF) and (A2, B2), which were expected according to the current model for echinocandin biosynthesis (cf. Figure 3 and Table 6). Notably, from the homotyrosine 3-hydroxylase knockout strain (ΔhT3H), seven additional pneumocandins were isolated, five of which had not been described previously. One compound, pneumocandin A4, had already been characterized before and is the most reduced pneumocandin main product produced by a G.lozoyensis strain (ATCC 20988). Even more degenerate is ‘compound 14’ (suggested name: pneumocandin B14) isolated from G. lozoyensis ΔhT3H. Compared to pneumocandin A0, its biosynthesis requires nine of 14 synthetic enzymes, including just two of the six oxygenases (PH, GH). Despite the degenerate structure, the antifungal activity of compound 14 was only slightly below average .
The strongly increased diversity in the last steps of pneumocandin biosynthesis, largely determined by stochastic processes and the availability of substrates accepted by promiscuous enzymes, is much better explained using the diversity-based model than the targeted oriented approach. However, it is questionable if all compounds synthesized are on standby for screening events, as the diversity-based model implies. As in other complex biosynthesis pathways, only few pneumocandins are produced in considerable concentrations; most are found only in trace amounts. Although a physiological function at such low concentrations should not be generally excluded, it is doubtful that these pneumocandins are potent enough to ensure a reasonable antibiotic activity. Thus, their biosynthesis appears to be more important here, not the final products themselves. In one way, pneumocandin biosynthesis as depicted in Figure 7 resembles a river with tributaries, which forms a large delta of pathways at its end. Only few pathways have a considerable substance flow; however, if the main stream is blocked, for example by inactivation of an enzyme or shortage of a substrate, others are ready to accept the metabolic flux. Consequently, in such a case, pneumocandin biosynthesis is not disrupted, but simply shifted toward a derivative, often without loss in overall production. Such a system not only provides biosynthetic stability in the case of an inactive enzyme, it also allows a maximum flexibility in evolution. Given a hazardous organism against which a minor byproduct has optimal activity, and not the main product, it requires only a few point mutations (possibly only one) or some regulatory effects to increase the byproduct production to effective concentrations. From that perspective, a compound library in the narrow sense is not presented for screening, rather a highly flexible biosynthetic system which is able to modify the product(s) rapidly through minor evolutionary events. Even so, some findings are not fully consistent with a diversity-based model. First, as mentioned before, in all wild-type echinocandin-producer strains analyzed so far, the main product is also the most oxidized metabolite, whose biosynthesis requires all synthetic enzymes. Thus, the genetic equipment of the clusters is focused on the production of a defined product. Furthermore, the diversity among echinocandin main products is limited, which suggests that a few structures are privileged by evolution. Second, there is no explanation as to why, for example, the incorporation of hydroxyproline 6 is very promiscuous in G. lozoyensis, while being very strict in other species. For instance, echinocandin production in A. pachycristatus breaks down when methylproline biosynthesis is disrupted , which clearly shows that module 6 of the NRPS is strictly specific for 4-methyl-3-hydroxyproline.
In summary, there is a remarkable structural diversity among the pneumocandins induced by distinct biocatalytic steps, which, for example, convert more than one substrate, generate more than one product from a single substrate or are simply omitted. Meanwhile, there are many steps that do not contribute to structural diversity and thus allow the setup of the highly conserved echinocandin backbone. Although these observations are not unusual for biosynthetic routes of secondary metabolites, the evolutionary models discussed here can only account for parts of pneumocandin biosynthesis. In particular, the diversity-based model provides a fundamental explanation for the otherwise barely reasonable formation of many pneumocandin congeners, even though this approach originally referred to secondary metabolism in general rather than specific pathways. Currently, there seems to be no theoretical model that combines the selective biosynthesis of core structures with the emerging diversity induced by certain enzymes typically involved into late ‘tailoring’ steps. A critical limiting point in the discussion of secondary metabolite evolution in greater depth and the development of new theoretical concepts is the poor knowledge of the actual function of such metabolites in the natural environment. Mostly, this is completely unknown; sometimes, as in the case of echinocandins, a pronounced bioactivity has been identified. The utilization of a secondary metabolite by its producer in the natural habitat, however, has been documented only in very few cases. Nevertheless, detailed information on the chemical ecology of an organism is necessary to draw more specific conclusions on the evolution of its secondary metabolites.
Although a number of questions are still open, pneumocandin biosynthesis belongs to the best studied in fungi. As the bioactivity of the products can be well examined, it constitutes an excellent model system for future studies on ecological aspects, which can also help to understand better the general principles of secondary metabolism.
I would like to thank Prof. Michael Müller for critical discussions and for reading the manuscript.
Stan CD, Tuchilus C, Stan CI. Echinocandins – new antifungal agents. Rev Med Chir Soc Med Nat Iasi 2014;118:528–36. Google Scholar
Balkovec JM, Hughes DL, Masurekar PS, Sable CA, Schwartz RE, Singh SB. Discovery and development of first in class antifungal caspofungin (Cancidas®) – a case study. Nat Prod Rep 2014;31:15–34. CrossrefGoogle Scholar
Groll AH, Schrey D, Walsh TJ. Echinocandins. In: Kauffman CA, Pappas PG, Sobel JD, Dismukes WE, editors. Essentials of clinical mycology. Springer New York, 2011:95–112. Google Scholar
Perlin DS. Mechanisms of echinocandin antifungal drug resistance. Ann NY Acad Sci 2015;1354:1–11. Google Scholar
Connors N, Pollard D. Pneumocandin B0 production by fermentation of the fungus Glarea lozoyensis. In: An Z, editor. Handbook of industrial mycology. Boca Raton, FL: CRC Press, 2004:515–38. Google Scholar
Adefarati AA, Hensens OD, Jones ETT, Tkacz JS. Pneumocandins from Zalerion arboricola. V. Glutamic acid-derived and leucine-derived amino-acids in pneumocandin A0 (L-671,329) and distinct origins of the substituted proline residues in pneumocandins A0 and B0. J Antibiot 1992;45:1953–57. Google Scholar
Adefarati AA, Giacobbe RA, Hensens OD, Tkacz JS. Biosynthesis of L-671,329, an echinocandin-type antibiotic produced by Zalerion arboricola – origins of some of the unusual amino-acids and the dimethylmyristic acid side-chain. J Am Chem Soc 1991;113:3542–45. CrossrefGoogle Scholar
Petersen LA, Hughes DL, Hughes R, DiMichele L, Salmon P, Connors N. Effects of amino acid and trace element supplementation on pneumocandin production by Glarea lozoyensis: impact on titer, analogue levels, and the identification of new analogues of pneumocandin B0. J Ind Microbiol Biotechnol 2001;26:216–21. Google Scholar
Cacho RA, Jiang W, Chooi Y-H, Walsh CT, Tang Y. Identification and characterization of the echinocandin B biosynthetic gene cluster from Emericella rugulosa NRRL 11440. J Am Chem Soc 2012;134:16781–90. CrossrefGoogle Scholar
Hüttel W, Youssar L, Grüning BA, Günther S, Hugentobler KG. Echinocandin B biosynthesis: a biosynthetic cluster from Aspergillus nidulans NRRL 8112 and reassembly of the subclusters ecd and hty from Aspergillus pachycristatus NRRL 11440 reveals a single coherent gene cluster. BMC Genomics 2016;17:570. Google Scholar
Jiang W, Cacho RA, Chiou G, Garg NK, Tang Y, Walsh CT. EcdGHK are three tailoring iron oxygenases for amino acid building blocks of the echinocandin scaffold. J Am Chem Soc 2013;135:4457–66. CrossrefGoogle Scholar
Chen L, Yue Q, Zhang X, Xiang M, Wang C, Li S, et al. Genomics-driven discovery of the pneumocandin biosynthetic gene cluster in the fungus Glarea lozoyensis. BMC Genomics 2013;14:339. CrossrefGoogle Scholar
Li Y, Chen L, Yue Q, Liu X, An Z, Bills GF. Genetic manipulation of the pneumocandin biosynthetic pathway for generation of analogues and evaluation of their antifungal activity. ACS Chem Biol 2015;10:1702–10. CrossrefGoogle Scholar
Chen L, Yue Q, Li Y, Niu X, Xiang M, Wang W, et al. Engineering of Glarea lozoyensis for exclusive production of the pneumocandin B0 precursor of the antifungal drug caspofungin acetate. Appl Environ Microbiol 2015;81:1550–8. Google Scholar
Youssar L, Grüning BA, Erxleben A, Günther S, Hüttel W. Genome sequence of the fungus Glarea lozoyensis: the first genome sequence of a species from the helotiaceae family. Eukaryot Cell 2012;11:250. CrossrefGoogle Scholar
Hibi M, Mori R, Miyake R, Kawabata H, Kozono S, Takahashi S, et al. Novel enzyme family found in filamentous fungi catalyzing trans-4-hydroxylation of L-pipecolic acid. Appl Environ Microbiol 2016;82:2070–7. CrossrefGoogle Scholar
Nordberg H, Cantor M, Dusheyko S, Hua S, Poliakov A, Shabalov I, et al. The genome portal of the Department of Energy Joint Genome Institute: 2014 updates. Nucleic Acids Res 2014;42:D26–D31. CrossrefGoogle Scholar
de la Cruz M, Martín J, González-Menéndez V, Pérez-Victoria I, Moreno C, Tormo JR, et al. Chemical and physical modulation of antibiotic activity in Emericella species. Chem Biodivers 2012;9:1095–113. CrossrefGoogle Scholar
Dreyfuss MM, Tscherter H. Antibiotic s 31794/F-1, 1979. US Patent 4173629, 1978-06-28. Google Scholar
Zou SP, Zhong W, Xia CJ, Gu YN, Niu K, Zheng YG, et al. Mutagenesis breeding of high echinocandin B producing strain and further titer improvement with culture medium optimization. Bioprocess Biosyst Eng 2015;38:1845–54. CrossrefGoogle Scholar
Benz F, Knüsel F, Nüesch J, Treichler H, Voser W, Nyfeler R, et al. Stoffwechselprodukte von mikroorganismen 143. Mitteilung. Echinocandin B, ein neuartiges polypeptid-antibioticum aus Aspergillus nidulans var. Echinulatus: Isolierung und bausteine. Helv Chim Acta 1974;57:2459–77. CrossrefGoogle Scholar
Traber R, Keller-Juslén C, Loosli H-R, Kuhn M, Von Wartburg A. Cyclopeptid-antibiotika aus Aspergillus-arten. Struktur der echinocandine C und D. Helv Chim Acta 1979;62:1252–67. CrossrefGoogle Scholar
Keller-Juslén C, Kuhn M, Loosli HR, Petcher TJ, Weber HP, von Wartburg A. Struktur des cyclopeptid-antibiotikums sl 7810 (= echinocandin B). Tetrahedron Lett 1976;17:4147–50. Google Scholar
Bills GF, Yue Q, Chen L, Li Y, An Z, Frisvad JC. Aspergillus mulundensis sp. Nov., a new species for the fungus producing the antifungal echinocandin lipopeptides, mulundocandins. J Antibiot 2016;69:141–48. CrossrefGoogle Scholar
Roy K, Mukhopadhyay T, Reddy GC, Desikan KR, Ganguli BN. Mulundocandin, a new lipopeptide antibiotic. I. Taxonomy, fermentation, isolation and characterization. J Antibiot 1987;40:275–80. CrossrefGoogle Scholar
Kanasaki R, Sakamoto K, Hashimoto M, Takase S, Tsurumi Y, Fujie A, et al. Fr209602 and related compounds, novel antifungal lipopeptides from Coleophoma crateriformis no. 738. J Antibiot 2006;59:137–44. CrossrefGoogle Scholar
Iwamoto T, Fujie A, Sakamoto K, Tsurumi Y, Shigematsu N, Yamashita M, et al. Wf11899a, wf11899b and wf11899c, novel antifungal lipopeptides 1. Taxonomy, fermentation, isolation and physicochemical properties. J Antibiot 1994;47:1084–91. Google Scholar
Kanasaki R, Abe F, Kobayashi M, Katsuoka M, Hashimoto M, Takase S, et al. Fr220897 and fr220899, novel antifungal lipopeptides from Coleophoma empetri no. 14573. J Antibiot 2006;59:149–57. Google Scholar
Strobel GA, Miller RV, Martinez-Miller C, Condron MM, Teplow DB, Hess WM. Cryptocandin, a potent antimycotic from the endophytic fungus Cryptosporiopsis cf. quercina. Microbiology 1999;145:1919–26. CrossrefGoogle Scholar
Noble HM, Langley D, Sidebottom PJ, Lane SJ, Fisher PJ. An echinocandin from an endophytic Cryptosporiopsis sp. And Pezicula sp. In: Pinus sylvestris and Fagus sylvatica. Mycol Res 1991;95:1439–40. CrossrefGoogle Scholar
Dreyfuss M. Neue Erkenntnisse aus einem pharmakologischen Pilz-screening. Sydowia 1986;39:22–36. Google Scholar
Kanasaki R, Kobayashi M, Fujine K, Sato I, Hashimoto M, Takase S, et al. Fr227673 and fr190293, novel antifungal lipopeptides from Chalara sp. No. 22210 and Tolypocladium parasiticum no. 16616. J Antibiot 2006;59:158–67. Google Scholar
Peláez F, Collado J, Platas G, Overy DP, Martín J, Vicente F, et al. Phylogeny and intercontinental distribution of the pneumocandin-producing anamorphic fungus Glarea lozoyensis. Mycology 2011;2:1–17. CrossrefGoogle Scholar
Bills GF, Platas G, Peláez F, Masurekar P. Reclassification of a pneumocandin-producing anamorph, Glarea lozoyensis gen. Et sp. Nov., previously identified as Zalerion arboricola. Mycol Res 1999;103:179–92. CrossrefGoogle Scholar
Morris S, Schwartz R, Sesin D, Masurekar P, Hallada T, Schmatz D, et al. Pneumocandin D0, a new antifungal agent and potent inhibitor of pneumocystis carinii. J Antibiot 1994;47:755–64. Google Scholar
Masurekar PS, Fountoulakis JM, Hallada TC, Sosa MS, Kaplan L. Pneumocandins from Zalerion arboricola. II. Modification of product spectrum by mutation and medium manipulation. J Antibiot 1992;45:1867–74. CrossrefGoogle Scholar
Schwartz RE, Giacobbe RA, Bland JA, Monaghan RL. L-671,329, a new antifungal agent. 1. Fermentation and isolation. J Antibiot 1989;42:163–67. Google Scholar
Hu Z-C, Peng L-Y, Zheng Y-G. Enhancement of echinocandin B production by a UV- and microwave-induced mutant of Aspergillus nidulans with precursor- and biotin-supplying strategy. Appl Biochem Biotechnol 2016;179:1–14. CrossrefGoogle Scholar
Demain A, Fang A. The natural functions of secondary metabolites. In: Fiechter A, editor. History of modern biotechnology I. Berlin, Heidelberg: Springer, vol. 69, 2000:1–39. Google Scholar
Tkacz J, Giacobbe R, Monaghan R. Improvement in the titer of echinocandin-type antibiotics: a magnesium-limited medium supporting the biphasic production of pneumocandins A0 and B0. J Ind Microbiol 1993;11:95–103. Google Scholar
About the article
Published Online: 2016-10-05
Published in Print: 2017-01-26
Declaration of interest: The author reports no declarations of interest.