Transcriptional response to prolonged perchlorate exposure in the methanogen Methanosarcina barkeri and implications for Martian habitability

Observations of trace methane (CH4) in the Martian atmosphere are significant to the astrobiology community given the overwhelming contribution of biological methanogenesis to atmospheric CH4 on Earth. Previous studies have shown that methanogenic Archaea can generate CH4 when incubated with perchlorates, highly oxidizing chaotropic salts which have been found across the Martian surface. However, the regulatory mechanisms behind this remain completely unexplored. In this study we performed comparative transcriptomics on the methanogen Methanosarcina barkeri, which was incubated at 30˚C and 0˚C with 10–20 mM calcium-, magnesium-, or sodium perchlorate. Consistent with prior studies, we observed decreased CH4 production and apparent perchlorate reduction, with the latter process proceeding by heretofore essentially unknown mechanisms. Transcriptomic responses of M. barkeri to perchlorates include up-regulation of osmoprotectant transporters and selection against redox-sensitive amino acids. Increased expression of methylamine methanogenesis genes suggest competition for H2 with perchlorate reduction, which we propose is catalyzed by up-regulated molybdenum-containing enzymes and maintained by siphoning diffused H2 from energy-conserving hydrogenases. Methanogenesis regulatory patterns suggest Mars’ freezing temperatures alone pose greater constraints to CH4 production than perchlorates. These findings increase our understanding of methanogen survival in extreme environments and confers continued consideration of a potential biological contribution to Martian CH4.

Mg(ClO 4 ) 2 , and Na(ClO 4 ) reduced total net CH 4 production by 48%, 32%, and 24%, respectively, relative to the perchlorate-free control (Fig. 2). Significant reduction in CH 4 production rates were observed across all conditions at 0 °C with respect to their 30 °C counterparts, and each perchlorate condition at 0 °C showed a statistically significant decrease in CH 4 for at least one time point relative to the 0 °C control (Fig. 2). No CH 4 production was observed in the media blank controls (data not shown). Cultures were monitored via optical density measurements at 600 nm (OD 600 ), but perchlorate-amended media experienced precipitate formation which made obtaining reliable growth data difficult (Fig. S3).
Energy conserving hydrogenase subunit F, echF, which supplies reduced ferredoxin to fmd, was down-regulated in the presence of Mg(ClO 4 ) 2 and Na(ClO 4 ) at 30 °C, whereas most subunits of ech were significantly up-regulated in 30 °C Ca(ClO 4 ) 2 conditions (Fig. 3). Among other hydrogenases demonstrating significant up-regulation was methanophenazine-dependent hydrogenase, vht, specifically, the large subunit vhtA and cytochrome b subunit vhtC (Fig. 3). At 0 °C, ech hydrogenases were not differentially regulated in the presence of perchlorates relative to the perchlorate-free control (Fig. 3).
Despite H 2 being the only reducing equivalent provided for the production of CH 4 in our incubations, the addition of Ca(ClO 4 ) 2 , Mg(ClO 4 ) 2 , and Na(ClO 4 ) resulted in significant up-regulation of all genes in the mono-, di-, and trimethylamine pathways and associated membrane permeases in the 30 °C perchlorate treatments (Fig. 3, Tables S7, S9, S11).
The terminal step encoding methyl-coenzyme M (CH 3 -CoM) reductase subunit alpha (mcrA) was down-regulated at 30 °C in Mg(ClO 4 ) 2 and Na(ClO 4 ) enrichments (Fig. 3, Table 1). Although this reduction in expression Table 1. Differential expression (Log 2 -fold change ± standard error) of M. barkeri MS genes implicated in the Wood-Ljungdahl pathway and hydrogenotrophic methanogenesis. LFC is relative to 30 °C perchloratefree control. Multiple entries indicate statistical significance in multi-copy genes (Wald test, p < 0.05). Abbreviations: n.s.d., no significant difference from control; W, tungsten active site; Mo, molybdenum active site. If gene is not listed, no significant differential expression was observed across all treatments. Full names of genes are referenced in the text and are also listed in Table S12. www.nature.com/scientificreports/ is consistent with the decreased CH 4 production rates observed in these treatments (Fig. 2), the Ca(ClO 4 ) 2amended M. barkeri MS, which generated the least CH 4 , showed no significant difference in expression of the mcr complex relative to the 30 °C perchlorate-free control (Fig. 3, Table 1). Furthermore, no elements of mcr were significantly differentially expressed at 0 °C in the perchlorate-free control relative to the 30 °C perchlorate-free control (Fig. 3) despite its far lower CH 4 production rate (Fig. 2). Therefore, the decreased expression of mcrA does not appear to be sufficient to explain the associated decrease in CH 4 production rates with perchlorate exposure at 30 °C or incubation at 0 °C. The carbon monoxide dehydrogenase/acetyl-CoA synthase complex (CODH/ACS), which plays a key role in both energy conversation and carbon fixation via the Wood-Ljungdahl pathway, demonstrated significant regulatory changes as a function of perchlorate exposure and temperature. The cooS subunit of CODH, which reversibly converts CO and CO 2 , was up-regulated in 30 °C Mg(ClO 4 ) 2 and Na(ClO 4 ) treatments (Fig. 3, Table 1). Carbon monoxide dehydrogenase subunit epsilon (cdhε), which recycles ferredoxin in the reversible conversion between CO and CO 2 , was up-regulated with Ca(ClO 4 ) 2 at 30 °C (Fig. 3, Table 1), but was down-regulated in the 0 °C Mg(ClO 4 ) 2 and Na(ClO 4 ) treatments (Fig. 3, Table 2). The 30 °C Ca(ClO 4 ) 2 treatment also demonstrated  Tables 1 and  2). Differential expression of genes in 0 °C perchlorate-amended transcriptomes are relative to the expression patterns of the 0 °C perchlorate-free control. Perchlorate-amended 30 °C and 0 °C perchlorate-free control transcriptomes are relative to the expression patterns of the 30 °C perchlorate-free control. Crossed white boxes indicate no statistically significant difference in expression. Heatmaps are excluded from genes exhibiting no significant differential expression across all conditions. Genes involved in non-hydrogenotrophic (H 2 /CO 2 ) pathways are encompassed in colored boxes with pathway directionality indicated by matching arrowheads (green: methylamines; pink: acetate; yellow: methanol). Full names of listed genes and metabolites are found in Tables S13 and S14, respectively. www.nature.com/scientificreports/ significant up-regulation of 5-H 4 SPT:corrinoid Fe-S protein methyltransferase (cdhγ), which plays a key role in the generation H 4 SPT and acetyl-CoA for biomass synthesis in the Wood-Ljungdahl pathway (Fig. 3, Table 1). Both copies of this gene were down-regulated at 0 °C in the Mg(ClO 4 ) 2 and Na(ClO 4 ) treatments (Fig. 3, Table 2). Both alpha and beta chains of ACS, respectively encoded by cdhα and cdhβ, were down-regulated in the 0 °C Na(ClO 4 ) treatment with respect to the 0 °C perchlorate-free control, while only cdhα was down-regulated in the 0 °C Mg(ClO 4 ) 2 treatment (Fig. 3, Table 2). Cdhδ, which encodes an iron-sulfur corrinoid protein, was also down-regulated at 0 °C in Mg(ClO 4 ) 2 and Na(ClO 4 )-supplemented incubations (Fig. 3, Table 2). We observed no significant differential expression of CODH/ACS complex genes in the 0 °C Ca(ClO 4 ) 2 treatment (Fig. 3).
We observed substantial up-regulation of ammonium transporters (amt) in the 30 °C perchlorate treatments, but no significant differences in amt expression were observed in any 0 °C treatments (Fig. 4). Ammonium measurements confirmed that bioavailable fixed N remained replete (mM-level concentration) across all conditions for the duration of the incubation (Table S3).
Down-regulation of sulfur-containing amino acids. In addition to the 20 common amino acids, M.
barkeri also encodes a 21 st residue, pyrrolysine, via the 'amber' stop codon UAG 63 . A comparison of the genes encoding amino acid synthesis proteins showed large negative log 2 -fold changes at 30 °C in Mg(ClO 4 ) 2 and Na(ClO 4 )-amended treatments with respect to cysteine-producing proteins cysteine synthase (cysK) and serine acetyltransferase (cysE) (Fig. 5a). Further examination of complete amino acid metabolic pathways (Figs. S4-S16) revealed that this pattern of substantial gene down-regulation was characteristic of not only cysteine, but also the other sulfur-containing amino acid methionine (Fig. 5b).

Discussion
In this study we took our inspiration from nearly two decades of research into Martian methane and associated astrobiology analogue studies to investigate how methanogens may survive under challenging conditions analogous of the Martian shallow subsurface, specifically under freezing temperatures and in the presence of chaotropic perchlorate salts. We tracked CH 4 production and changes in global gene expression in M. barkeri MS via RNA-Seq following 28 days' incubation at 30 °C or 0 °C, with and without 10 mM of dissolved Naperchlorate and 20 mM of dissolved Mg-or Ca-perchlorate. In accordance with prior findings 25,26 we observed quantifiable but inhibited methanogenesis in cultures supplemented with perchlorates (Fig. 2). For the first time we present transcriptomic evidence of the underlying biochemistry. Regulatory responses point to unambiguous shifts in amino acid synthesis and recycling, as well as mechanisms to combat increased osmotic stress-not unexpected reactions for an obligate anaerobe exposed to strongly oxidizing and chaotropic perchlorate salts. A surprising regulatory switch was observed in the methanogenesis pathway, with a significant up-regulation of methylamine-utilizing genes, despite these substrates not being present in the media ab initio. Metalloenzymes with molybdenum active sites, including (Mo)-formylmethanofuran dehydrogenase and (Mo)-Nase were among the most significantly up-regulated genes in perchlorate-amended incubations at 30 °C. We contextualize these  Evidence for selection against redox-sensitive amino acids under temperate conditions. Cysteine and methionine are exceptionally sensitive to oxidation by reactive radical species 64 . Both residues have shown strong binding affinities to both perchlorate and perchloric acid, resulting in the oxidation of methionine to methionine sulfoxide and cysteine to sulfonic acid 65 . The susceptibility of cysteine and methionine to react with perchlorate risks degradation of protein structure and function. The extensive and substantial down-regulation of cysteine and methionine metabolic pathways we observed in the presence of Na-and Mgperchlorates at 30 °C (Fig. 5) suggests a concerted effort by M. barkeri to reduce the synthesis and repair of these residues in the presence of these perchlorate species. A dearth of significant differential expression of amino acid-recycling genes (including those specific for cysteine and methionine) in both 0 °C and 0 °C perchlorate treatments demonstrates that freezing conditions are not sufficient to confer a decrease in transcriptional activity. It is clear that the synthesis and processing of essential amino acids in M. barkeri must proceed despite the metabolic stresses of frigid temperatures and strong oxidants.
A role for glycine betaine in cryo-and osmoprotection, and as a potential methylamine precursor. The regulatory patterns of enzyme complexes protecting against osmotic stress like opuAABC coincide with temperature and redox conditions. This operon is up-regulated at 30 °C with Na(ClO 4 ) and Mg(ClO 4 ) 2 , implying an increased need for cellular machinery to import glycine betaine to combat osmotic stress. In addition to being an osmoprotectant, glycine betaine is also cryoprotective 66 . While perchlorate salts are highly oxidizing, they also depress the freezing point of water. The down-regulation of opuAB in Mg(ClO 4 ) 2 -amended cultures at 0 °C might suggest a decreased cold shock response in M. barkeri, wherein glycine betaine uptake for its cryoprotective properties was not needed under freezing conditions. The up-regulation of methylamine-specific methanogenesis pathway genes in all three 30 °C perchlorate treatments was unexpected. Methylamines were not present in the media ab initio and ion chromatography mass spectrometry did not reveal the presence of monomethylamine in assessed samples (Table S3)-though if samples had ppb-levels of CH 3 NH 2 , it was diluted below the instrument's detection limit (see Materials and Methods). Free energy yields of methylamine methanogenesis reactions at 30 °C range from -91 to -143 kJ/ mol CH 4 -significantly less than the − 158 kJ/mol CH 4 of the hydrogenotrophic pathway (Table S4). It has been suggested that glycine betaine may be a potential precursor of trimethylamine (TMA) 67,68 . The observed up-regulation of monomethylamine and dimethylamine permeases mtmP and mtbP (Fig. 3) imply possible interactions between perchlorates and complex media constituents (e.g., perhaps yeast extract or casitone) that might yield methylaminated species for utilization by M. barkeri, though we acknowledge such interpretations are presently speculative as our ion chromatography analyses were not sensitive to the detection of dimethylamine, trimethylamine, or any other methylaminated species.
Up-regulation of Mo-containing enzymes suggest mechanism for H 2 -dependent perchlorate reduction. Fixed nitrogen was replete under initial incubation conditions (9.3 mM NH 4 Cl plus additional N from complex ingredients like yeast extract and casitone). The significant up-regulation of ammonium transporters (amt, Fig. 4), methylamine permeases (mtmP and mtbP, Fig. 3), and P-II repressors (nifI, Fig. 4) in 30 °C perchlorate treatments were consistent with observations of mM-level NH 4 + (Table S3) at the time of sample preservation for RNA-Seq. In M. barkeri, NH 4 + concentrations as low as 10 µM have been shown to be inhibitory to N 2 fixation 53-56 . It is therefore surprising to observe a simultaneous and substantial up-regulation of nitrogenase genes, an overwhelming proportion of which being associated with the molybdenum isoform, (e.g., nifDEHK, Fig. 4). Nitrogen fixation is an energetically expensive process, consuming at least 16 (and, by one calculation for methanogens 69 , perhaps more than 50) moles of ATP per mole of N 2 fixed 70 . The significant increase in expression of the nifI repressors in the 30 °C Mg(ClO 4 ) 2 and Na(ClO 4 )-amended treatments suggest a concerted effort to dedicate cellular energy towards signaling the shut off of Mo-Nase activity. What is the purpose of this transcriptional arms race?
Microbial perchlorate reduction has been identified throughout the Proteobacteria 35,71,72 , as well as halophilic 73 and hyperthermophilic Archaea 74 . The substrate-specific and Mo-containing perchlorate reductase (pcrAB), which reduces perchlorate to chlorite (ClO 2 -), and chlorite dismutase (cld), which dismutates ClO 2 into chloride (Cl -) and molecular oxygen (O 2 ), have been isolated and purified 35 , but homologues of these genes have not been found in the M. barkeri genome. However, chlorate-and perchlorate-binding affinities have been documented in structurally similar α-and β-subunits of other Mo-containing DMSO family reductases 35,75,76 . A plethora of inorganic and organic electron donors have also been implicated in microbial perchlorate reduction 35,77-79 including H 2 , sulphide (S 2 -), CH 4 , and yeast extract, which are available constituents in the incubations analyzed in this study. The biochemistry of microbial perchlorate reduction is complex and not yet fully understood. Nonetheless, the following environmental conditions must be met in order for perchlorate reduction to proceed: (1) dissolved O 2 concentrations must be < 2 mg L −1 ; (2) NO 3 must be completely consumed; and (3) molybdenum must be bioavailable as molybdate, MoO 4 2− , to synthesize Mo cofactors 35,80 in perchlorate-reducing metalloenzymes. All three conditions were coincidentally satisfied in this study's incubation conditions. Perchlorate reduction has been observed in sterile methanogenic media under an 80:20 H 2 :CO 2 atmosphere 25,26 and apparently exacerbated in media with the psychrotolerant methanogen Methanobacterium arcticum M2, leading to the suggestion that perchlorate could be utilized as a possible electron acceptor in a reversal of methanogenesis -the anaerobic oxidation of methane, AOM 25 (Table S3). Anion measurements in the perchlorate-amended incubations did yield an increase in total Cl -, but only enough to account for 35.5% Mg(ClO 4 ) 2 , 5.6% Ca(ClO 4 ) 2 , and 44.8% of Na(ClO 4 ) apparently reduced (Table S3). We note that due to its co-elution with Cl -, we were not able to ascertain the presence of chlorate (ClO 3 -). Chlorite (ClO 2 -) was not detected in any samples (Table S3). Additional resources and measurements in a future study are necessary to quantify ClO 3 and other potential intermediates (e.g. chlorinated organics) to fully assess the fate of ClO 4 -.
Sodium sulphide, Na 2 S, a common reducing agent in anaerobic media, has been previously ruled out as a significant contributor to abiotic perchlorate reduction in methanogenic media (medium MSH in Kral et al.) 26 . Yeast extract is a complex and undefined carbon source, and its inclusion in sterile MSH media has been associated with increased rates of perchlorate reduction compared to a minimal salts medium (medium MM in Kral et al.) 26 . In this study, preliminary measurements of perchlorate concentrations in sterile DSMZ 120a media failed to detect conclusive perchorate reduction (Table S3). More work is necessary to fully explore the composition and role of yeast extract as a potential reducing agent.
The free energy yield of H 2 oxidation coupled to perchlorate reduction (Eq. 2) is substantial (∆G˚' = − 289 kJ/ mol H 2 ) but the reaction is kinetically sluggish at ambient conditions and requires a metallic catalyst to overcome its large activation energy 81 .
Nickel (Ni) has long been known as an excellent hydrogenation catalyst 82 and is an essential cofactor for hydrogenase activity in methanogens 83 , but an essential role in microbial perchlorate reduction has never been suggested and it is not obvious from our transcriptomic data that perchlorates elicit a universal response by Ni active sites in hydrogenases (e.g. echE, vhtA, and frhα in Fig. 3). Instead, given perchlorate reduction's known Mo dependency 35,84 and the additional empirical evidence presented here, we cannot ignore the most parsimonious explanation: the substantial up-regulation of (Mo)-Nase (Fig. 4) and (Mo)-formylmethanofuran dehydrogenase (Fig. 3, Table 1) at 30 °C must be associated with perchlorate reduction. Certainly, a thorough investigation is warranted in order to elucidate finer details of the biochemistry.
The methylamine methanogenesis pathway as a mechanism for energy conservation. In methylamine methanogenesis, H 2 is recycled by the partial reversal of the methanogenesis pathway, generated via the oxidation of F 420 H 2 by frhα and the oxidation of ferredoxin by echF (green arrows in Figs. 3). H 2 generated by the oxidation of F 420 H 2 and ferredoxin diffuses across the membrane to be oxidized by vhtA in an energy conserving scheme to recycle methanophenazine (MP) [85][86][87][88] . Ferredoxin oxidation by echF also results in the translocation of 2H + through ech, contributing to the production of a proton gradient (high outside the cell) 88 . Evidence for the generation of this proton gradient might be indicated in our incubations based on drops in pH (up to 0.77 pH units) (Table S5). Based on up-regulation of the genes for methylamine methanogenesis, we attribute the differential expression of ech, vht, and frh hydrogenases (Fig. 3) and ferredoxin (Tables S10, S12, S14) as putative sources of a proton motive force in the 30 °C Ca(ClO 4 ) 2 -supplemented incubations.
We theorize that the thermodynamic spontaneity of H 2 -dependent perchlorate reduction (in the presence of a Mo catalyst) results in it outcompeting the endergonic first step of hydrogenotrophic methanogenesis (Eq. 3; ∆G˚' = + 16 kJ/mol H 2 ) 89 (Fig. 6a).
Likewise, we imagine H 2 intermediates generated during energy conserving steps of methylamine methanogenesis are siphoned off to reduce perchlorate in lieu of reducing MP via vht (∆G˚' = − 289 kJ/mol H 2 for perchlorate reduction 90 vs. ∆G˚' = − 50 kJ/mol H 2 for MP reduction 91,92 ) (Fig. 6b). Such interpretations warrant further investigation, but do fit within previously reported narratives of observed decreases in CH 4 production in perchlorate-supplemented methanogenic cultures 25,26 . Do all perchlorates trigger universal regulatory responses? Calcium-, magnesium-and sodium perchlorate were investigated in this study as they have been detected in active Recurring Slope Lineae (RSL) on Mars 43 . Until now, studies investigating methanogen survival in high concentrations of perchlorate salts have been unable to ascertain the specific metabolic responses behind observed decreases in CH 4 production, and whether these metabolic responses are universal regardless of the perchlorate's constituent cation. Our work demonstrates that, for M. barkeri MS transcriptomes assembled after 28 days of incubation, methanogenesis regulatory patterns are consistent across all three perchlorate conditions with methylamine gene utilization and a handful of (Mo)-fmd subunits at 30 °C (Fig. 3). Otherwise, global expression patterns in Ca(ClO 4 ) 2 treatments were somewhat distinct from Na(ClO 4 ) and Mg(ClO 4 ) 2 conditions, which, on the whole, exhibited similar transcriptomic responses at both temperatures (Figs. 4, 5; Tables 1, 2). Transcriptomes from additional time points can help clarify whether these discrepancies in expression are reflective of distinct metabolic responses to Ca-perchlorate versus Mg-and Na-perchlorate, or simply different stages of metabolism following prolonged www.nature.com/scientificreports/ exposure to perchlorates in general. We infer the latter is more likely given the novel and ubiquitous response of genes in the methylamine methanogenesis pathway.
Implications for astrobiology and Martian methane. These findings better constrain our growing understanding of how microbial life responds to prolonged exposure under extreme conditions characterized by strong oxidants, freezing temperatures, and osmotic stress-e.g., a Mars Special Region 93 potentially conducive to microbial habitability. Notably, our study shows that major metabolic disruption by perchlorates at 30 °C is not reflected to the same degree at 0 °C, which is more appropriately representative of the conditions at Mars Special Regions 93 . This finding offers new perspectives to contextualize observations of diffuse Martian CH 4 emissions, particularly in situ measurements made by the Curiosity rover 5,6 . This work offers an in-depth glimpse into the remarkable adaptability of methanogens to survive under oxidative stresses that mimic the www.nature.com/scientificreports/ challenging conditions of the Martian subsurface. The inferences made from this study provide many exciting opportunities for further research to better understand methanogen ecophysiology under extreme stress that may offer terrestrial insight into the enigmatic Martian methane story.

Materials and methods
Materials and culture conditions. Methanosarcina barkeri wild-type strain MS (ATCC 51582) was grown hydrogenotrophically ( Balch tubes were inoculated with 1 mL cell suspensions containing ~ 4 × 10 7 cells and incubated in batches of 9 biological replicates plus one autoclave-sterilized media blank for 28 days (Fig. S17). For each experimental treatment (perchlorate-free control, NaClO 4 , Mg(ClO 4 ) 2 , or Ca(ClO 4 ) 2 ), optical density measurements (OD) and CH 4 production were monitored weekly in parallel experiments incubated at 30 °C or 0 °C (Fig. 2). OD measurements were performed spectrophotometrically at 600 nm using a Genesys 30 Visible Spectrometer (Thermo-Scientific Corp., Madison, WI USA). At the end of the incubation experiment, direct cell counts were performed on enrichment aliquots fluorescently stained with Acridine Orange (AO) dye (catalogue #318,337, Sigma-Aldrich Chemical, Co., Milwaukee, WI USA) (50 µM final concentration) following an established procedure 96  Compositional analyses of cations, anions, and headspace gas is available for reference in Table S3.
RNA isolation and purification. After 28 days' incubation, culture and blank volume contents were briefly vortexed and aseptically transferred into sterile 15 mL Falcon tubes (Corning Inc., Corning, NY USA) at incubation temperature inside a Coy anaerobic chamber (Coy Laboratory Products Inc., Grass Lake, MI USA). Tubes were centrifuged at 3,000 × g for 1 min using an IEC Centra CL2 centrifuge (Thermo Electron Company, Milford, MA USA) to pellet cells. Media was poured off for pH and electrical conductivity measurements, leaving behind 1 mL. To ensure quantifiable RNA yields downstream, the nine biological replicates from each condition were consolidated into three sets of three samples each for extraction (3 mL pelleted cell suspension/ tube). RNAlater solution (ThermoFisher Scientific, Waltham, MA USA) was added to a final volume of ~ 12 mL. Samples were left to equilibrate at incubation temperature for 3 h, transferred to 4 °C for 24 h, and then stored at -80 °C until overnight shipment on dry ice to Princeton University for RNA extraction. Samples preserved at -80 C in RNAlater were thawed on ice in a sealed container before contents were transferred to 50 mL Falcon® tubes. An equal volume (12 mL) of nuclease-free water (Qiagen, Hilden, Germany) was added to RNAlater-preserved samples, briefly vortexed, and centrifuged at 5000 gx for 10 min using a Sorvall Legend XI centrifuge (ThermoFisher Scientific, Waltham, MA USA). The supernatant was subsequently discarded, and RNA was extracted following a modified protocol from a Zymo Quick-RNA Miniprep Plus Kit (Zymo Research, Irvine, CA USA). RNA lysis buffer and nuclease-free water were added to each sample in a 5:1 ratio, and sterile 0.7 mm garnet bashing beads (Qiagen, Hilden, Germany) were added to facilitate mechanical lysis during subsequent vortexing. Samples were then vortexed for 1 min and centrifuged at 10,000 × g at 4 °C for 1 min using an Eppendorf 5810R (Eppendorf, Hamburg, Germany) to pellet cell debris. The supernatant containing total nucleic acids was transferred to a yellow Spin-Away column (Zymo Research, Irvine, CA USA) fitted in a 2 ml collection tube. Samples were centrifuged at 10,000 × g for 1 min using an AccuSpin Micro 17 (ThermoFisher Scientific, Waltham, MA USA) to separate out genomic DNA. Following centrifugation, the Spin-Away filter was discarded and the filtrate was collected from the column for RNA purification. www.nature.com/scientificreports/ Total nucleic acids were then precipitated by adding 1 mL of chilled absolute ethanol. Pellets were mixed by pipetting and then transferred to green Zymo-Spin IIICG column filters fitted in clean collection tubes. Samples were centrifuged at 10,000 × g for 30 s to collect precipitated nucleic acids on the column, subsequently treated with 400 µL RNA wash buffer, and centrifuged for 30 s at 10,000 gx. The wash buffer was discarded and columns were treated with 80 µL DNAse I reaction mixture (per 80 µL: 5 µL DNAse 1 [1 U/µL], 8 µL 10X DNAse I reaction buffer [Zymo Research, Irvine, CA USA], 3 µL nuclease-free water, 64 µL RNA wash buffer [Zymo Research, Irvine, CA USA]) to degrade trace genomic DNA and left to incubate on ice in the dark for 15 min. DNAse-treated samples were then centrifuged at 10,000 × g for 30 s. The reaction buffers were discarded, column filters were washed three times with 400, 700, and 400 µl of RNA Prep Buffer (Zymo Research, Irvine, CA USA), centrifuging twice for 30 s at 10,000 × g and at 16,000 × g for 2 min for the final wash. Total RNA was then eluted into sterile, nuclease-free PCR tubes on ice using nuclease-free water pre-heated to 95 C. Total RNA was quantified using a Qubit hs RNA assay kit coupled to a Qubit 2.0 analyzer (ThermoFisher Scientific, Waltham, MA USA) following the manufacturer's protocol. RNA quality was subsequently assessed using a 2100 Bioanalyzer (Agilent Technologies Inc., Santa Clara, CA USA). RNA samples were kept at -20 °C until library preparation and sequencing.
All RNA extraction steps were performed using nuclease-free pipette tips (Corning Inc., Corning, NY USA) in a UV-sterilized laminar flow hood. All surfaces, pipettes, and gloves were wiped down at each step with RNaseZap solution (ThermoFisher Scientific, Waltham, MA USA) to minimize potential RNase contamination. Parallel extraction blanks of extraction kit reagents and blank growth media co-incubated with M. barkeri enrichment cultures were performed to ensure cleanliness of the extraction procedure and sterility of uninoculated growth media.
RNA library prep and RNA-Seq. Ribosomal RNA was depleted from total RNA using a Ribo-Zero Bacterial rRNA Removal Kit (Illumina, Inc., San Diego, CA USA) following the manufacturer's instructions and using the provided universal probe sequence. One-directional library prep was performed for each treatment and its constituent 3 sequencing replicates using the Nextera DNA Flex Library Prep Kit (Illumina, Inc., San Diego, CA USA) on the automated Apollo 324 system (Takara Bio USA, Inc., Mountain View, CA, USA). RNA-Seq was carried out for 318 cycles on two lanes of a NovaSeq SP Flowcell (Illumina, Inc., San Diego, CA USA) (1 × 150 bp) at the Princeton University genomics core facility.
Annotation and comparative transcriptomics. Quality filtering of single-end reads was performed using fastp v.0.12.6 97 , removing reads < 50 nt, containing > 1 Ns, Phred quality scores < 30, and sample barcode sequences. Using Bowtie2 v.2.3.2 98 , retained, quality-filtered reads from each experiment were mapped to coding sequence (CDS) regions subset from the complete M. barkeri MS reference genome obtained from NCBI GenBank (NZ_CP009528.1). CDS-mapped reads were then sorted, indexed, and processed for extraction from the sequenced transcriptome using Samtools v.1.5 99 and BEDTools v.2.17.0 100 . Gene annotation was performed using NCBI BLASTn v.2.2.29 + 101 against a reference M. barkeri MS CDS assembly database generated using option -makeblastdb. Protein assignment was determined as the entry with the greatest sequence identity alignment with the query sequence, the lowest E-value, and largest bit score. Differential expression analysis of investigated treatments relative to the 30 °C and 0 °C perchlorate-free controls, within-group (i.e. biological replicate) variance estimation, and FPM were performed using DESeq2 102 . Metabolic pathway involvement of identified genes was determined by referencing the Kyoto Encyclopedia of Genes and Genomes (KEGG) 103 .

Data availability
RNA-Seq data is available at NCBI GenBank under accession number PRJNA635445 (Sequence Read Archive SRP265010). The source code of all transcriptomic analyses presented in this manuscript are available upon request.