Donald Phan

Index
Year
2026
Field
Water technology
Status
Project
2 of 4

Enzymes against biofilm

Water scarcity becomes increasingly more pertinent, but biology can’t really desalinate seawater, so we can tackle the problem next to it. A survey and three models narrowed that problem to one object, a membrane spacer carrying an enzyme that stops bacteria coordinating.

Donald Phan
Sole researcher

Chapter 01 · The problem

Biology can’t desalinate water.

That was the first thing the survey settled. Pulling salt out of water fights the entropy of mixing, so it takes work, and an enzyme can only speed up a reaction that already runs downhill. Catalysts change rates, not equilibria.

Fig. 1Energy to desalinate one cubic metre of seawater
  1. Physical floor1.0635,000 ppm feed, 50% recovery
  2. Practical minimum1.56A single stage, same feed
  3. Modern plants2.5–4.0Energy is 30–50% of their operating cost
  4. Pumping ions with ATP≈17One ATP per ion, before feeding a single cell

Reverse osmosis already runs within two to three times of the physical floor, and much of the gap is pumps rather than membranes. Doing the same job with metabolism would cost about five times what a modern plant uses, before a single cell had been fed. Where biology has competed head-on, it has lost: aquaporin membranes have not reached a full-scale seawater plant in about twenty years.

But biology owns the problem sitting next to desalination.

Every seawater plant is built from these.A spiral-wound membrane element, about a metre long and twenty centimetres across. A plant making 100,000 cubic metres of water a day runs roughly 7,700 of them.

Unrolled, it is a sandwich wound round a tube.Seawater is pushed along the element between sheets of membrane. Fresh water crosses the membrane and spirals in to the central tube; what is left, the brine, carries on out of the far end.

Between the membranes sits the feed spacer.A diamond mesh of extruded polypropylene, under a millimetre thick, that holds the channel open. It is the cheapest part of the element, and the part biofilm grows on.

Bacteria in the feed settle on the mesh and multiply.Once there are enough of them they switch on together and pour out a sticky gel. The gel narrows the channel, and the pressure it takes to push water through climbs.

The obvious weapon is off the table.Chlorine kills bacteria, and it destroys the membrane too: polyamide degrades above about 0.1 ppm of continuous chlorine. So every plant runs the same awkward cycle.

Chapter 02 · What the industry does

Dechlorinate, foul, clean, repeat.

With chlorine ruled out, plants manage biofilm on a schedule. They watch the pressure drop along the feed channel, and when it has climbed 15% above where it started, they stop and clean.

Feed channel

ΔP +0.0%

Cleans 0Quorum

Day 0

Day zero: a clean channel.A few millimetres of one feed channel. The membrane lies below, the spacer’s two layers of filaments cross above it, and the top membrane is cut back so you can see in. The gauge reads the rise in pressure drop.

Bacteria settle and multiply.For weeks the pressure barely moves. Then the colony crosses its quorum threshold and switches on together, the gel builds fast, and at 15% the plant cleans.

Cleaning never gets all of it.Caustic and acid washes strip out most of the biofilm, and what is left regrows from a head start, so each interval is shorter than the last. Well-run plants clean two to four times a year; in the Arabian Gulf, monthly to fortnightly.

The real cost is the membrane.Across seven full-scale plants in the Netherlands, fouling took about 24% of a reverse-osmosis plant’s operating cost. The largest part was not cleaning chemicals but membrane elements replaced early.

Biology has been aimed at this before.

Quorum quenching (stopping bacteria from coordinating, rather than killing them) was proposed against membrane fouling in 2002. It works. Nobody has ever been able to buy it.

Departures

Quorum quenching against membrane fouling · since 2002

  1. 2009Free enzyme, dosed into the tankDiluted, digested and washed out, and none of it can be recovered.
  2. 2009Enzyme on magnetic carriersStable, but the worst on cost, and the carriers lost 30–40% of their volume in use.
  3. 2012Live bacteria, encapsulatedThe biology works: bacteria in beads stretch the time between cleans about 2.2×. The carriers cost too much to scale.
  4. 2013Gene&Green TKThe one dedicated company. Thirteen years, nothing for sale; its other product line breaks down nerve agents.
  5. 2017Enzyme bonded to nanofibre3% of its activity left after 20 days.
  6. 2025Salt-tolerant bacteriaConventional quenching fails above 2% salt, and seawater is 3.5%. New marine consortia cope; commercial use is still an open question.
  7. 2026Engineered enzyme on the spacerThis project.
Seventeen years of papers since the first, and no full-scale installation

Every one of these failed at something around the enzyme: the carrier, the bead, the organism, the dosing. None failed at the chemistry. The mechanism even passed its hardest test: in control reactors running empty beads, the biological effect was much the larger of the two. No major water company has publicly evaluated it.

Chapter 03 · What I’m proposing

Put the enzyme on the spacer.

Engineer the enzyme, not the carrier.

Graft it to the feed spacer, not into a bead.

Sell it as a spacer, not as a dosing regime.

The field spent seventeen years building better carriers around a fragile enzyme, when the enzyme was the thing that needed fixing. Make it tough enough to graft onto the spacer while the element is made, and the scaffolding disappears: no carrier to pay for, nothing to wash out, no live organism, nothing to dose.

Cells
—
Signal
—
Quorum: not yet
Enzyme

30% of a monolayer

60%

No enzyme

kcat/KM
—
Da
0
Delay, fresh
1.0×
Delay, aged
1.0×

Below the 2.2× bead benchmark

Bacteria vote with a molecule.Each cell leaks a little signal, an acyl-homoserine lactone. Crowded onto a surface, the signal builds up, and when it crosses a threshold the colony switches on together and starts making gel.

A lactonase cuts the signal’s ring.Breaking that ring is downhill, so here a catalyst is exactly the right tool: the opposite of asking an enzyme to desalinate. Grafted to the spacer, it cuts the signal before it can build up.

One number decides how much it helps.The Damköhler number, Da, compares how fast the surface destroys signal with how fast diffusion brings it in. The enzyme divides the signal at the surface by (1 + Da), so the colony needs (1 + Da) times as many cells to switch on.

Only two published enzymes are fast enough.Four orders of magnitude separate the candidates, and the most heat-stable are the slowest: SsoPox W263I survives 87.8 °C and is 300 times slower than GcL. So choose on speed, and engineer the stability in afterwards. Try them against the benchmark.

What it is made of

Substrate
Extruded polypropylene feed spacer
Commodity: buy it
Active molecule
An AHL-lactonase, which cuts the lactone ring of the signal
Scaffolds exist; none engineered for this duty
Durability
Surviving 3.5% salt, cleaning from pH 2 to 12, surfactants and months of flow
The actual work. Nobody has done it
Attachment
Covalent grafting to inert polypropylene
Known in principle; an earlier attempt lost 97% of its activity in 20 days
Product
A coated spacer roll, sold to whoever winds elements
Does not exist yet

What it is not

  • Not a membrane. It never touches the layer that does the separating.
  • Not a dosed chemical. Nothing is consumed or topped up.
  • Not a live organism. That is what stopped the bacterial approach in drinking water.
  • Not a desalination technology. It moves no ions.

Chapter 04 · What it solves

The prize is membrane life.

Quenching does not stop bacteria arriving or sticking. It stops them coordinating, so the gel that blocks the channel builds more slowly and the plant reaches its cleaning trigger later.

Today’s spacer

ΔP +0.0%

Cleans 0Quorum

Quenching spacer

ΔP +0.0%

Cleans 0Quorum

Day 0

Enzyme

Same feed, same clock.On the left, today’s spacer. On the right, the same mesh carrying JydB at 30% of a monolayer, where the colony has to grow nearly five times denser before it switches on.

Fewer cleans, longer-lived elements.On the bridge model’s default settings, today’s spacer is cleaned 39 times over a five-year element life and the quenching spacer 9 times. Plants do not pay much for cleaning. They pay for replacing elements early.

How much time that buys hangs on one unmeasured number.While the biofilm is still growing exponentially, a five-fold higher threshold buys about 1.6 generation times, three days on a sixty-day cycle. Once nutrients have capped its growth, it buys close to five times the interval. Move where growth levels off and watch the answer run from 77% fewer cleans to none.

About two cents a cubic metre

On default inputs a quenching spacer is worth $0.0197 per cubic metre of water, twice the one-cent kill threshold I set before starting: about $720,000 a year for a plant making 100,000 m³ a day. The spacer’s own cost barely registers. A $25 premium per element, spread over a five-year life, is a tenth of a cent per cubic metre.

Fig. 6What a quenching spacer is worth · prize model, a plant making 100,000 m³ a day

$0.0197net, per m³ of water

Clears the $0.010 kill threshold 2.0×

  1. Operating cost$0.5500
  2. × fouling share$0.1320
  3. × biological share$0.0594
  4. × share removed$0.0208
  5. − spacer cost$0.0197
  6. Kill threshold$0.0100

A year, one plant
$720,000
Cleans avoided
2.1 of 6 a year
Spacer cost
$0.0011/m³

Two of those inputs have never been measured: how much of the fouling bill is biological, and how much of it the spacer removes. Each alone swings the answer from below the threshold to three or four times it. So the go or no-go rests on two measurements, not on a business case.

Chapter 05 · What nobody has measured

Five numbers decide it.

Ranked by how far each one moves the answer. None of them is known yet, and every one can be measured.

  1. Where biofilm growth levels off

    The crossover density decides whether the enzyme’s delay buys logarithmic time or proportional time: the difference between a marginal product and a strong one. It may also explain why published delays scatter from 1.7× to 6×: two regimes, reported as one number.

    Settled byStaged biomass sampling in a flow cell

  2. Whether it survives cleaning

    Two published results bracket it twentyfold: 3% of activity left after 20 days, and over 60% after 250. Neither put the enzyme through a clean-in-place cycle. Nobody has.

    Settled byCoupons cycled through pH 2, pH 12 and surfactant

  3. How much enzyme fits

    Loading trades directly against survival. At 30% of a monolayer, JydB has to keep 32% of its activity to beat the bead benchmark; at ten monolayers’ worth, only 3%.

    Settled byLoading measured on activated polypropylene

  4. Biofouling’s share of the bill

    The one question that needs plant data rather than a model. What triggered each clean, and how often elements were replaced, settles it.

    Settled byOne operator’s cleaning log and replacement history

  5. Speed in seawater

    Every published rate was measured in dilute buffer. The two candidates with the best salt and heat tolerance, Aii20J and AiiA S1-5, have no published kinetics at all.

    Settled byRates measured in 0.6 M salt

Kill criteria, set before starting

  • The prize comes in under $0.01 per m³: stop.
  • No enzyme keeps its activity for about six months of clean-in-place: stop.
  • Grafting costs more than the biocide it replaces: stop.

The fastest progress available is one phone call to a plant operator and one flow-cell experiment. Neither needs protein work to start.

Behind the story

The chapters above are the argument, cut down. The working material is a landscape survey of biology and water scarcity, about 6,500 words and 59 sources, and three interactive models built alongside it: a prize model that prices the spacer, a spec model for how much enzyme it needs, and a bridge model that joins the two. The figures on this page run the same equations with the same defaults.

Sources

All pages opened and read, as of 20 September 2026.

Thermodynamics and membrane theory

Biomimetic and artificial water channel membranes

Aquaporin A/S financials

Quorum quenching

Markets and economics

Adjacent biology

Enzyme kinetics and durability

Reaction–diffusion and simulation