A microfluidics platform for combinatorial drug screening on cancer biopsies

Abstract

Screening drugs on patient biopsies from solid tumours has immense potential, but is challenging due to the small amount of available material. To address this, we present here a plug-based microfluidics platform for functional screening of drug combinations. Integrated Braille valves allow changing the plug composition on demand and enable collecting >1200 data points (56 different conditions with at least 20 replicates each) per biopsy. After deriving and validating efficient and specific drug combinations for two genetically different pancreatic cancer cell lines and xenograft mouse models, we additionally screen live cells from human solid tumours with no need for ex vivo culturing steps, and obtain highly specific sensitivity profiles. The entire workflow can be completed within 48 h at assay costs of less than US$ 150 per patient. We believe this can pave the way for rapid determination of optimal personalized cancer therapies.

Fig. 1

Microfluidic setup. a Chip design. A total of 16 syringes with aqueous samples are connected to the inlets in the microfluidic chip via tubing (10 with compounds, 2 with medium to generate single drug and control conditions, 2 for barcoding, 1 for the cell suspension, 1 with Caspase-3 substrate to detect apoptosis). Other 2 inlets in the microfluidics chip are used for carrier oil (FC-40) and mineral oil. The braille display unit is used to control the valves (red coloured circles) and regulate the flow coming from the aqueous phase syringes, resulting in different combinations. Plugs are collected in a tube connected to the outlet. b Experimental setup. Microfluidic chip mounted onto a Braille display aligning the microfluidic channels with the braille valves. c, d Cross section of a valve mounted on the Braille display in the open and closed configuration. e Combinatorial plugs production. Single compounds and pairwise combinations are automatically generated. First barcode (BC) plugs are generated followed by the corresponding assay plugs for each condition. f Storage of plugs. Plugs are stored in PTFE tubing for incubation and readout purposes. g Array of binary barcodes in a microfluidic channel. Plugs contain two different dyes (bright colour indicating a “1”, dark color indicating a “0”) were used to generate binary numbers 40–50

Fig. 2

General workflow of data acquisition and processing. a Different conditions are generated sequentially with multiple replicates (aqueous plugs) produced for each condition and collected in a tube. The whole experiment (run) is repeated multiple times and data for each run are processed separately and then combined. b Example of a plug sequence and corresponding readout. Each aqueous plug is generated by mixing the following components: cells (with orange-fluorescent dye to verify the proper mixing of all components), caspase-3 substrate and one or two compounds. Plugs of each tested condition are preceded by binary barcode plugs to encode the corresponding identification number (high concentration of blue fluorescent dye = 1, low concentration = 0, followed by an end of barcode signal)

Fig. 3

Microfluidic combinatorial drug screening in cell lines. a Boxplot of the sequence of conditions across multiple replicates for BxPC3 cells (z-scores of Caspase-3 activity) with control conditions in purple and conditions which are not significantly (Wilcoxon rank-sum test; p-value < 0.05) better than the control with light grey borders and dots. For each condition, replicates are shown as dots, and summary statistics are represented using a horizontal line for the median and a box for the interquartile range. The whiskers extend to the most extreme data point, which is no >1.5 times the length of the box away from the box. b Heatmap representation of the same conditions as in a (BxPC1 cell line) using median value across 6 replicates. Colored red scale starting from z-score equal to 0 (i.e., median activation across all conditions) while negative values are represented in grey. Conditions which are not significantly better than the control are marked with an x. c Heatmap representation for AsPC1 cells. d Overall drug efficacy for each cell line, computed as median z-score across all conditions (single or combinatorial perturbations) involving the drug. e Top 10 cell line-specific conditions for each cell line

Fig. 4

Validation of cell line-specific drug combinations in tissue culture experiments. a In all boxplots the horizontal line represents the median, the box the interquartile range and the whiskers extend to the most extreme data point which is no >1.5 times the length of the box away from the box. Conditions significantly higher than zero are marked (** one-tailed t-test; p-value ≤ 0.01). BxPC3 show a strong activation in the microfluidic system when treated with the combination of MK-2206 and PHT-427, which is not seen when treated with the single drugs. No strong activation is shown for AsPC1 treated with the same drugs. b Same behaviour is confirmed in 96-well plates where the drug combination shows a much higher signal than the two single drug samples for BxPC3 cells, while it stays at the basal level for AsPC1 cells. Error bars represent the standard error of the mean from two replicates. c, d Similarly, the combination of ACHP and Gefitinib is potent in AsPC1 cells but not in BxPC3 cells, both in microfluidic plugs (c) and in a 96-well plate format (d)

Fig. 5

In vivo validation of drug combinations in mouse xenograft models. a Percentage increase of tumour size (mm³) with respect to the first day of treatment (day 0) for AsPC1 and BxPC3, respectively. Statistical comparisons are reported with the corresponding p-value (if <0.1) or as ‘ns’ (not significant) when p-value >0.1. b Percentage variation with respect to the vehicle is shown to directly compare the two cell lines when treated with Gemcitabine or combination of MK2206 and PHT-427. For both panels, p-values are computed using one-tailed t-test and error bars represent the standard error of the mean. The experiment started with 5 mice per each group, error bars are not reported for the condition Gefitinib + ACHP after the third day, where only one mouse survived

Fig. 6

Microfluidic combinatorial drug screening of patient biopsies. a Workflow for patient samples. Functional drug screening on biopsies from human patients: each biopsy is dissociated to a single cell suspension, which is perturbed with different compounds using the microfluidics platform. b Overall drug efficacy for each cell line, computed as median z-score across all conditions (single or combinatorial perturbations) involving the drug. c Heatmap representation of efficacy of single and combinatorial drug perturbations for each patient using colored red scale starting from z-score equal to 0 (i.e., median activation across all conditions) while negative values are represented in grey. Conditions which are not significantly better than the control are marked with an x. The corresponding cell in the matrix is marked with ‘o’ in case of unmeasured conditions (two of the 10 drugs were not screened for the biopsy of liver metastasis) or technical issues (e.g., conditions with plugs showing unexpected dilutions of the orange marker dye added to the cell suspension). d Top 10 most effective drugs or combinations for each patient. Grey scale is used to map the specificity of each combination with light grey representing very unspecific conditions (i.e., effective in all patients where the condition was tested) and black representing highly specific conditions (effective only in one patient). e Comparison of efficacy of different patient samples to all perturbations with drug combinations ordered from left to right based on the number of patient samples in which they are effective. Drug targets are also shown in dark grey