Scientists want to improve photosynthesis to help feed the world, but it isn't so simple

Scientists screened about 15,000 bacterial Rubisco variants, revealing promising changes and unexpected obstacles to improving photosynthesis.

Plants look like experts at turning air and sunlight into food. But a key enzyme in that process, Rubisco, is slow and sometimes captures the wrong gas.

Scientists have now tried to improve a faster bacterial version of that enzyme, and the single changes that helped often stopped helping once they were combined.

Even the computer models built to predict such changes struggled to say which combinations would work.

The team screened about 15,000 altered versions of the enzyme inside living photosynthetic cells. What came out is a long list of which changes help and which hurt.

The list comes with a limit: the enzyme was bacterial, not a plant's, and any use in a crop would need further testing in plant cells and whole plants first.

Paul Hudson, a professor of systems biology at KTH Royal Institute of Technology in Stockholm, led the work with Ute Hoffmann and Axel Knave, who share first authorship.

Colleagues at SciLifeLab, the University of the Balearic Islands in Spain and Kobe University in Japan took part.

Bacteria supplied the enzyme

Every plant carries Rubisco, and so do some bacteria.

Sunlight supplies the power, and the enzyme captures carbon dioxide from the air and attaches it to a sugar, the first step in building the rest of the plant.

When Rubisco captures oxygen instead, it produces a byproduct the cell has to break down and recycle, which costs energy.

Plant Rubisco is a 16-piece assembly that is hard to produce and alter in the lab. Some bacteria carry a simpler form, built from identical copies of one piece, that fixes carbon faster.

Versions like it have been proposed as replacements for the plant enzyme in crops.

Hudson's team chose one from a bacterium called Gallionella. This type of Rubisco evolved mainly in places with little oxygen.

Chemists call capturing carbon dioxide carboxylation and capturing oxygen oxygenation, and the fast bacterial enzyme does more of both.

"But they also catalyze the oxygenation reaction more readily, meaning that they are less selective for carbon dioxide," Hudson said.

"Our idea was to take these fast Rubiscos and make them even faster, or at least reduce their oxygenation rate while retaining a high carboxylation rate."

Testing variants in bulk

The team put the altered enzymes into Synechocystis, a photosynthetic bacterium, or cyanobacterium.

Its own Rubisco had been switched almost entirely off, so each cell could grow only as fast as the bacterial Rubisco it carried allowed.

Each altered gene also carried a 20-letter DNA tag. These tags let researchers track which variants became more common as the cells grew together in one pooled culture.

The library had two parts. One tested individual changes at selected positions in the enzyme.

The other combined changes at seven positions, with 16,200 possible combinations in its design.

None of the positions was a guess. Most came from two programs that compare the enzyme with thousands of related Rubiscos and flag the spots most likely to tolerate change.

The pooled cells then grew for 12 generations in bioreactors, with or without oxygen in the gas feed. The cyanobacteria grew more slowly whenever oxygen was present.

Single changes often helped

Cells carrying about 8% of the single-swap variants grew faster than cells carrying the starting enzyme, compared with under 1% of the multi-change variants. The helpful swaps clustered in a few places.

Two of them were in a flexible loop beside the enzyme's active site, a stretch thought to help Rubisco tell carbon dioxide from oxygen.

Another helpful swap was at the seam where two copies of the enzyme meet, beside a tunnel that runs between the two active sites.

Swaps there helped more when oxygen was present.

One short stretch couldn't be changed at all: it forms a helix above the opening where the enzyme's sugar partner enters and leaves, and every swap there made growth worse.

To check the tag readings, the team grew six of the best variants on their own. Cells carrying each of the six variants outgrew cells carrying either the starting enzyme or the natural enzyme.

The cells held about the same amount of each enzyme, which argues against the idea that they were simply making more Rubisco.

Combinations mostly failed

The obvious next step is to combine the winners, and it mostly didn't work. On average, the more changes a variant carried, the worse the cells holding it grew.

Cells carrying one variant with three particular changes grew more slowly than cells carrying the starting enzyme. Cells carrying a fourth change on its own also grew more slowly.

Put all four together, and the cells grew noticeably better.

The computer models struggled with combinations too. Three programs that score a mutation by how it fits the pattern across related Rubiscos did reasonably well on single swaps.

On variants with several changes, their predictions pointed the wrong way, largely because of one swap they kept recommending that hurt every variant it appeared in.

A model trained on the team's own data did better at predicting the combinations already in the dataset.

But when the team built 11 new variants it recommended, each with six to 12 changes, eight of them didn't support growth at all. Only two beat the starting enzyme.

The best enzyme in the whole study, called Ad1, came from the simpler approach of adding proven changes together. Give Ad1 one more swap that had helped on its own, and growth collapsed.

Less reaction with oxygen

The team purified Ad1 and a six-change variant called C6 and measured them in the test tube with radioactive carbon dioxide.

Both reacted with oxygen more slowly than the starting enzyme, and Ad1 was choosier about carbon dioxide over oxygen.

Carbon-fixing speed changed in opposite directions, with C6 about 25% faster than the starting enzyme and Ad1 about 15% slower.

There was a cost. Both variants needed more carbon dioxide around them to work at full speed, so their overall efficiency at fixing carbon was lower too.

The better growers the team tested were also generally less heat-stable than the starting enzyme, which the authors wrote may reflect a looser fit between the two halves.

Inside the cells, the Ad1 strain held less of the sugar Rubisco works on and more of the product it makes. That fits a smaller bottleneck at that step.

The strain also grew better than the starting enzyme's cells in ordinary air, not only in the carbon-rich bioreactors.

Challenges ahead for crop use

The cells were cyanobacteria, not plants, and the enzyme was bacterial.

Moving a bacterial Rubisco into a crop would also mean moving the machinery that concentrates carbon dioxide around it, which the authors call an outstanding challenge.

The starting enzyme was itself a slower version of the natural one, and even Ad1 still fixed carbon more slowly than the natural enzyme.

The pooled screening also ran at 5% carbon dioxide, far above the level in air, which may have hidden differences in how the enzymes handle scarcer carbon dioxide.

And the team measured only two enzymes in detail, so it can't yet say that less reaction with oxygen is what sped up growth.

Next, Hudson's team wants to cut the oxygen reaction further in the promising variants and work out which parts of the enzyme help it choose carbon dioxide.

With funding from the European Union, the researchers are also testing new enzymes in plant cells and in cyanobacteria engineered to produce useful chemicals.

They have publicly shared their analysis code and results, including successful and unsuccessful changes, so others can train their own models.

The full study was published in the journal Science Advances.

-- -

Check us out on EarthSnap, a free app brought to you by Eric Ralls and Earth.com.