Isolation of Kitasatospora from Blue Spring Cave

Actinobacteria were isolated from cave sediments obtained from Blue Spring Cave in White County, Tennessee (Fig. 1a) via dilution plating on agar media (See Methods). One putative actinobacterial strain developed into an orange lawn with white mycelia on ISP-2 agar and formed cylindrical aerial hyphae with smooth spore chains dotted with protruding nodules (Fig. 1b and Supplemenatary Fig. S1). The 16S ribosomal gene sequence was used to infer the likely evolutionary history of this species by bootstrapping 1000 maximum likelihood projections in MEGA 7 [22] of 16S ribosomal DNA sequence alignment with closest relatives predicted by EZBiocloud’s 16S-based [23] ID search algorithm and 16S sequences of closest type strains from Streptomyces, Micromonospora, and Amycolatopsis (Fig. 1c). EZBioCloud identified Kitasatospora psammotica (99.93% similarity, basonym Streptomyces psammotica) [24, 25] as the closest evolutionary relative consistent with the phylogenetic tree (Fig. 1c). The percent similarity of 16S DNA was given the subspecies (ssp.) designation carrieae (Fig. 1c) in recognition of Lonnie Carr, a long-standing cave conservationist and owner of Blue Spring Cave, generously providing access to this gated 30+ mile system to researchers and explorers for decades.

Fig. 1

Caves are an ecological niche for new actinomycetes. (a) Passage within Blue Spring Cave. (b) Scanning electron micrograph of aerial hyphae and smooth tendrils with dull protrusions that lack clearly defined sporangia. White arrowhead in inset B indicates dull protrusion. (c) The evolutionary history was inferred by using the Maximum Likelihood method and Tamura-Nei model. The tree with the highest log likelihood (-3932.17) is shown. Initial tree(s) for the heuristic search were obtained automatically by applying Neighbor-Join and BioNJ algorithms to a matrix of pairwise distances estimated using the Maximum Composite Likelihood (MCL) approach, and then selecting the topology with superior log likelihood value. The tree is drawn to scale, with branch lengths measured in the number of substitutions per site. This analysis involved ten nucleotide sequences. There were a total of 1537 positions in the final dataset. Evolutionary analyses were conducted in MEGA 7

Multiplexed Activity Profiling (MAP) affords high-throughput, dose-dependent interrogation of bioactive metabolites

A MAP protocol was used to test a dilution series of crude metabolomic extracts from K. psammotica ssp. carrieae cultured with chemical and biological stimuli to identify dose-dependent responses of diverse mammalian cell status markers (Fig. 2a). K. psammotica ssp. carrieae was cultivated under ten distinct stimulus conditions, including solid and liquid media (ISP-2 and Bennett’s media), in the presence of competitive microorganisms (B. subtills, R. wratislavensis, T. pneumonia and E. coli), in the presence of sub-inhibitory concentrations of antibiotics (rifampin and streptomycin) and in the presence of lanthanide salts (LaCl3) [7, 9, 20, 21, 26]. The resultant metabolomes were extracted by adding hydrophobic resin HP-20 to cultures and extracting cells and resin with methanol, diluted 64-fold across four wells in microtiter plates, and evaporated to yield an extract dilution array (Fig. 2a). Metabolites were resuspended in cell culture media containing MV-4-11 AML cells at a density of 1 million mL−1 and incubated overnight for 16 h.

Fig. 2

Multiplexed activity profile of single cells challenged with metabolomes from actinomycetes cultured in diverse stimuli. (a) Harvested actinomycetes were incubated in the presence of diverse stimuli and then secondary metabolites extracted for bioactivity dose response in MV-4-11 cell line. Following incubation, cells were viability stained with Ax700, fixed with PFA, permeabilized with methanol, and fluorescently barcoded. Fluorescent cell barcoding is used to conserve well position through processing and analysis with NHS-functionalized, primary amine reactive Pacific Blue and Pacific Orange fluorescent dyes. Eight concentrations of PB are arrayed with six concentrations of PO to create 48 unique fluorescent identities that encode each well position. Barcoded cells are pooled and stained with a fluorescent antibody cocktail to measure molecular readouts with flow cytometry. Cells are digitally reassigned to well position via fluorescent barcode with DebacodeR algorithm. Extracts, well position, fluorescent barcode, and molecular readouts are combined to produce a multiplexed activity profile to assess bioactivity. (b) Quality control gating scheme of MAP assay. Collected events were gated by forward and side scatter to identify cells, forward scatter height for single cells, and Ax700 for viability. Six concentrations of pacific orange are encoded in the viable, singlets cells and each pacific orange concentration encodes eight pacific blue concentrations to identify cell populations by fluorescent barcode. A single cell population is tracked by the red gates and is identified in the debarcoded cell populations. (c) Multiplexed activity profile of S. psammotiucs ssp carrie cultured in diverse stimuli. Each black dot represents a single cell response at the observed dose and are compared to specific positive control compounds (red box) to infer single cell marker shifts versus vehicle control (blue box). This organism produced a cytotoxic compound that did not increase the number of apoptotic cells measured as the arcsinh ratio of median fluorescent intensity versus vehicle control but exhibited cytotoxic effects visualized by decreased viable cell populations. V = Vehicle, A = Aphidicolin, E = Etoposide, N = Nocodazole, R = Rapamycin, S = Staurosporine, PB = Pacific Blue, PO = Pacific Orange, BN = Bennet’s media, IS = ISP2, IA = ISP2-agar, IB = ISP2-B. subtilus, IL = ISP2-LaCl3, IR = ISP2-Rifampin, IW = ISP2-Rhodococcus wratislaviensis, IC = ISP2-Escherchia coli, IN = ISP2-Streptomycin, IT = ISP2-Tsukamorella pneumonia

After incubation cells were stained with a fluorescent Alexa 700 (Ax700) dye to label non-viable cells and measure overt cytotoxicity, then fixed in paraformaldehyde to memorialize cell status, and finally permeabilized in methanol to enable fluorescent antibody and fluorescent dye infiltration for detection of intracellular targets including cCAS3 (apoptosis), γH2AX (DNA damage), DNA (cell cycle), p-S6 (translation) and p-Hh3 (proliferation). Cells were fluorescently barcoded with an array of N-hydroxysuccinimide ester-functionalized amine (NHS) reactive fluorescent dyes composed of eight concentrations of Pacific Blue (one per row, A-H) and six concentrations of Pacific Orange (one per column, 1-6) to yield 48 well-specific fluorescently labeled cell populations (Fig. 2a). The barcoded cells were pooled and stained with a fluorescent module composed of antibodies and a DNA intercalator to detect bioactivity. Barcode and marker signal intensity were then acquired simultaneously with flow cytometry (Fig. 2a). Following signal acquisition, intact, singlet cell populations were gated by forward and side scatter parameters then gated for viable, Ax700 negative cells (Fig. 2b). Cells were debarcoded with the DebarcodeR algorithm to separate cell populations by fluorescent intensity of each barcode dye for digital reassignment to the original plate well (Fig. 2b). Conventionally, non-viable cells are excluded from analysis due to inconsistent reactions with fluorescent NHS-dyes and fluorescent antibodies that confound debarcoding and overall data interpretation.

Cells were phenotypically evaluated by inspecting bioactivity trends to determine bioeffector potential of each extract. K. psammotica ssp. carrieae extracts markedly reduced viable cell populations with relatively low functional marker shifts cross-referenced to bioactivity observed in positive control conditions: Staurosporine – cCAS3, Etoposide – γH2AX, Aphidicolin – DNA (G1), Rapamycin – p-S6, Nocodazole – p-HH3, DNA (G2) (Fig. 2c). For example, metabolites harvested from ISP-2 liquid culture exhibited viable cell loss at 1.2 mg mL−1 and 5 mg mL−1 extract doses as a consequence non-viable cell removal (Fig. 2c, IS). An increase in cytotoxicity was observed in viable cells remaining in effected wells (e.g. Figure 2c, BN, IW) and demonstrated less fluorescent signal of cell status markers compared to staurosporine or etoposide controls (Fig. 2c). Moreover, extracts taken from various stimulus conditions did not invoke marker changes or overt cytotoxicity including co-culture with B. subtilis and exposure to sub-lethal doses of rifampin antibiotic (Fig. 2c, IB and IR). Taken together, selective stimulus can stimulate or decrease biosynthesis of cytotoxic agents by K. psammotica ssp. carrieae that permeabilized cell membranes within 16 h of challenge (Fig. 2c). Moreover, this effect was observed at extract doses as minimal as 0.31 mg mL−1 (Fig. 2c), which motivated the decision to isolate and elucidate the bioactive agent(s) from K. psammotica ssp. carrieae. It is worth noting that the control compound staurosporine elicits an apoptotic cCAS3 response that can exhibit diminished cell populations relative to other compound controls as apoptotic cells with advanced membrane degradation are labeled non-viable by Ax700 and excluded from analysis (Fig. 2c). Conversely, etoposide challenged cells exhibited robust γH2AX DNA damage response without concomitant cell loss due to membrane disruption (Fig. 2c).

Multiplexed Activity Metabolomics (MAM) detects bioactivity in discrete metabolomic fractions

To identify the cytotoxic secondary metabolites, the K. psammotica ssp. carrieae extracted metabolome was fractioned by reverse phase chromatography connected to a split-flow polarity-switching electrospray mass spectrometric analyzer (ESI-MS) and a UV/VIS photodiode array detector (Fig. 3a). The ESI-MS portion is sacrificed to acquire mass spectra while the eluate from the UV/Vis is collected at a rate of 1 fraction min−1 of effluent per well to compare bioactive fractions with complementary spectral data and inform isolation of bioactive secondary metabolites. The K. psammotica ssp. carrieae metabolomic array was evaporated in microtiter plates in vacuo and resuspended in cell-containing media for incubation and subsequent processing with fluorescent cell barcoding, fluorescent staining, and debarcoding to identify well-specific bioactivity aligned with coordinated, discrete regions of the mass spectra (Fig. 3a and S2).

Fig. 3

Multiplexed activity metabolomics identifies candidate bioactive compounds from single-cell data. (a) Whole metabolomes from active MAPs are fractionated into a metabolomic array with split flow (3:1) HPLC/UV/MS system. One-minute metabolomic fractions are collected in 48 wells, dried in vacuo, and resuspended in media-containing cells. Cells are incubated, fixed in PFA, permeabilized in methanol, fluorescently barcoded, pooled, stained, flowed, and debarcoded and compared to marker shifts and spectral data to determine well fractions with potential for isolating a bioactive secondary metabolite. (b) Each metabolomic fraction is evaluated independently for phenotypically disturbed cell populations. The single-cell response profile of S. carrieae metabolomic fractions reveals cell loss in fractions 20 and 22. (c) Active metabolomic fractions are cross-referenced against complementary mass spectral data (total ion current – orange trace) to identify candidate natural products for isolation. Candidate ion currents are extracted to observe temporal relationships to observed bioactivity. Extracted ion currents of 625.39 (blue trace) and 661.39 (red trace) highlight candidate bioactive compounds that elute at minutes 20 (blue trace) and 22, respectively

MV-4-11 cells incubated for 16 h with control compounds staurosporine and etoposide exhibited robust shifts for apoptotic and DNA damage markers respectively. Fraction-dependent responses at 16-h incubation were evaluated by comparison to positive and vehicle controls then quantified as the arcsinh ratio of percent cells positive for each marker verse vehicle control (Fig. 3b). While the fluorescent signal of cCAS3 and γH2AX were elevated in well 20 they were decreased in well 22 despite abundant cell death in each well (Fig. 3b). The lack of marker response for DNA, p-HH3, and p-S6 (Fig. S2) suggested signaling-independent cytotoxicity for these cellular processes. Overall cytotoxicity was determined by the percent viable cells in each well compared to vehicle control. Condition 20 and 22 demonstrated reductions of viable cells to 4.4% and 17.2% respectively compared to vehicle control (Fig. S3).

The bioactivity profiles were compared to liquid chromatography/mass spectrometry (LC/MS) data to identify and extract well-localized ion currents to observe relative abundance of metabolites for isolation. This identified two putative secondary metabolites with an m/z 625.4 in fraction 20 and m/z 639.4 in fraction 22 (Fig. 3c). The UV/Vis spectral data of these secondary metabolites demonstrated local UV maxima from 320–360 nm (Fig. 3c), consistent with the family of polyene antibiotics. Inspection for compounds with similar UV spectra identified two additional but less active polyenes in fractions 14 and 19 with positive extracted ion currents of m/z 627.4 and m/z 611.4 respectively (Fig. S3).

Isolation and structural elucidation of filipins

K. psammotica ssp. carrieae culture was scaled to afford isolation of the four polyenes. Cultures were extracted using HP-20 resin and metabolites purified via LH-20 chromatography and preparative HPLC. Established Fieser-Kuhn rules [27, 28] indicated that candidate m/z 639.4 likely contained five conjugated olefins. Further 1D NMR studies were also consistent with this metabolite containing multiple C-O methine functionality (Fig. S4 and Table S1). Based upon these parameters, literature searches identified the candidate compound as the known polyene macrolide metabolite filipin II(1), which was confirmed by subsequent NMR studies including correlated spectroscopy (COSY), heteronuclear multiple-bond correlated spectroscopy (HMBC), and heteronuclear single quantum coherence spectroscopy (HSQC) (Fig. 4, S4 and Table S1). The decreased mass of candidate m/z 611.4 indicated this metabolite lacked two methylene functional groups in comparison to the assigned structure of compound 1 (Fig. 4). Subsequent NMR analysis of this metabolite demonstrated the core filipin structure was intact, and that the pendant alkyl chain lacked two methylenes (Fig. 4, S5, and Table S2). These assignments confirmed the second polyene macrolide metabolite isolated as a the previously described member of the filipin family, chainin(2) (Fig. 4)[29, 30].

Fig. 4

Structural assignment, exact mass, and spectral profiles of isolated filipin metabolites. (1) Filipin II; 639.41082 (2) Chainin; 611.37952 (3) Filipin XV; 627.37444 (4) Filipin IX; 625.39517. Blue = COSY correlation, Red = HMBC correlation, Purple = COSY and HMBC correlation

Following the identification of these previously reported members of the filipin family, two lower-abundance metabolites were isolated with m/z of 627.4 and 625.4 respectively. The m/z of the more polar of these metabolites (627.4) indicated the addition of a hydroxyl functional group relative to compound 2 (Fig. 4 and S6). In addition to sharing local UV maxima of 320–360 nM of the previously isolated filipin members, further analysis by 2D NMR revealed an intact filipin core and indicated an unusual hydroxylation pattern at carbon 3’ (3 Fig. 4 and Table S3). This assignment, combined with high-resolution mass spectrometry analysis validated the molecular formula of this metabolite (theoretical: 649.3564 (M + Na), observed: 649.3582 (M + Na), and confirmed its identity as a previously unreported member of the filipin family herein referred to as filipin XV (3, Fig. 4) which was also light yellow in color. We were unable to selectively derivatize the C3’ for the purpose of resolving the stereochemistry. The final metabolite isolated possessed one additional methylene group relative to compound 2, based upon its m/z of 625.4. 2D NMR analysis (COSY, HMBC, HSQC) of this metabolite demonstrated that the core filipin structure was intact, and that the additional methylene functionality was located on the pendant alkyl chain (Fig. 4, S7, and Table S4). The validation of the molecular formula of this metabolite (theoretical: 647.3771 (M + Na), observed: 647.3770 (M + Na)) confirmed the identity of this metabolite as filipin IX (4, Fig. 4).

Filipin-induced cell permeation is caspase-3 independent

The staurosporine and filipin injury phenotypes shared similar losses of viable cells (Fig. S3). Since staurosporine-induced apoptosis also yields a degree of non-viable cell exclusion in this system (Fig. 3 and S3), it was possible that filipins engaged apoptosis to account for overall cell loss by non-viable exclusion prior to 16-h time point signal measurement. This was tested by challenging cells with 10 µM of each filipin compound, staurosporine, or vehicle control and sampling cell populations across time points to measure cCAS3 induction and permeation by flow cytometry. At each timepoint, cells were stained with Ax700 to detect non-viable cells prior to fixing, then permeabilized with methanol and stained with fluorescent cCAS3 antibody to measure apoptosis. By 60 min, 80% or more of cells treated with 1, 2, and 4 exhibited cell population shifts towards the Ax700 positive, non-viable phenotype independent of cCAS3 activation (Fig. 5a). This phenomenon was observed at all timepoints measured after 60 min (Fig. 5 and S8). Compound 3; however, did not induce significant cCAS3 or non-viable marker shifts at any timepoint which indicated hydroxylation of the 3’ carbon on a four-carbon pendant alkyl chain prevents filipin-induced cell permeation.

Fig. 5

Filipin-induced death is cCAS3 independent and pendant alkyl chain dependent. (a) Single cell contour plots of apoptotic marker cCAS3 and Ax700 viability stain. Cells were treated for indicated time then stained with Ax700, fixed, permeabilized, and stained for cCAS3. Compounds 1, 2, and 4 disrupted cell membranes without activating cCAS3. (b) Cells were challenged with each filipin and stained with propidium iodide (cell viability, DNA) and quinacrine (intracellular ATP) cocktail. Compounds 1, 2, and 4 exhibit robust PI labeling (white arrows) and concomitant loss of ATP. Compound 3 does not enhance PI infiltration of loss of ATP (white arrowhead). (c) Quantification of percent positive PI or quinacrine positive cells versus vehicle control. Cells were challenged with 10 μM of each compound for the indicated time. *p < 0.05 vs vehicle control. Scale bar = 25 µm; Red, PI = DNA; Green, quinacrine = ATP

These observations were confirmed with a dimetric fluorescent microscopy assay that utilized propidium iodide (PI), a red fluorescent DNA intercalator that only penetrates cells with compromised cell membranes, and quinacrine, a green fluorescent intracellular ATP detection agent that is lost following membrane permeation [31, 32]. MV-4-11 cells treated with 10 µM of 3 presented robust quinacrine staining and few PI positive cells comparable to vehicle control (Fig. 5b, c) Alternatively, 10 μM of 1, 2, and 4 all permeabilized cell membranes as confirmed by PI positive staining and subsequent loss of quinacrine staining (Fig. 5b, c). Staurosporine positive control was positive for cCAS3 induction, PI infiltration, and quinacrine signal loss while vehicle control remained cCAS3 and PI negative and quinacrine signal positive.

(Fig. S9). In combination, flow cytometry and fluorescent microscopy confirmed that diminished cell populations observed in MAM are consistent with the bioactivity of 1 and 4. The observed cytotoxicity of 2 (Fig. 5) was unexpected and most likely failed to deplete viable cells in MAM because of insufficient fractionated well concentrations (Fig. 3b and S3). Moreover, all compounds share identical polyene, macrolide chemical scaffolds with varied alkane pendant lengths. Only compound 3, which carries a hydroxyl group on C 3’, was found to be non-cytotoxic (Figs. 4, 5).