# KTH screens 15,000 bacterial rubisco variants in cyanobacteria; beneficial changes often fail combined

- Published: 2026-11-15 10:16 UTC
- Section: Life & Health › Biotechnology
- Tags: rubisco engineering, bacterial rubisco, Gallionella rubisco, Synechocystis PCC 6803, barcoded mutagenesis library, epistasis, carboxylation and oxygenation, photosynthesis improvement
- URL: https://realtech.media/news/rt-82428774ca1cf7daa79d9f38_kth_screens_15/

**Rubisco variants from Gallionella were tested in living Synechocystis cells; single helpful substitutions often lost their effect when combined. Results are from bacteria, not plants.**

![KTH screens 15,000 bacterial rubisco variants in cyanobacteria; beneficial changes often fail combined](https://realtech.media/media/2026/11/fee8dec90454f0f7.jpg)
*Earth.com*

A team led by Paul Hudson at KTH Royal Institute of Technology in Stockholm screened about 15,000 variants of a bacterial rubisco in living photosynthetic cells. The results were published in Science Advances. The enzyme was the form II rubisco from Gallionella sp., and the host was the cyanobacterium Synechocystis sp. PCC 6803 [1].

Rubisco captures carbon dioxide in photosynthesis but also reacts with oxygen, which costs the cell energy. Fast bacterial forms are less selective for carbon dioxide. The authors aimed to make such enzymes faster or to reduce oxygenation while keeping high carboxylation [1].

## Growth-coupled screen

The cells' own rubisco had been almost entirely switched off, so growth rate depended on the introduced enzyme. Each variant carried a DNA barcode, which allowed variants to be tracked in a pooled culture. The library contained single-site and multi-site variants [1].

Substitutions that improved fitness clustered in loop 6 near the active site, at the dimer interface, and in potential gas tunnels [1]. According to the research summary, about 8% of single-swap variants outgrew the starting enzyme, against under 1% of multi-change variants. One short helix near the substrate opening tolerated no changes: every swap there reduced growth.

## Combinations and models

Combining winners mostly did not work: on average, the more changes a variant carried, the worse the cells grew. The dataset contains examples of epistasis, and the authors note it can serve to benchmark computational models of rubisco [1]. Of 11 new variants proposed by a model trained on the data, eight did not support growth and two beat the starting enzyme. The best variant, Ad1, came from adding proven changes together.

In vitro, Ad1 and a six-change variant, C6, both showed reduced catalytic efficiency for oxygenation (kcat/Ko). Carboxylation turnover (kcat C) increased in one of them [1]. According to the summary, C6 was about 25% faster than the starting enzyme and Ad1 about 15% slower. Both needed more carbon dioxide to reach full speed.

## Limits

The work was done in cyanobacteria with a bacterial enzyme, so any use in crops would need testing in plant cells and whole plants. Screening ran at 5% carbon dioxide, far above air levels, and only two enzymes were characterised in detail. The team is now testing new enzymes in plant cells and in cyanobacteria engineered to produce chemicals.

## Sources

1. [Science Advances](https://doi.org/10.1126/sciadv.aee9222)
2. [Earth.com](https://www.earth.com/plants/scientists-want-to-improve-photosynthesis-to-help-feed-the-world-but-it-isnt-so-simple)

---

Source: RealTech, realtech.media — https://realtech.media/news/rt-82428774ca1cf7daa79d9f38_kth_screens_15/