Picodroplet partitioned whole genome amplification of low biomass samples preserves genomic diversity for metagenomic analysis

Abstract

Background: Whole genome amplification (WGA) is a challenging, key step in metagenomic studies of samples containing minute amounts of DNA, such as samples from low biomass environments. It is well known that multiple displacement amplification (MDA), the most commonly used WGA method for microbial samples, skews the genomic representation in the sample. We have combined MDA with droplet microfluidics to perform the reaction in a homogeneous emulsion. Each droplet in this emulsion can be considered an individual reaction chamber, allowing partitioning of the MDA reaction into millions of parallel reactions with only one or very few template molecules per droplet.

Results: As a proof-of-concept, we amplified genomic DNA from a synthetic metagenome by MDA either in one bulk reaction or in emulsion and found that after sequencing, the species distribution was better preserved and the coverage depth was more evenly distributed across the genomes when the MDA reaction had been performed in emulsion.

Conclusions: Partitioning MDA reactions into millions of reactions by droplet microfluidics is a straightforward way to improve the uniformity of MDA reactions for amplifying complex samples with limited amounts of DNA.

Keywords: Amplification bias; Droplet microfluidics; Metagenomics; Multiple displacement amplification; Whole genome amplification.

Fig. 1

Illustration of the method. a DNA is denatured with alkaline solution and neutralized, and the MDA reaction mix is prepared. Aliquots of these solutions are mixed either directly for a bulk reaction in a PCR tube or in a microfluidic chip b to generate an emulsion with c homogeneous picoliter-sized droplets. The generated emulsion is collected in a PCR tube plugged by a PDMS plug to allow maintained droplet stability during long-term incubation

Fig. 2

Amplification yields from MDA reactions in emulsion and standard bulk reactions from different starting template concentrations. Final double stranded (ds) DNA concentrations were measured after breaking the emulsion and collecting the aqueous phase from all droplets of the reaction

Fig. 3

Proportion of properly paired reads mapping to respective reference genome. Absolute numbers of reads are listed in Additional file 1: Table S3

Fig. 4

Coverage breadth of mapped reads and assembled contigs. Semi-transparent bars at the back show the percentage of the genomes that are covered at least once when reads subsampled to include the same total amount of data for all samples were mapped to the reference genomes. Dense bars at the front show the proportion of the genomes that are covered with de novo assembled contigs from the same data set

Fig. 5

Lorenz curves showing the cumulative fraction of mapped bases plotted as a function of the cumulative fraction of the genome that is covered at least once for a T. roseus, b C. akajimensis, and c P. stutzeri. Prior to analysis, the data was subsampled to include the same amount of data, corresponding to an average 5× coverage depth for each respective genome, for all samples