Introduction

In recent decades, natural-product screening programs devoted an enormous effort to discovering new microbial secondary metabolites with interesting biological activities. One of the most important steps in the drug discovery process from microbial natural products is the selection of new microorganisms, such as members of the actinobacteria group. The selection criteria applied to wild-type actinomycete strains evolved in response to these needs and include a vast array of morphological, chemo-taxonomic and molecular fingerprinting methods (Peláez and Genilloud 2003). In the past decade, multiple molecular fingerprinting methods were developed for Streptomyces and other members of the actinomycete group, based on random amplifications of repetitive sequences from microbial genomes (Anderson and Wellington 2001). Alternatively, the amplification of specific ribosomal gene sequences (Jensen et al. 1993; Welsh and McClelland 1991) or the analysis or restriction patterns of rDNA PCR products (Cook and Meyers 2003) were frequently used to evaluate diversity and to distinguish among large actinomycete populations.

There are several examples where PCR screens for genes associated with secondary metabolism are used to evaluate the biosynthetic potential of actinomycetes. These include non-ribosomal peptide synthetases (NRPS), modular and aromatic polyketide synthases (PKS-I, PKS-II), hydroxymethylglutaryl coenzyme A reductases and aminoglycoside resistance genes (Metsa-Ketela et al. 1999; Anderson et al. 2002; Sigmund et al. 2003; Ritacco et al. 2003; Ayuso and Genilloud 2004). The drawback to these methods is that they are not effective at “dereplicating” or identifying unique isolates without extensive sequencing of the PCR products obtained. It is not currently known whether the detection of high amplification frequencies in a given actinomycete population reflects or not a potential to produce secondary metabolites with biological activity. If so, combining the data derived from traditional molecular fingerprinting methods and new fingerprinting approaches based on the metabolic potential to produce secondary metabolites would enable screening efforts to focus on the most talented groups, increasing the chances of finding secondary metabolites with interesting biological activities.

For this purpose, we initially evaluated the potential bias of the PCR amplification approach when targeting conserved sequences within biosynthetic domains, before developing a fingerprinting approach based on the restriction analysis of amplified metabolic gene sequences. This method was later applied to study the occurrence of these biosynthetic systems in a collection of wild-type actinomycetes isolated from tropical soil samples where the presence of PKS-I, PKS-II and NRPS sequences could be related to the antimicrobial activity observed in some of the microbial groups tested. Here, we discuss the application of this approach as an alternative dereplication method for actinomycetes on the basis of their potential to produce bioactive compounds.

Materials and methods

Bacterial strains and cosmids

The wild-type actinomycetes used in this study were isolated from three tropical soils collected on Martinique (Windward Islands, Central America). Soils were serially diluted and plated onto soil extract- and humic acid-based agar media supplemented with 20 μg/ml nalidixic acid and cycloheximide; and colonies were isolated under a stereoscope after observation of their distinct morphology.

The type strains used were Saccharopolyspora erythraea MA6655(ATCC 11635), S. hygroscopicus NRRL5491 (ATCC 29253) and S. avermitilis MA6847 (ATCC 31267). Cosmids that contained the avermectin biosynthetic pathway (pVE855, pVE859, pVE 923, pVE 924) were obtained from D.J. MacNeil (MacNeil et al. 1992).

All strains were grown at 28°C on YME agar medium (0.4% yeast extract, 1% malt extract, 0.4% glucose, 0.2% Bacto-agar) and ATCC-2 liquid medium (0.5% yeast extract, 0.3% beef extract, 0.5% peptone, 0.1% dextrose, 0.2% potato starch, 0.1% CaCO3, 0.5% NZamine E).

DNA manipulation

DNA extraction

Total genomic DNA from the different actinomycetes used in this study was recovered and purified as described by Innis et al. (1990).

PCR primers

Consensus ketosynthase (KS) PCR primers KS-BEF (5′- CCGCGCGAGGCGCTGGCCGTCGAC- 3′) and KS-BER (5′-CCGCGCCGGGCGGGGGTCTCGTCGTTCGGCATCAGCGGCACCAACGCG-3′) were designed to universally detect KS domains from PKS-I systems. Degenerate PCR primers target specifically NRPS adenylation domains (A3, A7R), PKS-I KS domains and methyl malonyl transferase domains (K1F, M6R), as described by Ayuso and Genilloud (2004). A second degenerate KS primer, K2R (5′-CVTTCGGVVTCAGCGGSACBAA-3′) derived from KS_BER, was used in combination with K1F to amplify specifically a shorter fragment of KS domains. PKS-II sequences were amplified specifically using the primers KSα and KSβ (5′-TSGRCTACRTCAACGGSCACGG-3′, 5′-TACSAGTCSWTCGCCTGGTTC-3′), designed to target conserved sequences in KSα and KSβ domains (A. Ayuso, unpublished data). Primers G1 and L1 were used to amplify the variable ribosomal spacer regions between the 16S and 23S rDNAs (Jensen et al. 1993). All primers were supplied by ECOGEN.

PCR amplifications

PCR amplifications were performed in a Peltier PTC-200 thermal cycler in a final volume of 50μl containing 10% of extracted DNA, 0.4 μmol of each primer, 0.2 mmol of each of the four dNTPs (Roche), 5 μl of extracted DNA, 1 unit of Appligene Taq polymerase (with its recommended reaction buffer) and 10% of DMSO. PKS-I amplifications with primers KS-BE F/R were performed according to the following profiles: 5 min denaturation at 95°C and 35 cycles of 1 min at 94°C, 1 min at 62°C and 1 min extension at 72°C, followed by 10 min at 72°C, 5 min denaturation at 95°C and 15 cycles of 1 min at 94°C, 1 min at 62°C and 1 min extension at 72°C, followed by 10 min at 72°C. NRPS, PKS-I and PKS-II amplifications with degenerate primers were performed according to the following profile: 5 min at 95°C and 35 cycles of 30 s at 95°C, 2 min at either 55°C (for K1F/M6R, K1F/K2R), 58°C (for KSα/KSβ) or 59°C (for A3F/A7R) and 4 min at 72°C, followed by 10 min at 72°C. Amplification products were analyzed by electrophoresis in 1% agarose gels stained with ethidium bromide.

G1/L1 PCR amplifications were performed as described by Hirsch and Sigmund (1995). PCR products were analyzed by electrophoresis in 4–20% acrylamide gradient mini-gels (Criterion TBE gels, BioRad).

RFLP analysis

PCR products were cloned using the TOPO TA cloning kit (Invitrogen). Restriction analysis of the clones was performed using HinfI or Sau3A1 according to Sambrook and Russel (2001) and separation on 4–20% gradient polyacrylamide gels (Criterion precast gel 4–20% TBE; Bio-Rad).

Cosmid library construction

A total of 3.0 g (wet weight) of Streptomyces sp. strain DC328 mycelia was used for genomic DNA isolation. Genomic DNA was prepared for cloning by random shearing and then end-repair was used to generate blunt ends; and sized DNA was then ligated into pWEB::TNC (Epicentre Technologies) as per the manufacturer’s instructions. The DNA was then packaged using MaxPlack packaging extracts and the cosmid was used to infect Escherichia coli EPI100-T1, according to the instructions of the manufacturer (Epicentre Technologies). A total of 1,800 cosmid-containing clones were isolated and individually frozen at −80°C in microtiter dish wells.

DNA sequencing

Cloned products were sequenced using universal primers M13R-28 and M13F-20 in an ABI Prism dye terminator cycle sequencing kit (Perkin Elmer).

Evaluation of antimicrobial activity

In vitro antimicrobial susceptibility tests were performed using the following panel of strains: Bacillus subtilis MB964 (ATCC 6633) and clinical isolates Staphylococcus aureus MB 5393, E. coli MB4926 and Candida albicans MY1055, all from the Merck Culture Collection. The inoculum and assay plates for bacteria and yeast were prepared as described by Suay et al. (2000). Agar plugs from wild-type actinomycete strains cultivated in YME agar for at least 7 days were applied to the surface of the assay plates, which were incubated at either 28°C (yeast) or 37°C (bacteria). Growth inhibition zones were measured after 24 h of incubation.

Data analysis of G1/L1 PCR amplifications and HinF1 restriction fingerprinting

G1/L1 PCR amplification patterns and restriction fingerprinting patterns were analyzed with the BioNumerics ver. 2.5 program (Applied Maths). A similar matrix was generated using the Pearson correlation moment; and cluster analysis was performed with the UPGMA algorithm. Sequence database searching was done using BLAST (Altschul et al. 1990). Sequence alignments were done using CLUSTAL W (Thompson et al. 1994) and phylogenetic analysis was done in PHYLIP (Felsenstein 1981), using the neighbor algorithm. The analysis was bootstrapped using 100 replicates.

Results

Development of the type I PKS-I KS consensus PCR screen

The previously described degenerate primers K1F and M6R targeted specifically to PKS-I KS domains and methylmalonyl transferase domains in actinomycetes and had shown the limitations of their use to detect the frequently occurring malonyl-transferase domains in PKS-I biosynthetic pathways (Ayuso and Genilloud 2004). To broaden the detection range, we focused exclusively on KS domains and designed two new pairs of KS PCR primers. One pair was non-degenerate (KS-BE F/R) and encompassed a region of 1,014 nt. The second was degenerate (K2R/K1R) and amplified a 250 bp region. The specificity of both pairs of primers was evaluated using the erythromycin biosynthetic cluster of Saccharopolyspora erythraea MA6655 and the rapamicin biosynthetic system of Streptomyces hygroscopicus NRRL 5491 respectively as model systems. After amplification of Saccharopolyspora erythraea genomic DNA (1 μg) with primers KS-BE F/R, PCR fragments were cloned and analyzed by restriction analysis with Sau3A1. The erythromycin biosynthetic pathway from S. erythraea contains six KS domains, from which only one was detected among the two distinct clones selected according to their restriction profile. Further PCR optimization was done, which included reducing the number of PCR cycles and reducing the DNA template concentration (Table 1). Ultimately, 11 KS domains were obtained, which included the previously detected KS domains and two additional new ones (Fig. 1). Similarly, genomic DNA from Streptomyces hygroscopicus was amplified with the degenerate primers KS1F/KS2R and cloned. Twenty-five clones were selected for analysis and 16 different KS sequences were identified, of which two were from a rapamycin cluster and the rest were uncharacterized KS domains.

Table 1 Detection of KS domains in the four model systems studied
Full size table
Fig. 1
figure 1

Sequence diversity of KS domains isolated from S. erythraea by PCR and comparison of the KS domains obtained from S. erythraea with KS domains from the erythromycin biosynthetic pathway. A KS domain from the rapamycin biosynthetic cluster of Streptomyces hygroscopicus was included as a root (RapA3; accession number X86780).The bootstrap values for 100 replicates are given. Asterisks represent novel KS domains. The accession numbers for the Saccharopolyspora erythraea KS domains are: S. erythraea KS1 AY678081, S. erythraea KS2 AY678082, S. erythraea KS3 AY678080, S. erythraea KS4 AY678079, S. erythraea KS5 AY678073, S. erythraea KS6 AY678078, S. erythraea KS7 AY678077, S. erythraea KS8 AY678076, S. erythraea KS9 AY678075, S. erythraea KS10 AY678074, S. erythraea KS11 AY678083

Full size image

To investigate the distribution of PKS-I KS domains within actinomycete isolates further, we compared the PCR detection of KS domains from isolates having cosmid libraries with the KS domains obtained by direct PCR of the source DNA. To do this, we used Streptomyces avermitilis, which has eight documented PKS-I pathways which include avermectin and oligomycin (Omura et al. 2001), and a soil isolate, Streptomyces sp DC328, which had an unknown polyketide production capacity but was positive by PCR using the KS-BE PCR primers. Eleven KS domains were detected by screening Streptomyces avermitilis chromosomal DNA, of which one was from the avermectin biosynthetic pathway, seven from the oligomycin biosynthetic pathway and an additional three from uncharacterized PKS-I pathways in this strain. Analysis of the four overlapping cosmids that contained the avermectin biosynthetic pathway enabled the detection of six of the 12 KS domains (Table 1). Finally, we also tested this approach with Streptomyces sp. DC328. Chromosomal DNA was screened by PCR using the KS-BE PCR primers, the PCR products were cloned and 30 clones were characterized by restriction digestion with Sau3A1. After sequencing, five unique KS domains were identified. A systematic approach was taken to further evaluate this strain for KS domains by screening a cosmid library that contained 1,800 cosmids; and 49 were positive for KS domains by PCR (Table 1). This represented 2.6% of the library. PCR products were cloned and 250 clones were analyzed by restriction digestion. Eleven cosmids were identified that contained multiple KS domains. Single KS domains were detected from the remaining 38 cosmids that were positive for KS domains by PCR (Fig. 2a). The different KS domains identified by restriction analysis were all sequenced and found to represent 25 KS domains, including the five domains that were also detected by PCR of chromosomal DNA (Fig. 2b). The diversity of novel KS domains from Streptomyces sp. DC328 was compared with KS domains from characterized PKS pathways (Fig. 2c)

Fig. 2
figure 2figure 2

Evaluation of PKS-I screen using Streptomyces sp. DC328. a example of multiple-restriction profiles obtained from amplification products from a single cosmid (i), compared with single-restriction profiles that were obtained for 38 of the 49 positive cosmids (ii). Each gel represents the results for a single cosmid. b Dendrogram of 20 of the unique restriction profiles obtained from Streptomyces sp. DC328, using Sau3A1 digests. c Cluster analysis of the KS sequences obtained for Streptomyces sp. DC328 compared with known PKS-I KS domains. Streptomyces sp. DC328 sequences are in numbered red boxes; and all the PKS-I KS domains from characterized pathways are color-coded with the key given in the figure

Full size image

Detection of PKS-I, PKS-II and NRPS sequences in wild-type actinomycetes

Taking into account the inherent limitations of the direct PCR method, we focused our study on a wild-type actinomycete population of 329 strains isolated from three tropical soil samples collected in Martinique, to evaluate the diversity of biosynthetic sequences that could be detected by standard amplification in a microbial population from a tropical environment (Table 2).

Table 2 PCR detection of PKS-I, PKS-II and NRPS biosynthetic sequences and antimicrobial activities in wild-type soil actinomycetes. For taxonomy, isolates were characterized to the genus level, using micromorphological characters. Antimicrobial test panel: BS Bacillus subtilis, SA Staphylococcus aureus, EC Escherichia coli, CA Candida albicans
Full size table

Isolates were screened for the presence of NRPS, PKS-I and PKS-II sequences by specific amplification of total DNA. All three pathways were detected at a high frequency in the Streptomyces, Micromonosporaceae and Nocardiaceae populations (Table 2). The detection levels of PKS-I sequences varied according to the specificity of the pair of primers used, K1F/K2R or K1F/M6R. However, the ratios between both pairs remained quite similar in the different actinomycetes groups. Some PKS and NRPS sequences were also detected among the few members of the families Thermomonosporaceae and Streptosporangiaceae included in the study and in many of the sterile actinomycete strains that could not be assigned to a defined taxonomic group (Table 2).

Antimicrobial activity and presence of biosynthetic gene sequences

The production of antimicrobial activity by the soil isolates was evaluated against a test panel of three bacteria (B. subtilis, Staphylococcus aureus, E. coli) and a yeast (C. albicans). Over half (62%) of the Streptomyces showed some antimicrobial activity against at least one of the test strains, mostly against the Gram positive test strains B. subtilis and Staphylococcus aureus (55%). The activity levels against E. coli and C. albicans were 25% and 14%, respectively. In contrast, very little activity was observed among representatives of the family Micromonosporaceae (1.3%) and no activity was obtained within the family Nocardiaceae. We also detected the production of activity in both members of the Thermomonosporaceae and in one of the two strains of the family Streptosporangiaceae (Table 2).

To determine any existing relationship between the presence of NRPS and PKS sequences and the potential to produce antimicrobial compounds, isolates were divided into two groups of active and inactive isolates and were analyzed for the presence of these biosynthetic gene sequences. We observed that higher detection rates were obtained for the target pathways in the group of active strains than in the group of inactive cultures (Table 3). In the case of Streptomyces, the percentages of positive NRPS and PKS-I amplifications were almost two-fold higher in the active compared with the inactive group; and this difference was not extended to PKS-II sequences (Table 3). This correlation was not seen for the Micromonosporaceae and Nocardiaceae isolates, due to their low performance in the bioactivity screen (Table 3).

Table 3 PCR detection of NRPS, PKS-I and PKS-II biosynthetic systems in wild-type soil actinomycetes. Wild-type strains were divided into two groups: active strains, which included strains showing antimicrobial activity against at least one of the four control strains tested, and inactive strains, which included strains for which no activity was observed in the antimicrobial assays
Full size table

Metabolic PCR fingerprinting and diversity of wild-type actinomycetes

To evaluate the diversity of the wild-type actinomycete population used in the study, we used the amplification of the variable spacer region between ribosomal genes 16S and 23S with primers L1/G1 to generate reproducible fingerprints that allowed strain discrimination below the species level (Hirsch and Sigmund 1995). We compared the amplification patterns of wild-type Streptomyces and members of the families Micromonosporaceae and Nocardiaceae containing NRPS, PKS-I and PKS-II sequences; and we built three dendrograms. The strains were grouped into 14 clusters of closely related strains with above 75% similarity in both Streptomyces and the Micromonosporaceae, whereas only two clusters were outlined in the Nocardiaceae (data not shown).

An alternative fingerprinting approach based on restriction analysis of the amplified biosynthetic sequences was explored to determine whether isolates could be simply characterized for their biosynthetic potential. HinfI restriction patterns were studied for the NRPS, PKS-I and PKS-II amplification products of the strains included in the three major groups in study. We observed in almost all cases a high pattern heterogeneity among the strains included in the clusters defined on the amplification of ribosomal sequences, suggesting a quite diverse biosynthetic potential even among closely related strains. To define any potential relatedness among the restriction patterns observed within each of the microbial groups studied, we built three dendrograms for each of the major taxonomic groups studied, on the basis of a similarity matrix generated by the addition of the three restriction fingerprints (Figs. 3, 4, 5).

Fig. 3
figure 3

Diversity of wild-type Streptomyces. Dendrogram showing similarity, based on the HinfI restriction patterns of NRPS, PKS-I and PKS-II amplification products. The L1/G1 ribosomal fingerprint cluster assignment is indicated for each isolate. Solid lines Duplicate strains with similar restriction profiles, dashed lines unrelated isolates with similar restriction profiles. RightThe antimicrobial activity pattern is shown by squares for each isolate

Full size image

Most of the wild-type strains analyzed present distinct NRPS and PKS restriction patterns and correspond to different isolates, in spite of their closely related ribosomal fingerprint patterns. Nevertheless we observed in the three major groups studied some cases of similar restriction patterns in ribosomally related isolates (Figs. 3, 4, 5). The similarity of restriction patterns was also paralleled by a similar antimicrobial profile in Streptomyces, suggesting the presence of duplicate isolates within the population studied. In addition, in Streptomyces and the Micromonosporaceae we observed cases of high similarity of restriction patterns between unrelated strains assigned to different ribosomal clusters (Fig. 3, 4).

Fig. 4
figure 4

Diversity of wild-type Micromonosporaceae. Dendrogram showing similarity, based on the HinfI restriction patterns of NRPS, PKS-I and PKS-II amplification products. The L1/G1 ribosomal fingerprint cluster assignment is indicated for each isolate. Solid lines Duplicate strains with similar restriction profiles, dashed lines unrelated isolates with similar restriction profiles, arrows strains with antimicrobial activity against S. aureus MB5393 and B. subtilis MB954

Full size image
Fig. 5
figure 5

Diversity of wild-type Nocardiaceae. Dendrogram showing similarity, based on the HinfI restriction patterns of NRPS, PKS-I and PKS-II amplification products. The L1/G1 ribosomal fingerprint cluster assignment is indicated for each isolate. Solid lines Duplicate strains with similar restriction profiles

Full size image

Discussion

In this investigation, we evaluated the specific detection of KS domains from four model systems and demonstrated that, despite the abundance of these target domains within the strains studied, one can only obtain a proportion of the domains that are present, leaving the metabolic potential of the isolates underestimated as a result of the bias introduced by the PCR method itself (Ishii and Fukui 2001; Suzuki and Giovannoni 1996). Even using the most optimized procedures, as in the case of Streptomyces avermitilis, only half of the 12 avermectin domains were detected despite using cosmids that contained the pathway as templates. Many novel KS domains were detected using these model systems, highlighting the abundance of putative novel PKS pathways. To further evaluate this, novel KS domains from Streptomyces sp. DC328 were compared with published KS domains from characterized macrolide biosynthetic pathways, using a phylogenetic approach (Fig. 2c). The trend for characterized PKS pathways is for KS domains within a module to cluster together. The hypothetical KS domains detected in this study appeared to form distinct clusters, but one cannot use this as a predictive tool for KS pathways without a more extensive analysis of the cosmids from which they were detected. Direct PCR analysis has the obvious limitations that it may underestimate diversity and may detect irrelevant pathways. Despite this, we obtained data that demonstrate that the metabolic potential of these isolates was clearly under-represented.

We went on to use our optimized procedures to evaluate our wild-type population. Previous studies focused on how similar methods can be used to assess the metabolic potential of members of the Streptomyces genus (Sosio et al. 2000; Metsa-Ketela et al. 2002). This study confirmed these findings and expanded our knowledge about the distribution of NRPS, PKS-I and PKS-II pathways in members of the families Micromonosporaceae and Nocardiaceae and in other minor population components, where few producers have been described.

The results we derived from a comparison of the microbial populations with and without antimicrobial activity suggest that, in strains of the genus Streptomyces, a clear relationship can be established between the occurrence of biosynthetic gene sequences and the production of antimicrobial activities. This correlation was not seen for the other major families studied. It should be considered that not all the sequences detected with our primers are necessarily involved in the synthesis of industrially relevant bioactive metabolites and that the amplification products can reflect the presence of other genes involved in the biosynthesis of other types of metabolites, such as pigments or structural components of the microbial cell (Portevin et al. 2004). It is also important to note that biological activities of interest are governed by the screen to which the isolate is subjected. In this investigation, we looked for antimicrobial activity, because actinomycetes produce compounds that have anti-tumor, anti-helmintic and immunosuppressive properties, to name but a few (Anderson et al. 2000). We can use the detection of a secondary metabolite pathway as an indicator of metabolic potential, but the strain must still be grown in a way that will express the pathway and be subject to the right biological assays. Studies of sequenced bacterial genomes of known strains such as Streptomyces coelicolor A(3)2 and S. avermitilis have also demonstrated numerous silent pathways for which concerted cloning and fermentation efforts have demonstrated production and activity (Chen et al. 2000; Challis and Hopwood 2003; Hopwood 2003; Knight et al. 2003; Zazopoulos et al. 2003).

The approaches discussed in this study help to identify isolates with a potential talent. We also investigated a rapid method to fingerprint strains using PCR products obtained from metabolic screens as an alternative characterization tool to be applied not only in the study of distinct isolates among ribosomally related groups, but also in the identification of enriched populations with different biosynthetic potentials to be studied for the production of antimicrobial compounds. Our results show the diversity of restriction amplification profiles can suggest the extreme richness in these sequences within each individual strain in the three microbial groups studied. The examples of unrelated wild-type strains sharing similar PCR restriction fingerprints could reflect the occurrence of horizontal gene transfer events within members of a given microbial community, as was characterized by Sigmund et al. (2003) for hydroxy-3-methylglutaryl coenzyme A reductase pathways in natural actinomycete populations.

A concentrated effort is still required to unravel the regulation of how these pathways are expressed, along with fermentation studies, so that this as yet untapped biotherapeutic resource can be harnessed from these taxa.