Actinobacteria from Antarctica as a source for anticancer discovery

Although many advances have been achieved to treat aggressive tumours, cancer remains a leading cause of death and a public health problem worldwide. Among the main approaches for the discovery of new bioactive agents, the prospect of microbial secondary metabolites represents an effective source for the development of drug leads. In this study, we investigated the actinobacterial diversity associated with an endemic Antarctic species, Deschampsia antarctica, by integrated culture-dependent and culture-independent methods and acknowledged this niche as a reservoir of bioactive strains for the production of antitumour compounds. The 16S rRNA-based analysis showed the predominance of the Actinomycetales order, a well-known group of bioactive metabolite producers belonging to the Actinobacteria phylum. Cultivation techniques were applied, and 72 psychrotolerant Actinobacteria strains belonging to the genera Actinoplanes, Arthrobacter, Kribbella, Mycobacterium, Nocardia, Pilimelia, Pseudarthrobacter, Rhodococcus, Streptacidiphilus, Streptomyces and Tsukamurella were identified. The secondary metabolites were screened, and 17 isolates were identified as promising antitumour compound producers. However, the bio-guided assay showed a pronounced antiproliferative activity for the crude extracts of Streptomyces sp. CMAA 1527 and Streptomyces sp. CMAA 1653. The TGI and LC50 values revealed the potential of these natural products to control the proliferation of breast (MCF-7), glioblastoma (U251), lung/non-small (NCI-H460) and kidney (786-0) human cancer cell lines. Cinerubin B and actinomycin V were the predominant compounds identified in Streptomyces sp. CMAA 1527 and Streptomyces sp. CMAA 1653, respectively. Our results suggest that the rhizosphere of D. antarctica represents a prominent reservoir of bioactive actinobacteria strains and reveals it as an important environment for potential antitumour agents.


Results
Bacterial community associated with Deschampsia antarctica rhizosphere. In this study, a total of 249.176 high-quality 16S rRNA reads were recovered after the QC filter step (Table S1-Supplementary Material). Sequences were assigned to 229.631 OTUs, and the rarefaction curves reached the plateau phases, confirming the adequate sequencing depth of the samples. The bacterial community structure revealed by the higherranking taxa classification was congruent among the different rhizosphere samples (Fig. 2).
The average taxonomic signatures of all samples from sites Morro da Cruz, Torre Meteoro Direito e Torre Meteoro Esquerdo (MDC, TMD and TME, respectively) showed that Actinobacteria (34%) was the most abundant phylum, followed by Chloroflexi (21%) and Proteobacteria (20%) (Fig. 2a). Members of the phyla Acidobacteria and Verrucomicrobia were detected at lower frequencies (< 5%). The 'unclassified' sequences contributed to 7% of the dataset, showing that the understanding of the plant-microbiome interaction is still limited and  www.nature.com/scientificreports/ biotic biodegradation and metabolism, DNA replication and repair, membrane transporter, and biosynthesis of other secondary metabolites were significantly enriched (p < 0.05) in the Actinobacteria dataset when compared to the total bacterial community (Fig. 3).
isolating, culturing and identifying actinobacteria. A total of 72 actinomycetes were isolated with the media and procedures employed in this study. The isolates belonged to the genera Actinoplanes, Arthrobacter, Kribbella, Mycobacterium, Nocardia, Pilimelia, Pseudarthrobacter, Rhodococcus, Streptacidiphilus, Streptomyces and Tsukamurella from 6 families in the phylum Actinobacteria. Genomic analysis based on 16S rRNA sequences was used to select 30 isolates as representative strains of the accessed actinomycetes diversity. The nearest type strains, the percentage of identity and the GenBank access number are presented in Table S2-Supplementary Material. The strains are deposited in the Collection of Microorganisms of Agricultural and Environment Importance (CMAA)-Embrapa, Brazil (Fig. 4).
phylogenetic analysis. According to the phylogenetic reconstruction based on the 16S rRNA gene similarity the bioactive actinobacterial strains were affiliated with the genus Streptomyces. The CMAA 1527 strain was closely related to Streptomyces aurantiacus NBRC 13017 T , while CMAA 1653 was related to Streptomyces fildesensis DSM 41987 T (Fig. 6), although they form a distinct phyletic branch towards the phylogenetic tree.
Bioassay-guided fractionation and structural identification of compounds. The crude extracts obtained from Streptomyces sp. CMAA 1527 and Streptomyces sp. CMAA 1653 showed high antiproliferative activity. Both extracts were analysed by LC-MS and presented different chemical profiles (Fig. 7). The major peak in Streptomyces sp. CMAA 1527 crude extract was analysed by LC-MS at the retention time of 13.15 min and had mass spectra with m/z 826 (Fig. 7a). The extract was fractionated by semi-preparative HPLC, resulting in 4 grouped fractions (FR-1, FR-2, FR-3 and FR-4) that were subjected to a biological assay.  Table 1 shows the antiproliferative potential of fraction FR-3, the most bioactive fraction according to human cancer cell line panel. FR-3 was submitted to structural identification experiments, although other fractions showed lower biological activities (data not shown).   Figure S3-Supplementary Material). Cinerubin B was identified based on spectral data such as HRESIMS, 1 H NMR, COSY, gHMBC, and gHMQC. The compound was isolated as a red amorphous powder that was soluble in methanol (MeOH) and trichloromethane (CHCl 3 ) with UV absorption maxima at 490 nm. The 1 H NMR spectrum of cinerubin B ( Figure S4-Supplementary Material) showed signs of characteristic chemical shifts of anthracycline class; the signals of phenolic hydrogen were at δ 12.99 (s, 1H), δ 12.83 (s, 1H), and δ 12.28 (s, 1H). There were also signs of aromatic hydrogen at δ 7.75 (s, 1H) and δ 7.33 (d, J = 4.41 Hz, 2 H), as well as a broad doublet for a carbinolic hydrogen at δ 5.27 (dl, J = 2.50 Hz, 1 H). Other observed signals that are characteristic of this compound refer to three methoxyl group hydrogens at δ 3.71 (s, 3H), the hydrogen signals from an ethyl group at δ 1.76 (q, J = 7.0 Hz, 2H) and δ 1.09 (t, J = 7.0 Hz, 3H), and the methyl hydrogen signals belonging to an N-dimethyl group at δ 2.16 (s, 6H). In addition, the presence of three anomeric hydrogens at δ 5.49 (sl, 1H), δ 5.12 (dl, J = 2.9 Hz, 1H), and δ 5.21 (dl, J = 2.9 Hz, 1H) indicated the presence of a trisaccharide moiety in the chemical structure of the compound. The 2D correlations from the HMBC and COSY spectra (Table S4-Supplementary Material) were used to assign all signals present in the cinerubin B structure ( Fig. 8 and Figure S5-Supplementary Material).
In order to identify other anthracyclines the LC-MS/MS analysis using neutral loss methodology was performed. The chemical structure of the predominant compound produced by Streptomyces sp. CMAA 1527 (cinerubin B; m/z 826.3298 u) showed a neutral loss of 240 u, which refers to a charge-remote hydrogen rearrangement with consequent loss of disaccharide cinerulosyl-2-deoxyfucosyl to form the product ion m/z 586 as a base peak, which it was attributed to pyrromycin. This methodology allowed us to show the presence of other bioactive compounds in the crude extract ( Figure S6-Supplementary Material) but with lower concentrations than cinerubin B since it was not possible to perform the isolation process. These compounds were identified as 1 or 11-hydroxysulfurmycin B (m/z 854), auramycin B (m/z 812) or N-desmethylcinerubine B (m/z 812) and 1-deoxycinerubin B (m/z 810) which have been attributed to the lower bioactivities detected at the others fractions ( Figure S7, S8 and S9, respectively-Supplementary Material). Furthermore, it is noteworthy that these These ions provided abundant structural information, and their identification was performed. HRESIMS/MS data were identical to those previously reported in the literature 44 . As a result of the fractionation procedures, FR-5, FR-6 and FR-7 were recovered, within the FR-7 portion as the main bioactive fraction ( Fig. 5d and Table 1). Thus, the compound present in the fraction FR-7 was identified as actinomycin V (C 62 H 84 N 12 O 17 ) (Fig. 9). The other fractions obtained by semi-preparative HPLC did not show bioactivity in the experiments carried out in this study.

Discussion
The analysis of the rhizosphere microbiota in cold extreme habitats is crucial to understand the ecological roles and unlocking the biotechnological potential of these environments. Actinobacteria was also the dominant phylum in the rhizosphere of other D. antarctica retrieved from the maritime Antarctica region, as well as soils from cold and desert environments [45][46][47][48] . This prevalence may be explained by their spore-forming ability, UV radiation tolerance and other environmental adaptations that ensure survival in harsh conditions. Actinobacterial genera are known for their great potential to produce several bioactive substances 47,49,50 , although they also have the ability to degrade and use complex organic compounds and for their bioremediation of contaminated soils 51 . In fact, the predictive functional analysis using the Actinobacteria-derived 16S rRNA gene sequences showed  52 also showed the genetic potential of Actinobacteria for C metabolism and membrane transport systems, which supports our PICRUSt functional predictions. The results obtained by 16S rRNA gene sequencing coupled to predictive functional analysis were useful to reveal diversified Actinobacteria pathways to produce biologically active metabolites. In doing so, we reported the presence of many pathways related to the biosynthesis of antibiotics such as streptomycin, novobiocin, phenylpropanoids, pyridine alkaloids, stilbenoids, neomycin, vancomycin, and tetracyclines, indicating a great bioactive potential of the bacterial community associated with D. antarctica (Table S5-Supplementary Material). Despite advances in environmental sequencing (metagenomics) and single-cell genomics, the screening procedures of novel molecules remain largely related to the microorganisms culturing 53 . Thus, efforts to isolate new psychrotolerant strains were performed, resulting in a diverse collection of Actinobacteria revealed by the 16S rRNA gene sequencing. However, recent studies have found that discrepancies in terms of chemical and morphological features were possible in microorganisms belonging to same taxonomy, as determined for identical 16S rRNA gene sequences 54 . This finding was especially observed when studying bacteria from geographically distant origins or associated with different host species, in accordance to the ecovar concept [55][56][57] . Therefore, although the limitations of 16S rRNA gene have been widely recognized for screening of bioactive strains, we have assumed that the rhizosphere of D. antarctica could lead to high grade of specialism, with a lot of clonal populations. This fact was also observed in the 16S rRNA sequences of all isolated strains (data not shown), and supported by previous reports of Antarctic microbiota, which demonstrated that the type of habitat dramatically constrained the bacterial community composition 45,58 .
Although the Actinobacteria diversity has been underestimated through cultivation-dependent techniques, the isolation was able to access undiscovered species with the potential to produce novel bioactive compounds [59][60][61] . Some of them were not detected in microbiome analysis, possibly belonging to the rare community. Indeed, previous studies reported that cultivation techniques are still unable to access the real microbial diversity found in natural environments. The cultivation of recalcitrant microorganisms from extreme environments is difficult, especially the rhizosphere of Antarctic plants. However, the use of different culture media and growing conditions may be tested to improve the recovery of a major diversity of Actinobacteria members 14,62 . The low sequence identity registered in this study indicates the potential to identify novel species that require a polyphasic taxonomic approach for proper characterization and description. Based on phylogenetic, phenotypic For current purposes, antiproliferative screening using P. aphanidermatum CMAA 243 T as a model organism resulted in 17 bioactive strains. The bioassay was followed the protocol successfully applied to the primary screening against the Oomycota Pythium, which contain in its hyphal walls and cell membrane, proteins, lipids   64 . Although both strains have been presented in the literature, they have not been listed as potential producers of antitumour compounds. Cinerubin B is an important compound belonging to the anthracycline class. Anthracyclines are a special class of antibiotics widely used as antitumour agents, and the compounds daunomycin, doxorubicin, adriamycin, and aranciamycin anhydride have received great attention since their discovery 65 . The pronounced activity presented in the secondary metabolites of Streptomyces sp. CMAA 1653 can be clearly justified by the presence of actinomycin V (m/z 1,269.6170 u). The product ions at m/z 956, m/z 857, m/z 657, m/z 399 and m/z 300 corresponded to the loss of the Val-Pro-Sar-MeVal chain and were similar to those observed for actinomycin D 44 . Moreover, the product ion at m/z 558 indicated the presence of another Pro-Sar-MeVal residue in structure, as well as the product ion at m/z 459, which was characteristic of the core nucleus structure of actinomycins 44,66 . Actinomycins have been used as chemotherapeutic agents for the treatment of a variety of cancers and are produced by many Streptomyces strains 24 . They are a family of bicyclic chromopeptide lactones that attach two pentapeptide lactones of nonribosomal origin. Its use dates approximately 70 years ago, acting under several types of malignant human tumours, including nephroblastoma (Wilms' tumour) and childhood rhabdomyosarcoma [67][68][69] , and has attracted attention for its potential to assist in the development growth of multi-drug resistance strains and inhibition of HIV-1 reverse transcriptase 44,70,71 . conclusions This study reveals that the root-associated bacteria of Deschampsia antarctica Desv. from Antarctic ecosystems are a rich source of molecules with antitumour properties. The isolation methods employed in this study were able to retrieve Actinobacteria taxa that were not detected in microbiome analysis, showing that the combination of both strategies may be useful for recovering both rare and abundant members of the Actinobacteria communities for biotechnological exploration. The secondary metabolites of Streptomyces sp. CMAA 1527 and Streptomyces sp. CMAA 1653 showed valuable antiproliferative activities against human cancer cells and therefore can contribute significantly to the development of drugs of technological importance. Cinerubin B and actinomycin V were identified in the bioactive fractions of Streptomyces sp. CMAA 1527 and Streptomyces sp. CMAA 1653, respectively, but many other potential bioactive compounds can still be explored from Antarctic Actinobacteria. Based on these results, one of the two native Antarctic plants represents an attractive and unique model for the study of symbiosis and the discovery of antitumour lead compounds by the associated microbiome.   74 . The PCR products were pooled in equimolar ratio and purified by the SizeSelect EX E-Gel electrophoresis system (Life Technologies Corporation) for DNA size selection. The recovered fragments were further purified using the Agencourt AMPure XP kit (Beckman Coulter-Brea, CA, USA). Purified library product was quantified using the Qubit Fluorometric Quantification. The emulsion PCR procedure and sample enrichment were performed on the Ion OneTouch 2 system using the Ion Template PGM OT2 400 kit (Life Technologies Corporation). The V2 316 chip was used for sequencing as instructed in the Ion Torrent platform manual (Personal Genome Machine-PGM, Life Technologies Corporation).

Methods
Statistical and bioinformatic analyses. Raw data obtained from Ion Torrent sequencing were converted to FASTA file and submitted to quality control (QC) using the Galaxy platform 75 , with the following QC parameters: quality score = 25; barcode size = 5; quality window = 50, and maximum number of homopolymers = 6. Sequences with low quality and lengths < 180 bp were removed. After QC, the remaining high-quality reads were analyzed by the Quantitative Insights Into Microbial Ecology (QIIME) version 1.9 software package 36 . Open reference operational taxonomic unit (OTU) were grouped at 97% sequence similarity OTU using UCLUST method and representative sequences of each OTU were taxonomically classified using the PyNAST alignment against GREENGENES (version gg_13_8) 16S reference database 37,76 . Chimeric sequences were detected and removed by the UCHIME algorithm 76 before the construction of the OTU table. Afterward, all OTUs consisting of one single sequence (singletons), chloroplasts and non-bacterial sequences were removed before taxonomic classification.
The accurate predictive analysis was performed to infer the functional contribution of rhizosphere-associated Actinobacteria in comparison with the total community. Then, the PICRUSt tool 41 was used to predict KEEG Orthology (KO) functional profiles of the bacterial community using the 16S rRNA dataset based on OTUs and reference genomes database [38][39][40] . For this, 16S reads affiliated to the Actinobacteria phylum were filtered after Greengenes database annotation. The final output of this workflow was quantified in terms of predicted function abundances per sample per OTU. Subsequently, the data were analyzed by the STAMP (Statistical Analysis of Metagenomic Profiles) software package v. 2.1.3 77 to evaluate biologically meaningful differences between rhizosphere-associated Actinobacteria and the total bacterial community. Statistical inferences were performed by G-test (w/Yates) + Fisher's, DP method: Asymptotic-CC 0.95 and p-value > 0.05 filter.
Isolation, culturing and identification of actinobacteria from rhizosphere. A pool of rhizosphere samples (0.25 g of each sample for three collected points; n = 9) was serially diluted (10 -4 , 10 -5 and 10 -6 ) in saline solution (NaCl 0.85%), and aliquots of 100 μL were plated on Starch Casein Agar (SCA) 78 , Humic Acid-Vitamin Agar (HVA) 79 and Yeast Extract Malt Extract Agar (YEME) 80 . All media were supplemented with cycloheximide and nystatin (25 μg/mL) and incubated at 16 °C for 6 weeks. Isolates from the Deschampsia antarctica rhizosphere with the typical morphology of Actinobacteria colonies were taken from the selective isolation plates. The pure cultures obtained by repeated streaking were maintained on Glucose Yeast Extract Agar (GYEA) at 4 °C and as mixtures of mycelial fragments and spores in 20% glycerol (v/v) at − 80 °C.
Genomic DNA extraction, PCR amplification and sequencing methods were performed according to procedures adopted in   34 . The 16S rRNA consensus contigs were obtained using PhredPhrap/ Consed 81 and were aligned by ClustalW against nearest corresponding sequences using EzBioCloud 82 and MEGA 7.0 42 . Phylogenetic trees were inferred using the neighbour-joining 83 tree-making algorithm drawn from the MEGA 7.0 package. Topologies of the resultant trees were evaluated by bootstrap analysis 84 based upon 1,000 replicates. cultivation, extraction, and isolation of bioactive compounds. The metabolites of Streptomyces sp. CMAA 1527 and Streptomyces sp. CMAA 1653 were obtained by growing cultures in Potato Dextrose Broth on a rotating shaker (180 rpm, 16 °C) for 10 days. After growth, the mycelium was separated by centrifugation (7,000 rpm, 15 min, room temperature) followed by filtration (0.22 μm filter membrane). Each liquid culture medium was extracted with ethyl acetate (1:3 v/v), and the organic phase was evaporated to dryness under vacuum (Büchi Waterbath B-480). From each crude extract, at least three aliquots were separated, and two of them were subjected to antifungal (oomycete Phytium aphanidermatum CMAA 243 T ) and antiproliferative (human tumour and non-tumour cell lines) evaluations. Aliquots of each crude extract (650 mg) were fractionated by semipreparative HPLC on a Shimadzu LC system (Shimadzu, Kyoto, Japan) equipped with a CBM-20A controller, an LC-6AD pump, a DGU-20A5 degasser, and an SPD-20A UV-vis detector. The chromatographic separation was carried out using a C 18 Zorbax Eclipse XDB column (250 × 9.4 mm, 5 μm; Agilent), and the mobile phase used was composed of 0.1% formic acid as system A and acetonitrile as system B. The gradient Scientific RepoRtS | (2020) 10:13870 | https://doi.org/10.1038/s41598-020-69786-2 www.nature.com/scientificreports/ elution programme was 0-40 min from 30 to 100% of system B at a flow rate of 4 mL/min. The chromatograms were monitored using wavelengths at 320 and 500 nm.
Screening of bioactive strains. Antifungal activity. Bioactive strains were screened as described by Santos and Melo (2016) 63 , using Pythium. aphanidermatum CMAA 243 T as a model organism. This bioassay was carried out as a primary screening for antitumour compounds from microorganisms. The antagonism assay was performed using sterile disc paper (0.5 cm diameter), crude extract solubilized in ethyl acetate (1 mg/mL) and potato dextrose agar medium. The plates were incubated at 28 °C for 24 h and evaluated according to the size of the inhibition zone, such as pronounced (+++), moderate (++), reduced (+) or absent (−).
Antiproliferative activity.  85 . First, all extracts were screened against three tumour cell lines (U251, MCF-7 and NCI-H460). Then, the most promising extracts were evaluated against the complete panel. For both experiments, each extract was solubilized in dimethyl sulfoxide (100 mg/ mL) followed by serial dilution in complete medium affording the final concentrations (0.25, 2.5, 25 and 250 μg/ mL). Doxorubicin was used as a positive control at final concentrations of 0.025, 0.25, 2.5 and 25 μg/mL and was diluted following the same protocol.
Each cell line was grown in 96-well plates (100 μL cells/well) for 24 h and exposed for to the extracts, doxorubicin (positive control) or complete medium (negative control) for 48 h. Before (T 0 plate) and after (T 1 plates) sample addition, cells were fixed with 50% trichloroacetic acid dyed with sulforhodamine B 0.4% in acetic acid 0.1%, and colorimetric evaluation was recorded at 540 nm. The difference in absorbance between the untreated cells before (T 0 plate) and after (T 1 plates) sample addition represented 100% cell proliferation. Negative values represented a reduction in the cell population related to the T 0 plate. The cell proliferation data were analysed using Origin 8.0 software (OriginLab Corporation) using sigmoidal regression to calculate the TGI and GI 50 values (sample concentrations required total cell growth inhibition or to elicit 50% of cell death, respectively).
Lc-MS analysis. The compounds present in the bioactive fractions were analysed using a Waters ACQUITY UPLC H-Class system coupled to the Xevo TQ-S tandem quadrupole (Waters Corporation, Milford, MA, USA) mass spectrometer with a Z-spray source operating in the positive mode and to the diode arrangement detector (DAD) from 220 to 600 nm, according to Crevelin et al. (2014) 31 . The samples were solubilized in MeOH and injected (5 µL) into a Zorbax Eclipse XDB-C18 column (150 × 4.6 mm, 3.5 µm particle size-Agilent, Santa Clara, CA, USA); the mobile phase used for gradient elution consisted of 0.1% formic acid as system A and acetonitrile with 0.1% formic acid as system B. The flow rate was 0.5 mL/min, and the gradient elution programme started with 30% B, increased to 90% B in the following 25 min, remained at 90% B for 5 min, and returned to the initial condition within the following 5 min. The source and operating parameters were optimized as follows: capillary voltage, 3.2 kV; cone voltage, 40 V; source offset, 60 V; Z-spray source temperature, 150 °C; desolvation temperature (N 2 ), 350 °C; desolvation gas flow, 800 L/h (mass range from m/z 150 − 1,200).
High-resolution mass spectrometry analysis. The exact mass of the compounds present in the bioactive fractions was determined by high-resolution mass spectrometry (HRESIMS) 31,86 on a mass spectrometer microTOFII-ESI-Q-TOF (Bruker Daltonics, Billerica, MA, USA) employing an infusion pump (Kd Scientific, Holliston, MA, USA) at a flow rate of 100 µL/min. The voltage of the capillary was 3,500 V, and the voltage of the end plate was − 400 V in positive ionization mode. The source and operating parameters were optimized as follows: drying gas temperature (N 2 ), 250 °C; a flow rate of 4 mL/min and a pressure of 0.4 bar. The tandem mass spectrometry experiments (MS/MS) with collision-induced dissociation were carried out using N 2 as the collision gas on the selected precursor ion at collision energy values ranging from 10 to 50 eV. For internal calibration, a solution of sodium trifluoroacetic acid (Na-TFA) at a concentration of 1 mg/mL was used. Data acquisition and analysis were performed using Compass Data Analysis 4.1 software (Bruker Daltonics Corporation). nMR analysis. Nuclear magnetic resonance (NMR) measurements were carried out on a Bruker DRX 500 spectrometer (Bruker Daltonics Corporation) operating at 500 MHz for 1 H and 125 MHz for 13 C (δ in parts per million relative to Me4Si, J in Hertz). The isolated sample (5 mg) was dissolved in deuterated chloroform (CDCl 3 , Sigma-Aldrich) and analysed by 1D and 2D NMR using a Shigemi 5 mm NMR microtube, as described in   86