Genetic research at the laboratory
Credit: Pgiam / Getty Images

A new study shows how a scalable microfluidic workflow, called Combi-seq, can screen hundreds of drug combinations in picolitre-size droplets for transcriptome changes as a readout for drug effects. Using this process, the group was able to predict synergistic and antagonistic effects of more than 400 drug pairs, as well as their pathway activities.

The work was led by Christoph Merten, former group leader at EMBL Heidelberg and presently at the Swiss Federal Institute of Technology Lausanne (EPFL). It was carried out in collaboration with the group of Julio Saez-Rodriguez, professor at the Medical Faculty of Heidelberg University, and a group leader at EMBL. The computational analysis was led by Bence Szalai.

The study was published in Nature Communications.

Merten pointed out key features of the system to Inside Precision Medicine:

  • It is able to generate droplets of 420 different chemical compositions (not just different concentrations, but rather different compounds/mixtures) within just 90 min, with about 500 replicates per composition.
  • It features a barcoding strategy to sequence the entire transcriptome of cancer cells in response to 420 different treatments in a single experiment
  • It provides not just a single phenotypic readout (e.g., a fluorescence signal upon apoptosis), but rather for a detailed cancer signaling map—allowing resistance mechanisms, biomarkers, and drug sensitizers to be discovered at large scale.

Many experts believe the next wave in anti-cancer therapeutics is likely to arrive in the form of drug combinations tailored specifically to a patient’s own, unique tumor cells. But since every patient’s tumor cells are different and continuously accumulate mutations, it is difficult to predict how they will react to a specific drug.

“When testing the effects of drugs, we are limited by the amount of tissue we can obtain from patient biopsies, which is generally low. This means that with conventional technology, it’s impossible to screen hundreds of drug combinations for their effect on patient tumor cells,” Merten said.

Present technologies can only test a handful of drug conditions at a time. They also usually provide purely yes-or-no answers, e.g., whether the cells live or die after treatment.

Combi-seq, the researchers say, overcomes these challenges through innovative use of microfluidics. Because of the low volume of liquids required, researchers can carry out large-scale experiments with very small sample volumes. In 2018, Merten’s and Saez-Rodriguez’s groups employed microfluidics to test 56 anticancer drug combinations in cancer cells from patients.

The present technique, established by former EMBL PhD student Lukas Mathur, takes this process even further. It works by precise microfluidic manipulation of cells in solution. The researchers used this to isolate cells in droplets, each of which was only around a tenth of a millimeter in diameter. In addition to a cancer cell, each droplet contained a specific drug combination and a DNA sequence called a ‘barcode’, used as a label for the applied treatment condition. After 12 hours of treatment, the researchers could pool the cells, collect their genetic material for sequencing (identified by the ‘barcodes’), and analyze the results.

Not only does Combi-seq massively scale up how many treatment combinations can be tested simultaneously, it also allows scientists to gather accurate transcriptomics data from drug-exposed cells. It does so by incorporating next-generation sequencing into the workflow. Instead of just telling us whether a cell lives or dies after drug treatment, this method provides a wealth of information about the cell’s response that medical professionals might be able to draw upon when determining treatment strategies.

“Using such transcriptomics data, we can make statements about how the signaling pathways in the cell react to the drug or about which genes are up- or down-regulated in response. This is so much more powerful than anything we have had previously,” Merten said.

“Generating these datasets for different patients for a tumor type and applying advanced computational methods on them can improve our understanding of why drugs often do not work and ultimately improve patient care,” said Julio Saez-Rodriguez.

Also of Interest