OUP user menu

An ice-binding protein from an Antarctic sea ice bacterium

James A. Raymond, Christian Fritsen, Kate Shen
DOI: http://dx.doi.org/10.1111/j.1574-6941.2007.00345.x 214-221 First published online: 1 August 2007


An Antarctic sea ice bacterium of the Gram-negative genus Colwellia, strain SLW05, produces an extracellular substance that changes the morphology of growing ice. The active substance was identified as a ∼25-kDa protein that was purified through its affinity for ice. The full gene sequence was determined and was found to encode a 253-amino acid protein with a calculated molecular mass of 26 350 Da. The predicted amino acid sequence is similar to predicted sequences of ice-binding proteins recently found in two species of sea ice diatoms and a species of snow mold. A recombinant ice-binding protein showed ice-binding activity and ice recrystallization inhibition activity. The protein is much smaller than bacterial ice-nucleating proteins and antifreeze proteins that have been previously described. The function of the protein is unknown but it may act as an ice recrystallization inhibitor to protect membranes in the frozen state.

  • sea ice
  • bacterium
  • ice-binding protein
  • Colwellia


Sea ice is the home of numerous species of bacteria (Bowman et al., 1997, 1998; Staley & Gosink, 1999; Brown & Bowman, 2001; Deming, 2002; Junge et al., 2002; Brinkmeyer et al., 2003; Groudieva et al., 2004; Auman et al., 2006). The ability to survive at temperatures as low as −20°C is a prerequisite for life in sea ice (Thomas & Dieckmann, 2002) and living bacteria have been observed in the liquid phase of sea ice at temperatures as low as −15°C (Junge et al., 2001). Among the many adaptations that have been described in polar psychrophiles (summarized by Walker et al., 2006), only a few have been associated with survival in sea ice. One such mechanism is thought to be secretion of extrapolymeric substances that accumulate in brine pockets in sea ice and help to preserve a liquid environment (Krembs et al., 2002). Another potential freezing tolerance mechanism is ice recrystallization inhibition (RI) activity (the ability to prevent the formation of larger grains of ice at the expense of smaller grains), which is thought to protect membranes from freezing injury. RI activity has been found in a number of freezing-tolerant soil bacteria (Walker et al., 2006; Wilson et al., 2006) and Antarctic lake bacteria (Gilbert et al., 2004). Although proteins with RI activity have been described in a number of cold-hardy plants (Smallwood et al. 1999; Griffith et al., 2005), little is known about the proteins responsible for bacterial RI activity. We isolated an Antarctic sea ice bacterium that produces an extracellular protein that causes the development of facets on growing ice crystals, a phenomenon that is usually associated with RI activity (e.g. Griffith & Yaish, 2004). This bacterium provided an opportunity to investigate the nature of bacterial RI proteins. Here we describe the gene encoding this protein and show that it is similar to genes encoding ice-binding proteins (IBPs) found in a snow mold and sea ice diatoms.

Materials and methods


Sampling was conducted on winter sea ice on the west side of the Antarctic Peninsula (65.187°S, 65.610°W) on 11 September 2002 on Southern Ocean Globec cruise LMG-02-05 (Hofmann et al., 2002). Water samples were obtained with a sterile 50-mL conical centrifuge tube (Falcon) using aseptic techniques from seawater infiltrating the snow on a ∼10-m-diameter pack ice floe. Temperatures within the flooded snow ranged between −1.7 and −1.8°C. Samples were immediately placed into 0–2°C darkened incubators until the ship reached Punta Arenas, Chile, and were then shipped to the Desert Research Institute (DRI) on ice (0°C). At DRI, samples were stored at 0–3°C until used. Chilled agar plates of marine broth (MB) 2216 medium (Difco) were inoculated with serial dilutions of the samples and incubated at 0°C (range −1 to +2°C). About 200 individual heterotrophic colonies were streaked three times and then individual colonies from the streaks were transferred to liquid MB, in which they were subsequently maintained. The isolates, consisting of mostly bacteria and a few fungi, were also frozen in glycerol at −80°C for longer term storage. The ∼200 isolates were placed in 96-well plates and incubated at 0–3°C with shaking. Growth rates were determined by measuring the absorbance at 660 nm every 12 h for 13 days using a Molecular Devices, Spectramax Plus plate reader. Twenty isolates with the highest growth rates were selected for further characterizations.

Ice-binding activity and recrystallization inhibition activity

Ice-binding activity was assayed with ice single crystals, i.e. ice crystals having a single crystallographic orientation, whose flat surfaces (basal planes) are perpendicular to the ice c-axis. The growth of such crystals was examined in the solutions to be tested (Raymond, 2000). The development of irregular, faceted fingers (dendrites) on the crystal edges and pits and other irregularities on the basal plane are indications of binding of an impurity to the crystal faces (Buckley, 1951).

RI activity of a solution was assayed by flash freezing 3-μL drops with a liquid-nitrogen-cooled glass slide (Raymond & Fritsen, 2001). The frozen drops were kept at subzero temperatures and periodically examined with a microscope to observe recrystallization

Ice-binding protein purification

At the University of Nevada, Las Vegas, supernatants of the spent culture medium of 17 of the 20 cold-growing isolates were screened for the presence of IBPs with the above-described ice-binding assay. The isolate with the greatest activity was selected for further study.

IBP was purified from the culture medium supernatant by exploiting the IBP's affinity for ice (Raymond & Fritsen, 2001) using three freeze–thaw cycles. In each cycle, the sample was adjusted to an osmolality of c. 300 mOsm kg−1 and frozen overnight in 500-mL centrifuge bottles at c. −15°C. The bottles were centrifuged upside down to expel brine and impurity proteins and then thawed at room temperature. The purified sample was confirmed to have ice-pitting activity. Purity was checked with two-dimensional (2D) electrophoresis with SYPRO ruby staining at the Proteomics Center, University of Nevada, Reno. De novo sequencing of peptides from the spot of interest was done by tandem MS/MS at the Molecular Structure Facility, University of California, Davis, as described previously (Janech et al., 2006). Many attempts to amplify products from the bacterial DNA using degenerate primers based on these peptides were made without success.


DNA was prepared with a Promega DNA kit and shotgun sequenced in one run by 454 Life Sciences (Branford, CT) as described (Margulies et al., 2005), yielding 218 000 DNA fragments of about 100 bp each. Fragments corresponding to the MS peptides, and other overlapping fragments, were found using tblastn and blastn and assembled with bioedit (http://www.mbio.ncsu.edu/BioEdit/page2.html). The sequence was confirmed by PCR amplification and sequencing.

The 16S rRNA gene was amplified and sequenced with universal primers (9bF, 5′-GRGTTTGATCCTGGCTCAG-3′; 519uF, 5′-CAGCMGCCGCGGTAATAC-3′; 1512uR, 5′-ACGGHTACCTTGTTACGACTT-3′). The 16S rRNA gene sequence was confirmed and extended to full length with the above described 454 DNA fragments. Identification of the start and end of the 16S rRNA gene was based on the start and end signals of the 16S rRNA gene of Escherichia coli K-12 (accession no. U00096).

Molecular mass and pI of the predicted protein were predicted with protein calculator v3.2 (http://www.scripps.edu/~cdputnam/protcalc.html). Signal peptides were predicted with SignalP (http://www.cbs.dtu.dk/services/SignalP/).

Expression of recombinant IBP

The IBP gene was amplified by PCR with primers 5′-ATGAAAACCTTAATAAGCAATTCGAAAAAAGTA-3′ and 5′-GTTATAGAGGAGCTTCTTCATATTGCTC-3′. The resulting 760-bp product was inserted in the pEXP-NT/TOPO vector (Invitrogen), which was then transformed into One-Shot competent E. coli cells (Invitrogen). Plasmids were purified from several clones with a Wizard Plus SV miniprep kit (Promega) and sequenced. A plasmid in which the IBP insert had the correct orientation and reading frame with respect to an N-terminal His tag was then incubated with the reaction mixture of the Expressway cell-free expression system (Invitrogen) for 4–6 h at c. 30°C. The supernatant from the reaction mixture (100 μL) was mixed with 100 μL water, and in some cases 5 μL 4.0 M NaCl, resulting in a final osmolality of about 600 mOsm kg−1, and then assayed for ice-binding activity as described above. A reaction mixture in which the DNA was replaced with an equal volume of water served as a negative control.


Ice-binding protein distribution and characteristics

Following incubation of the ∼200 isolates of sea ice bacteria at 0–3°C, 17 of the 20 fastest growing isolates were examined for ice-binding activity. Of these, one (strain SLW05) showed relatively strong activity, eight showed very weak activity and the rest showed no activity. The 16S rRNA gene sequence of SLW05 (accession no. DQ788794) was nearly identical (>99.7% identity) to the 16S rRNA gene sequences of several Arctic and Antarctic sea ice bacteria of the Gram-negative genus Colwellia, the closest being Colwellia sp. (accession no. DQ060396) from Arctic sea ice and Colwellia piezophila (accession no. AY771755) from the Arctic. Accordingly, we tentatively identify our isolate as Colwellia sp. SLW05. The Arctic marine sediment bacterium Colwellia psychrerythraea strain 34 H, whose genome has been sequenced (Methe et al., 2005), was one of the more distantly related Colwellia species (97% identity).

Spent culture medium of Colwellia sp. SLW05 cells contained a substance that caused growing ice to develop pits on the ice basal plane and irregularly faceted dendrites on the crystal edges (Fig. 1a and b). Ice growing in pure culture medium does not show these features (Fig. 1c). In addition, spent culture medium has RI activity (Fig. 2), i.e. an ability to prevent the growth of large grains of ice at the expense of small grains. The ability to deform the surface of ice and inhibit recrystallization are both indications of the presence of one or more IBPs. Thermal hysteresis, the difference between the freezing point and melting point, of the spent culture medium was <0.1°C, suggesting that no proteins in the spent medium have significant antifreeze activity.


Effects of Colwellia sp. SLW05 on ice and isolation of an ice-binding protein from the medium. (a, b) Two views of the basal plane of an ice single crystal growing in spent culture medium, showing faceted, irregular growth on the edges (a) and pitting (b). No such features develop in the presence of uninoculated culture medium (c). Scale bar in (a), 1 mm (approximately the same as in b and c). (d) Two-dimensional electrophoresis of ice-binding fraction of Colwellia sp. SLW05 culture medium. Arrow indicates IBP. Molecular weight ladder in kilodalton is shown on the left. pI scale is shown at the bottom. Portion of gel from pI 5 to pI 9 (not shown) has no significant spots.


Recrystallization inhibition activity of spent culture medium of Colwellia sp. SLW05. Drops of culture medium (bottom left) and 20 μg mL−1 bovine serum albumin (BSA) at the same osmolality (top left) were flash frozen and allowed to recrystallize for 24 h at −4.2°C. Scale bar, 1 mm.

IBP sequence

Ice affinity-purified IBP (prepared from cell-free spent culture medium) was confirmed to have ice-pitting activity and yielded two spots on a 2D electrophoretic gel (Fig. 1d). MS/MS de novo sequencing yielded two peptides [VTLTG(Q/GA)LAK and LSVNTGTTVDGR] for the lower spot (∼24 kDa, pI ∼4.0) and one peptide [(MN)FQYHELVP(AR/VK)] for the upper spot (∼53 kDa, pI ∼3.7). None of these sequences was found in the C. psychrerythraea 34 H genome (accession no. NC_003910). The two peptides from the lower spot weakly matched portions of an IBP-like domain in a hypothetical protein in the genome of Shewanella denitrificans (IYLTGGAQAK_ and ITVNTGASVNGR of accession no. YP_562921). The IBP domain of this protein was previously shown to be similar to IBPs from Antarctic sea ice diatoms and cold-adapted fungi (Janech et al., 2006). Attempts to use this sequence information to obtain the Colwellia DNA sequence by PCR were unsuccessful, and prompted us to sequence the whole genome by the 454 method (Margulies et al., 2005). Several Colwellia sp. SLW05 DNA fragments from the 454 sequence data matched the two peptide sequences from the lower spot, and by assembly of overlapping sequences were found to be closely located. With these sequences and other overlapping sequences, an ORF was identified that contained both peptide sequences (accession no. DQ788793; Fig. 3). The sequence was confirmed by PCR amplification and sequencing. No DNA fragments matching the sequence of the upper spot were found.


Predicted nucleotide and amino acid sequences of ice-binding protein of Colwellia sp. SLW05 and 5′ and 3′ untranslated regions. Underlined sequence indicates probable secretory signal. Double-underlined sequences are peptides identified by MS/MS sequencing. Codon in bold is stop signal.

The ORF consists of 759 nucleotides encoding 253 amino acids. The calculated molecular mass and pI were 26 350 Da and 4.61, respectively, in approximate agreement with the lower spot in Fig. 1(d). The reason for the lower observed pI (4.0) is unclear but it may be a result of phosphorylation of Ser, Thr or Tyr residues. The closeness of the apparent and predicted sizes of the protein suggest that it has little or no carbohydrate or lipid moieties. The protein appeared to have a 27-residue secretory signal, which is consistent with its presence in the spent culture medium. As was hinted at by the MS/MS peptides, the Colwellia IBP was very similar to several previously identified IBPs and hypothetical proteins with full IBP-like domains from a wide variety of organisms, including sea ice bacteria, a sea ice diatom and a snow mold. Representative sequences with 40–61% identities and 59–75% similarities to the Colwellia sequence are shown in Fig. 4.


Alignment of amino acid sequence of Colwellia sp. SLW05 IBP (accession no. DQ788793) with IBP sequences from the sea ice diatom Navicula glaciei (accession no. DQ062566; 40% identity, 59% similarity in the region of overlap) and the snow mold Typhula ishikariensis (AB109745; 42%, 60%), putative IBPs from the sea ice bacteria Polaribacter irgensii (ZP_01118128; 61%, 75%) and Psychromonas ingrahamii (ZP_01349469; 59%, 71%) and an IBP-like domain of a hypothetical protein from the marine bacterium Shewanella frididimarina (YP_749708; 52%, 69%).

To confirm that the gene encodes an ice-binding protein, the gene was expressed in an E. coli cell-free expression system. Reaction mixtures containing the IBP expression vector had ice-binding activity as shown by the development of irregular faceting on the edges of the crystals and pitting on the basal plane, whereas the negative controls did not induce such features (Fig. 5a). In addition, the reaction mixture had RI activity whereas the control did not (Fig. 5b). These results confirm that the IBP gene encodes an ice-binding protein, and they also strongly suggest that the ice-binding activity and RI activity of the spent culture medium (Figs 1 and 2) are due to the IBP. As the expression system can produce only simple proteins, these results show that the IBP does not require a lipid or carbohydrate moiety for activity.


Ice-binding and recrystallization inhibition activities of recombinant IBP. (a) Images of ice single crystals (ice c-axis normal to page) after growing for c. 1 h at −1.1°C in the supernatant of the reaction mixtures of a cell-free expression system without (left) and with (right) an IBP expression vector. The two upper panels represent reactions that were made at the same time with the same reaction mixture except for the presence of the IBP expression vector. The two lower panels represent a similar pair of reactions made on a different date. Ice-binding activity is indicated by irregular faceting on the edges of the crystal and pitting (arrow) on the basal plane, which are absent in the controls. Scale bar, 1 mm.Scale bars in other panels are approximately the same. (b) Recrystallization of flash-frozen supernatants of reaction mixtures with and without the IBP expression vector kept at −5.2°C for the indicated times. All panels have the same scale. Scale bar, 1 mm.

Homologs of a domain in the Colwellia sp. SLW05 IBP (VWIMQISGNLNQANAKRVTL) were found to be encoded in three other sequences in the Colwellia sp. SLW05 genome (with predicted amino acid lengths of 105, 121 and 303 residues). These sequences did not closely match any proteins in the databases, although weak matches (∼30% identity) were found for some hypothetical proteins of unknown function in the sea ice bacterium Psychromonas ingrahamii (accession nos. ZP_01349491, ZP_01349009 and ZP_01349976).


Cold-adapted bacteria produce a number of proteins that interact with ice, including several ice-nucleating proteins (INPs) with typical sizes of 120–180 kDa (Gurian-Sherman & Lindow 1993; Kawahara, 2002), and several antifreeze proteins (AFPs), i.e. proteins that noncolligatively depress the freezing point (Table 1). The Colwellia sp. IBP is much smaller than the INPs and AFPs, and unlike some of them has little or no lipid or carbohydrate moiety.

View this table:

Bacterial antifreeze proteins

SpeciesSourceTypeMol. wt
Pseudomonas putidaArctic rhizobacteriumLipoglycoprotein164INAXu et al. (1998); Muryoi et al. (2004)
Marinomonas primoryensisAntarctic saline lakeUncharacterized protein>1002RIA(Gilbert et al. 2004, 2005)
Pseudomonas fluorescensAntarctic soilUncharacterized protein80Kawahara et al. (2004)
Moraxella sp.Antarctic soilLipoprotein52No INAYamashita et al. (2002)
Flavobacterium xanthumAntarctic soilUncharacterized protein59RIAKawahara et al. (2007)
  • * INA, ice nucleating activity; RIA, recrystallization inhibition activity.

  • P. Davies, pers. commun.

The resemblance of the Colwellia IBP to the IBPs of sea ice diatoms and a snow mold raise the question of whether the eukaryotic IBPs were acquired from bacteria by horizontal gene transfer (HGT). One argument in favor of HGT is the finding of ‘restricted occurrence’ (Doolittle, 2002), in which a gene restricted to a small set of organisms in one domain is also found in a small set of organisms from a different domain. The IBPs seem to satisfy this criterion because among the hundreds of organisms whose genomes have been fully sequenced, IBP and IBP-like genes are found in only about a dozen that are irregularly scattered over the three domains of life. On the other hand, a rule of thumb for HGT is that amino acid sequence identities should be at least 60% (Doolittle, 2002), which is well above the 40–42% identities between the Colwellia IBP and the eukaryotic IBPs. Thus, independent evolution of the IBPs cannot be ruled out. IBP sequences from a greater number of organisms will be needed to determine to what degree HGT is involved in their appearance.

Other IBPs and IBP-like proteins that resemble Colwellia IBP are described by Janech et al. (2006). The latter include four prokaryotic hypothetical proteins of unknown function that contain an entire IBP-like domain. Curiously, the non-IBP parts of these proteins appear unrelated. None of these four prokaryotes is noted for being cold-hardy (although related species are well known to occur in sea ice) and so their IBP-like proteins may have functions unrelated to freezing tolerance. The function of a large hypothetical protein (68 kDa) from the Antarctic sea ice bacterium Psychroflexus torquis that contains a full IBP domain (accession no. ZP_01254171; submitted by Bowman et al.; 26% identity to Colwellia IBP) is also unknown. Identifying the functions of these proteins may elucidate the origins of IBPs.

It is also interesting that an IBP gene does not occur in the genome of C. psychrerythraea 34 H, a species isolated from presumably ice-free Arctic marine sediments. However, C. psychrerythraea has also been isolated from sea ice (Bowman et al. 1997; Groudieva et al. 2004). Similarly, hypothetical IBPs are present in the genomes of the Antarctic sea ice bacteria Polaribacter irgensii (submitted by A. Murray & J. Staley) and Psychromonas ingrahamii (submitted by A. Copeland et al.), but not in the genomes of the Antarctic marine bacteria Actinobacterium PHSC20C1 (accession no. NZ_AAOB00000000) or Pseudoalteromonas haloplanktis TAC125 (accession nos. NC_007481 and NC_007482). Together, these results suggest that the IBP has a role related to survival in an icy environment.

Because Colwellia IBP has RI activity, it may serve to protect bacterial membranes from the damaging effects of ice recrystallization. The IBP of the sea ice diatom Navicula glaciei is similar to the Colwellia IBP in sequence (40% identity, 59% similarity) and size, and both IBPs have secretory peptides and are found in the extracellular medium. Semipure Navicula IBP has RI activity and has been shown to increase the freeze–thaw survival of both Navicula glaciei cells and human red blood cells (Raymond & Janech, 2003; Kang & Raymond, 2004). In the latter case, the greatest protective effect was observed under conditions that favored recrystallization, suggesting that the protective effect was due to inhibition of recrystallization. RI activity has been associated with freezing tolerance in several species of soil bacteria (Wilson et al., 2006), and has been found to be a characteristic of numerous species of Antarctic lake bacteria (Gilbert et al., 2004) and Antarctic mosses (Doucet et al., 2000). The presence of IBPs in spent culture medium and the fact that they have secretory peptides raises the possibility that they function extracellularly, perhaps in the confines of brine pockets in sea ice.

Other or additional functions of the Colwellia IBP cannot be ruled out. For example, IBPs might help in some way to preserve brine pockets in sea ice, which have been shown to be critical to survival (Krembs et al., 2002). IBPs might also help cells to stick to sea ice (Raymond, 2000), which may provide a more favorable environment than the underlying water. Experimental evidence of such binding has recently been reported by Wilson et al. (2006) who found an RI-active soil bacterium that was about 30 times more likely to be embedded in growing ice than the RI-inactive E. coli.

It should be noted that about half of the sea ice isolates examined showed no signs of IBP activity, so an IBP is not necessarily essential for survival. For example, no IBP-like gene was found in the genome of Psychrobacter arcticus 273–4 (accession no. NC_007204) isolated from a 20 000–40 000-year-old Siberian permafrost core. It is possible, however, that one IBP-producing species of bacteria may provide protection for other species that do not produce an IBP, as was found to be the case with RI-positive soil bacteria (Walker et al., 2006).


We thank Betsy Kreidberg for culturing the sea ice bacterial strains used in this study. We thank Dr. Jae-Shin Kang for screening isolates for IBP activity and for additional Colwellia sp. SLW05 culturing. We thank Kathy Schegg of the Nevada Proteomics Center for 2D electrophoresis, the staff of the Nevada Genomics Center for sequencing of PCR products and Dr Chris Ross for assistance with Blast software. The latter three were each supported by an NIH IMBRE grant, which is gratefully acknowledged. This research was mainly supported with resources from Nevada's NASA EPCCoR cooperative agreement (NCC5-583) and partially supported by a grants from the National Science Foundation (OPP0088000 to J.A.R. and OPP9910098 and OPP9910098 and OPP0421514 to C.F.). We thank Drs M. Janech, B. Hedlund, and F. Van Breukelen for helpful suggestions and two anonymous reviewers for critical readings of the manuscript.


  • Editor: Gary King


View Abstract