The Bowman Lab

We just published the first analysis of a microbial community inhabiting natural frost flowers in the journal Environmental Microbiology Reports.  Our results are a bit surprising, and I’ll get to them shortly.  First, a brief look at what we did.  Frost flowers have received quite a bit of media attention lately.  If you missed the coverage frost flowers are delicate crystal structures that are nearly ubiquitous to the surface of newly formed sea ice (so long as the ice forms under a relatively cold atmosphere).  Salt, organic material, and bacteria from the source seawater are concentrated in frost flowers during the process of sea ice growth and maturation.

Although chemical and biological reactions are typically suppressed by low temperatures, the high concentrations of all these organic and inorganic materials, and the presence of ample energy (from sunlight), means that some ecologically interesting things might be happening here.  For example check out this paper by a group of atmospheric chemists connected with the OASiS (Ocean-Atmosphere-Sea Ice-Snow) project.  Among other things they report a high concentration of formaldehyde in frost flowers, probably the result of the photolysis (sun-driven breakdown) of larger organic molecules.

Our group is interested in what biological processes might be happening in frost flowers.  To develop testable hypothesis however, we needed to make some initial observations about the system.  The most basic observation that a microbiologist typically wants to make is an assessment of community composition; what bacteria are present and their abundance relative to one another.  To do this we relied on an analysis of the 16S rRNA gene, a commonly used taxonomic marker gene for prokaryotes.  By comparing the 16S rRNA gene sequences in frost flowers with those of known bacteria in a database, we can approximate the composition of the frost flower microbial community.

The small lead from which frost flowers and young sea ice were sampled in April, 2010. Not the prettiest picture of frost flowers around, but representative of 90 % of the days in Barrow.

What we found really surprised us, in fact it took several months to wrap our heads around the results.  The heatmap below shows the relative abundance of bacterial genera in our analysis, as determined by one of the four 16S rRNA gene identification methods we used.  Black indicates bacterial genera below detection, the remainder are scaled from white to blue to red (most abundant).  The columns are our samples; there are four frost flower (FF) and four young sea ice (YI) samples.  Several microbial genera, in particular the Methylobacteria, Rhizobium, and Mesorhizobium, are enriched within frost flowers relative to the underlying young sea ice.  These genera are all members of the order Rhizobiales, thus we posit that that Rhizobiales are, in general, enriched in these frost flowers.

Relative abundance of genera by microarray analyis of the 16S rRNA gene (Bowman et al. 2013). Genera belonging to the order Rhizobiales tend to be enriched in frost flowers relative to young sea ice.

This is pretty weird.  The Rhizobiales are certainly not unheard of in marine waters, but are virtually unreported in sea ice and we can’t find any report of them dominating a marine environment.  And they do truly dominate these frost flowers, in a more in-depth analysis of a single frost flower sample we found that 77 % of the 16S rRNA genes classified as Rhizobiales.  The environment where Rhizobiales do typically dominate is (ironically) on the roots of real flowers, or at least on legumes.  There they engage in a mutualistic relationship with the plant, providing fixed nitrogen in exchange for carbohydrates.

This fact might be a clue as to how Rhizobiales could come to dominate the sea ice surface.  Although there isn’t a lot of evidence for algae (aka phytoplankton trapped in the ice) in these frost flowers, there is for the underlying young ice.  Rhizobiales can be found in close associations with phytoplankton, just as they are with plants, so perhaps the Rhizobiales end up in ice because so many phytoplankton do.  Once in the ice, the now stressed phytoplankton and bacteria might end their relationship (enough stress will test even the closest of relationships), leaving the Rhizobiales free to transport to the ice surface during brine rejection (a key element of sea ice growth).

Of course this is just a hypothesis, but hypothesis development was our goal with this study.  We are left in the best possible situation, with an interesting observation and just enough data to develop further questions (okay, the best possible situation would be a definitive answer and a paper in Nature, but still…).  To refine those questions we are continuing to work with these samples, in particular with metagenomes obtained from one of our frost flower and one of our young ice samples.

In an earlier post I mentioned some odd microbiological observations that our group made during field work in Barrow, AK in 2010.  I also talked about how I’m hoping to repeat that observation this year, using the microscopy technique FISH.  In addition to collecting new data however, there is plenty of work left to be done on the 2010 samples.  I’m spending a lot of time right now working with two metagenomes derived from one young sea ice and one frost flower sample from the 2010 sample set.  A metagenome, as the name suggests, is a compilation of many different genomes.  Consider a liter of seawater.  It might contain around one billion bacteria, and therefor one billion bacterial genomes.  Although bacterial genomes can vary quite a bit even within a given bacterial “species”, for the sake of argument lets say that those one billion bacteria comprise 1000 bacterial species, representing 1000 different genomes.

We’d like to know something about what the bacterial assemblage in our liter of water is doing.  Are they photosynthesizing?  Consuming high molecular weight organic compounds?  Living free in the water or attached to particles?  Clues to all these questions, indeed to the entire natural history of each species, can be found within their genomes.  It isn’t at all practical however, to sequence all 1000 of these genomes (most belonging to species that you couldn’t even bring into pure culture in the lab without many years of work).  The solution?  Sequence all the DNA contained within the water, never mind which species it originally belonged to!

The resulting mess of sequence data is the metagenome, and allows for the least biased way of assessing the metabolic potential of a microbial assemblage.  To do this a researcher sifts through all the little bits of DNA, usually in the range of 50-100 bp, and assigns a putative function to the gene of origin based on similarity to a known gene in a database.  That’s great, but it would still be nice to know something about the metabolic potential of individual species in the assemblage.  This requires assembling the metagenome, something that was not possible until just a few years ago.  Given enough computer power, enough sequence data, and a low diversity assemblage (just a few species), researchers have been able to reconstruct entire microbial genomes from metagenomes.  One such research group is right here at the UW School of Oceanography, and recently published their metagenome derived genome (from an uncultured Euryarchaeota) in the journal Science.

With a lot of guidance from them I’ve established a pipeline for assembling my Barrow metagenomes.  I won’t get complete genomes, there isn’t nearly enough sequence data, but I might be able to create large enough contiguous sequence segments (contigs) to link some metabolic functions with specific bacteria in that environment.  You can check out my workflow here.  In brief I use a series of de Bruijn Graph assemblies and read-to-contig alignments to gradually build bigger contigs.  For data reduction I use a combination of standard trimming tools and digitial normalization.  The workflow page includes links to all of the tools I’m currently using.  The figure below shows the results from my first round assembly.  The x-axis is contig length in kmer (to convert to bp add 22).  The y-axis is coverage, or (roughly) the number of times that particularly contig is seen in the assembly.  In the first round I’ve got contigs out to 10,000 bases, long enough to code for several genes.  Not bad!

Ocov vs. contig length (in kmer, bp = kmer+22) following the first round assembly of PE Illumina reads using Velvet. The reads were trimmed and reduced by digital normalization prior to assembly.

The last couple of posts on this blog have been about 16S gene sequences, and how microbial ecologists use these sequences to identify different bacteria and determine evolutionary relationships.  The primary method for the latter is to build phylogenetic trees.  A phylogenetic tree is basically a “family tree” of a bacterial lineage.  A whole field of statistics exists that describes different ways of inferring the phylogeny behind a tree, and evaluating the confidence in a given tree.

Like any other statistical method what you feed into a tree-building algorithm determines the confidence you can have in the final product.  For sequence data a big factor is sequence length.  A large number of long, high quality, well aligned sequences will invariably produce a tree with high confidence scores that probably represents evolutionary reality.  The problem is that these days sequence lengths are rarely long (or of particularly high quality).  For most analyses we favor next-generation sequence data that produces a lot of short sequence “reads”.  You can do a lot with these reads, but what you can’t do is align them to one another (very well), or produce any kind of reasonable tree.

Enter pplacer.  Pplacer is a great program produced by Erik Matsen’s group at the Fred Hutchinson Cancer Research Center (“the Hutch”).  You can read the original paper on pplacer here.  What pplacer does is map short reads to an exisiting, high quality reference tree (created in advance by the user).  This allows the quantitative taxonomic analysis of next-gen sequence data at a resolution surpassing what can be achieved with the standard classification tools (such as RDP).

A “fat” tree produced from pplacer. Several hundred thousand 454 reads were classified down to the order level, reads from the target order were placed on a reference tree constructed from near full-length sequences from the RDP. The wide red “edge” represents a large number of placements within the genes Blastobacter.

Pplacer’s a great program, and with pre-compiled executables (mac and linux only) it’s a cinch to get up and running.  It is however, a rather finicky program.  Rarely has history documented an alignment parser that breaks on so many different characters.  Any punctuation in your sequence names for example, will cause pplacer to turn up its nose at your hard-won sequence data and refuse to cooperate.

I used to maintain an elaborate series of shell scripts for prepping my query and reference sequences for pplacer, laboriously modifying them for each new data set.  I finally got tired of doing this and wrote a wrapper script in Python to clean my ref and query fasta files and do the initial alignment.  The script (PlaceIt) can be found here.  The wrapper relies on the silva core set for alignment, using mothur, so it is only suitable for 16S analysis.  It would be easy to modify it to work with any other method of alignment (note that the Matsen group has a whole suite of scripts for working with pplacer, some of these are bound to be more suitable for many analyses).  If pplacer sounds useful for your work you should check it out!  And maybe PlaceIt can help you…

***NOTE***

For the code in this article to work you must use the reverse complement of the probe, not the probe itself.  I’ll correct it in the near future.

************

I’m preparing for some spring field work in Barrow, Alaska where I’ll be following up (hopefully) on some odd observations that we had there in 2010.  In particular I’d like to quantify some specific marine bacteria in Barrow sea ice.  The method that I’ll use to do this is called FISH, which stands for fluorescent in-situ hybridization.  FISH is a microscopy technique, so it differs from my normal sequence-based approach to evaluating community composition.

On the left, marine bacteria stained with DAPI, a non-specific stain that binds to DNA. On the right one specific clade is identified using FISH. The photo is courtesy of Matthew Cottrell to the original site at http://www.teachoceanscience.net/teaching_resources/education_modules/marine_bacteria/explore_trends/.

All we know about our mystery 2010 bacteria at this point are their classification based on partial 16S sequences.  Using this classification I’ve selected several FISH probes.  A FISH probe is a short (18-20 nucleotide) strand of DNA identical to some diagnostic region of a target clade’s 16S rRNA gene (a clade is an evolutionarily related group of organisms).  Since the actual 16S rRNA in the target ribosomes are complementary to the 16S rRNA gene, the probe – if it finds its way into a bacterial cell belonging to that clade – should stick to the ribosome.

Tagged on to one end of the probe is a molecule that fluoresces under green light.  Viewed in a microscope under green light bacteria that have taken up the probe, and in which the probe has adhered to the ribosome, will show up as tiny points of red light.  Assuming you know how much water you made the slide from, counting the tiny points of light tells you how many members of the target clade were present when the sample was taken.

FISH probes can be have a broad or narrow specificity.  There are probes for example, that tag virtually all Bacteria and Archaea (universal probes), just Bacteria or just Archaea (domain level probes), and down through the taxonomic rankings to probes that tag only for specific bacterial “species”.  The probes that I’ll be using target at approximately the family level  (if your high school biology is too far back, the taxonomic rankings go domain, phylum, class, order, family, genus, species).

I’m fortunate to have two metagenomic datasests from our 2010 field season as this gives me the opportunity to test the probes before going into the field.  Without testing I don’t really know if they’ll target the bacteria we observed in 2010.  A metagenome is a collection of short DNA sequences from the most abundant genes in the environment, in this case 10-20 million sequence reads.  Among that massive collection of gene fragments are fragments belonging to 16S rRNA genes, and some fraction of those belong to the bacteria that I’d like to count this year.  With a little Python it’s a simple thing to search the metagenome for sequence fragments that match the probe:

import re
probes = ['TCGCTGCCCACTGTC', 'CCGACGGCTAACATTC','CTCCACTGTCCGCGACC']
found_42 = []
found_44 = []

for each in probes:
     n = 0
     h = 0
     probe = re.compile(each)

     with open('overlapped_42.fasta','r') as fasta:
          for line in fasta:
               n = n + 1
               print n
               if re.search(probe, line) != None:
                    h = h + 1
     found_42.append(h)
     fasta.close()

print found_42

The “with open” statement was new to me, the advantage to using it here is that the sequence files (5-10 Gb) are not held in memory.  The script isn’t exactly fast, it took about 2 hours to search all the sequences in both metagenomes for all three probes.  You might notice that it wastes a lot of time looking at the sequence id lines in the fasta files, eliminating that inefficiency could probably cut down on time substantially.  The good news is that the number of probe hits matches really well with our expectations.  Over 2500 hits between the three probes for one of the metagenomes!  Had we used FISH in 2010 it is very likely that we would have seen an abundance of our target bacteria, hopefully we will this year…

Sidenote: Bacteria vs. bacteria… It may appear that my capitalization is inconsistent for bacteria.  Bacteria can refer, as a proper noun, to the domain Bacteria.  It can also be used as a common noun for prokaryotic life in general, in which case it is simply bacteria.  Unfortunately some have advocated rather strongly for the word bacteria to replace the word prokaryote on the grounds that the latter term suggests a closer evolutionary relationship between the Archaea and Bacteria than actually exists.  Given the strong ecological overlap between the Bacteria and Archaea it is essential to have a term that refers to both as an ecological unit, and prokaryote, while imperfect, is much less misleading than bacteria!

Sort of.  Over the last three years I’ve  been fortunate to participate in a working group on the origin of life, coordinated by John Baross.  John is the only faculty member in the group, the rest are graduate students from various departments (Earth and Space Sciences, Atmospheric Sciences, Microbiology, and Oceanography) participating in the Astrobiology Program.  Three members of the original group are now postdoctoral researchers at other institutions (Princeton, Stanford, and the Carnegie Institute).  The working group began as a seminar series.  Each week (over pizza and beer – critical ingredients for the origin of life) a different student, or pair of students, would discuss some aspect of the origin of life as it related to their area of expertise.  After a couple of years, and with a lot of encouragement from John, we decided to formalize our thoughts on the subject in a review article, now in publication in Geobiology.

The writing effort was led by Eva Stueken, a doctoral student in Earth and Space Sciences who made the biggest conceptual leaps for the group (check out Eva’s other recent paper in Nature Geosciences on the sulfur cycle of the early Earth).  In the review article we try to develop the concept of early Earth as a “global reactor”.  Often research groups within the origin of life community develop expertise on a specific geological environment and the synthesis therein of some key chemical precursors to life.  This kind of specialization is necessary; without in-depth knowledge it is impossible to advance understanding.  The pitfall is that it’s too easy to focus on your area of expertise and try to make the whole origin of life happen there whether it be a warm little pond, a hydrothermal vent, or a glacier.

The problem to this approach is that the early Earth wasn’t composed of a single geological environment, and there is no reason to think that all of the chemistry and selection necessary to produce life had to happen in once place.  Chemically and physically the early Earth was an incredibly diverse and dynamic place, with a range of conditions partitioned in microenvironments.  Hydrothermal vents, marine sediment, seawater, beaches, ice, dust grains in the atmosphere, volcanic pumice, and a thousand other sites each hosted many microenvironments just as they do today.  Instead of one site (e.g. a black smoker in a hydrothermal vent field) hosting all the different reactions required to produce life it seems more likely that different processes would occur in different places, wherever the conditions were most optimal for each process and the stability of the products  These products would have had many opportunities to mix, mingle, be selected, and react further in the oceans, land, and atmosphere of the Early Earth.

From Stueken et al 2013. It is reasonable to consider that the origin of life would have been the result of the synthesis, transport, and selection of chemical precursors across a range of environments. A select few environments are shown here, with associated mechanisms for transporting reaction products.

I started this entry with a note that this is not really a new model for the origin of life. There is not yet any complete model to replace, and what we develop in the review is more of a scaffold than an model.  We stop short of noting the specific reactions that would need to occur in the global reactor and the environments in which they would be partitioned, necessary elements for a working model of the origin of life.  Hopefully other researchers with specific expertise in these reactions will be able to use the scaffold to place these reactions in context, as the model itself undergoes further refinement and modification.

I’m continuing the theme of cost-saving lab tricks (see this post on the topic).  Over the summer my laptop died on me.  While putting everything back together I couldn’t find my key for the commercial synchronization software (Allway Sync) that I’d been using to backup my computer.  I could have migrated to the provided backup software in Windows 7 (or Timemachine, if I had a mac’nbook), but for some reason I have a strong mistrust of these programs.  It may be that I’m not an experienced enough user, but I want my backup files to be a human-navigable copy of my hard drive.  This way if my laptop is lost or suddenly fails I can plug the drive into any computer and instantly access what I need.  If my laptop fails as a result of a virus it also allows me to manually rebuild the system, hopefully avoiding a re-infection.

To avoid repurchasing third party synchronization software I decided to look for a solution that relies on windows commands.  After a little searching I discovered the windows command robocopy which, when the right options are selected, does the trick (the linux/unix/os command cp should work for those systems, though I haven’t dialed in the options yet).  My laptop has two hard drives, so I run robocopy for each as such:

ROBOCOPY C:\ F:\C_backup\ /zb /xo /E /Z /V /R:2 /W:0 /MIR /tee

ROBOCOPY E:\ F:\E_backup\ /zb /xo /E /Z /V /R:2 /W:0 /MIR /tee

The backup drive, one in my office and one at home, is always F.  The command is telling the system to write everything on C (and E) to a specified directory on F.  The zb option enables “backup mode” (if access is denied a read-only copy of the file is created), xo insures that only new or modified files are copied, E copies all subfolders, Z allows a restart if the connection is lost, V enables verbose mode (so you can see what is being copied), R:2 means that two attempts will be made to copy a file, if it cannot be accessed on the first attempt, w:0 indicates that the system should wait 0 seconds for the second try, MIR removes deleted files from the backup drive, and tee directs the log output to the screen.

To make the two commands easier to execute I wrote them into a batch file.  A batch file (analogous to a shell script in the unix family of operating systems) is a text file with the extension .bat.  You can create this in notepad or the text editor of your choice, and no header (such as the #!/bin/bash for bash shell scripts) is required.  Double clicking on the .bat file instantly executes the commands in it, in the order that they appear.

This worked well for a while.  Backup.bat lived on my desktop and I’d execute it whenever I felt the need to backup (typically daily, once at office and once at home).  Robocopy is fast and not very resource intensive allowing me to continue working while it ran.  However after a bit I decided I wanted to house the file in my Start menu for easier access.  I struggled with this before learning that the Start menu really doesn’t want to host a .bat file.  Fortunately there’s an easy solution – convert the .bat to a .exe.  There are a few free converters out there, I used this one.  My new backup.exe is recognized by the Start menu and allows me to quickly, freely, and transparently backup my system whenever I want.

NOTES ON TIME: When robocopy is run for the first time it takes a while, at least overnight for the average hard drive.  Subsequent backups on my system (two hard drives totally ~ 700 gb) takes ~2 hours, depending on the number of changes, the number of files in the source directory tree, and the size of the tree.  The process can be stopped just by closing the terminal window that pops up during execution, with no ill-effects.

Many thanks to Robert Krulwich, co host of NPR’s RadioLab, for a great article on frost flowers.  Reading through the comments I can see that there is a lot of public interest in this phenomenon.  I want to take the opportunity to clarify a couple of details that Robert didn’t have the time to go into in his article.

One comment to the article notes that frost flowers probably melt to 1-2 mililiters, not milimeters.  Mililiters is correct… the average frost flower is 2-4 cm tall and has a similar diameter.  They are mostly air however, and melt to a much smaller volume.  If you want to know more about the geometry of frost flowers check out these two papers:

Frost flower surface area and chemistry as a function of salinity and temperature

Specific surface area, density, and microstructure of frost flowers

There were a number of comments on the blog regarding how salt gets into frost flowers.  Frost flowers can be so salty that they are bitter to taste (like bitterns at a solar salt harvesting pond).  This seems counter-intuitive, frost flowers are derived from atmospheric moisture that has gone through a distillation process via evaporation at the ice surface (one comment noted that a small amount of salt can evaporate with the water vapor, this is true, but accounts only for a very small quantity).  The salt comes from the surface of newly formed sea ice, which is very salty due to the process of brine rejection during ice formation.

This brine at the ice surface is wicked upwards into the frost flowers by capillary action, as one reader correctly guessed (or at least this is our best understanding at the moment).  This process stops when all the available brine has been used up, or the frost flower melts under its own over-accumulation of salt (often as the day warms slightly) destroying the capillary flow.  Here’s a very brief video created from a slide in a presentation I made a while back that illustrates this:

One reader of the blog raised a very important issue regarding all this salt.  Its presence in the frost flowers (as in all sea ice) means that these structure are not solid.  Unless it is extremely cold (-54 C for sea ice) all saline ice is a porous matrix containing liquid brine and interspaced among solid, almost entirely salt-free crystals.  The more salt, the larger the volume of brine in the matrix.  This is what makes sea ice so interesting as an ecosystem, all that liquid brine at even very cold temperatures offers a habitat for a diverse array of microscopic life.  The following images of Arctic winter time sea ice (from Krembs et al. 2002) illustrate this occupation of sea ice pore spaces by microbial life.

I have an experiment that I want to do that involves growing bacteria in relatively large volumes (around 10 L) in a cold room for several weeks.  Typically in this sort of experiment you stir the cultures with a magnetic stir bar.  This keeps the oxygen level up and prevents the bacteria from falling out on the bottom of the vessel or adhering to the sides.  Unfortunately this is an unfunded side project and it is prohibitively expensive to purchase the nine commercial stir plates I would need to do this (at $300 a piece).  I decided to try to  build a stir plate alternative.

Electric motors and magnets are cheap, so I thought that I could epoxy a couple of circular rare Earth magnets to an electric motor and be done with it.  This creates a spinning magnet well enough, but the speed of the motor is way too fast for a stir bar in a culture vessel to keep up.  The speed can be reduced by dropping the voltage, but there are practical limitations to this if you want the motor to spin at 60-100 rpm (something like 0.6 % of what the motor was designed for, which would necessitate a ridiculously small voltage).

It turns out that there is an easier way to control the speed of the motor, using something called pulse-width modulation.  A pulse-width modulator outputs “pulses” of the input voltage.  The longer the pulse, the higher the average voltage received by the motor, and the faster the motor spins.  Here’s an image from the robotics site micromouse that illustrates the process.

I searched high and low for an off-the-shelf pulse width modulator and couldn’t find one.  Fortunately there are two relatively straightforward ways to build one.  The elegant way uses a microcontroller and requires programming a simple piece of firmware (necessitating some knowledge of C).  This seemed a little advanced for a first attempt, but good instructions can be found here (I think the LED light example would transfer to this application without modification).  I opted for the uglier analogue circuit method using a kit made by Velleman (DC to Pulse Width Modulator K8004), available online from Radioshack (but not in stores).

I have very little experience with soldering, but after a couple of false starts and some guidance from YouTube things went pretty quick.  Total assembly time on the kit was around 4 hours.  I could probably do it again in less than half the time.  Here’s what the assembled board looks like wired into a power source (I used a 19V drill battery) and my electric motor with magnet  (I won’t show you the backside with my ugly soldering effort).  I had very little hope that it would actually work as described, but was pleasantly surprised!Here’s the whole prototype, spinning a beaker of water at low speed.  The speed controls are pretty twitchy, but manageable.  What I’ll try to do next is connect a series of nine electric motors (with magnets) to the single PWM, built into a wood frame sturdy enough to support the incubation vessels.  Should be pretty straightforward and I think the increased resistance on the circuit will make speed control even easier.

Final cost for the project?  Kit = $30 (x 1), Motor = $3 (x 9), Magnet = $3 (x 18), all together = $111.  That’s a significant savings over the $3K the commercial version would have cost…

Well, at least on another blog.  Thanks to Dr. Kim Martini, a physical oceanographer at the University of Alaska at Fairbanks, for including linked photos from this blog in an article on frost flowers for Deep Sea News.  Dr. Martini’s article can be viewed here:

http://deepseanews.com/2012/11/the-icy-plumage-of-the-arctic/?utm_source=dlvr.it&utm_medium=twitter

Dr. Martini’s article isn’t the only place where frost flowers have appeared in print recently.  Writing in the journal Polar Biology, Aslam and co-authors recently reported the results of what must have been a very challenging study of frost flowers grown in a laboratory tank – a much larger and more natural experimental setup than our early laboratory studies.  The Aslam study was particularly interesting in that natural seawater was used for the experiments, trucked to the lab in Germany in a food-grade commercial tanker from the North Sea.  The North Sea is far from being a polar sea, but the hope is that this water is similar enough to polar seawater to provide insight into how organic compounds in seawater are distributed in newly formed sea ice.

Frost flowers growing on the surface of laboratory sea ice, from Aslam et al. 2012.

In agreement with our findings for laboratory and natural frost flowers Aslam and colleagues found that organic material and bacterial cells are concentrated near the surface of newly formed sea ice, in liquid brines and in frost flowers.

What we still don’t know is what, if anything, all these bacteria and organic compounds at the ice surface are doing.  Microbial metabolism decreases with decreasing temperature, and it is very cold at the surface of young sea ice, but not cold enough to make microbial activity here impossible.  And all that organic material at the ice surface can actually help bacteria survive there, by protecting them from high salt and freezing conditions.

The high concentration of organics also help overcome the poor efficiency of bacterial enzymes at low temperature.  The more organic substrate an enzyme has to interact with the less it needs to function efficiently to still get the job done.  Imagine that you’ve spilled m&ms all over your kitchen table and are trying to eat them as quick as you can.  Your fingers are very cold (you are trying to save on your heating bill) and you don’t have your normal dexterity.  The more m&ms on the table the more likely it is you will find ones that you can actually pick up.  Now fill your kitchen with saltwater the salinity of the Dead Sea, add a UV lamp bright enough to give you a sunburn, and repeat the experiment.  You are now experiencing the life of a marine bacterium at the surface of newly formed sea ice.