The bacterial strain #4 (hereafter referred to as S. glacialimarina TZS-4_T) was previously isolated from a Baltic Sea ice sample (Luhtanen et al., 2014). Two other Shewanella strains, S. frigidimarina ACAM 591 and S. baltica LMG 2250, were purchased from the Leibniz Institute DSMZ-German collection of microorganisms under catalog number DSM-12253 and DSM-9439, respectively. All strains were grown on a solid medium containing rich 25% w/v marine broth [abbreviated as rMB; 7.5g peptone (Sigma-Aldrich), 1.5g yeast extract (Fisher Bioreagents), and 9.35g marine broth (BD-Difco) in 1L of ddH₂O (Milli-Q, Merck)] agar (15g/L) and stored in 30% v/v glycerol stock at −80°C.

Biolog GENIII microplates were used for initial phenotypic characterization, providing a standardized platform to test 94 different phenotypic characteristics, including 71 carbon utilization assays and 23 chemical sensitivity tests. Samples were prepared and tested according to the manufacturer’s instructions. All assays were performed in triplicates.

The bacteria were grown in both rMB media and rMB agar in the presence of different antibiotics, such as ampicillin (100μg/ml in ddH₂O), kanamycin (50μg/ml in ddH₂O), tetracycline (10μg/ml in ddH₂O), and chloramphenicol (25μg/ml in 99.5% EtOH). The growth on antibiotic media was observed after 24h, and results were interpreted as resistant (+) or susceptible (−).

The genomic sequence of S. glacialimarina TZS-4_T was determined using the Pacific Biosciences PacBio Sequel sequencing technology at the DNA Sequencing and Genomics Laboratory (Helsinki Institute of Life Science, University of Helsinki). Phylogenetic trees based on maximum likelihood, neighbor-joining, and maximum parsimony methods using the consensus sequence of the seven gene copies of 16S rRNA were constructed on MEGA software with 100 replicates as bootstrap value. To infer genome-based phylogeny, we constructed a concatenated bacterial core-gene set phylogenetic tree by the up-to-date bacterial core-gene set (UBCG) method (Na et al., 2018). A model phylogenetic tree was depicted for species delineation based on average nucleotide identities (ANI) score using the fastANI (ver. 1.31) and OrthoANI tools (ver. 0.90; Lee et al., 2016; Jain et al., 2018). OrthoANI calculated the ANI score of query organism against a group of reference genomes and generated a distance matrix. The phylogenetic tree with its heat-map shows the relationship among the genome sets. We used AAI-profiler¹ (Medlar et al., 2018) as an additional method to predict the taxonomic identity of the S. glacialimarina TZS-4_T. Open reading frames were predicted using GenMark.hmm prokaryotic version 3.25 (Besemer, 2001) applying the same search parameters as used for S. frigidimarina NCIMB400. The genome sequence has been submitted to GenBank under the accession number CP041216.

The size and shape of S. glacialimarina TZS-4_T were determined by transmission electron microscopy (TEM). The cells were grown in rMB media at 15°C until they reached an OD₆₀₀ of 0.6, which corresponds to approximately 10⁸ cells/ml. Later, 2μl of the culture was incubated on carbon-coated Cu mesh grids for 1min and then negative stained with filtered, neutral 2% uranyl acetate for 15s. TEM examination was conducted using a JEOL JEM-1400 microscope (Jeol Ltd., Tokyo, Japan) operating at 80kV and equipped with Gatan Orius SC 1000B bottom-mounted CCD camera (Gatan Inc., United States).

For scanning electron microscopy (SEM), cells were grown to OD₆₀₀ of 0.6, as described in section Growth Conditions. Cells were harvested from 10ml of culture by centrifugation (5min, 3,200g, 4°C) and resuspended in the original volume with phosphate buffer saline (PBS). Cells were fixed with 2.5% v/v glutaraldehyde at 4°C for 20h. After fixation, the cells were washed twice with PBS using centrifugation and resuspended in half of the original volume of PBS. Cell suspension (100μl) was spread on concanavalin A-coated glasses, fixed with osmium tetroxide, and dehydrated through ethanol series (50, 70, and 96%). Dehydrated cells were incubated in hexamethyldisilazane overnight and coated with platinum. Finally, SEM examination was conducted using FEI Quanta 250 Field Emission Gun (FEG) Scanning Electron Microscope.

Starter culture of S. glacialimarina TZS-4_T was used to inoculate a 50ml culture as described in section Growth Conditions and grown until it reached OD₆₀₀ of 0.8. Samples were prepared at different temperature conditions, i.e., 5, 15, and 25°C, and the pellet from 5ml culture was washed five times with PBS. Finally, the washed pellet was used for the analysis of fatty acid (FA) composition by gas chromatography. The bacterial pellets were subjected to transmethylation by heating with 1% v/v H₂SO₄ in methanol under nitrogen, and the formed FA methyl esters were extracted with hexane (Christie, 1993). The samples were dried with anhydrous Na₂SO₄ and concentrated. The FA methyl esters were identified based on their mass spectra recorded by GCMS-QP2010 Ultra (Shimadzu Scientific Instruments, Kyoto, Japan) and compared to the spectra of several authentic standard mixtures (including Supelco 47,080-U Bacterial Acid Methyl Ester BAME Mix) and published reference mass spectra (Christie, 2019). Quantitative composition was determined by using a Shimadzu GC-2010 Plus with flame ionization detector. The responses were corrected by using the theoretical correction factors for this detector (Ackman, 2007). Both gas chromatographs were equipped with a Zebron ZB-wax capillary column (30m, 0.25mm ID and film thickness 0.25μm; Phenomenex, Torrance, CA, United States). The FA compositions are expressed as mol% profiles, and the FAs are marked using the following abbreviations: (carbon number):(number of double bonds) n-(position of the first double bond calculated from the methyl end) (e.g., 16:1n-7 for palmitoleate). The identified iso and anteiso branches and the 3-OH substituents of the FAs were indicated. The FA components exceeding 0.1mol% at least in one sample were listed, and the remaining were summed as trace FAs. The fold change analysis of unsaturated FA relative concentrations was performed w.r.t 15°C samples as reference. SD was calculated using propagation of error.

RNA isolation was performed using acidic phenol:bromochloropropane (BCP; Chomczynski and Mackey, 1995; Gregorova et al., 2020). The cells were grown from the starter culture in 50ml of rMB until it reached OD₆₀₀ of 0.8. The cells were harvested at 3200g, 4°C for 10min and stored in −80°C. The pellet was resuspended in 4ml of 0.9% w/v NaCl solution followed by 4ml of acidic phenol (Sigma) and 800μl of BCP (Acros organic). Glass beads were added to break the bacterial cells and vortexed for 10min at room temperature (RT). The lysate was centrifuged at 10000g for 10min at RT, and aqueous phase was collected. The aqueous phase was re-extracted twice with phenol:BCP (2ml of acidic phenol and 400μl of BCP). Total RNA from the aqueous phase was precipitated by adding 2.5 vol of 99.6% v/v EtOH, at −20°C overnight and pelleted by centrifugation at 10000g for 20min at 4°C. The RNA pellets were air-dried and resuspended in RNAase-free ddH₂O water. RNA concentration was measured using a Nanodrop 2000 spectrophotometer (Thermo Scientific) and run on 2% w/v agarose, Tris-borate EDTA (TBE) gel with of Midori green (4μl per 100ml of 2% agarose) for quality assessment. The images were captured on Gel-doc XR (Biorad).

Primers were designed for each gene using the IDT PrimerQuest tool² (Supplementary Table ST-1). The purified RNA (30μg) was DNAase treated with 7U of RQ1 RNAase-Free DNase (Promega). cDNA was synthesized using 200U of Maxima reverse transcriptase primed with 1μl random hexamers (0.2μg/μl) (Thermo Scientific) and 3μg of DNase-treated RNA. Both DNAase treatment and reverse transcription reactions were performed according to the manufacturers’ instructions. cDNA synthesis was validated by PCR (using primers listed in Supplementary Table ST-1), and the products were analyzed on a 2% agarose gel (Supplementary Figure SF-1). Quantitative PCR reactions were performed using Perfecta SYBR green FastMix, Low ROX (Quantabio) consisting of 5μl of 2× Perfecta mix, 0.4μl of 2μM primer mix (fwd and rev primers), adjusted to 10μl of final volume with ddH₂O water. Control samples were included to check amplification arising from contaminating genomic DNA and from the primer-dimer formation. All the samples were run in technical triplicates on Quantstudio 3 Real-time PCR system (Thermo Fisher Scientific). The transcript was amplified using the following conditions: 95°C for 3min followed by 50cycles of 95°C for 30s, 62°C for 30s, and 72°C for 30s. A melting curve profile was generated to determine the formation of a single amplification product. Primer efficiency for all the genes was determined using the formula Efficiency (%)=100×(10^-1/slope-1) with 5-fold dilution series of template cDNA. Comparison of fold change between different target genes was using ∆∆C_T method, and gyrase A (gyrA) gene was selected as reference for normalization (Livak and Schmittgen, 2001). Data analysis and statistical tests for RT-qPCR were performed on GraphPad Prism 9.

The RT-qPCR primers (Supplementary Table ST-1; housekeeping genes gyrA, 16S rRNA, and recA, and target genes nanA and nanH) were tested on cDNA using end-point PCR. The PCR reaction conditions were 95°C for 2min followed by 35cycles of 95°C for 30s, 62°C for 30s and 72°C for 30s. The amplification temperature of the primer pairs was confirmed using gradient PCR from 60 to 64°C (data not shown). The products of PCR reactions were analyzed on a 2% agarose TBE gel. The images were captured using a Gel-doc XR (Biorad).

GenBank accession numbers of Shewanella whole genome sequences used for fastANI and UBCG analysis were as follows: NZ_CP050313.1, NZ_CP047422.1, NZ_CP018456.1, NZ_CP033575.1, NZ_CP046378.1, NC_008700.1, NC_017571.1, NC_017579.1, NC_009052.1, NC_009665.1, NC_009997.1, NC_011663.1, NC_016901.1, NZ_CP028730.1, NZ_CP028355.1, NZ_LS483452.1, NZ_CP022358.1, NZ_CP045857.1, NC_007954.1, NZ_CP041783.1, NC_008345.1, NC_010334.1, NZ_CP020472.1, NZ_CP020373.1, NZ_CP034015.1, NC_009092.1, NZ_CP041153.1, NZ_CP022272.1, NZ_CP036200.1, NC_004347.2, NC_009901.1, NC_011566.1, NZ_CP041036.1, NZ_CP014782.1, NC_017566.1, NC_009438.1, NZ_CP046329.1, NZ_LR134321.1, NZ_LR134303.1, NZ_CP028435.1, NC_009831.1, NC_008577.1, NZ_CP048031.1, NZ_CP022089.2, NZ_CP039928.1, NC_008321.1, NC_008322.1, NZ_CP015194.1, NZ_CP041329.1, NZ_CP041151.1, NC_008750.1, NZ_CP023019.1, NZ_CP041614.1, NC_014012.1, and NC_010506.1.

Bacterial strain S. glacialimarina TZS-4_T was originally isolated from the Baltic Sea ice outside of Tvärminne Zoological station, Hanko, Finland, and cultivated on Zobell media containing undefined Baltic Sea water (Luhtanen et al., 2014). To standardize the growth media and remove any influence of seasonal fluctuation in seawater composition, we investigated if marine broth (MB) could be used to cultivate S. glacialimarina TZS-4_T. First, the salt concentration of the Baltic Sea water used was determined to be 6.9g/L, which is consistent with brackish coastal water. We then compared the growth in Zobell media and in rMB, which has a defined composition with a total salt concentration of ~7.8g/L. rMB was found to be well suited for S. glacialimarina TZS-4_T and supports a faster growth rate and a higher cell density at stationary phase (Supplementary Figure SF-2). Based on the favorable growth characteristics, rMB was adopted for all subsequent experiments in this study.

Genomic DNA sequencing of S. glacialimarina TZS-4_T was performed on a PacBio Sequel platform and assembled using hierarchical genome-assembly process (HGAP) analysis (Chin et al., 2013), achieving a coverage depth of 90.54×, indicating a high read confidence. The number of PacBio sequence reads was 57,211 with a N50 read length of 16,906bp. Its genome comprises a single circular chromosome of 4,456kb with a GC content of 41.12%, featuring 3,906 predicted open reading frames (ORFs) with 3,725 protein-coding sequences (Figure 1).³ There are a total of 118 RNA coding genes, 91 of which encode for tRNA and 22 for rRNA, of which eight encode for 5S rRNA, seven for 16S rRNA, and seven for 23S rRNA. In addition, two complete and one incomplete prophage segments were identified (Figure 1) by phispy tool on the RAST server and PHASTER (Arndt et al., 2016) with hits to predicted proteins of, e.g., phage 1/44, which was previously reported to infect the cold-active Shewanella sp. #44 strain (Luhtanen et al., 2014; Senčilo et al., 2015).

Next, we wanted to position S. glacialimarina TZS-4_T in regard to other Shewanella species. To this end, we constructed phylogenetic trees based on the 16S rRNA gene by maximum likelihood, neighbor-joining, and maximum parsimony methods with 100 replicates as bootstrapping and compared the tree topologies. The trees included 29 Shewanella strains representing 13 different phenotypic sub-groups with Vibrio and Pseudomonas strains as out-groups. The phylogenetic analyses yielded equivalent results (Supplementary Figure SF-3), with maximum likelihood showing that S. glacialimarina TZS-4_T forms a differentiated branch with S. aestuarii SC18 (97.25%) and S. denitrificans OS217 (96.53%) branching next to it (Figure 2A). Similar results were obtained using AAI-profiler which compares the predicted proteome of an organism to Uniprot database (Supplementary Figure SF-4; Medlar et al., 2018). Limiting the phylogenetic analysis to the 92 most conserved genes yields a similar distribution, with cold-active Antarctic and Arctic Sea bacteria, such as Shewanella sp. Arc9-LZ, S. aestuarii SC18, S. polaris, and S. frigidimarina clustering alongside S. glacialimarina TZS-4_T (Supplementary Figure SF-5).

To obtain independent confirmation, two bioinformatic tools – fastANI and orthroANI – were used to calculate the average nucleotide identity (ANI) values for S. glacialimarina TZS-4_T in relation to other Shewanella species. For fastANI analysis, the comparison was made against whole genome sequences (WGS) of 55 Shewanella strains available in NCBI. The result showed that S. aestuarii yielded the highest ANI score with 611 matching sequences S. glacialimarina TZS-4_T, followed by S. Arc9-Lz and S. frigidimarina (Supplementary Table ST-2). Further comparison was done using orthroANI on a limited subset of nine closely related Shewanella species, which delineated S. glacialimarina TZS-4_T (ANI=75.31) along with S. aestuarii (ANI=76.08; Figure 2B). The ANI scores derived from fastANI and orthroANI are both below 95%, which is currently considered as the benchmark value for classifying bacterial species (Richter and Rosselló-Móra, 2009). Taken together, this implies that S. glacialimarina TZS-4_T is genetically divergent and constitutes a distinct Shewanella species, which clusters with other cold-active marine Shewanella species.

To determine the characteristics of S. glacialimarina TZS-4_T, we grew the bacteria on rMB-agar where they form smooth, circular, and convex colonies with a diameter of 1–2mm and a milky white appearance. Gram staining confirmed that S. glacialimarina TZS-4_T is a Gram-negative, rod-shaped bacterium. Interestingly, after 3days of growth the colonies turn reddish-brown. A similar phenomenon was observed when the bacteria were grown for up to 29h in rMB-broth at 15 and 25°C (Supplementary Figure SF-6). We also found that pigment formation is not affected by light intensity (data not shown), but the cells are less pigmented at low growth temperatures (Supplementary Figure SF-6). This suggests that the pigment may have a role in temperature or UV-radiation adaptation, as was previously proposed for a S. frigidimarina strain isolated from Antarctic glacier snow (Martin-Cerezo et al., 2015), where the red pigment in question was identified as cytochrome c3. Indeed, a closer inspection of the S. glacialimarina TZS-4_T genome revealed genes encoding for a putative cytochrome c3 family protein.

Next, we examined the physiology of S. glacialimarina TZS-4_T using transmission electron microscopy (TEM). This showed a rod-shaped bacterium 0.5μm in diameter and ~2μm in length (Figure 3A) with clearly visible long polar flagella (Figure 3B). We also observed various bead-shaped extensions surrounding the bacteria. To visualize these, SEM analysis was performed, exposing a granular structure extending from the outer membrane of the bacteria (Figure 3C; highlighted in the insert). These extensions resemble a previously identified nanowire-like structure in S. oneidensis MR-1 (Gorby et al., 2006). The nanowires function as channels for extracellular electron transport in bacterial communities, allowing electron transfer from external surfaces, such as oxidized metals (Pirbadian et al., 2014). Genes encoding for nanowire-like structures, such as omcA/mtrC, were also identified in the S. glacialimarina TZS-4_T genome, but the potential function of these structures is unclear and remains to be determined.

To further study the phenotypic characteristics of S. glacialimarina TZS-4_T, we performed a number of standardized panel tests, comparing the results to the related psychrophilic bacterium S. frigidimarina (type strain ACAM 591) and the mesophilic reference bacterium S. baltica (type strain LMG 2250; Table 1). Shewanella glacialimarina TZS-4_T is a motile, facultatively anaerobic bacteria that utilizes glucose and maltose as its sole source of carbon. However, the bacterium does not produce H₂S, which is a characteristic feature of many Shewanella strains associated with fish spoilage in cold storage. Furthermore, S. glacialimarina TZS-4_T is also capable of cytochrome oxidase and catalase production.

Next, to determine the growth rate, we explored different physiological conditions, such as temperature, pH, salinity, and hydrogen peroxide induced oxidative stress. We observed that S. glacialimarina TZS-4_T is psychrophilic and grows from low to ambient temperatures in a range between 0 and 25°C, with an optimum at 15°C (Figure 4A; Supplementary Figure SF-2). However, growth at 0°C requires preconditioning (growth at 4°C) of the starter cultures, as those grown at 15°C do not cope with the cold-shock and yield no growth at freezing conditions. During the first 6h post-inoculation, the growth rate is the slowest at 0 and 4°C, whereas it is the fastest at 25°C. On the other hand, there is an earlier cessation of growth at high temperature, as the cell density recorded at 24h post-inoculation is lower at 25°C than at 15°C (Figure 4A). Accordingly, at 24h we also observed a lower number of viable cells at 25°C (~1 ×10⁹cfu/ml) than at 15°C (~8 ×10⁹ cfu/ml), and the cells lose their ability to grow altogether at 28°C (data not shown). Indeed, a significant loss of viability is also observed during the 24–48h post-inoculation window (Supplementary Figure SF-2), which is consistent with the culture entering the death phase. S. glacialimarina TZS-4_T has a propensity for neutral pH conditions with an optimum at pH 7.0, although it tolerates pH ranging from 5.0 to 9.5 (Figure 4B). It also has a high NaCl tolerance, with 35 and 55g/L NaCl reducing growth but not abolishing it (Figure 4C). Since the oxygen solubility from the environment is increased at low temperature (Karbowiak et al., 2009), we analyzed the ability of the cells to tolerate reactive oxygen species by supplementing the growth media with different concentrations of hydrogen peroxide (H₂O₂; Figure 4D). The cells grow in the presence of 1mM H₂O₂, but 2mM and 4mM H₂O₂ concentrations resulted in the complete cessation of growth.

Finally, we investigated the antibiotic resistance of S. glacialimarina TZS-4_T and found it to be susceptible to chloramphenicol, kanamycin, and tetracycline, but resistant to ampicillin (Table 1). Indeed, sequence analysis revealed that S. glacialimarina encodes resistance genes for ampicillin, including class-D β-lactamase and metallo-β-lactamase. Other virulence-associated factors, such as putative T1SS-secreted agglutinin RTX (cytotoxin), putative collagenase genes, and clp proteases, were also identified from the genomic sequence.

To further investigate the cold adaptation of S. glacialimarina TZS-4_T, we analyzed the FA modulation as it is a well-known strategy of cold-active bacteria to maintain membrane fluidity and viability. Our analysis revealed that palmitoleic acid (16:1n-7) was the most prevalent FA component of S. glacialimarina TZS-4_T, constituting 32–37mol% of the total FAs, with the highest values detected at 5°C. In general, iso-13:0 and iso-15:0 accounted for more than 10mol% and were followed by 12:0, 14:0, and 16:0 that constituted close to 5% (Supplementary Table ST-3). A significant increase of certain unsaturated FAs, including 7-tetradecenoic acid (14:1n-7), eicosatetraenoic acid (20:4n-3), stearidonic acid (18:4n-3), and eicosapentaenoic acid (20:5n-3), was observed at low temperatures (Figure 5). Indeed, at 5°C the ratio of 20:5n-3 had raised close to 6mol%. A previous study on S. electrodiphila MAR441 showed similar temperature-dependent regulation of the relative concentration of certain n-3 FAs that have been associated with improved membrane fluidity and increased cell viability at low temperatures (Zhang and Burgess, 2017).

Interestingly, S. glacialimarina TZS-4_T features a complete sialic acid metabolism gene cluster in its chromosome, including nanA that encodes for the major catalytic enzyme N-acetylneuraminate lyase and the sialidase encoding gene nanH. Upon blastn search, two distantly related Shewanella species, S. pealeana and Shewanella sp. YLB-09, also contain a similar sialic acid metabolism gene cluster (Figure 6A). Phylogenetic comparison using nanA revealed that among Shewanella, only these species carry this gene cluster, which is otherwise commonly found among pathogenic bacteria, such as Vibrio cholerae and Yersinia pestis (Figure 6B). Since other Shewanella species, such as S. putrefaciens, have been reported to cause serious health disorders among freshwater fish (Paździor et al., 2019), we hypothesize that the sialic acid metabolism gene cluster may play a role as a putative pathogenicity factor in S. glacialimarina TZS-4_T. To determine if the nanA and nanH genes are transcribed at normal growth conditions, we performed an RT-qPCR expression analysis at 5, 15, and 25°C. Our results show that nanA and nanH are stably expressed at 5 and 15°C, although a significant upregulation occurs at 25°C w.r.t 15°C (Figure 6C). Furthermore, nanA and nanH expression remains stable throughout the exponential growth phase and only a slight, albeit not statistically significant downregulation is observed at the stationary phase for cells grown at 5°C (Supplementary Figure SF-7). This shows that the sialic acid enzymes genes are transcribed in S. glacialimarina TZS-4_T and there is a significant temperature-dependent expression profile. However, the upregulation seen at 25°C suggests that the sialic acid metabolism gene cluster most likely does not further cold-active growth or a role in food spoilage during cold storage.