Conserved genomic and amino acid traits of cold adaptation in subzero-growing Arctic permafrost bacteria

Abstract

Permafrost accounts for 27% of all soil ecosystems and harbors diverse microbial communities. Our understanding of microorganisms in permafrost, their activities and adaptations, remains limited. Using five subzero-growing (cryophilic) permafrost bacteria, we examined features of cold adaptation through comparative genomic analyses with mesophilic relatives. The cryophiles possess genes associated with cold adaptation, including cold shock proteins, RNA helicases, and oxidative stress and carotenoid synthesis enzymes. Higher abundances of genes associated with compatible solutes were observed, important for osmoregulation in permafrost brine veins. Most cryophiles in our study have higher transposase copy numbers than mesophiles. We investigated amino acid (AA) modifications in the cryophiles favoring increased protein flexibility at cold temperatures. Although overall there were few differences with the mesophiles, we found evidence of cold adaptation, with significant differences in proline, serine, glycine and aromaticity, in several cryophiles. The use of cold/hot AA ratios of >1, used in previous studies to indicate cold adaptation, was found to be inadequate on its own. Comparing the average of all cryophiles to all mesophiles, we found that overall cryophiles had a higher ratio of cold adapted proteins for serine (more serine), and to a lesser extent, proline and acidic residues (fewer prolines/acidic residues).

psychrophile, psychrotroph, cold adaptation, subzero, permafrost, amino acid composition

INTRODUCTION

Permafrost accounts for 27% of all soil ecosystems on earth and is home to active and diverse microbial communities, despite extreme conditions characterized by permanent sub-zero temperatures, low water activity and often oligotrophic conditions (Goordial et al. 2012). The identification of active microbial communities in permafrost has fueled interest in understanding the adaptations that allow these organisms to sustain their growth and survival in such harsh environments. Reduced enzyme activity, protein denaturation and misfolding, inhibitory secondary RNA/DNA structures and increases in membrane rigidity are challenges that must be overcome at low temperatures (Bakermans et al. 2012; Mykytczuk et al. 2013; De Maayer et al. 2014). Noted adaptations to cold found in psychrophiles include, but are not limited to, using cold shock proteins and helicases to stabilize DNA/RNA processes, membrane modifications to increase unsaturated and branched chain fatty acids, synthesis of carotenoids and antifreeze proteins, stimulation of osmotic and oxidative responses, differential expression of isozymes with different temperature optima and broad changes in translational and ribosomal processes (Bakermans et al. 2012). In this paper, we use the term psychrophile to encompass both eurypsychrophiles and stenopsychrophiles (Feller and Gerday 2003; Cavicchioli 2016); eurypsychrophiles are defined as organisms that possess a broad growth range (Tmax >20°C), and who, while capable of growth at very low temperatures (−15°C), usually retain optimum growth rates at temperatures above 20°C, while stenopsychrophiles exhibit narrower growth ranges (Tmax <20°C), and usually grow best at temperatures ranging between 5°C and 15°C (Cavicchioli 2006). The term cryophile is used here to denote specifically those psychrophiles that can grow at subzero temperatures (Seckbach 2013; Vester, Glaring and Stougaard 2015; Goordial et al. 2016).

Ensuring that adequate flexibility is retained is a crucial adaptation of cold active proteins to preserve function at low temperatures. Cold adapted proteins have been shown to possess a number of amino acid (AA) changes that impart increased flexibility, including fewer salt bridges, a lower content of proline residues, fewer hydrogen bonds, a reduced Arg/(Arg + Lys) ratio and increased serine and glycine content (Aghajari et al. 1998; Russell et al. 1998; Georlette et al. 2000; Huston, Methe and Deming 2004; Collins, Gerday and Feller 2005; Ræder et al. 2008; Metpally and Reddy 2009). For example, the elongation factor 2 protein of the eurypsychrophilic archaea Methanococcoides burtonii, isolated from the permanently cold Ace Lake in Antarctica, exhibits greater structural flexibility as a result of fewer salt bridges, less densely packed hydrophobic cores and a reduction in proline residues in surface loops (Thomas and Cavicchioli 2000). A lipase from the Antarctic bacterium Psychrobacter immobilis B10 also contains a low number of salt bridges and proline residues, along with a reduced proportion of arginine to lysine residues, a small hydrophobic core and a very small number of aromatic-aromatic interactions (Arpigny, Lamotte and Gerday 1997). Differences in k_cat and K_m values, corresponding to increased catalytic efficiency at lower temperatures, have been measured in certain psychrophilic enzymes when compared to their meso- or thermophilic homologs (Feller et al. 1992; Georlette et al. 2000; Thomas, Kumar and Cavicchioli 2001).

Several studies have carried out comparative genomic analyses between psychrophiles and mesophiles and/or thermophiles in order to examine differences in AA composition, which may impart greater flexibility of proteins in cold adapted microorganisms. Of these, only four focused on microorganisms that exhibit growth at subzero temperatures and in each case focused on a single organism (Methé et al. 2005; Ayala-del-Río et al. 2010; Mykytczuk et al. 2013; Goordial et al. 2016). Previous genomic analyses include those focused on psychrophilic protein comparisons against large nonredundant protein databases and comparisons against select specific mesophiles (Saunders et al. 2003; Rabus et al. 2004; Methé et al. 2005; Grzymski et al. 2006; Metpally and Reddy 2009; Ayala-del-Río et al. 2010; Zhao et al. 2010; Mykytczuk et al. 2013; Goordial et al. 2016). Methods for comparison have included those that compare total AA counts or composition across an entire genome between mesophiles and psychrophiles, to those that compare AA counts only between previously matched homologous proteins. Results from these studies have been inconsistent. While some identified a number of changes in AA composition or proportions in psychrophilic genomes, consistent with those identified in psychrophilic proteins (Saunders et al. 2003; Ayala-del-Río et al. 2010; Zhao et al. 2010), others have found limited changes (Methé et al. 2005; Mykytczuk et al. 2013; Goordial et al. 2016), and in some cases, none (Rabus et al. 2004; Yang et al. 2015). It remains unclear whether these AA substitutions are universal adaptations that can be conclusively identified on a large scale across genomes in psychrophiles. It is also of interest to determine if there are different adaptations associated with psychrophiles capable of growth at sub-zero temperatures (cryophiles), compared with those psychrophiles that have narrower growth ranges and that may not be adapted to the challenges associated with subfreezing temperatures. To our knowledge, very few studies on AA cold adaptive traits in psychrophiles have looked specifically at cryophilic psychrophiles. Furthermore, in those studies that perform comparisons using distantly related organisms, it is not always apparent if the differences observed are the result of cold adaptation or simply phylogenic divergence. Comparisons using organisms in the same genus, or neighbor genus, provide a more stringent analysis, as we expect these members to share close common ancestry, and therefore observed AA differences between psychrophiles and their close mesophilic relatives are more likely to be the result of cold adaptation.

In the present study, we selected five subzero-growing (cryophilic) strains (Table 1), previously isolated from the Canadian high Arctic (Steven et al. 2007a, 2008; Lacelle et al. 2011) and spanning several phyla (ia, Proteobacteria and Firmicutes), for genome sequencing and annotation, which were used to perform in-depth genomic and AA comparative analyses. These organisms were isolated from permafrost environments that are harsh, often oligotrophic, with low water activity (aw = ∼0.85) and ambient in situ temperatures of ∼−17°C (Gilichinsky, Soina and Petrova 1993; Steven et al. 2007b, 2008). Survival of microbes in these environments is hypothesized to be possible due to the presence of very thin brine films (Gilichinsky et al. 2003; Steven et al. 2006), where higher solute and salt concentrations act as freezing point depressants, allowing water to remain liquid. It is expected that microorganisms adapted to these environments would have mechanisms that allow them to grow under both cold (subzero) and high salt conditions. Canadian Arctic permafrost salt concentrations have been estimated at ∼14.6 g/kg (Steven et al. 2007a), while salt concentrations in Siberian cryopeg brine lenses have been measured at 170–300 g/L (17–30%) (Gilichinsky et al. 2003). We compared the genomes of our cryophilic strains from these environments to closely related mesophilic relatives in the same genus. With these new genome sequences, we aim to increase our potential for understanding and identifying the AA basis of cold adaptation in sub-zero growing organisms, including potentially conserved or important adaptive traits. In addition, we also investigate the presence of known cold adaptation genes as well as genomic redundancy, which is the presence of multiple gene copies, in each cryophile.

Table 1.

Permafrost and ice wedge strains selected for sequencing and cold adaptation analysis.

Strain	Location isolated	Sample type	Temp °C	Salinity (%)	Phylum	Reference
Actinotalea sp. KRMCY2a	Moose Lake, Yukon	Ice Wedge	−5–25	0–8	Actinobacteria	This study & Lacelle et al. (2011)
Polaromonas sp. Eur3 1.2.1	Eureka, Nunavut	Permafrost	−5–22	0–3	Proteobacteria	This study & Steven et al. (2008)
Paenisporosarcina sp. Eur1 9.01.10a	Eureka, Nunavut	Permafrost	−5–25	0–8	Firmicute	This study & Steven et al. (2007a)
Methylobacterium sp. AL-11	Eureka, Nunavut	Permafrost	−5–25	0–5	Proteobacteria	This study & Steven et al. (2008)
Kocuria sp. KROCY2a	Old Crow, Yukon	Ice Wedge	−5–25	0–19	Actinobacteria	This study & Lacelle et al. (2011)

^aActinotalea sp. KRMCY2, Paenisporosarcina sp. Eur1 9.01.10 and Kocuria sp. KROCY2 were originally identified as Cellulomonas sp. KRMCY2, Sporosarcina sp. Eur1 9.01.10 and Microbacterium sp. KROCY2, respectively, but were renamed following genome sequencing and 16S rRNA gene analysis (see text).

MATERIALS AND METHODS

Organism selection and identity

Actinotalea sp. KRMCY2, Paenisporosarcina sp. Eur1 9.01.10, Methylobacterium sp. AL-11, Kocuria sp. KROCY2 and Polaromonas sp. Eur3 1.2.1. were isolated from permafrost cores from Eureka, Nunavut or from ice samples from the Yukon, as previously described (Steven et al. 2007a, 2008; Lacelle et al. 2011). These strains were selected for genomic sequencing due to their ability to grow at sub-zero temperatures under laboratory conditions (Table 1). Based on 16S rRNA gene sequencing Kocuria sp. KROCY2, Paenisporosarcina sp. Eur1 9.01.10 and Actinotalea sp. KRMCY2 were initially identified as belonging to the genus Microbacterium, Sporosarcina and Cellulomonas, respectively (Steven et al. 2007a; Lacelle et al. 2011). However, recent 16S rRNA gene analyses using the EzBioCloud quality-controlled database (Yoon et al. 2017) has identified the closest taxonomically valid type strains for Kocuria sp. KROCY2, Paenisporosarcina sp. Eur1 9.01.10 and Actinotalea sp. KRMCY2 to be Kocuria palustris DSM 11 925(T) (100% similarity), Paenisporosarcina indica PN2T(T) (99.16% similarity) and Actinotalea ferrariae CF5–4(T) (98.13% similarity), respectively (Table S1, Supporting Information). Thus, we have renamed our strains accordingly. The closest taxonomically classified type strains for Polaromonas sp. Eur3 1.2.1 and Methylobacterium sp. EUR3 AL-11 were Polaromonas jejuensis NBRC 106 434(T) (99.04% similarity) and Methylobacterium radiotolerans JCM 2831(T) (99.93% similarity), respectively.

Growth conditions

Cultures were maintained on R2A agar or liquid (BD Difco, Sparks, Maryland) for Actinotalea sp. KRMCY2, Paenisporosarcina sp. Eur1 9.01.10, Methylobacterium sp. AL-11 and Kocuria sp. KROCY2 and ½ R2A for Polaromonas sp. Eur3 1.2.1. Growth temperatures were described previously (Steven et al. 2007a, 2008; Lacelle et al. 2011) on agar only and tested in this study using liquid cultures at −5°C, 0°C, 5°C, 10°C, 22°C and 25°C in Tryptic Soy Broth (TSB; BD Difco, Sparks, Maryland), 1/10 TSB, ¼ TSB, R2A or ½ R2A and supplemented with varying concentrations (2%–7%) of sucrose, glycerol or NaCl at subzero temperatures. Salinity tolerance was determined from liquid cultures grown on R2A amended with NaCl in 1%–2% increments from 0% to 20%. Previous phylogenic classification was verified by comparing the 16S rRNA gene sequence of each organism to the EzBioCloud quality-controlled database for comparison to valid taxonomically classified type strains (Yoon et al. 2017).

DNA extraction

Strains were grown to stationary state at room temperature (∼23°C) on TSB (BD Difco). Genomic DNA was isolated using the Epicentre MasterPure Gram Positive DNA Purification Kit (Epicentre, Madison, Wisconsin) as per the manufacturer's instructions. Extracted DNA was checked for purity and quantified with the NanoDrop 1000 (Thermoscientific, Wilmington, Delaware) and sent for whole genome sequencing according to DOE Joint Genome Institute (JGI) standards.

Genome sequencing and assembly

The draft genomes of Kocuria sp. KROCY2, Actinotalea sp. KRMCY2, Polaromonas sp. EUR3 1.2.1 and Methylobacterium sp. EUR3 AL–11 were generated at the DOE JGI using the Pacific Biosciences (PacBio) technology. Pacbio SMRTbell libraries were constructed for each isolate and sequenced on the PacBio RS platform, which generated 166 364 filtered subreads totaling 486.7 Mbp for Kocuria sp. KROCY2, 253 745 filtered subreads totaling 785.8 Mbp for Actinotalea sp. KRMCY2, 150 954 filtered subreads totaling 668.5 Mbp for Polaromonas sp. EUR3 1.2.1, and 369 217 filtered subreads totaling 928.2 Mbp for Methylobacterium sp. EUR3 Al-11. All general aspects of library construction and sequencing performed at the JGI can be found at http://www.jgi.doe.gov. The raw reads were assembled using HGAP (version: 2.3.0) (Chin et al. 2013). The final draft assembly contained one contig in one scaffold for Kocuria sp. KROCY2, Actinotalea sp. KRMCY2 and Polaromonas sp. EUR3 1.2.1 and produced an average input read coverage of 186.4X, 179.6X and 108.4X, respectively, and seven contigs in seven scaffolds for Methylobacterium sp. EUR3 AL-11 with input read coverage of 158.9X.

The draft genome of Paenisporosarcina sp. EUR3 2.2.2 was generated at the DOE JGI using the Illumina technology. The Illumina shotgun library and long insert mate pair library were constructed and sequenced using the Illumina HiSeq 2000 platform (Bennett 2004). 22 752 872 reads totalling 3412.9 Mb were generated from the shotgun and 40 430 662 reads totalling 3679.2 Mb were generated from the long insert mate pair library. All general aspects of library construction and sequencing performed at the JGI can be found at http://www.jgi.doe.gov. All raw Illumina sequence data were passed through DUK, a filtering program developed at JGI, which removes known Illumina sequencing and library preparation artifacts (Li, Copeland and Han 2011). Filtered Illumina reads were assembled using AllpathsLG (PrepareAllpathsInputs: PHRED 64 = 1 PLOIDY = 1 FRAG COVERAGE = 150 JUMP COVERAGE = 25; RunAllpathsLG: RUN = std pairs TARGETS = standard VAPI WARN ONLY = True OVERWRITE = True) (Gnerre et al. 2011). The final draft assembly contained 18 contigs in 5 scaffolds. The final assembly is based on 3412.4 Mb of Illumina Std PE, 3679.1 Mb of Illumina CLIP PE post filtered data, which provides an average 2026.1X Illumina coverage of the genome.

Genome annotation

Genes were identified using Prodigal (Hyatt et al. 2010), followed by a round of manual curation using GenePRIMP (Pati et al. 2010) for finished genomes and draft genomes in fewer than 10 scaffolds. The predicted CDSs were translated and used to search the National Center for Biotechnology Information (NCBI) nonredundant database, UniProt, TIGRFam, Pfam, KEGG, COG and InterPro databases. The tRNAScanSE tool (Lowe and Eddy 1996) was used to find tRNA genes, whereas ribosomal RNA genes were found by searches against models of the ribosomal RNA genes built from SILVA (Pruesse et al. 2007). Other non-coding RNAs such as the RNA components of the protein secretion complex and the RNase P were identified by searching the genome for the corresponding Rfam profiles using INFERNAL (Nawrocki and Eddy 2013). Additional gene prediction analysis and manual functional annotation was performed within the Integrated Microbial Genomes platform (http://img.jgi.doe.gov) developed by the Joint Genome Institute, Walnut Creek, CA, USA (Markowitz et al. 2009).

Genome and AA analyses

The genomes of close relatives (genus or neighboring genus) to the cryophilic strains, which have been annotated as mesophiles, were collected from publicly available databases (JGI), and were used for comparative analyses to the cryophilic genomes (Table S2, Supporting Information). Functional gene assignments, including COG and KEGG categories and pathways, were obtained from the JGI website for all the genomes, cross-referenced and used for comparative analyses. AA cold adaptation analyses was carried out as described previously (Goordial et al. 2016), with some modifications. Briefly, custom databases were created using the mesophilic genomes, and each cryophilic genome was compared to the mesophilic databases using stand-alone BLASTP (Camacho et al. 2009) with a cut-off e-value of 1e–15. The top BLASTP match protein sequence from each mesophile genome for each cryophilic protein was used for further analyses. Only non-hypothetical proteins and their mesophilic matches (full proteins) were included and assessed for cold adaptation at the AA level for the following indices previously associated with cold adaptation: arginine-to-lysine ratio; frequency of acidic, polar uncharged, and charged residues; proline, glycine, serine residues; aromaticity; aliphacity; and grand average of hydropathicity (GRAVY). The values for each AA trait were averaged from the selected mesophilic proteins, and this average was compared to the cryophilic protein using a one-sample t-test. Cryophilic proteins that were found to be significantly different from the average of the mesophilic proteins (P < 0.05) for each index were then assigned as ‘cold adapted’ or ‘hot adapted’ depending on the direction of change. We considered proteins cold adapted for each index if the direction of change was significantly lower for proline, R/K (arginine/lysine) ratio, acidic, charged, aliphatic, aromatic and hydrophobicity (GRAVY) indices and significantly higher for serine, glycine and polar uncharged indices (Table S3, Supporting Information). This analysis was repeated using the mesophile genomes, in the place of the cryophile genome, as controls and for comparison purposes. The proportions, in the total genome, of cold and hot adapted genes for the cryophiles and their mesophile controls were compared using a chi-square test with Bonferroni correction (P ≤ 0.0167) to determine if the differences seen were significant. All AA analyses were carried out using in-house python scripts publicly available at https://github.com/ColdAdaptationScripts.

RESULTS AND DISCUSSION

Genome and strain properties

High-quality permanent drafts were obtained for all five cryophilic genomes (Table 2). Kocuria sp. KROCY2 had the smallest genome at 2.8 Mbp, while the largest genome (7.2 Mbp) belonged to Methylobacterium sp. EUR3 AL-11 (Table 2). Polaromonas sp. Eur3 1.2.1 and Actinotalea sp. KRMCY2 had genomes of similar size at 4.4 Mbp and 4.5 Mbp, respectively, while Paenisporosarcina sp. Eur1 9.01.10 has a slightly smaller genome (3.5 Mbp). Consistent with the genome sizes Methylobacterium sp. EUR3 AL-11 had the highest number of predicted genes with 7001, followed by Polaromonas sp. Eur3 1.2.1, Actinotalea sp. KRMCY2 and Paenisporosarcina sp. Eur1 9.01.10 with 4238, 4145 and 3438 genes, respectively, and lastly Kocuria sp. KROCY2 with 2480. Close to 80% of all protein coding genes could be assigned function predictions for all five organisms except for Methylobacterium sp. EUR3 AL-11, which was closer to 70% (Table 2). The GC content also varied greatly from 38% in Paenisporosarcina sp. Eur1 9.01.10 to 72% for Actinotalea sp. KRMCY2. Methylobacterium sp. EUR3 AL-11 and Kocuria sp. KROCY2 both had 71% GC content, very similar to Actinotalea sp. KRMCY2. Polaromonas sp. Eur3 1.2.1 GC content was slightly lower at 60%. While GC rich regions have been noted in some psychrophiles (De Maayer et al. 2014), overall, we found GC content in our cryophiles to be consistent with other members of their respective genera (Table S2, Supporting Information). Interestingly, Polaromonas sp. Eur3 1.2.1 had a fairly high % (∼10%) of horizontally transferred genes compared to the other cryophiles (1%–4%) (Table 2). However, we found similar levels to that of Polaromonas sp. Eur3 1.2.1 in other members of this genus (Table S2, Supporting Information).

Table 2.

Genome properties of sequenced permafrost strains.

Organism	Polaromonas sp. Eur3 1.2.1	Actinotalea sp. KRMCY2	Kocuria sp. KROCY2	Methylobacterium sp. EUR3 AL-11	Paenisporosarcina sp. Eur1 9.01.10
Name in JGI database	Polaromonas sp. Eur3 1.2.1	Cellulomonas sp. KRMCY2	Microbacterium sp. KROCY2	Methylobacterium sp. EUR3 AL-11	Sporosarcina sp. Eur3 2.2.2
Genome size (Mbp)	4.4	4.5	2.8	7.2	3.5
%GC	60	72	71	71	38
Gene count	4303	4207	2480	7001	3503
Protein coding genes	4238	4145	2415	6907	3438
16S rRNA	2	1	3	6	3
Protein coding genes with function prediction	3471	3205	1931	4913	2711
with enzymes	1234	1000	786	1393	905
with KEGG pathways	1281	1121	817	1653	962
with KEGG orthology	2235	1907	1305	2798	1821
with COG	2915	2687	1710	4206	2314
with Pfam	3582	3359	2014	5100	2850
Horizontally transferred count (percent %)	417 (9.69)	185 (4.4)	122 (1.74)	25 (1.01)	57 (1.63)
JGI Taxon ID	2619618817	2545824564	2540341240	2546826724	2528768230
NCBI taxon ID	1305734	1304865	1305732	1305730	1305836
Sequencing status	Permanent draft	Permanent draft	Permanent draft	Permanent draft	Permanent draft
Release date	30/06/2015	13/11/2013	24/09/2013	02/12/2013	08/08/2013
High quality	Yes	Yes	Yes	Yes	Yes
Bioproject accession	PRJNA195644	PRJNA195883	PRJNA195886	PRJNA195879	PRJNA195884

We attempted to determine subzero growth rates for all the cryophilic strains; however, although chosen because of their ability to grow on solid R2A agar supplement with 7% sucrose at −5°C (Steven et al. 2007a, 2008; Lacelle et al. 2011), we struggled to find liquid media adequate for growth at low temperatures in several strains. R2A media supplemented with 7% sucrose is susceptible to freezing in liquid form, and alternative liquid media tests have yet to be as successful for growth. In addition, we found both Methylobacterium sp. EUR3 AL-11 and Actinotalea sp. KRMCY2 susceptible to forming large clumps of cells in liquid cultures, further complicating growth curve measurements. Figure S1 (Supporting Information) shows growth on agar plates at subzero temperatures for those organisms for which obtaining subzero growth curves proved problematic. Paenisporosarcina sp. Eur1 9.01.10 grew well at −5°C with a generation time of 14.6 days, although it grew optimally at 25°C (Table S4, Supporting Information). This growth rate is similar to that observed for other cryophiles at −5°C (Mykytczuk et al. 2013; Goordial et al. 2016). Only Polaromonas sp. Eur3 1.2.1. exhibited characteristics of a stenopsychrophile with a faster generation time at 10°C (3.1 days) than at 22°C (3.4 days). It also showed similar generation times at 0°C (3.6 days) and was incapable of growth above 22°C (Table S4, Supporting Information). Several members of the genus Sporosarcina and Polaromonas are known to be psychrophilic, having been isolated from numerous polar environments (Irgens, Gosink and Staley 1996; Yu et al. 2008; Margesin et al. 2012; Wang et al., 2015, 2016; Yan, Xiao and Zhang 2016).

Genome comparisons and cold adaptation genes

All five organisms possess genes known to be important in cold adaptation and cold growth including cold shock proteins, RNA helicases, osmotic and oxidative stress, carotenoid synthesis, translation factors and membrane and peptidoglycan modifications (Table 3). Although the presence of these genes is not unique to psychrophiles, we also found them to be present in mesophiles, their abundance was higher in the cryophiles in many cases. Polaromonas Eur3 1.2.1 possesses 13 copies of universal stress protein (UspA family) genes, produced in response to numerous environmental stressors (Kvint et al. 2003), and 12 copies of Na+/H+ antiporters. While also present in mesophilic Polaromonas strains, we found the MnhD subunit of Na+/H+ antiporters to be higher in copy number in Polaromonas sp. Eur3 1.2.1 (six) than in mesophilic relatives (one/two) (Fig. 1A). Similarly, a Na+/melibiose symporter was higher in gene abundance in Paenisporosarcina sp. Eur1 9.01.10 than mesophiles (Fig. 1C). Kocuria sp. KROCY2, able to grow in up to 19% salt, possesses nine copies of choline-glycine betaine transporter (Table 3), for uptake of the compatible solute glycine betaine, several more than the mesophiles we used in comparative analyses (Fig. 1B). Four of the five cryophiles also possess the genes for synthesis of trehalose, another well-established compatible solute; 10 copies of the first enzyme in trehalose synthesis, trehalose-6-phosphate synthase, are found in the genome of Actinotalea sp. KRMCY2 (Table 3). Trehalose and glycine betaine are important compatible solutes in cold adapted organisms, and along with sodium transporters, have roles in osmoregulation and stability (Thomas, Kumar and Cavicchioli 2001; Doyle et al. 2012). Salt tolerance would be necessary for survival in permafrost habitats consisting of thin films of salty water. The permafrost cryophilic and halotolerant Planococcus halocryophilus is also known to have multiple copies of genes involved in glycine betaine transport (Mykytczuk et al. 2013).

Figure 1.

COG categories and gene functions with higher abundances of copies in the cryophilic genomes when compared to mesophiles. A) Polaromonas sp. Eur3 1.2.1; B) Kocuria sp. KROCY2; C) Paenisporosarcina sp. Eur1 9.01.10; D) Methylobacterium sp. EUR3 AL-11; and E) Actinotalea sp. KRMCY2. The blue column closest to the Y-axis represents copy numbers for the cryophiles in each case. Letters for the COG categories are given in parenthesis after the gene description: C) Energy production and conversion; D) Cell cycle control, cell division, chromosome partitioning; E) Amino acid transport and metabolism; G) Carbohydrate transport and metabolism; H) Coenzyme transport and metabolism; I) Lipid transport and metabolism; J) Translation, ribosomal structure and biogenesis; K) Transcription; L) Replication, recombination and repair; M) Cell wall/membrane/envelope biogenesis; N) Cell motility O) Post-translational modification, protein turnover, and chaperones; P) Inorganic ion transport and metabolism; R) General function prediction only; S) Function unknown T) Signal transduction mechanisms; U) Intracellular trafficking, secretion, and vesicular transport; V) Defense mechanisms; X) Mobilome: prophages, transposons.

Table 3.

Gene functions and COG or Enzyme Commission (EC) categories present in psychrophilic genomes with known or predicted roles in cold adaptation and growth.

Category and Gene name	COG/EC ID	Polaromonas Eur3 1.2.1	Actinotalea KRMCY2	Kocuria KROCY2	Methylobacterium EUR3 AL-11	Paenisporosarcina Eur1 9.01.10
Cold shock, stress and HSP proteins
Cold shock protein, CspA family	COG1278	1	1	2	6	4
Universal stress protein, UspA family	COG0589	13	0	1	4	3
Co-chaperonin GroES (HSP10)	COG0234	1	2	1	1	1
Chaperonin GroEL (HSP60 family)	COG0459	2	1	2	1	1
Molecular chaperone DnaK (HSP70)	COG0443	3	0	1	2	1
Molecular chaperone IbpA, HSP20 family	COG0071	5	1	0	4	0
Molecular chaperone, HSP90 family	COG0326	1	1	0	1	0
Molecular chaperone GrpE (heat shock protein)	COG0576	1	3	1	1	1
Ribosomal 50S subunit-recycling heat shock protein	COG1188	1	2	1	1	1
DNA replication and repair
DNA gyrase/topoisomerase I, subunits A&B	COG0187; 0188	4	4	3	4	4
Recombinational DNA repair protein RecR	COG0353	1	0	1	1	1
RecA-superfamily ATPase, KaiC/GvpD/RAD55 family	COG0467	1	1	0	2	0
RecA/RadA recombinase	COG0468	1	2	1	1	1
DNA repair ATPase RecN	COG0497	1	6	1	1	1
Superfamily II DNA helicase RecQ	COG0514	1	0	2	2	2
Rad3-related DNA helicase	COG1199	3	4	1	0	2
Membrane and peptidoglycan alteration
3-oxoacyl-(acyl-carrier-protein) synthase	COG0304	2	0	4	4	1
3-oxoacyl-[acyl-carrier-protein] synthase III (KASIII)	COG0332	1	1	3	2	1
3-oxoacyl-[acyl-carrier-protein] reductase	EC:1.1.1.100	12	3	3	6	5
Glycosyltransferase involved in cell wall biosynthesis	COG0438; 0463	17	10	19	39	7
3-hydroxyacyl-CoA dehydrogenase	COG1250	8	1	1	5	4
Fatty-acid desaturase	COG1398; 3239	1	1	1	0	1
D-alanyl-D-alanine carboxypeptidase	COG1686; 2027	2	1	1	4	3
Carotenoid biosynthesis
Phytoene dehydrogenase-related protein	COG1233	0	2	3	4	1
Phytoene/squalene synthetase	COG1562	2	1	1	4	1
Polysaccharide capsule
Capsular polysaccharide biosyn. protein EpsC	COG1086	1	0	1	1	1
Capsule polysaccharide export protein	COG3524	0	0	0	3	0
Capsular polysaccharide biosynthesis protein	COG3944; 4421	0	0	3	1	1
Exopolysaccharide biosynthesis protein	COG4632	0	1	0	0	0
Osmotic stress
ABC proline/glycine betaine transport, ATPase component	COG1125; 4175	1	1	2	2	1
ABC proline/glycine betaine transport, permease	COG1174; 4176	2	0	4	3	1
Choline-glycine betaine transporter	COG1292	0	0	9	0	1
Osmoprotectant binding protein	COG1732	1	0	3	1	1
Choline dehydrogenase or related flavoprotein	COG2303	1	1	1	7	1
Trehalose-6-phosphate synthase	COG0380	2	10	1	1	0
Trehalose-6-phosphatase	COG1877	1	1	1	1	0
Maltooligosyltrehalose synthase	COG3280	0	1	1	1	0
Na+/proline symporter	COG0591	1	1	1	2	3
Na+/H+ antiporters	COGs: 3004; 1055; 0025; 0651; 1009; 1320	12	5	7	11	5
Oxidative stress
Catalase	COG0753	2	0	2	1	3
Peroxiredoxin	COG0678; 1225	3	3	2	4	3
Glutathione peroxidase	COG0386	2	1	1	2	1
Spermidine synthase	COG0421	1	1	0	0	1
Thioredoxin reductase	COG0492	2	1	2	4	5
Glyoxylase or related hydrolase, β-lactamase superfam II	COG0491	6	4	4	9	7
Toxin/Antitoxin modules
Ser/Thr kinase RdoA, MazF antagonist	COG2334	2	1	0	2	1
Antitoxin component of MazEF	COG2336	1	1	0	0	0
mRNA-interferase, toxin component of MazEF	COG2337	2	1	0	1	1
mRNA interferase YafQ, toxin component of YafQ-DinJ	COG3041	2	1	0	2	0
Antitoxin component of RelBE or YafQ-DinJ	COG3077	1	0	0	2	0
Translation and transcription factors
tRNA-dihydrouridine synthase	COG0042	3	1	1	2	2
tRNA A37 threonylcarbamoyladenosine dehydratase	COG1179	1	0	0	0	1
Translation elongation factor EF-Tu, GTPase	COG0050	2	1	1	2	1
Translation elongation factor EF-G, GTPase	COG0480	3	1	1	1	2
Translation initiation factor IF-2, GTPase	COG0532	1	1	1	1	1
Translation initiation factor IF-3	COG0290	1	1	1	1	1
Transcription antitermination factor NusA	COG0195	1	1	1	1	1
Transcription termination factor NusB	COG0781	1	2	1	1	1
Superfamily II DNA and RNA helicase	COG0513; 1061	7	4	5	6	3
Superfamily II DNA or RNA helicase, SNF2	COG0553	0	0	1	0	2
Superfamily II RNA helicase	COG4581	0	2	1	0	0

All five strains have multiple copies of superfamily II DNA/RNA helicases, which includes the DEAD box helicases, shown to be crucial for stabilizing RNA at lower temperatures (Kuhn 2012). In addition, most cryophiles possessed toxin/antitoxin (TA) modules, with both Actinotalea sp. KRMCY2 and Methylobacterium sp. EUR3 AL-11 having copies of genes for TA systems not seen in their mesophilic counterparts (Fig. 1D and E). TA systems are important regulators of persistence and stress responses in certain bacteria, including Mycobacterium tuberculosis and Escherichia coli (Tiwari et al. 2015; Sauert et al. 2016), and one such system was found to be strongly induced at −10°C in Planococcus halocryophilus (Raymond-Bouchard et al. 2017). Paenisporosarcina sp. Eur1 9.01.10 and Actinotalea sp. KRMCY2 both have higher copy numbers of the DNA binding transcriptional regulators of the LacI/PurR family then the mesophiles we analyzed (Fig. 1C and E). This family contains several members and includes repressors of the lactose operon, purine nucleotide synthesis and trehalose operon. Interestingly, transposase gene numbers were higher in abundance in four of the five strains, Polaromonas sp. Eur3 1.2.1, Kocuria sp. KROCY2, Actinotalea sp. KRMCY2 and Methylobacterium sp. EUR3 AL-11 compared to their respective mesophilic relatives (Fig. 1A, B, D and E). Transposase activity has been shown to be linked to the stress response, and therefore, may be a mechanism in which these organisms gain genomic plasticity and a selective advantage necessary to survive and colonize extreme cold environments (de Saro et al. 2013; Shen et al. 2014). Increases in gene copy numbers have also been theorized to be an adaptive feature in and of itself by offering an advantage through selective expression of cold or hot adapted gene copies at different temperatures (Maki, Yoneta and Takada 2006; Mykytczuk et al. 2013). This is believed to confer advantages during cold growth given that selectively upregulated gene copies are likely to function better at their induced temperature, and thus allows organisms to have an expanded growth temperature range, such as P. halocryophilus, which grows from −15°C to 37°C.

AA adaptations

To examine AA substitutions that could favor increased flexibility, we performed genome wide comparative AA analyses using the predicted protein sequences for each cryophile. We determined the number of significantly cold or hot adapted genes in each cryophilic genome for specific AA indices associated with cold adaptation (Fig. 2 and Table S3, Supporting Information). Previous studies have used the ratio of cold/hot proteins as a standard to determine cold adaptation for a given trait and deemed those with cold/hot ratios >1, signifying enrichment of cold adapted proteins, as ‘cold adapted’ (Ayala-del-Río et al. 2010; Mykytczuk et al. 2013). For comparative purposes and as a control to check if these cold/hot ratios are specific to cold adaptation in cryophiles, we also determined the number of hot and cold adapted genes and cold/hot ratios for two mesophiles for each cryophile (Fig. 2 and Fig. S2, Supporting Information). As these results highlight, cold/hot ratios of >1 can be seen in all of the mesophiles for certain traits, and we therefore argue that using a ratio cutoff of 1 is not enough on its own to indicate cold adaptation for a given trait in psychrophilic genomes. We instead focused on determining significant differences in proportions of hot and cold genes in the cryophiles vs the mesophiles as determined by chi-square analysis with Bonferroni correction (P ≤ 0.0167) used to correct for multiple comparisons (Fig. 2; Figs S2 and S3, Supporting Information). Of note, most of the proteins in both cryophile and mesophile genomes were classified as neutral with no significant changes in AA content (Fig. S3, Supporting Information). Cold/hot ratios were therefore determined for the remaining proteins that showed significant changes and could be classified as hot or cold adapted.

Figure 2.

Amino acid traits of cold adaptation in cryophilic proteins. Left side graphs: significantly (P < 0.05) hot and cold adapted proteins identified in each cryophile (A-E) for each amino acid trait measured: aliphatic index, R/K ratio, acidic residues, aromaticity, GRAVY and proline, serine, glycine, polar and charged residues. Right side graphs: ratio of cold/hot adapted proteins identified in each cryophile (name bolded; black bar) and in two mesophilic relatives of each cryophile. Proportions of cold adapted proteins that were significantly different in the cryophile genome when compared to both mesophilic genomes are starred.

While we did not find the proportions of cold and hot adapted genes in the cryophilic genomes to be significantly different from the mesophiles for many of the amino acid traits, there were noteworthy exceptions. Polaromonas sp. Eur3 1.2.1 had significantly higher and lower proportions of cold and hot adapted genes (Table S3, Supporting Information), respectively, for the index proline then the mesophiles, with an overall much higher cold/hot ratio (Fig. 2A), indicative that more cryophilic proteins possess overall fewer proline residues than their mesophilic counterparts. Actinotalea sp. KRMCY2 was also significantly cold adapted for proline (Fig. 2C). Since prolines are covalently bonded to the nitrogen atom of the peptide group, they impose constraint on rotations of the peptide backbone and create rigid kinks in peptide chains, thus contributing to overall reduced flexibility of proteins. A reduction in proline residues has also been observed, when compared to mesophiles, in the genomes of the cold marine bacteria Shewanella halifaxensis and Shewanella sediminis (Zhao et al. 2010) and in a number of cold adapted enzymes, including the psychrophilic α-amylase from Alteromonas haloplanctis and the 3-isopropylmalate dehydrogenase from the eurypsychrophile bacterium Vibrio sp. 15 (Wallon et al. 1997; Aghajari et al. 1998).

Both Kocuria sp. KROCY2 and Actinotalea sp. KRMCY2 had higher proportions of cold adapted proteins for serine (i.e. more serine residues) than the mesophiles, while Actinotalea sp. KRMCY2 and Methylobacterium sp. EUR3 AL-11 had significantly higher proportions of cold adapted proteins for glycine (more glycine) (Fig. 2B, C and D). Due to its small size, glycine provides more conformational freedom, thus increasing flexibility, and serine, as a polar uncharged amino acid, may compensate partly for the general reduction in charged residues in psychrophilic proteins (Saunders et al. 2003). Serine and glycine have been shown to be higher in a number of psychrophilic proteins, including an Antarctic citrate synthase (Russell et al. 1998) and a DEAD-box RNA helicase (Linding et al. 2003). Increases in serine residues is one of the more commonly observed traits and has also been found in a uracil-DNA-N-glycosylate (Ræder et al. 2008), a thermolysin (Adekoya et al. 2006), and the genomes of Colwellia psychrerythraea and marine Glaciecola species (Methé et al. 2005; Qin et al. 2013). Significant cold adaptation was also observed for aromaticity (fewer aromatic residues) in Kocuria sp. KROCY2 and Methylobacterium sp. EUR3 AL-11, and for the R/K ratio, with an overall lower number of arginines to lysines, in Polaromonas sp. Eur3 1.2.1. The bacterial lipase from the Antarctic psychrophile P. immobilis B10 features a very small number of aromatic-aromatic interactions (Arpigny, Lamotte and Gerday 1997), and proteins present at the cell surface of Planococcus halocryophilus at −10°C have significantly fewer aromatic residues then surface proteins present at higher temperatures (Ronholm et al. 2015).

In addition to serine and glycine, Actinotalea sp. KRMCY2 showed the greatest amount of significantly cold adapted indices with six out of the 10 being significantly higher for cold adaption from the mesophiles. However, for the other cryophiles, only a few of the indices showed evidence of cold adaptation, and in the case of Paenisporosarcina sp. Eur1 9.01.10, none showed significant cold adaptation (Fig. 2E). Many members of the genera Paenisporosarcina and Sporosarcina are psychrophilic, as discussed above, and therefore it is possible that most members of these genera possess characteristics of cold adaption, even in those characterized as mesophiles in the literature. This would explain why little to no significant cold adaptation is detected in these organisms.

By comparing the average cryophile ratios for each amino acid adaptive trait with the average of the mesophile ratios for those traits, we were able to highlight some differences in several amino acid indices. A comparison of all cryophiles with all mesophiles revealed higher average ratios of cold/hot adapted genes in the cryophiles for serine, and slightly higher for acidic, proline, polar, charged and R/K indices (Fig. 3 and Table S3, Supporting Information). We did not, however, find these results to be significantly different by t-test (P ≤ 0.05), though the results were close for proline (P = 0.07), serine (P = 0.1) and acidic residues (P = 0.1). This lack of significant difference is in large part the result of noticeable variation in cold/hot ratios between the different genera and phyla for the specific amino acid indices we studied. Amino acid cold adaptive strategies may be taxa specific and vary greatly between individual proteins. Indeed, to date, not every amino acid feature discussed here has been found in a single genome or cold-active protein, rather most tend to possess only a few of these traits (Rodrigues and Tiedje 2008). Contradictory findings have also been reported with certain amino acid changes, such as glycine, appearing to be increased in some psychrophilic proteins (Russell et al. 1998; Linding et al. 2003) but decreased in certain psychrophilic genomes (Zhao et al. 2010). Thus, the adaptive strategies employed by psychrophiles may vary from organism to organism. These adaptations may not be present in all psychrophiles, as is the case with Desulfotalea psychrophila, which did not show detectable amino acid substitutions indicative of cold adaptation in its genome (Rabus et al. 2004), and in our case with Paenisporosarcina sp. Eur1 9.01.10. The cold-active isocitrate dehydrogenase from psychrophilic Colwellia maris also did not appear to show any characteristic amino acid changes associated with psychrophilic proteins (Watanabe et al. 2005).

Figure 3.

Average ratio of cold/hot adapted proteins for all cryophiles and all mesophiles for each amino acid trait measured.

Proteins with only a single or very few amino acid changes, which is enough to have significant impacts on function, would not be as easily detectable in these large-scale analyses since a single change is less likely to result in a protein begin detected as significantly cold adapted when compared to mesophiles. In addition, changes may, in some cases, only occur in a subset of proteins and not result of large-scale changes across the entire genome/proteome. The overall large number of proteins in this study that were neither cold or hot adapted would add weight to this theory. In Psychrobacter arcticus, noted adaptations appeared to be more prevalent in proteins involved in cell growth and reproduction (Ayala-del-Río et al. 2010). Changes may also be conformational in nature, affecting secondary, tertiary and quaternary structure, either through changes in location of the amino acids in the sequence, which would not affect overall amino acid counts, or through one or two amino acid changes, which would be difficult to detect in this analysis.

Lastly, and perhaps most importantly, the specific location of the amino acids, not simply total numbers, may be especially important. The citrate synthase from Antarctic strain DS2-3R showed a reduction in salt bridges and proline residues specifically in the loop regions of the protein, while increases in glycine and serine occurred close to catalytic sites (Russell et al. 1998). In the psychrophilic Alteromonas haloplanctis α-amylase, it is the core of the protein that shows reduced numbers of proline and arginine (Aghajari et al. 1998). Nevertheless, while we did not find significantly conserved traits across the cryophiles, nor observed noticeable differences in cold adaptation between cryophiles for many amino acid features, our results do highlight noteworthy trends in amino acid properties of cryophilic proteins, and did identify traits that are significantly different from mesophilic relatives for several of the cryophiles, especially with regards to glycine, proline and serine. Changes in these amino acids are amongst the most consistently observed cold adaptive traits in studies focusing on cold adaption of psychrophilic proteins, and these changes seem to span broad phyla.

CONCLUSION

We found evidence for the presence of many cold adaptation genes in the genomes of cryophiles reported in this study, as well as genomic redundancy (higher gene copy numbers), theorized to be an important feature of cold adapted organisms. We identified significant cold adaptation in the genomes, to maintain protein flexibility at low temperatures, in four of the five cryophiles for a subset of the amino acid traits we studied, including proline, glycine, serine and aromaticity. Comparing all cryophile genomes from our study to all mesophile genomes in our database, we found trends favoring lower proline and acidic residues, and higher serine content in cryophilic proteins. It is possible that changes in content of these specific amino acids may be especially important to increase protein flexibility during subzero growth. Further comparisons between cryophiles, psychrophiles and mesophiles would need to be done to determine if this is the case. On the whole, however, we could not easily detect cold adaptations on a genome-wide scale for many of the amino acid indices we investigated, and we found a significant degree of variability between the genera we studied. This may indicate that, while some overlap exists, overall, cold-adapted microbes use different approaches to adapt to low temperatures. In addition, since amino acid adaptations may be specific to certain subsets or categories of proteins, we believe that individual comparative analyses focusing on single proteins and taking into account not only total amino acid numbers, but the location of specific amino acids within the sequence, the overall potential conformation of the protein, and, if possible, functional assays, are more likely to provide an accurate overall representation of cold adaption in psychrophilic proteins. Overall, this study has increased our understanding of the genetic basis of cold adaptation in permafrost microbes, and thus helps up better understand how these organisms behave and grow at subzero temperatures.

SUPPLEMENTARY DATA

Supplementary data are available at FEMSEC online.

ACKNOWLEDGEMENTS

Sequencing was performed as part of a U.S. Department of Energy Joint Genome Institute Community Sequencing Program Project (Quartely; ID 1015712). Thanks to Nicole Shapiro, Tanja Woyke, Marcel Huntemann, James Han, Amy Chen, Nikos Kyrpides, Victor Markowitz, Krishna Palaniappan, Natalia Ivanova, Natalia Mikhailova, Galina Ovchinnikova, Andrew Schaumberg, Amrita Pati, Dimitrios Stamatis, Tatiparthi Reddy, Henrik P. Nordberg, Michael N. Cantor, and Susan X. Hua of JGI.

FUNDING

This work was supported by fellowships from the Natural Sciences and Engineering Research Council (NSERC) CREATE Canadian Astrobiology Training Program (CATP) and NSERC Canadian Graduate Scholarship to IRB, and NSERC Discover Grant Program, Northern Research Supplement grant, and a Polar Continental Shelf Project grant to LGW. The work conducted by JGI is supported by the Office of Science of the U.S. Department of Energy under Contract No. DE-AC02-05CH11231.

Conflict of interest. None declared.