Microdroplet-based one-step RT-PCR for ultrahigh throughput single-cell multiplex gene expression analysis and rare cell detection

Gene expression analysis of individual cells enables characterization of heterogeneous and rare cell populations, yet widespread implementation of existing single-cell gene analysis techniques has been hindered due to limitations in scale, ease, and cost. Here, we present a novel microdroplet-based, one-step reverse-transcriptase polymerase chain reaction (RT-PCR) platform and demonstrate the detection of three targets simultaneously in over 100,000 single cells in a single experiment with a rapid read-out. Our customized reagent cocktail incorporates the bacteriophage T7 gene 2.5 protein to overcome cell lysate-mediated inhibition and allows for one-step RT-PCR of single cells encapsulated in nanoliter droplets. Fluorescent signals indicative of gene expressions are analyzed using a probabilistic deconvolution method to account for ambient RNA and cell doublets and produce single-cell gene signature profiles, as well as predict cell frequencies within heterogeneous samples. We also developed a simulation model to guide experimental design and optimize the accuracy and precision of the assay. Using mixtures of in vitro transcripts and murine cell lines, we demonstrated the detection of single RNA molecules and rare cell populations at a frequency of 0.1%. This low cost, sensitive, and adaptable technique will provide an accessible platform for high throughput single-cell analysis and enable a wide range of research and clinical applications.


Supplementary Figures
. Droplet characterization and rare droplet enumeration.
(A) Diameter of droplets collected from one experiment to ensure droplet diameter fell around 124 µm.
Small samples of droplets were collected on a glass slide during droplet generation and imaged using an inverted microscope. The diameter of the droplets were determined using ImageJ. Centerline of the box plot shows the median; box limits indicate the 25th and 75th percentiles as determined by R software; whiskers extend 1.5 times the interquartile range from the 25th and 75th percentiles, outlier is represented by a dot. n = 68 sample points.
(B) Our image analysis platform was able to detect and enumerate microdroplets with fluorescence signal down to ratios of 1 in 10,000. Microdroplets were generated using pure PBS or fluorophore conjugated antibodies diluted in PBS. The fluorescent droplets were mixed with droplets without fluorophores at ratios of 1 in 100, 1,000, and 10,000. The mixtures were then imaged using automated fluorescence microscopy to estimate the proportions of fluorescence positive droplets. Figure S2. Inclusion of T7 bacteriophage single-stranded DNA binding protein, gene 2.5 protein (gp2.5) partially rescued qRT-PCR from lysate-mediate inhibition.
(A) Full-length gel electrophoresis image of Figure 2C. The same dilution factor was applied to samples labeled with the same number.
(B) Gata3 mRNA was analyzed in 10,000 EL4 cells using an in-house RT-qPCR mix with increasing gp2.5 concentrations. An agarose gel electrophoresis image visualized the increasing yield of Gata3 products as gp2.5 was increased from 0 μg to 2.0 μg per 10 μL reaction.
(C and D) Gata3 mRNA levels in 20 and 20,000 EL4 cells were analyzed using the in-house reaction, and CellsDirect kit with or without the addition of gp2.5 (10 μL reaction). A 2% agarose gel visualizes the end products of the RT-qPCR with 20,000 EL4 cells (C) and the qPCR amplification curves are shown in D. Combinations of IL-7Rα, Gata3 and EPCR IVTs were spiked into 10 μL RT-qPCRs to determine if the singleplex parameters could quantify vastly different target abundances. Genes analyzed in triplex reactions (red curves) were compared directly to their three respective singleplex controls (black curves) that were performed in the same experiment. Green lines indicate the thresholds for CT values.

RT-qPCR.
Four different combinations of IVT of the target genes were assembled, representing the extreme transcript abundance differences expected in single cells (right). These combinations were quantified using the optimized triplex RT-qPCR mix, and compared directly to the singleplex reactions. Thus, each triplex reaction was compared to three individual singleplex reactions. C T values of each reaction is shown at the bottom right corner of its qPCR plot. The user input parameters that may affect the result of the assay, such as the average number of cells each droplet contains (λ) and the cellular composition of the sample mixture (c), into the simulation. The content and gene signature of each droplet in an experiment is then assigned randomly based on the input parameters, generating a simulated dataset. The simulated data is analyzed using our deconvolution model to produce a prediction w of the cellular composition of the sample. The experiment is repeated to generate a large number of predictions to evaluate their precision and the accuracy by comparing them to the input cellular composition c.

Microfluidic device fabrication
The droplet generator designs were modeled with AutoCAD (

RNA purification and IVT
Primers used to generate the cDNA template for in vitro transcription were designed to flank the region targeted by the primers used in the multiplex reaction (Supplementary Table S2). IVTs are 700-800 bp segments of the mRNA transcript around the exon-exon boundaries targeted for subsequent RT-qPCR.
The forward primers used to generate the cDNA template included a minimal T7 promoter sequence to allow the cDNA to be transcribed by T7 RNA polymerase, and the reverse primers contained a poly(T) sequence for the generation of transcripts with a poly(A) tail.
The generation of the cDNA by PCR was carried out in a 100 μL reaction consisting of 10 mM TrisHCl pH 8.3, 50 mM KCl, 1.5 mM MgCl2, 0.1 mg/mL BSA, 0.05% Triton X-100, 200 μM dNTPs, and a 250 nM pair of forward and reverse primers. The assembled reactions were heated to 94°C before 5 U of Taq polymerase were added. The addition of Taq polymerase at 94°C prevented the amplification of non-specific products at lower temperatures. All PCRs were performed on an Applied Biosystems 9700 instrument using the following thermal cycling conditions: 5 minutes at 94°C, followed by at least 25 cycles of 94°C for 15 seconds, 60°C annealing for 30 seconds and 72°C for 1 minute. Afterward, reactions were held at 72°C for 5 minutes, and then 4°C until they were loaded onto 2.0% agarose gels containing GelRed (Biotium) for analysis. The PCR products were then purified using the Qiagen QIAquick PCR Purification Kit (column-based). The purified cDNA was quantified using a NanoDrop ND-1000, and stored at -20°C until it was needed for in vitro transcription.
In vitro transcription was performed using the MEGAscript T7 Transcription Kit (Invitrogen). 200 ng of purified cDNA was added to each reaction. The reactions were incubated at 37°C for 4 hours. 1 μL of DNase provided in the kit was added to digest the cDNA template, and the synthesized RNA was purified using TRIzol Reagent (Invitrogen). The purified transcripts were quantified using a Nanodrop ND-1000.
IVT copy numbers were approximated based on molecular weight calculations performed using OligoCalc (Northwestern University). IVTs were diluted to 10 10 copies, snap-frozen on dry ice, and stored at -80°C as 2-5 μL aliquots in PCR tubes for single use.