The fly that lets winter in

Go as far south as land goes and the insects very nearly run out. In maritime Antarctica, in mats of wet moss near the shore, there is a wrinkled grub several millimetres long. Belgica antarctica is a non-biting midge, a true fly in the order Diptera, and its larva does nearly all the living on this page.

It is the southernmost recorded free-living insect, and the only free-living insect endemic to maritime Antarctica. Both words are doing work. Free-living means it makes its own way rather than riding on a host, and parasitic lice and fleas can occur farther south than it does. Endemic means it lives here and nowhere else, which rules in this midge but not Parochlus steinenii, a midge native to the same region but not only there.

It is a fly that, in practice, does not fly. The adult's wings are reduced to stubs. It is also missing the halteres that other flies use as gyroscopes to stay balanced in the air. It walks across wet ground instead of taking to the air. The larvae live in wet coastal moss, algae, and rotting organic matter, often 10 to 20 m from the shore.

On Cormorant Island in January 2018, one good plot held a measured maximum of 38,850 larvae per square metre. Lay a sheet of A4 paper on moss that crowded and you are covering roughly 2,400 of them. The number belongs to the plot that was sampled, not to all patches of moss. It still changes what you picture: the single wrinkled larva under a moss stem is one of a crowd living beneath your boots.

Most of its life is spent as a larva, and much of that is spent frozen. A two-year cycle is common, though not the only one. Like other insects, the larva grows in stages called instars, each one ending in a moult; this midge passes through four of them. In the common pattern it meets its first winter around the second instar and its second winter as a fourth-instar larva. The back half of that fourth instar, labelled L4-2, enters obligate diapause: a set pause in its development the larva cannot skip, even if conditions turn mild.

Even the good season is cold. Southern summer arrives while the northern half of the world is in winter. Across five sites watched from mid-December into late March, the moss and soil averaged 3.0 to 5.6 C. That is the summer figure, not a winter one, and it covers the midge's best months. When the moss and soil rise above freezing, the larvae move and feed.

Then, in the common pattern, the adults come out together, all at once. They walk, they mate, they lay eggs in a blob of jelly, and they never eat. An adult lives roughly 7 to 14 days. Two winters of getting ready, spent in a week or two, and nearly the whole life story belongs instead to the larva and to the wet ground around it.

The larva freezes solid, and then it is fine. It is freeze-tolerant in summer and in winter alike, not only in some prepared season. Ice forms in the extracellular spaces, the gaps around and between its cells, while the cells themselves stay whole. When the thaw comes, liquid water returns to the moss and movement returns to the larva, and it carries on from where it stopped.

It is not a brief freeze either. A temperature logger buried in occupied moss on Torgersen Island recorded one winter from mid-to-late April through mid-November. The ground stayed frozen without a break for roughly seven to eight months, mostly sitting near -1 to -3 C. A larva in that moss would have spent the better part of a year frozen solid.

A second record, taken 1 cm down, found 0 to -2 C for more than 300 days, with -7 C reached on only two occasions. Dull, in other words, and nowhere near as savage as the word Antarctica suggests. The wet moss is what buffers it. These two records come from different places and are not readings of the same winter, but together they show what a larva's winter can be: long, frozen, and buffered.

Drying a larva out under controlled conditions can leave it better at surviving a freeze later, which is a strange thing for drought to do for you. In one test, larvae collected in summer spent 48 hours at 98.2 percent relative humidity and +4 C. That is damp air by any normal standard. It still pulled water out of them slowly: they lost about 30 percent of their osmotically active water, the part free to move in and out rather than bound up inside the tissue.

Then came the cold. All of the slowly dried larvae survived three days at -10 C. Of the larvae that had not been dried, fewer than one in ten survived even two days at -10 C -- a shorter test, and still far worse results. The obvious answer is wrong, too. Drying left the supercooling point near -9 C in both groups, so the dried larvae were not holding ice off any better than the others. The protection did not come from holding ice off. What did produce it, the study did not establish.

Three different things in this chapter all involve losing water, and they are worth keeping apart. Plain desiccation is simply drying out. Prior drying is the rehearsal above, where losing water first leaves a larva better able to survive a freeze later. Cryoprotective dehydration is the third and the strangest: a cold larva sitting near ice loses water to that ice, bit by bit, and by giving up the water it never freezes at all. Only that last one is offered as a whole-winter route, and it is the one this page ends by questioning.

When Kelley's team sequenced this midge, they found a haploid genome of about 99 Mb, roughly 99 million letters of DNA. For an insect that is very small. What it was missing is the interesting part: hardly any repeating stretches, few of the parasitic sequences that copy or move themselves around a genome, called transposable elements, and unusually short introns, the non-coding gaps that sit inside genes. Their 2014 paper described it as the smallest sequenced insect genome reported in 2014.

That record has since fallen. A 72.1 Mb assembly for Xenos peckii, a twisted-wing parasite, was published in 2024. It is tempting to link the tidy genome to the hard life, and that temptation should be resisted. A stripped-down genome and freeze tolerance turn up in the same insect here, but nothing in the work shows that one produced the other.

The way the larva handles stress is spare in its own way. Heat-shock proteins are a cell's repair kit for trouble, normally ramped up sharply when something goes wrong. In the larvae Rinehart's team tested, hsp70, hsp90, and a small heat-shock protein were being made all the time. The alarm was already sounding, and the heat and cold shocks they tried did not make it louder. That was true of the larvae in those particular tests; the adults did not behave the same way.

Cryoprotective dehydration is real, and this midge can do it. In the founding laboratory experiment, fourth-instar larvae held near a physically separate piece of ice gave up their water to it, never froze, and lived. The same paper found something that muddies the story: in wetter soil, ice touching a larva simply seeded ice inside it, a process called inoculative freezing.

So which route is even available turns on moisture. In the wet moss where these larvae really do spend the winter, ice presses against the larva from all sides and seeds freezing directly, leaving little room for the dehydration route. Dry conditions with the ice held physically apart, of the kind a laboratory can arrange, leave that route open. Nobody has published a moisture line where one takes over from the other -- there is no published universal cutoff -- and rather than invent one, the laboratory below refuses to answer in the middle.

Someone finally ran the two side by side. Yoshida's team took second-year diapausing fourth-instar larvae, 55 per treatment, and held them at -5 C for six months: one group frozen, one group cryoprotectively dehydrated. The frozen larvae won, and not by a little. Of the frozen group, 41.8 percent were alive at six months and 23.6 percent went on to become adults during the following 50-day scoring window. Of the dehydrated group, 9.1 percent were alive, and not one of them became an adult.

Read those numbers carefully before spending them. The treatments were held at one constant temperature, and the dehydrated one was severe. The two groups were also cooled in different ways, and becoming an adult depends on breaking diapause as well as on surviving. So these are not field survival rates, and they are not a verdict on all ways of drying a larva out. What they do rule out is the tidy story -- that the clever move is to dry out so that nothing can freeze. Given six months of it, the larvae that surrendered to the ice were the ones still alive at the end.