EcoSym

Starved Biofilters and Dormancy

For what nitrifying bacteria are and how they make a living, see Microbes. For how a tank builds a biofilter in the first place, see Starting the Cycle.

The thing every keeper has noticed

A sponge filter sits in a bucket for three weeks while you rescape the tank. You put it back and nothing happens: no ammonia, no nitrite, the fish are fine. A quarantine tank runs empty for two months between fish and still handles a new arrival on the first day. You finish a fishless cycle, then life gets in the way and you do not buy the fish until a month later, and the cycle is still there waiting. Most people who have kept fish for a few years have run this experiment by accident.

This should be surprising. Nitrifying bacteria live on ammonia. Take the ammonia away and you have taken away the only thing they eat, and a population that is not eating should shrink until there is nothing left of it. That is not what happens, and the reason is worth understanding. It is why "keep your media wet" is real advice, and it is why a heavily planted tank that has driven its own ammonia down to zero has not thereby dismantled its own biofilter.

Why a starved population should die, and why this one does not

The standard picture of a bacterial population is a chemostat. Cells grow at a rate set by how much food is available, and they lose biomass at a rate set by two things that do not depend on food at all. One is maintenance, the energy a cell spends staying alive. The other is mortality, the cells that die. When food is plentiful, growth wins and the population climbs. As food runs down, growth slows until it can no longer pay for maintenance plus mortality, and past that point the population shrinks every day without limit.

That picture is accurate for cells floating free in the water. Where it breaks is that it treats maintenance and mortality as constants, and in a nitrifying biofilm neither of them is.

Maintenance falls when growth stops

The measurement that settles this uses an unusual piece of equipment called a retentostat. An ordinary chemostat washes cells out along with the water, so you can never tell whether a shrinking population is dying or simply leaving. A retentostat keeps every cell and feeds them at a falling rate, which drives the growth rate towards zero while the cells are still alive and countable.

Tappe and colleagues (1999) put Nitrosomonas europaea and Nitrobacter winogradskyi through exactly this. Two of their results matter here. First, as growth approached zero the maintenance demand fell to somewhere between a third and an eighth of what the same organisms' ordinary chemostat behaviour predicted. A cell that has stopped growing is running a different and much cheaper metabolism. Second, across weeks of starvation they saw no significant cell death at all. The cells were idling rather than dying.

Attachment is what makes the difference

The other measurement explains why this shows up so reliably in aquariums and so unreliably in a bottle of bacteria. Batchelor and colleagues (1997) starved Nitrosomonas europaea two ways: as a suspension of free-floating cells, and as a biofilm attached to a surface.

The suspension degraded badly. Unstarved, it took about nine hours to get going once it was handed ammonia. After 42 days of starvation it took 153 hours, more than six days of lag before nitrite production resumed at all.

The biofilm, starved for the same 43 days, resumed with no measurable lag whatsoever. Some cells were lost along the way, but whatever survived started working immediately.

That is the difference between the water and the media. A biofilm is a dense community of cells embedded in a secreted EPS matrix, packed closely enough to signal to each other and sheltered from most of what would otherwise kill them. The bacteria in your water column behave the way the textbook says. The ones in your filter and substrate do not, and that is where almost all of your nitrification lives.

Viruses need a host that is building something

There is a third loss, and for a starved biofilm it is the largest of the three. Bacteria are killed by bacteriophage, viruses that infect a cell, replicate inside it, and burst it. In a well-fed population that is a substantial drain. In a starved one it very nearly stops, for two independent reasons.

Nabergoj and colleagues (2018) counted how many virus particles come out of an infected cell across a range of host growth rates. At the fast end they got 89 per cell. At the slow end they got 8. The relationship is close to linear in the host's growth rate, and below a threshold growth rate the virus stops replicating altogether. A virus has to build itself out of the host's own machinery, so a cell that has powered down is a poor factory and eventually not a factory at all.

Separately, the EPS matrix that shelters biofilm cells from grazers shelters them from viruses too. Melo and colleagues (2020) found infected cells were markedly rarer in the parts of a biofilm where the matrix was densest, and that mechanically breaking the biofilm apart improved phage killing by well over an order of magnitude. The matrix is a physical barrier that most phage do not get through.

What the model does

Every surface-attached nitrifier in the simulator carries a running measure of how much ammonia has been reaching it lately. Three of its loss terms fall towards a floor as that measure drops: maintenance, base mortality, and the yield a virus gets from an infected cell. Nothing else changes. The rate at which viruses encounter cells still scales with how densely the bacteria are packed, exactly as before.

"Lately" is doing real work in that sentence. The measure is an average over about two days rather than a reading of the moment, and it has to be. A tank fed once a day has essentially no ammonia in it for twenty-two hours out of twenty-four, no matter how heavily it is stocked, so a bacterium that judged its situation instant by instant would spend most of every day concluding it was starving. Real bacteria do not, because powering down is a change in which genes are being read and that takes days to happen and days to undo. Two days is long enough to see past a feeding schedule and short enough to notice when the food genuinely stops.

The transition is placed well below the ammonia a bioload leaves standing. A stocked tank at equilibrium holds somewhere in the range 0.02 to 0.35 mg/L of ammonia depending on how heavily it is stocked, and across all of that the relief is effectively absent: a tank with fish in it has a fed biofilter and gets the same numbers it always did, as does a tank in the middle of a cycle. The relief comes on below about 0.01 mg/L, which is the territory of a tank with no bioload or one whose plants are taking everything. Down there, staying alive costs a starved biofilm three to six times less than staying alive costs a growing one.

Three things the model deliberately does not do.

It does not throttle the actual work. A starved biofilm oxidises every ammonia molecule that reaches it at the full per-cell rate. There are simply fewer cells to reach. Slowing the enzyme as well would count the same shortfall twice.

It does not add a delay on the way back. It would be natural to model dormancy as a separate state that cells have to be woken out of, the way the simulator models resting eggs in Daphnia. The measurements rule that out. Batchelor's biofilms had no lag after six weeks, and Bollmann and colleagues (2005) recovered a nitrifier's full ammonia-oxidising activity within 30 to 60 minutes of adding ammonium, after as much as two weeks with nothing to eat. Cells starved three days needed about half an hour; cells starved a week needed about fifty minutes. On the timescale of an aquarium that is instantaneous.

It does not apply to the water column. Free-floating nitrifiers keep the ordinary chemostat treatment and wash out, because that is what Batchelor measured them doing.

Ammonia comes back first, nitrite comes back last

The two halves of the biofilter do not survive starvation equally, and this is where the behaviour becomes specific enough to recognise in your own tank.

Tappe's retentostat pulsed both organisms after starving them. The ammonia oxidiser, after 17 days without food, was back to half its maximum rate within an hour. The nitrite oxidiser, after 35 days, showed no detectable oxygen consumption at all for the first five hours, and needed four separate nitrite pulses over about eighty hours to reach the same halfway mark. The authors drew the obvious conclusion: this asymmetry is why nitrite turns up in the environment when ammonia becomes available again.

Elawwad and colleagues (2013) saw the same thing at reactor scale and on a longer clock. Nitrifying biofilm carriers parked for 97 days got their ammonia oxidation back in under a week. Their nitrite oxidisers declined faster during the idle period and took roughly seven weeks to recover.

The simulator gives the nitrite oxidiser a shallower dormancy than the ammonia oxidiser for this reason, and that produces a specific, checkable prediction. If you stock a tank that has been sitting idle for a long time, what you should see is not an ammonia spike. Ammonia gets handled. What you may see is a nitrite reading a week or two later, as the ammonia oxidisers get back to work faster than the nitrite oxidisers can clear what they produce. In the simulator's test rig, going from three months idle to nine months idle roughly doubles the nitrite excursion and barely moves the ammonia one.

There is a third guild, and it comes through best of all. Comammox Nitrospira does the whole conversion inside one cell and has an extraordinarily tight grip on ammonia, half-saturating around fifty times lower than the classic ammonia oxidisers. In water where an ammonia oxidiser has effectively stopped growing, comammox is still comfortably fed. It gets no special dormancy setting in the model because it does not need one. Its whole niche is famine, and a long-idle tank is its kind of tank.

What this means for your tank

A cycle you established does not evaporate because you waited. If you fishless-cycled and then took a month to buy the fish, you still have a cycle. Continuing to dose ammonia during the gap does no harm and leaves you with a slightly larger biofilter, but it is not the difference between having one and not.

Keep your media wet and it will keep. Everything above describes a biofilm that is starving. It does not describe a biofilm that is drying out, which is a completely different injury and one the simulator does not model. Desiccation destroys cells outright rather than powering them down. A sponge in a bucket of tank water is doing the experiment described here. A sponge on a shelf is doing a much worse one.

A planted tank has not lost its biofilter just because ammonia reads zero. Fast-growing plants and floating cover out-compete nitrifiers for ammonium and can hold the water column near zero indefinitely. The biofilter in such a tank is small, because standing stock tracks the ammonia supply rather than the surface area available, but it is there and it responds. A trickle of ammonia is enough to sustain a population that costs very little to maintain, which is why the same tank that reads zero on a test kit can still absorb a fish.

A long-idle tank still deserves slow stocking. None of this makes an idle biofilter equal to a working one. It carries less biomass, and the fewer cells there are, the longer it takes to build back up to a new load. Adding one fish to a tank that sat empty for nine months is a very different proposition from adding six, and the simulator will show you the difference.

What the model does not claim

The measurements this is built on run to about three months. Elawwad's 97-day carriers are the longest properly controlled starvation-and-recovery experiment we could find, and the simulator's nine-month behaviour extrapolates from it, bounded on the other side by the hobby observation it was built to reproduce. Treat a year-long idle period as a qualitative answer rather than a quantitative one.

The model also does not represent drying, freezing, chlorine, or the medication that killed the biofilter in the first place. It represents one situation: a submerged, oxygenated biofilm with nothing to eat.

Finally, none of this changes how big the biofilter eventually becomes. Standing stock at equilibrium is set by how much ammonia arrives per day, not by how the population got there, so a tank that idled nine months and a tank that never idled at all settle to the same biofilter under the same bioload. The whole cost of starvation is in the transient: how large the excursion is on the way back, and how long it lasts.

References

  • Batchelor, S.E., Cooper, M., Chhabra, S.R., Glover, L.A., Stewart, G.S.A.B., Williams, P. & Prosser, J.I. (1997). Cell density-regulated recovery of starved biofilm populations of ammonia-oxidizing bacteria. Applied and Environmental Microbiology 63(6): 2281–2286.
  • Bollmann, A., Schmidt, I., Saunders, A.M. & Nicolaisen, M.H. (2005). Influence of starvation on potential ammonia-oxidizing activity and amoA mRNA levels of Nitrosospira briensis. Applied and Environmental Microbiology 71(3): 1276–1282.
  • Elawwad, A., Sandner, H., Kappelmeyer, U. & Koeser, H. (2013). Long-term starvation and subsequent recovery of nitrifiers in aerated submerged fixed-bed biofilm reactors. Environmental Technology 34(5–8): 945–959.
  • Kits, K.D., Sedlacek, C.J., Lebedeva, E.V., Han, P., Bulaev, A., Pjevac, P., Daebeler, A., Romano, S., Albertsen, M., Stein, L.Y., Daims, H. & Wagner, M. (2017). Kinetic analysis of a complete nitrifier reveals an oligotrophic lifestyle. Nature 549: 269–272.
  • Melo, L.D.R., Pinto, G., Oliveira, F., Vilas-Boas, D., Almeida, C., Sillankorva, S., Cerca, N. & Azeredo, J. (2020). The protective effect of Staphylococcus epidermidis biofilm matrix against phage predation. Viruses 12(10): 1076.
  • Nabergoj, D., Modic, P. & Podgornik, A. (2018). Effect of bacterial growth rate on bacteriophage population growth rate. MicrobiologyOpen 7(2): e00558.
  • Tappe, W., Laverman, A., Bohland, M., Braster, M., Rittershaus, S., Groeneweg, J. & van Verseveld, H.W. (1999). Maintenance energy demand and starvation recovery dynamics of Nitrosomonas europaea and Nitrobacter winogradskyi cultivated in a retentostat with complete biomass retention. Applied and Environmental Microbiology 65(6): 2471–2477.