Neglected tropical diseases are tropical infections that are especially common in low-income populations in developing regions of Africa, Asia, and the Americas. These diseases are caused by a variety of pathogens such as viruses, bacteria, protozoa, and helminths. Of these diseases, Sleeping Sickness and Chagas disease are caused by parasitic flagellated protozoa, Trypanosoma brucei and T. cruzi, respectively [1, 2]. To treat the diseases, some anti-trypanosomals including suramin, pentamidine, eflornithine, melarsoprol, benznidazole, and nifurtimox have been used [3, 4], but some problems in their use remain to be unsolved; the drugs are expensive, toxic, and difficult to administer, and more seriously, parasite resistance to them is increasing. Therefore, safer and highly effective anti-trypanosomal drugs are urgently required.

Actinoallolides were discovered as anti-trypanosomal macrolides from the culture broth of an endophytic actinomycete, Actinoallomurus fulvus MK10-036 [5]. Actinoallolide A displays potent and selective in vitro activity against T. brucei rhodesiense STIB900 and T. cruzi Tulahuen C4C8 with IC50 values similar to those of the therapeutic drugs mentioned above [5]. Actinoallolide A is a 12-membered macrolide, which have a 5-membered hemiacetal moiety inside the ring and a side chain including constitutive asymmetric centers [5]. Although macrolides are a class of natural products consisting of a macrocyclic lactone ring [6] and biosynthesized by the joining together of simple acyl-CoA [7], there has been no published report of the biosynthesis of actinoallolides to date. Here, we describe the identification and heterologous expression of the actinoallolide biosynthetic gene cluster.

We found that another A. fulvus strain, K09-0307, also produces actinoallolide A (Scheme S1 and Figure S1). In this study, we used the strain K09-0307 to identify the gene cluster for actinoallolide biosynthesis. We have obtained a draft genome sequence of the strain K09-0307. The total size of the sequence was 8.7 Mb and a putative actinoallolide biosynthetic gene cluster was predicted by the antiSMASH web server [8]. The gene cluster spans a contiguous 53 kb DNA region (DDBJ accession number LC326402) that comprises seven genes encoding three polyketide synthases (PKSs) (aalA1, aalA2, and aalA3), cytochrome P450 (aalB), acyl-CoA dehydrogenase (aalC), crotonyl-CoA reductase (aalD), and TetR family regulator (aalR) (Table 1, Fig. 1a). The PKSs are composed of one loading module and 10 extension modules (Fig. 1b). The signature amino acid residues of acyltransferase (AT) domains predicted the substrates (Figure S2); malonyl-CoA in the modules 6 and 10, methylmalonyl-CoA (or propionyl-CoA) in the loading module and modules 1, 2, 3, 4, 5, 7, and 8, and ethylmalonyl-CoA in module 9. The PKSs have some modification domains in each module; eight ketoreductase (KR) domains in modules 1, 2, 3, 4, 5, 6, 7, and 8, four dehydratase (DH) domains in modules 3, 4, 7, and 8, and two enoyl reductase (ER) domains in modules 4 and 8 (Fig. 1b). Following the fingerprint rule of the KR domain [9], KR2 and KR6 were classified as the A1-type, which give L-configuration at the beta position of the thioester (Figure S3). KR3, KR4, KR5, KR7, and KR8 were classified as the B1-type, which give a D-configuration. We confirmed that the stereochemistry deduced from the fingerprint is fully identical to that of actinoallolide A (Fig. 1b). KR1 could not be classified into any type because the fingerprint of KR1 was similar to both the A1 and B1 types (Figure S3). Although KR1 has catalytic amino acids and the NADPH binding motif, it seems to be inactive, as predicted from the structure of actinoallolide A that has a ketone at C-21. The proposed pathway for actinoallolide biosynthesis is described in Fig. 1b. Following the formation of polyketide backbone by AalA1, AalA2, and AalA3, the macrolactone is formed by a thioesterase (TE) domain of AalA3. The cytochrome P450 AalB might be involved in hydroxylation at C-6 and subsequent nucleophilic addition to a ketone at C-3 presumably gives a unique 5-membered hemiacetal in the macrolactone.

Table 1 Proposed functions of open reading frames in the actinoallolide biosynthetic gene cluster
Full size table
Fig. 1
figure 1

a Actinoallolide biosynthetic gene cluster and relative regions of two cosmids, 2D5 and 5C5. b Proposed pathway for actinoallolide biosynthesis

Full size image

To clone the actinoallolide biosynthetic gene cluster, we screened a cosmid library of the A. fulvus K09-0307 genomic DNA by PCR. Consequently, cosmids 2D5 and 5C5, which carried left and right parts of the gene cluster, were isolated (Fig. 1a). In order to assemble both parts of the gene cluster, four DNA fragments were prepared (Figure S4); 14 kb XhoI fragment of 2D5 (fragment 1), 31 kb BstBI and SpeI fragment of 5C5 (fragment 2), 12 kb BstBI fragment of 2D5 containing 40 bp overlapping sequences with fragments 1 and 2, respectively (fragment 3), and 7 kb PCR product of pYIK1 containing 40 bp overlapping sequences with fragments 1 and 2, respectively (fragment 4). The four DNA fragments were assembled by Gibson assembly [10] to yield pYIK1-aal, which carried the entire actinoallolide biosynthetic gene cluster (Figures S4 and S5).

An integrase gene, attP, oriT, and an apramycin resistance gene were inserted into pYIK1-aal. The resulting vector pYIK3-aal was introduced into S. coelicolor M1152 by conjugation. S. coelicolor M1152/pYIK3-aal was incubated in YD medium (yeast extract 1% and glucose 1%) for 7 days and the ethyl acetate extract of the culture broth was analyzed by LC-HRESIMS. However, none of the actinoallolides were produced by S. coelicolor M1152/pYIK3-aal. The gene cluster contains a TetR family regulator gene, aalR. TetR is known as a repressor that controls the expression of tetracycline-resistance genes [11]. It seems likely, therefore, that expression of the actinoallolide biosynthetic gene cluster is controlled by AalR.

In order to express the gene cluster, we used the constitutive promoter ermE* of pYIK1. The 55 kb MfeI and SpeI fragment of pYIK1-aal was assembled with the PCR product of pYIK1 to yield pYIK1-aal2, which had ermE* promoter at 100 bp upstream from the start codon of aalA1 (Figure S5). Following insertion of an integrase gene, attP, oriT, and an apramycin resistance gene, the resulting vector pYIK3-aal2 was introduced into S. coelicolor M1152 by conjugation. S. coelicolor M1152/pYIK3-aal2 was incubated in YD medium for 7 days and the ethyl acetate extract of the culture broth was analyzed by LC-HRESIMS. The analysis identified a compound with m/z 587.3538 ([M + Na]+; retention time 17.4 min), which exactly matched the m/z value of actinoallolide A (calcd for C32H52O8Na: 587.3554; Fig. 2). Furthermore, its MS/MS fragmentation pattern also matched that of actinoallolide A (Figure S6). These results confirmed that the gene cluster is responsible for the biosynthesis of actinoallolide. However, the concentration of actinoallolide A in the culture broth of S. coelicolor M1152/pYIK3-aalM (0.012 mg L−1) was much lower than that in the culture broth of A. fulvus K09-0307 (1.8 mg L−1). To improve the heterologous production, optimization of the culture condition and/or the promoter would be needed.

Fig. 2
figure 2

LC-HRESIMS analysis of actinoallolide production. Extracted ion count chromatoghram from m/z 587.3 to 587.4. A compound with m/z 587.3538 [M + Na]+, which exactly matched the m/z value of actinoallolide A (calcd for C32H52O8Na+: 587.3554), was detected at 17.4 min in the ethyl acetate extract of culture broth of S. coelicolor M1152/pYIK3-aal2

Full size image

The plasmid pYIK3-aal2 does not contain the crotonyl-CoA carboxylase/reductase gene aalD. Crotonyl-CoA carboxylase/reductase catalyzes conversion of crotonyl-CoA to ethylmalonyl-CoA, which is one of the polyketide synthase extender units [12]. However, AalD is dispensable for the biosynthesis of actinoallolide via heterologous expression in S. coelicolor M1152 (M145 Δact Δred Δcpk Δcda rpoB[C1298T]). The crotonyl-CoA carboxylase/reductase SCO6473, a homologous protein of AalD (identity/similarity: 75%/84%), was found in the genome of S. coelicolor M145 by a BLAST search. SCO6473 might work as an alternative to AalD.

In conclusion, we first identified the actinoallolide biosynthetic gene cluster using a combination of genome analysis and a heterologous expression approach. More promising actinoallolide analogs may be obtained by genetic engineering of the gene cluster.