All experimental work was performed under standard laboratory safety conditions, and all radiological, chemical, and biological safety precautions were observed following rules and regulations established for respective research institutions.

The ascomycetous and basidiomycetous yeasts used in this study and their isolation sites are presented in Table 1.

Sixty environmental samples were collected between 2001 and 2015 as a part of a larger study. These samples represent desert sands (Arizona, Nevada, New Mexico); dried plant debris from deserts (Arizona, Nevada, New Mexico); water, sediments, and soil from abandoned mines and mine drainages (coal mine in Maryland, silver and gold mines in Colorado, mercury mine in Idrija, Slovenia); hot springs (Colorado; Radenci, Slovenia); water and sediments from acidic river (Rio Tinto, Spain); and radioactive waste storage tanks (Uniformed Services University of the Health Sciences, Maryland). One gram of each environmental sample was resuspended in 10 ml of MQ purified water and allowed to settle for 30 min. One milliliter of the supernatant was added to 10 ml of the oligotrophic medium AM (complex Acidiphilium Medium) (San Martin-Uriz et al., 2014) adjusted to pH 2.3 with HNO₃, and incubated in a shaker incubator (200 rpm) at 25°C for 4 days. One hundred microliters were then spread on AM plates (pH 2.3) and incubated at 25°C under 36 Gy/h. After 3 days of continuous CIR, the plates were inspected for growth. Single colonies were re-inoculated on fresh AM solid medium.

Throughout this work, CIR exposures specified under 36 Gy/h (~22°C) were performed in a ¹³⁷Cs irradiator (GammaCell 40, J. L. Shepard and Associates). For all other CIR exposures, we used a second adjustable dose rate ¹³⁷Cs irradiator (Mark 1 Model 68 A, J. L. Shepard and Associates), also at ~22°C. Acute exposures were performed in a ⁶⁰Co irradiator (10 kGy/h) (J. L. Shepard and Associates) at 0°C.

The minimum pH and the highest Hg²⁺, merbromin, Cr⁶⁺, and Cr³⁺ concentrations supporting growth were determined in liquid AM and Yeast Extract-Peptone-Dextrose (YPD)² medium. The overnight (O/N) cultures pre-grown at optimal temperatures were washed twice in sterile MQ and used to inoculate fresh liquid media adjusted for pH with HNO₃, and/or supplemented with different concentrations of heavy metals to a final OD₆₀₀ ~0.1. The strains were incubated in a shaker incubator, 200 rpm, at optimal temperatures. After inoculation, the OD₆₀₀ was measured every 24 h for 1 week.

Maximum growth temperature and anaerobic growth were determined by observing colony formation on solid YPD medium incubated at various temperatures (25–65°C; temperature maxima); and for anaerobic growth, at a given strain's optimal temperature, in the presence or absence of atmospheric oxygen for 1 week.

Survival following acute forms of γ-radiation was determined on solid YPD medium by colony forming unit (CFU) assay as described previously (Daly et al., 2004). The ability of cells to grow under CIR on YPD pH 7.0 and AM pH 2.3 was monitored visually. The ability of a strain to form biofilms was tested in 96-well microtiter plates, as described by others (O'Toole, 2011), with eight replicate wells for every strain and each condition, and eight wells for blank controls. Pulsed-field gel electrophoresis (PFGE) with MD1149 genomic DNA was performed as described previously (Saracli et al., 2003).

The OD₆₀₀ of an O/N culture of MD1149 in the Yeast Mold Broth (YM, Difco) was adjusted to 0.05 in modified Hommel's Minimal Salts (HMS; 0.3% (NH₄)₂SO₄, 0.05% NaCl, 0.07% MgSO₄, 0.04% Ca(NO₃)₂, 0.04% K₂HPO₄, 0.25% KH₂PO₄, 0.06% yeast extract, 5% glucose). At the indicated time (2, 4, 6, and 8 days), OD₆₀₀ was assessed, and 10 ml of culture were centrifuged twice at 5,000 × g to obtain spent liquid medium (SLM).

Organic acids in SLM were identified and measured using a Waters Xevo G2-XS QTOF mass spectrometer (Waters Corporation, Milford, MA USA) coupled with a Waters Acquity H Class chromatography system. Organic acids were separated on a Waters Acquity UPLC HSS C18 1.8 μm 2.1 × 100 mm column using a modification of a previously published method (Fernández-Fernández et al., 2010). Mobile phases were methanol (solvent A) and water with 0.5% formic acid (solvent B). The separation method was as follows: initial, 90% B; 0.1 min, 90% B; 6 min, 70% B; 6.1 min, 90% B; 12 min, 90% B. The flow rate was 125 μl/min. The column compartment thermostat was set at 35°C, and the autosampler tray temperature was maintained at 4°C. Detection was accomplished by mass spectrometry with the electrospray ion source operating in negative ion, resolution mode. Data acquisition was performed using MassLynx Version 4.1 data acquisition software (Waters Corp.), with MS^e data-independent centroid acquisition and leucine enkephalin lockmass correction.

Liquid chromatography mass spectrometry (LC-MS) and liquid chromatography tandem-mass spectrometry (LC-MS/MS) settings were as follows: Low Energy, 50–1000 Da, 6 V collision energy; High Energy, 50–1000 Da, collision energy ramp from 10–40 V; Scan time: 0.5 s; Source: 120°C; Desolvation, 450°C; Desolvation gas flow: 800 l/h; capillary voltage: 1.90 kV; Sampling cone voltage: 40 V. Experimental samples were compared with authentic standards (Sigma-Aldrich, St. Louis, MO) to identify organic acids present in cell culture media. Quantitation was performed with Waters TargetLynx software with quadratic calibration curve fitting.

MD1149 was first identified at the genus level based on micro- and macro morphology and assimilation test (YT MicroPlate™, BIOLOG Inc.), and then to the species level using genetic molecular identification (Mohamed et al., 2014).

Total DNA was isolated from MD1149 using the Wizard Genomic DNA Purification Kit (Promega, Madison, WI, USA) and quantified by NanoDrop 2000 (Thermo Fisher Scientific).

The ITS1-5.8S rDNA-ITS2 and 18S rDNA sequences were matched to the GenBank non-redundant nucleotide database with the BLASTN algorithm (Altschul et al., 1990). MD1149 and related sequences were analyzed for similarity within the Geneious software package (Kearse et al., 2012) by using MUSCLE alignment (Edgar, 2004). The aligned sequences of representative strains were used to construct a phylogenetic tree with the PhyML 3.0 software (Guindon et al., 2010) with approximate likelihood-ratio test for branch supports, and with six substitution rate categories. The substitution model, alpha parameter of the gamma distribution and the proportion of invariable sites, was estimated by jModelTest 2.0 (Darriba et al., 2012).

The draft genome was generated using a combination of Illumina and 454 technologies. Two short-insert paired-end libraries, a fragment, 625-bp insert size (2 × 300 bp reads) and an overlapping fragment, 405-bp insert size (2 × 300 bp reads) were sequenced using version 3 chemistry on the MiSeq (Illumina, Inc., San Diego, CA, USA) (Bennett, 2004). Two large-insert paired-end libraries (8-kbp and 20-kbp insert size) were constructed and sequenced on the 454 GS FLX (Roche/454 Life Sciences, Branford, CT, USA) (Margulies et al., 2005). The draft data was assembled de novo with CLC Genomics Workbench v9.0 (QIAGEN Aarhus, Denmark). Repetitive sequences were identified using RepeatMasker (Smit et al., 2013–2015) and RepBase library (Jurka et al., 2005). The genome assembly completeness was evaluated with the Benchmarking Universal Single-Copy Orthologs (BUSCO 1.22) (Simão et al., 2015) software using the dataset for fungi.

For pairwise genome alignments, the following genomes were used: Rhodotorula sp. (Goordial et al., 2016), R. mucilaginosa (Deligios et al., 2015), R. glutinis (Paul et al., 2014), R. toruloides (Zhang et al., 2016), R. graminis (Firrincieli et al., 2015), Puccinia graminis (Duplessis et al., 2011). The genome alignments of contigs longer than 100 kbp were calculated with the PROmer algorithm, as implemented in MUMmer 3.23, and plotted with the MUMmerplot utility (Kurtz et al., 2004) as described by Hane et al. (2011).

MD1149 total RNA was isolated, then pooled and sequenced from cells grown under the following conditions: YPD (25°C, O/N and 3 days), YNB + 2% glucose (25°C, O/N and 3 days), YNB + 2% glucose (6°C and 37°C, O/N), YNB + 2% glycerol (25°C, O/N), AM pH 2.3 (25°C, O/N, 0 and 36 Gy/h), and AM pH 7 (25°C, O/N, 0 and 36 Gy/h). The RNA was isolated from MD1149 using RiboPure RNA Purification Kit for Yeasts (Thermo Fisher Scientific). RNA integrity was assessed by Fragment Analyzer (Advanced Analytical Technologies Inc., Ankeny, IA, USA). Sequencing libraries were prepared from 500 ng of total RNA input using the TruSeq Stranded mRNA Library Preparation Kit (Illumina) with barcoded adapters. Sequencing libraries yield and concentration were determined using the Illumina/Universal Library Quantification Kit (KAPA Biosystems, Wilmington, MA, USA) on the Light Cycler 480 (Roche Diagnostics Co., Indianapolis, IN, USA). Library size distribution was determined using the Fragment Analyzer™ (Advanced Analytical Technologies Inc.). Clustering and sequencing was performed on the NextSeq 500 (Illumina) with paired-end reads of 75 bp length.

RNAseq reads were quality trimmed with Sickle (Joshi and Fass, 2011) and aligned to the assembled genome with TopHat 2.1.1 (Kim et al., 2013). The alignment was then used for the transcriptome assembly with Trinity 2.2.0 (Haas et al., 2013) in Genome Guided mode with jaccard clipping and a maximum intron length of 1,500 bp. Protein-coding and tRNA genes were annotated using MAKER 2.31.8 (Campbell et al., 2014). The complete Swissprot database was used as evidence, along with the database of BUSCO, a set of Basidiomycete fungal proteomes and the sequenced transcriptome of MD1149. Three gene predictors were used in the MAKER pipeline: SNAP (Korf, 2004; Campbell et al., 2014), GeneMark-ET (Lomsadze et al., 2014), and Augustus (Stanke and Waack, 2003). This Whole Genome Shotgun project has been deposited at DDBJ/ENA/GenBank under the accession PJQD00000000. The version described in this paper is version PJQD01000000.

Predicted protein sequences from the MD1149 genome were processed through automated functional annotation using the PSAT metaserver (Leung et al., 2016), which ran EFICAz 2.5 (Kumar and Skolnick, 2012), blastp against the KEGG, MetaCyc, BRENDA, and STRING databases (Caspi et al., 2014; Chang et al., 2015; Szklarczyk et al., 2015; Kanehisa et al., 2016), and Interproscan (Jones et al., 2014). Additionally, protein sequences were processed through online servers running SignalP 4.0 (Petersen et al., 2011) and TMHMM 2.0 (Krogh et al., 2001). The pepstats utility (EMBOSS suite) was used to calculate protein properties (Rice et al., 2000).

GO annotations were also quantified and then projected into the full GO hierarchy using slim-o-matic (Courtot et al., 2016) and reviewed using Protege (Munsen, 2015). High-level categories were selected, into which the MD1149 GO annotations were mapped (slimmed). Categories corresponding to GO molecular function and biological process were quantified and expressed as percent of annotations across the MD1149 genome.

Predicted proteins from MD1149, R. graminis, R. sp. JG-1b, and R. toruloides were compared by all-against-all blastp at identity cutoff 95% and query coverage ≥95% using CGP³, and post-processed to identify fasta sequences unique or in common among the species and putatively duplicated within each species. Sequence logos of conserved nucleotide positions in all introns of median length were drawn using WebLogo 3 (Crooks et al., 2004).

In order to find a suitable candidate for bioremediation of acidic radioactive environmental waste sites, we first screened a variety of aquatic and terrestrial environments (desert sands, acid mine drainages, soils, and water samples) for strains that are both acid- and CIR-resistant. This strategy yielded only one strain, named MD1149, isolated from a sediment sample from an abandoned acid mine drainage facility in Maryland, USA (39°31′34.22″N, 79°1′12.16″W). MD1149 is a red-pigmented, unicellular, non-sporulating, ovoidal, obligately aerobic, budding yeast (Figure 1A, Table 1), which became pleomorphic under 36 Gy/h (Figure 1B). Phylogenetic analysis based on the internal transcribed spacer (ITS) and small subunit rRNA (SSU) sequences identified MD1149 as the basidiomycetous yeast R. taiwanensis [closely related to the type strain BCRC 23118(T) = CBS 11729(T)] (Figure 2), and confirmed by its micromorphological, macromorphological, and physiological characteristics (data not shown). Based on PFGE, the genome size of MD1149 was estimated to be >13 Mbp (Figure 1F). MD1149 was deposited with the Microbial Culture Collection EX as EXF-12971. Since other environmental samples screened did not yield additional acid- and CIR-resistant strains, we extended our study by including 26 distinct yeasts from the Microbial Culture Collection EX (Table 1).

Like most of the tested yeasts in this study, MD1149 was capable of growing luxuriantly at pH 2.3 and 7.0 under 36 Gy/h (Table 1). However, it was the only strain capable of growth under 66 Gy/h (Figure 1C).

Survival assays yield a radiation resistance metric named D₁₀, which represents the acute radiation dose (Gy) giving 10% CFU survival (Daly et al., 2007, 2010; Sharma et al., 2017). Among tested strains, the most resistant was Saccharomyces cerevisiae EXF-5294 (D₁₀, 3.2 kGy), and the most sensitive was Debaryomyces hansenii (D₁₀, 0.3 kGy). The D₁₀ of MD1149 is 2.5 kGy and ranks among the most radiation-resistant yeasts identified both for acute and chronic exposures (Table 1). Importantly, the radiation resistance of MD1149 increased with decreasing pH, from D₁₀ 0.8 kGy at pH 7.0 to D₁₀ 2.5 kGy at pH 2.3 (Figure 1D).

Optimal and maximum growth temperatures for the strains are reported in Table 1. Pichia kudriavzevii was the most thermotolerant and could grow at 45°C. The temperature maximum of MD1149, which optimally grows at 20–25°C, was 32°C. The most temperature-sensitive strains were R. benthica and R. larynges, which could grow at 25°C or below.

Over the course of 7 days, growth of strains in YPD and AM media adjusted to different pH values was montiored spectrophotometrically. Growth was considered as increasing when the OD₆₀₀ rose above 0.1. The pH minima for growth of the yeasts are presented in Table 1. A full pH-dependent growth response curve for MD1149 is presented (Figure 1E). As shown in Table 1, the pH minima supporting growth of the yeast in rich (YPD) or oligotrophic (AM) media were very similar. Rhodotorula calyptogenae was the only strain that could not grow in YPD at low pH, whereas it grew well in AM at pH 2.5. Remarkably, growth responses of MD1149 and Pichia kudriavzevii in AM and Rhodosporidium kratochvilovae in YPD showed that their pH-minima approximate 1.5 (Figure 1E, Table 1).

The most common metal contaminants at DOE sites are U, Sr, Cs, Tc, Cr, Pb, and Hg. Among these, U, Tc, Hg, and Cr are significantly less mobile when reduced, and are capable of being immobilized by microorganisms (Daly, 2000). We tested yeasts for their resistance to Hg and Cr: mercury in the form of HgCl₂ (Hg²⁺) and merbromin (organo-Hg), and chromium in the form of CrCl₃ (Cr³⁺) and K₂Cr₂O₇ (Cr⁶⁺). Table 1 summarizes heavy metal tolerances for strains grown in oligotrophic medium (AM) instead of YPD; YPD contains phosphates and myriad small organic molecules (e.g., peptides) that can mask metal toxicity (e.g., Mergeay, 1995). As expected, Hg²⁺ in HgCl₂ is considerably more toxic than Cr³⁺ and Cr⁶⁺. The two strains most resistant to Hg²⁺, Cr³⁺, and Cr⁶⁺, were MD1149 and R. kratochvilovae (Figure 3 and Table 1), which could grow in AM supplemented with 50 μM HgCl₂, and at significantly higher concentrations of Cr³⁺ and Cr⁶⁺. In contrast, most strains were resistant to millimolar concentrations of Hg when added as merbromin. Wickerhamomyces anomalus and Candida pseudolambica could grow in liquid medium supplemented with 3 mM Cr³⁺, whereas MD 1149 was resistant only to 0.5 mM Cr³⁺ (Table 1). The growth responses of MD1149 in AM supplemented with increasing concentrations of HgCl₂ or K₂Cr₂O₇ were distinct. Whereas increasing the concentration of Hg²⁺ increased the length of the lag-phase before the onset of exponential growth, increasing the concentration of Cr⁶⁺ slowed the growth of MD1149 (Figures 3A,B). Furthermore, in contrast to Cr⁶⁺/Cr³⁺, we showed that Hg²⁺ had a significant detrimental effect on MD1149 growth under CIR or not (Figures 3C,D).

Biofilms are very important in bioremediation, since they offer sorption sites for many divalent cations that are toxic, and thus prevent their migration in the environment. The biofilm-forming capacity in yeasts was estimated with crystal violet assay (O'Toole, 2011) after 24 h incubation (in case of MD1149 it was additionally monitored over 5 days) at pH values 2–6, in the presence and absence of chronic gamma-radiation (36 Gy/h), and in oligotrophic (AM) and rich medium (YPD). This assay was performed on 8 parallels for each strain. The results are summarized in Figures 4, 5. The average absorbance of negative control (no inoculum) was subtracted from each measurement. Difference in A₅₇₀ between the negative control and the sample above 0.2 was interpreted as an indication of biofilm formation.

Out of 27 strains, 3 strains were unable to form biofilms: Cryptococcus liquefaciens, Debariomyces hansenii, and S. cerevisiae EXF-5281. For the remaining strains, biofilm forming capacity was strongly dependent on the species and physical parameters. For most strains, biofilm formation was inhibited by CIR. However, under specified conditions, biofilm formation in 7 species (C. pseudolambica, M. fruticola, P. kudriavzevii, R. benthica, S. bayanus, S. cerevisiae, and S. kudriavzevii) was moderately enhanced under CIR, based on A₅₇₀ values (Figure 5). In the absence of CIR, pH values below 4 stimulated biofilm formation in 4 yeasts, but inhibited biofilms in the remainder. As pH values decreased to 2.3, in the presence or absence of CIR, MD1149 increasingly formed dense biofilms (Figure 4).

While monitoring the growth of MD1149 in YM medium (Figure 6A), we noted unusually high OD₆₀₀ values in stationary phase cultures compared to growth in YPD or AM. This was not due to high cell concentrations in YM, but instead was caused by secretion of metabolites that absorb at 600 nm, which accompanied the drop in pH from 6 to 2.0–2.5 (Figure 6A). These metabolites are now being further investigated.

The rapid pH drop suggested that MD1149 produced significant quantities of organic acids, theoretically in excess of 10 mM depending on pK_a values of the acids present. Therefore, we analyzed the SLM by LC-MS to identify any excreted organic acids. We detected the presence of at least six organic acids by LC-MS and LC-MS/MS as well as elemental composition prediction using Waters MassLynx 4.1 software. These acids included citric, homoaconitic, homocitric/homoisocitric (constitutional isomers), malic, rhodotorulic, and succinic. It is noteworthy that organic acids with available reference spectra (citric, malic, and succinic) matched the precursor and product ions from LC-ESI-QTOF (liquid chromatography-electrospray ionization-quadrupole-time of flight-mass spectrometry) spectra published in MassBank (Horai et al., 2010). Furthermore, we procured standards for three of the six organic acids detected (citric, malic, and succinic) and quantitated their abundance (Figure 6B). The combined total concentration of these three acids was ~4 mM, consistent with the idea that these three acids contributed to the decrease in media pH, and that the other identified organic acids (which we were unable to characterize) contribute to the full pH change.

The estimated size of the genome assembly (without mitochondrial DNA) was 19.58 Mbp, and the final assembly of 181 scaffolds is based on 19.935 Gbp of draft sequence data, which provides 230× coverage of the genome. We identified a total of 26 scaffolds containing either 5′ or 3′ tandem DNA repeats with a sequence TTAGGG, which correspond to the most prevalent telomeric repeats (Teixeira and Gilson, 2005). Based on this result, we can conclude that the genome of MD1149 is organized in at least 13 chromosomes.

The size of the assembled mitochondrial genome was 38.20 kbp, slightly less than 40.39 kb reported by Zhao et al. (2013a). An alignment of both mitochondrial genomes showed that the sequences were largely syntenic (Figure 7D).

The content of GC pairs was 40.85% in the mitochondrial genome and 61.69% in the nuclear DNA. This finding is comparable to R. glutinis (61.87%) and R. mucilaginosa (60.54%), but lower than in R. graminis (67.76%), which has one of the most GC-rich genomes among available fungal genomes. The number of repetitive sequences was relatively low at 1.49%.

The number of genes annotated in the genome was 7,122. The genome completeness was estimated by searching the predicted proteome for 1,438 groups of BUSCO. We found 91% complete matches, 7% were fragmented and 2% were missing. More than 97% of MD1149 genes contained introns (Table 2), with an average of 6.2 exons and 5.2 introns per gene (Figure 8C, Table 2) (R. graminis: 6.2). The median length of introns was 69 bp (R. graminis: 101 bp), and they contained the typical 5′ and 3′ consensus sequences (Figures 8A,B). The median length of the exons was 151 bp (Figure 8D). The average length of the predicted proteins was 531, and their amino acid composition and isoelectric points were comparable to those of other Rhodotorula spp. (Figure 9).

When compared to related species, the distributions of gene families were similar. Only 71 predicted MD1149 proteins and 4 OrthoMLC groups were unique (Figures 7A,B). Although the number of duplicated genes in MD1149 was similar to that in Rhodotorula sp. JG-1b, it was much lower than in R. graminis and R. toruloides (Figure 7C). GO-slim analysis revealed the expected distribution of functional categories (Figures 7E,F) for MD1149 genes. To better understand the remarkable radiation resistance of MD1149, we further analyzed the genome for the presence of genes involved in homologous DNA recombination, non-homologous end joining, oxidative stress response, Mn homeostasis, heavy metal resistance, and hydrolases; results are presented (Table 3). The set of genes and their copy number are comparable to other fungi.

The pairwise genome alignments showed a high level of macrosynteny between MD1149, R. mucilaginosa and Rhodotorula sp. JG-1b (Figure 10). With R. toruloides, R. glutinis, and R. graminis the order of the alignable regions was mixed, but the genomic rearrangements appear to have occurred within the same DNA molecules and not between them (Figure 10). This form of evolution is known as mesosynteny, and it was previously thought to be restricted only to filamentous ascomycetes (Hane et al., 2011).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.