Many streptomycetes make insecticidal metabolites
To compare the prevalence of specialized metabolites that act on prokaryotes, lower eukaryotes and higher eukaryotes we created small molecule extracts from 56 Streptomyces strains and tested them for inhibitory activity against Escherichia coli, Bacillus subtilis, Saccharomyces cerevisiae, Candida albicans and Drosophila melanogaster. To avoid bias based on known specialized metabolites in the well-established model systems we selected streptomycetes randomly from the Wright Actinomycetes Collection (34). All of the strains we used were wild isolates that had not been previously characterized—at the outset of this work the genome sequences and specialized metabolic potential of all of them were unknown.
Focusing first on the microbial screens (Fig. 1A) we assessed the ability of extracts to inhibit >90% growth and found that 25 extracts were active against B. subtilis, two were active against E. coli, 8 were active against S. cerevisiae and 5 were active against C. albicans. This is consistent with many previous screens (35) and confirms that antibacterial activity against Gram-positive bacteria is very common while activity against lower eukaryotic fungi and Gram-negative bacteria is less so.
To investigate the bioactivity of the extracts against Drosophila, we seeded twenty newly hatched first instar larvae into tubes containing a control food source or a food source supplemented with Streptomyces extract. We monitored their progression through to pupation and eclosion to adult flies for 14 days. Extracts from 7 strains WAC-240, -237, -303, -288, -210, -211 and -213 had potent inhibitory activity against the growth and development of D. melanogaster such that no larvae developed into adult flies. In their place we observed dead, desiccated larvae with arrested growth at various stages of development. One potent insecticidal extract was derived from strain WAC-288 (Fig 1B). This work demonstrates that insect-toxic specialized metabolites are relatively common; they are similar in number to those that are active against Gram-negative bacteria and lower fungi.
Toxicity of Streptomyces spores
We then asked whether spores of streptomycetes that produce insect-toxic extracts are themselves harmful to Drosophila. We chose 6 Streptomyces strains that had generated fly-toxic extracts (WAC-211, -213, -237, -240, -288 and -303) and another 6 (WAC-173, -175, -183, -190, -287 and -302) that generated non-toxic extracts and conducted feeding experiments with fly larvae. We prepared spores from these 12 strains, washed them twice in PBS to remove media and then added each of them to fly growth medium. We seeded the spore-treated and control tubes with larvae and allowed them to feed and develop for 14 days. The contrast between the two groups was dramatic (Fig. 2A): spores of strains that did not made toxic extracts (WAC-173, -175, -183, -190, -287 and -302) had little or no effect on fly development: 60% to 100% of the larvae in each culture proceeded through normal development to generate healthy adults after 14 days. In contrast, spores of strains that generated toxic extracts (WAC-211, -213, -237, -240, -288 and -303) had lethal effects on the larvae such that all of them had arrested development and died prior to developing to adulthood. WAC-288 spores were particularly toxic to larvae. We observed reduced mobility of the spore-fed larvae within 3-6 hours of ingestion and a rapidly worsening condition such that all larvae were motionless, desiccated and non-viable within ~24 hours.
To determine whether this effect could be observed in fruit flies other than domesticated Canton-S D. melanogaster strains, we repeated the spore-feeding experiment with 6 outbred fruit fly strains from the D. melanogaster Genetic Reference Panel (DGRP) (36). As shown in Sup. Fig. S1, the result was the same: the six outbred D. melanogaster strains were equally sensitive to WAC-288 spores such that all of the larvae were killed prior to completing development. We further repeated the experiment with D. virilis, D. suzukii, D. yakuba, D. simulans and D. pseudoobscura. Again the result was the same: the larvae of all five species died in the presence of the Streptomyces spores; none of them completed development to adults. This is the first demonstration that live streptomycetes can poison a multicellular eukaryote.
Cosmomycin D is the causative agent of killing by WAC-288
To understand the basis of this toxicity we isolated the insecticidal compound from WAC-288 using bioactivity-guided fractionation. The compound absorbed light at λ = 494 nm, a characteristic of anthracycline molecules (37). It had a parent ion mass-to-charge ratio of m/z 1189.5869 [M + H]+ from which we calculated a chemical formula of C60H89N2O22. We carried out tandem mass spectrometry on this compound and observed fragments m/z 1071, 941, 831 and 701 [M + H]+ (Fig. 1C, Sup. Fig. S2). This mass fragmentation pattern has been reported previously for the red-pigmented compound cosmomycin D (13, 38-41). As a complementary approach, we sequenced the WAC-288 genome and analyzed the 7.4 Mbp sequence using AntiSMASH that predicts specialized metabolite biosynthetic gene clusters. We found that the strain is predicted to encode 24 biosynthetic gene clusters for specialized metabolites (Sup. Table S1). One of these biosynthetic gene clusters is predicted to generate cosmomycin D (Fig. 1D). This is based on a high degree of sequence homology in all encoded proteins (97%) as well as a similar gene organization to the known cosmomycin D biosynthetic gene cluster in S. olindensis (Sup. Fig. S3) (42).
To determine whether the cosmomycin D biosynthetic genes were responsible for the toxicity of WAC-288 spores to fly larvae we constructed mutations in orf1219 (encoding a PadR-like regulator), orf1222 (encoding a β-keto acyl synthase) and orf1245 (encoding a predicted cluster-situated regulator) (Sup. Fig. S4, Table S2). We confirmed that the three mutants were defective in producing cosmomycin by LC-MS (Sup. Fig. S5) and compared the capacity of their spores to kill fly larvae to the parent strain. The result (Fig. 2B, C) indicated that all three mutants had lost their ability to kill larvae. The yields of viable adult flies were identical to the negative control in all three cases with 70%-95% mature flies. In contrast, spores from the wild type parent, WAC-288, killed all of the larvae in the culture. This confirms that cosmomycin D is responsible for the insecticidal activity of WAC-288 spores.
To determine whether this phenomenon occurs with other well-characterized streptomycetes we compared it to the effect of spores of S. avermitilis, the producer of avermectin, an inhibitor of invertebrate locomotion (43, 44). As a control we used SUKA-22, an S. avermitilis mutant that is unable to produce this compound (45) (Fig. 2D). Consistent with the mode of action of avermectin, S. avermitilis spores were also toxic to the fly larvae. The phenotypic effect however, was distinct from that of WAC-288. In this case we observed complete paralysis of all larvae within 10 minutes of ingestion: the larvae ceased locomotor movement, a well-known effect of avermectin. Consistent with the cause of this being avermectin, SUKA-22 had no effect on larval phenotype, survival or development. Taken together, these data demonstrate that the spores of streptomycetes that produce insecticidal compounds are toxic to invertebrates and confirms that this toxicity is due to specific specialized metabolites.
The mechanism of spore-associated lethality is determined by the toxic metabolite
Cosmomycin D is a DNA intercalator and many of these molecules cause cell death in eukaryotes. This, for example, is why doxorubicin, a molecule that shares the same anthracycline scaffold as cosmomycin D is used as a chemotherapeutic drug (46). In mammalian cells, DNA damage leads to the activation of the apoptosis initiator caspase-9 and executioners: caspase-7 and caspase-3. This process also occurs in Drosophila melanogaster however it is carried out by Dronc, a capsase-9 homolog, and two caspase-3 homologs Dcp-1 and DrICE (47).
To determine whether WAC-288 spores induce the activation of caspase proteins we carried out an experiment in which we fed spores of the cosmomycin-D producing parent or the cosmomycin-D defective ∆cosD-orf1222 mutants to third instar larvae. We dissected out the digestive tracts of the larvae 6 hours after feeding and them and stained them with the antibody #9661, which binds to the activated caspase-9 homolog Dronc in Drosophila melanogaster (48). The results demonstrated that feeding spores of WAC-288 resulted in the activation of Dronc primarily in cells of the posterior end of the midgut and hindgut, suggestive of a cell death-like phenomenon in this region of the larvae’s digestive tract (Fig. 3A-E). We observed the same phenomenon when we fed pure cosmomycin D to larvae (Fig. 3F). In contrast, feeding with the ∆cosD-orf1222 mutant spores did not result in Dronc activation in cells of the digestive tract (Fig 3G). This suggests that WAC-288 kills fly larvae by compromising the cells of their midgut and hindgut digestive tracts.
To determine whether this effect is shared by S. avermitilis we carried out the same experiment comparing the effect of wild type S. avermitilis and WAC-288. We again found that a phenotypic effect of feeding with S. avermitilis was much quicker than with WAC-288 – the larvae were paralyzed within minutes. In marked contrast to the effect of WAC-288, dissected digestive tracts from larvae that had fed on wild type S. avermitilis for the same duration did not display cell death-like activity (Fig 3H, I). This is consistent with the distinct mechanism of action of avermectin, which is muscle paralysis via the inhibition of the invertebrate-specific glutamate-gated chloride channel at neuromuscular junctions (51). These data demonstrate that the mechanisms of invertebrate killing by Streptomyces spores are determined by specialized metabolites produced by each species.
Chemical attraction by 2-methylisoborneol leads flies to a spore-contaminated food source
WAC-288 and S. avermitilis, produce the volatile compound 2-methylisoborneol (2-MIB). Indeed, the production of 2-MIB is widespread and highly conserved throughout the actinobacterial phyla (Sup. Fig. S7). While the unrelated volatile terpene geosmin is known to influence insect behaviour via repulsion (52) and attraction (31, 32), the biological effect of 2-MIB on flies has not be described. We therefore tested 2-MIB for effects on adult flies in a T-maze and found that low concentrations (2 µg/mL at the source) of compound attracted them (67% of total flies preferred 2-MIB, p<0.001) whereas high concentrations (2x103 µg/mL at the source) repelled them (4.7% preferred 2-MIB, p<0.001) (Fig. 4B, Sup. Video 1).
We then carried out an experiment to determine whether flies were attracted to spore-contaminated food sources and whether this was due to the production of 2-MIB. Flies were placed in a closed container with two food sources: one source was a control that lacked Streptomyces spores, the other contained either wild type WAC-288 or a mutant unable to produce 2-MIB. Flies were allowed to choose between the different food sources for 24 hours with their choice being evident by their final physical location at the experimental end point as well as the location where they had laid their eggs. We also observed the subsequent success of their progeny in both conditions by clearing the adult flies and allowing the eggs to hatch and undergo development. We observed more flies near the medium containing wild type WAC-288 spores compared to medium lacking added bacteria (76% wild type, 24% control, p<0.05). This was consistent with the attractive properties of 2-MIB at lower concentrations of pure molecule (2 µg/mL). In contrast, in a similar experiment in which the food contained the 2-MIB defective strain, this preference was absent (55% ∆2-mib-orf919, 44% control, p-value: not significant) (Fig. 4C).
The consequence of the attraction to the wild type strain was significant. Any eggs that were laid in the uncontaminated medium grew normally and generated adult progeny, as expected. In contrast, all of the progeny that were deposited and hatched in the presence of WAC-288 died prior to pupation (Sup. Fig. S8). In this case therefore, the consequence of attraction to 2-MIB was a complete failure to generate viable progeny.
To determine whether this attractive effect is widespread in the Drosophila genus we repeated this food choice experiment with D. melanogaster outbred lines as well as five other distantly related Drosophila species (Sup. Fig. S9). As with the domesticated laboratory strain of D. melanogaster, the outbred strain (DGRP cross 1), D. virilus, D. yakuba, D. simulans and D pseudoobscura were attracted towards the WAC-288 contaminated food source. The one exception was D. suzukii which exhibited a slight but not statistically significant repulsion from the WAC-288 contaminated food source. Therefore, most outbred and domesticated species of Drosophila are attracted to 2-MIB producing streptomycetes including those that produce compounds that are exceptionally harmful to them and their larvae.