Vanderbilt Institute of Chemical Biology



Discovery at the VICB







Co-Culture of Microorganisms Triggers Synthesis of Novel Natural Products



By: Carol A. Rouzer, VICB Communications
Published:  July 24, 2015



Culture of the actinomycete Nocardiopsis with bacteria of different genera leads to striking metabolic changes and the discovery of a novel class of natural products.


A large proportion (60-70%) of drugs currently on the market are natural products or derived from natural products. Thus, natural products represent a rich source of structural scaffolds from which new drugs may be discovered. Consequently, the search for novel “secondary metabolites” - compounds that do not play a role in the routine structural and energy metabolic functions of an organism - is an ongoing enterprise. Genomic studies of microorganisms have identified numerous gene clusters likely to code for enzymes that synthesize secondary metabolites, and in some cases, the structures of the metabolites can even be predicted. These studies suggest that there is a wealth of natural product chemical diversity that remains untapped, but the identification of new secondary metabolites is not a trivial pursuit. Traditionally, this has been accomplished by isolating new microorganisms, growing them in large fermentors, and then attempting to isolate novel compounds, often present in minute amounts, from the complex mixtures of metabolites. This approach is not only inefficient, it usually fails to exploit the majority of the microorganism’s synthetic potential for the simple reason that the gene clusters that code for the metabolites of interest are not active under routine culture conditions. Now, Vanderbilt Institute of Chemical Biology members Brian Bachmann and John McLean report the stimulation of novel natural product synthesis by co-culture of different species of microorganisms and the use of their Molecular Expression Dynamics Investigator (MEDI) approach to efficiently identify metabolites of interest. This approach has already led to the discovery of the polyketide natural products ciromicins A and B [D. K. Derewacz, et al. (2015) ACS Chem. Biol., published online June 17, DOI:10.1021/acschembio.5b00001].


Prior genomic investigations had shown that the actinomycete Nocardiopsis sp FR40 possesses numerous gene clusters that should enable it to produce a diverse range of secondary metabolites. However, the only natural products isolated from this microorganism at the time of the study were the apoptolidins. To simplify their search for new natural products, the investigators used a genetically modified strain of Nocardiopsis sp FR40, that cannot synthesize the apoptolidins, Nocardiopsis sp FR40 ΔApoS (NF). They then cultured NF alone and in the presence of one of four other bacterial species from different genera, Escherichia coli (EC), Bacillus subtilis (BS), Tsukamurella pulmonis (TP), and Rhodococcus wratislaviensis (RW). In each case, the second bacterial species was added as a small inoculum, so that the NF metabolome would predominate. In fact, at the end of the six day co-culture period, only NF could be detected in the cultures, indicating that it had become the dominant species.


The investigators used ultra performance liquid chromatography-ion mobility-mass spectrometry (UPLC/IM-MS) to analyze their fermentation mixtures. In addition to the co-cultures, they also analyzed fermentation broths from monocultures of each of the microorganisms. The result was 22.5 gigabytes of raw data, which they normalized and processed to identify “features”, defined as an ion with a unique m/z and retention time. They then subjected the resulting 1,693 features to the MEDI algorithm, which produces a self-organizing map (SOM). An SOM is a form of artificial neural network that is created on a grid of nodes (in these experiments a 25 x 26 node grid was used). Each node is represented mathematically by a multidimensional vector with each dimension corresponding to a parameter measured in the experiment. A comparable vector is assigned to each feature identified in the UPLC/IM-MS data. In this case, the vector parameters were the intensities of the features measured under each culture condition. Initially, the node vectors on the SOM grid are randomly assigned. During the “training” process, each feature is assigned to the node on the grid having the vector that most closely matches the vector of the feature. As the assignments are made, the node vectors are also modified to more closely match those of the assigned features. The result of this process is that features with similar vectors - that is features that responded to the experimental conditions in similar ways - are assigned to the same region of the map. Once training is complete, the sum of the intensities of all the features at each node is represented on a heat map, where colors ranging from blue to red correspond to intensities ranging from low to high, respectively (Figure 1). Heat maps representing feature intensities corresponding to each individual culture condition instantly reveal differences in features among the various conditions. More importantly, subtraction of the monoculture metabolome heat maps from the corresponding co-culture heat map readily identifies those features that change most dramatically as a result of combining the two species. The investigators designated the areas of the SOM containing these features as regions of interest (ROIs) for further study.



Figure 1.
Heat maps generated from each biculture (left) aligned with the heat maps from the corresponding monocultures (center two) and the resulting heat map generated by subtracting those of the monocultures from that of the biculture (right). ROIs corresponding to biculture-induced responses are encircled and numbered in white. Figure reproduced by permission from D. K. Derewacz, et al. (2015) ACS Chem. Biol., published online June 17, DOI:10.1021/acschembio.5b00001. Copyright 2015 American Chemical Society.


The MEDI analysis showed that co-culture with RW or TP induced a greater metabolic response in NF than co-culture with BS or EC. The investigators were interested to find that a surprisingly large number (about 20%) of features were up-regulated at least two-fold in at least one of the co-cultures, and about 14% of the features were only detected at significant levels in the co-cultures. These findings demonstrated that co-culture was a very effective way to induce the biosynthesis of novel metabolites by NF. Furthermore, a significant number of up-regulated features were unique to a single co-culture, suggesting that these metabolites resulted directly from the interaction of the two species of microorganism rather than a nonspecific effect, such as a reduction in nutrient availability.


To further confirm the results of their MEDI analysis, the researchers subjected their data to principal component analysis (PCA), a multivariate statistical approach. PCA also creates multidimensional vectors from the data and then identifies linear combinations of variables that summarize the dataset in terms of “principal components” that capture the variance in the data. Frequently, the results are graphed with each feature plotted in terms of the first principal component - the component that accounts for the greatest amount of variability in the data - on the abscissa, and the second principal component (accounting for the second greatest amount of variability) on the ordinate. This usually results in a clustering of features according to conditions that cause variability in the data set. PCA results (Figure 2) confirmed that culture conditions were a major source of data variability with features from the monocultures and co-cultures well separated along either the first or second principal component axis or both. Furthermore, the researchers discovered that the features that contributed most to the differences identified by PCA were also the most intense features in the ROIs on their MEDI heat maps. Thus, the two approaches yielded similar results. However, the investigators also noted that MEDI is much more likely than PCA to detect features present at low concentration, a major advantage to the MEDI approach.




Figure 2. Results of PCA of data comparing monocultures of each indicated bacterium and NF versus co-cultures of that same bacterium with NF. Figure reproduced by permission from D. K. Derewacz, et al. (2015) ACS Chem. Biol., published online June 17, DOI:10.1021/acschembio.5b00001. Copyright 2015 American Chemical Society.



Genomic data suggested that NF should produce natural products of the polyene macrolactam type of polyketide. Thus, the investigators searched the co-culture metabolites for compounds having the expected polyene UV/Vis absorbance. They found one such metabolite, with absorbance maxima at 207 and 290 nm and an accurate mass of 515.275 Da. This metabolite was up-regulated in all co-cultures but was particularly prominent in co-cultures of NF with RW and TP. A second metabolite of the same mass, but lacking the characteristic polyene chromophore, was also present in the co-culture fermentations. To identify the metabolite, the researchers scaled up their co-culture of NF with RW and isolated the polyene compound from the resulting fermentation broth. High resolution mass spectrometry yielded a molecular formula of C28H38N2O7 for the purified polyene, and its structure was determined by a combination of COSY, TOCSY, HSQC, HMBC, and NOESY NMR spectroscopy, confirming that it was, indeed, a novel natural product of the polyene macrolactam class. The researchers named the new compound ciromicin A (Figure 3).




Figure 3. Structures of ciromicin A, ciromicin B, and vincenistatin.


Although the second new compound of mass 515.275 Da was not present in the large-scale fermentation in sufficient quantity for purification, the investigators were intrigued to discover that ciromicin A converted to this metabolite upon exposure to light. Maximal conversion occurred in response to light of wavelength 300 nm, but this also resulted in the formation of additional isomers. Use of light with a wavelength of 400 nm yielded only the single metabolite of interest. Extensive NMR studies yielded its structure, and it was named ciromicin B (Figure 3).


The investigators used genomic data from NF to identify the putative gene cluster responsible for ciromicin biosynthesis. This enabled them to propose a ciromicin biosynthetic route, which is similar to that of vincenistatin (Figure 3), a previously characterized polyketide natural product. Since the structure of ciromicin A is similar to that of vincenistatin, the investigators hypothesized that it might also share vincenistatin’s cytotoxicity. They tested this hypothesis using the MV-4-11 human leukemia cell line, confirming that both ciromicin A and ciromicin B are cytotoxic with IC50’s of 8.1 μM and 9.3 μM, respectively, compared to 0.24 μM for vincenistatin. Neither compound exhibited antibacterial or antifungal activity in tests against a limited number of microorganisms.


The pyrrolidinol substructure in ciromicin A is novel among polyketide natual products that have been reported thus far, although structures similar to that of ciromicin B are known. The diastereoselective conversion of ciromicin A to ciromicin B is a heretofore unprecedented reaction, and its mechanism will be an interesting subject for further investigation.


These studies confirm that co-culture of microorganisms is an effective way to trigger the production of novel natural products. This is consistent with the hypothesis that the purpose of many secondary metabolites is to enable microorganisms to survive in an environment where they encounter a wide range of other species. The results also demonstrate the utility of the MEDI approach for the analysis of the large metabolomics datasets that are produced in the search for new natural products. We look forward to learning more about the chemistry, biochemistry, and pharmacology of the ciromicins, and to additional applications of co-cultures and MEDI analysis to the discovery of even more novel compounds.








The Vanderbilt Institute of Chemical Biology, 896 Preston Building, Nashville, TN 37232-6304, phone 866.303 VICB (8422), fax 615 936 3884
Vanderbilt University is committed to principles of equal opportunity and affirmative action. Copyright © 2014 by Vanderbilt University Medical Center