A Raman-based strategy finds more ethanol-tolerant microbes from Baijiu fermentation

http://english.cas.cn/newsroom/research-news/202607/t20260729_1179357.shtml

http://10.1016/j.biortech.2026.135397

Scientists from CAS QIBEBT have developed a Raman-activated cell sorting strategy for the high-throughput discovery of ethanol-tolerant microorganisms based on their metabolic activity rather than growth.

Conventional microbial screening generally follows an “isolate first, test later” workflow, in which cells are cultivated under selective pressure, isolated, and evaluated. This approach is slow, labor-intensive and biased toward fast-growing or abundant microbes. This is important in solid-state baijiu fermentation, where rising ethanol concentrations stress microbial communities, while valuable but low-abundance active microbes are easily overlooked.

Therefore, the researchers shifted the criterion from whether a cell could grow under ethanol stress to whether it maintained active metabolism. They combined heavy-water (D2O) labeling with single-cell Raman spectroscopy and a high-throughput Raman-activated flow cytometer known as FlowRACS. Metabolically active cells incorporate deuterium from D2O into newly synthesized biomolecules, producing a characteristic carbon-deuterium (C-D) Raman signal. Based on this signal, the researchers derived a carbon-to-deuterium ratio (CDR) and a Raman tolerance index (RTI) to quantify the in-situ ethanol tolerance of individual cells.

The platform functionally sorted 2,400 cells per hour with an accuracy of 91.3%. One sorting round followed by culture from pit mud pretreated with 8% ethanol yielded six pure strains, all with an RTI above 50%, as opposed to only two among nine isolates from conventional agarose-plate screening. Overall, this strategy increased strain-mining efficiency 4.5-fold and phenotypic validation 7-fold, reducing the time required for key steps from days to hours.

Genomic and transcriptomic analyses revealed distinct tolerance mechanisms. Lactiplantibacillus plantarum (RTI = 85.1%) enhanced lactate dehydrogenase expression by 3.4-fold while maintaining high alcohol dehydrogenase activity, indicating adaptive redox balancing. Staphylococcus epidermidis (RTI = 62.2%) activated lipid synthesis and glycerol metabolism genes, indicating cell envelope remodeling. Both strains were rare in the original microbial community, demonstrating that function-based sorting recovers key microbes that are missed by abundance- or growth-based approaches.

The study could also be extended to acid, salt, and solvent tolerance, which would aid in building robust industrial chassis strain libraries. This work represents the latest progress of the iMAPS Consortium (in-situ Metabolic Atlas Projects @ Single-cell; www.iMAPS.info).

more insights

https://j.people.com.cn/n3/2026/1116/c95952-20519299.html

The first train for Qingdao Metro Line 7 rolled off the production line on the 12th in Chengyang District, Qingdao, Shandong Province. Featuring smart sensing, smart control, smart operation, and smart maintenance, the train serves as a prime example of advanced rail transit equipment powered by artificial intelligence (AI).

Developed jointly by CRRC Qingdao Sifang Co., Ltd. (“CRRC Sifang”) and the Qingdao Metro Group, this fully automated, smart train is built upon the standardized technical platform for Chinese subway rolling stock. It was designed specifically to meet the smart operation requirements of the Qingdao Metro system. The train has a maximum operating speed of 100 km/h, consists of a six-car configuration, and has a maximum passenger capacity of 1,862.

The train integrates “Zhuolun”—an industrial-grade large language model developed by CRRC. Leveraging AI-driven enhancements, the train achieves “GoA4” (Grade of Automation 4) fully automated operation.

In terms of smart sensing, the train is equipped with advanced “eyes” and an “AI brain.” The vehicle is equipped with over 2,600 monitoring and measurement points. By leveraging both an onboard high-performance computing platform and a ground-based computing center, a smart system that integrates sensing and computation has been established. This system collects real-time data on the vehicle’s status as well as its internal and external environment, utilizing AI algorithms to provide early warnings for potential malfunctions and to perform intelligent analysis of anomalies. Through advanced AI-powered video analysis, the system can intelligently identify irregularities—such as issues with doors, windows, or displays, as well as unusual passenger behavior—enabling a rapid response.

Researchers from the CAS Qingdao Institute of Bioenergy and Bioprocess Technology (QIBEBT), together with collaborators, have developed a Raman-activated cell sorting platform that combines rapid cell sorting with sensitive measurements of weak Raman signals and small-volume cell collection.

Called Pit-RACS, the optical-tweezer-assisted platform operates in two modes on the same microfluidic chip, allowing researchers to adjust the sorting strategy according to the strength of the Raman signal.

Raman-activated cell sorting can identify living cells by their chemical fingerprints without fluorescent labels. Yet researchers have faced a trade-off: fast, flowing systems struggle to measure weak Raman signals, while systems that allow longer measurements sort cells slowly. Collecting the selected cells in a large volume can also make subsequent experiments difficult.

According to the researchers, cells with strong resonance Raman signals are measured and sorted as they flow through the device. In this mode, they sorted up to 836 ± 1 events per minute with a 10-millisecond spectral acquisition time.

For weaker non-resonance signals, optical tweezers move individual cells into a cell-free buffer stream. This allows for longer measurements without interference from nearby cells. This method sorted approximately 50 cells per minute. Switching modes requires no hardware changes.

The platform can also temporarily store selected cells and release them together in approximately 10 microliters of liquid. In the reported tests, the collected cells were over 97% pure and could proceed to cultivation or genomic analysis without an additional concentration step.

“Raman signals vary greatly between biological targets. We designed the two modes so researchers can choose the measurement time their question requires while using the same sorting platform,” said Prof. MA Bo, corresponding author of the study.

To demonstrate industrial strain screening, the researchers sorted a mutant library of Yarrowia lipolytica and identified a strain with 77.4% higher β-carotene yield than the control. In another demonstration, they combined heavy-water labeling with Pit-RACS to enrich phosphate-solubilizing microorganisms from river water. They used the concentrated output directly for metagenomic sequencing, revealing 12 species of phosphate-solubilizing bacteria across nine genera.

Back to …