IMIS

Frost flowers for microbial analyses were removed from the pond site into sterile 1 L plastic bags using an ethanol‐rinsed spatula. A second scraping over the same surface area yielded the corresponding, operationally defined brine skim, i.e., the surface slush layer. Samples of sea ice were also collected, as described above, along with the samples of seawater and snow from the surrounding area. Samples of frost flowers, the underlying surface slush layer, and snow were melted directly over the shortest possible period (always <12 h, with sample temperature remaining at ≤0°C), while sea ice samples were melted into sterile 0.2 µm filtered brine according to the isohaline approach described by Ewert et al. [2013].
Study Extent: The thin‐ice station, POLY I (74°13.905′N, 20°07.701′W, 29–30 cm thick on 22 March, snow covered with varying thickness), was situated in a recurrent winter polynya region about 3 km off the landfast ice edge.
Method step description:

1. An area of ~2.5 × 7 m was opened near POLY I at 16:00 geomagnetic time (GMT) on 22 March to expose the ocean to the atmosphere (hereinafter referred to as the “pond” site). The opening of the pond was done using a handheld ice saw by cutting smaller segments that then were pushed to the side underneath the ice cover. A time‐lapse camera was installed at the pond site to document the development of frost flowers as the ice formed in situ. Half of the pond was reopened on 24 March at 15:00 GMT, i.e., after about 47 h of the initial pond opening. At this time, the initial ice was ~12 cm thick. The recurrent polynya at this location will occur as open water (as evidenced by satellite imagery just prior to our arrival) or with a young ice cover (like we experienced); frost flowers are known to occur regularly on this polynya ice.
2. DNA was extracted from the different sample types for amplification and sequencing of the 16S ribosomal ribonucleic acid (rRNA) gene using the phenol‐chloroform method, as in Bowman et al. (2013). One patch of frost flowers was sampled to obtain the upper centimeter portions separately from the basal portions. The V3–V5 regions of the 16S rRNA gene were amplified using primers 357F and 926R for 30 cycles. An aliquot of the amplified material, along with positive and negative controls, was visualized on a gel to insure proper fragment length. Amplicons were purified using the GeneJet Purification Kit (Fermentas) and submitted to the Tufts University Sequencing Center, where amplicons underwent a second round amplification for 10 cycles using barcoded primers 517F and 967R. Second round amplicons were gel purified prior to library construction. Sequencing was conducted on the 454 FLX platform (Roche) using titanium chemistry.