Abstract
Anti-cancer therapies often exhibit only short-term effects. Tumors typically develop drug resistance causing relapses that might be tackled with drug combinations. Identification of the right combination is challenging and would benefit from high-content, high-throughput combinatorial screens directly on patient biopsies. However, such screens require a large amount of material, normally not available from patients. To address these challenges, we present a scalable microfluidic workflow, called Combi-Seq, to screen hundreds of drug combinations in picoliter-size droplets using transcriptome changes as a readout for drug effects. We devise a deterministic combinatorial DNA barcoding approach to encode treatment conditions, enabling the gene expression-based readout of drug effects in a highly multiplexed fashion. We apply Combi-Seq to screen the effect of 420 drug combinations on the transcriptome of K562 cells using only ~250 single cell droplets per condition, to successfully predict synergistic and antagonistic drug pairs, as well as their pathway activities.
Introduction
Despite major progress over the last decades, cancer remains a major cause of death. Our increased molecular understanding of the molecular basis of cancer has led to the development of targeted therapies. These therapies have so far provided limited efficacy and only in a small subset of patients1, despite major efforts to characterize patients genomically to find response biomarkers.
An approach that holds the promise to improve this situation is to complement large-genomic profiling in basal conditions with measurements after perturbing cancer cells with drugs2. While many approaches can be used to perform drug screenings, they are often low in throughput3, cost and time extensive4, and/or require large amounts of cells5, which together strongly restricts the number of potential drugs that can be screened per tumor biopsy. This limitation gets more pronounced when considering drug combinations due to the sheer number of potential combinations, which increases exponentially with the number of tested drugs.
Due to limited screening capacities, computational approaches to model drug–drug interactions have been developed6. While models on drug efficacies improved over the past years by an increase in available data resources, predictions on drug responses remain challenging and limited to well-characterized systems such as cell lines, thereby limiting their translatability into clinics. Among the different data types, gene expression states of cells were shown to be highly predictive of drug response7. Additionally, data repositories of drug-induced transcriptional changes, such as LINCS8, have proven to be a valuable resource. While there are already perturbation screening platforms available in plates for bulk9,10 and single-cell11,12 transcriptomics, they usually require large numbers of cells per tested condition, and they have not been used for screening drug combinations. Therefore, integrating transcriptomic readouts into a miniaturized combinatorial drug screening platform with the potential to screen tumor biopsies will enable more relevant predictions and increase our understanding of the mode of action of synergistic and antagonistic drug–drug interactions.
Droplet-based microfluidics, which uses picoliter to nanoliter-sized droplets as reaction vessels to perform cellular screens, provides a promising approach to achieve this goal. Due to the miniaturization over several orders of magnitude as compared to conventional plate-based screens, the number of drugs or drug combinations can be massively upscaled while working with low input cell numbers13. We previously demonstrated the first step in this direction by integrating Braille valves into a droplet microfluidic system to generate drug combinations in so-called plugs (~500 nl large droplets) stored sequentially in tubings14. Plugs were used to directly screen 56 combinatorial treatment options on pancreatic tumor biopsies to find the most potent drug pairs using a phenotypic apoptosis readout. While our previous approach provided the first proof of concept in directly screening patient material, the still relatively large volumes of 500 nl limited the number of drug pairs tested. Furthermore, an apoptosis assay provides only a single endpoint readout with limited insights into the drug pairs’ mode of action, which could significantly improve our understanding and the predictability of drug combinations to tackle resistance mechanisms.
To overcome these limitations, we present here a microfluidic platform that allows to perform highly multiplexed screens of hundreds of drug combinations in an emulsion of picoliter-sized droplets. By introducing a deterministic combinatorial barcoding approach, where sets of two barcodes encode drug pairs, we managed to screen all conditions in a highly multiplexed fashion, without the need to keep any spatial order (e.g., wells, plug-sequence). Since the DNA barcodes were designed for whole transcriptome analysis of cells after drug perturbation, we were additionally able to perform massively parallelized gene expression-based profiling of drug combinations. We demonstrated that the presented Combi-Seq approach can be applied to determine the impact of drugs on cell viability and cellular signaling, thus providing a high-throughput approach to discover synergistic drug pairs and decipher their mode of action.
Results
Microfluidic workflow to generate drug combinations in picoliter-sized droplets
Multiplexed combinatorial drug screens were performed in single-cell droplets by encapsulating drugs together with DNA barcode fragments (Fig. 1a). Each pairwise drug combination was encoded by a unique combination of two DNA barcode fragments, which together provided a priming site for reverse transcription (poly-dT) and PCR….
