Kobe University and Shimadzu automate metabolome sampling of microbial cultures

Kobe University and Shimadzu said on October 20, 2026 that they had developed an automated metabolome analysis system joining microbial culture, sampling, quenching of metabolism inside the cells, centrifugation, metabolite extraction and dispensing into analysis plates. Metabolome analysis reads a cell's state from the sugars, amino acids and energy-related compounds inside it. Kobe University says broth reaches the tube that stops the metabolic reactions about 13 seconds after leaving the vessel, and that the machine keeps to its set sampling times at night and in the early morning. The stated use is picking strains for producing pharmaceutical raw materials, food ingredients, biofuels and chemicals; neither partner has shown use in GMP manufacturing or quality control testing. Samples go into 96-well plates, and the throughput of one run, the time per sample and any plan or price for a product are unpublished. The system stands at the DBTL smart cell development hub set up at Kobe University under a NEDO project, where companies can use it.

The lag between drawing a sample and stopping it decides the value

For cells still in culture, the result turns on the time between drawing the sample and stopping the metabolic reactions. Some metabolites inside cells change within seconds, so a sample that keeps reacting gives a different value from the one that was in the vessel. Labs therefore mix the sample with a chilled solution right after drawing it, a step called quenching. Measuring sugars and acetate in the cell-free supernatant is a different job: inside the cells, the handling between drawing and stopping is part of the measurement conditions.

Sampling, chilling, centrifugation, removal of the supernatant and addition of extraction solvent have mostly been done by hand, so the lag varied with the operator and the number of samples. The two say the hard part of time-course work was processing many samples the same way in a short time, and that night and early-morning sampling needed an experienced person on standby. They describe the system as the world's first to automate culture through sample preparation as one unit (as of October 2026, according to Kobe University).

Handing the work to the machine did not add much variation. The paper reports a coefficient of variation averaging 6.9% when one extract was measured eight times on the instrument, and 8.8% when eight samples drawn from the same broth were each processed automatically after quenching, an increase of 1.9 points. Both are averages over the 76 compounds detected in every replicate. A barcode ties each sample to the vessel, the sampling time and its place on the plate.

What it changes when you choose a strain

What changes is the evidence you hold when two strains look alike in growth and one has to be chosen for production. Four strains of E. coli used widely in research and in material production were measured to 19 hours after the start of culture. Even where the cells were multiplying alike, the amounts and the movement of sugars, amino acids and energy-related compounds differed by strain: the W strain held relatively high intracellular branched-chain amino acids, and in the C strain, sugar phosphates and intermediates of the MEP pathway, the route toward terpene synthesis, rose together.

The work was published online in the American Chemical Society journal Analytical Chemistry on October 19, 2026 and was chosen for the cover. The partners expect the recorded processing times, temperatures and sample IDs to make metabolome data comparable across researchers and institutions. Development started in a NEDO project on production techniques for high-performance products using plants and other organisms (fiscal 2016 to 2021) and continued after it ended.

For a team outside Japan, this is not a catalog product but a machine at a Kobe university hub open to companies, with quenching time and preparation steps fixed so strains and media can be compared on intracellular metabolites under the same handling. Before building it into a project, ask the throughput of a run, the time per sample, and whether your own strain and medium stay inside the same variation.