Comparative analysis of droplet-based ultra-high-throughput single-cell RNA-seq systems

Background

Single-cell RNA sequencing (scRNA-seq) has been crucial in the study of biological heterogeneity and in the characterisation of rare cell types (1). Multiple different technologies can be used to obtain single cells for RNA sequencing, spanning from fluorescence-activated cell sorting (FACS) (2) to both continuous-flow (3) and droplet-based microfluidics.

Droplet microfluidics enables the high-throughput generation of water-in-oil droplets (4). By co-encapsulating both cells and barcode-containing beads/hydrogels, one can label all the RNA from each cell with a unique barcode, such that after pooling and sequencing, each read can be mapped back to its cell-of-origin. The use of droplet microfluidics has enabled a massive increase in throughput with a reduction in cost. The three main methods for droplet-based single-cell sequencing are InDrop (5), Drop-seq (6) and 10X genomics, which differ in bead type, bead manufacturing, barcode design, cDNA amplification, amongst others. In a recent preprint, Zhang, Li, Liu, Chen and colleagues compared the three droplet-based methods to assess their respective strengths and weaknesses.

 

Comparison of InDrop, Drop-seq and 10X genomics

The three systems operate by similar principles, where bead-embedded primers with cell barcodes are used to capture RNA from each cell. These primers have a similar structure – all contain PCR handles, cell barcodes, unique molecular identifiers and a poly-T section to capture mRNA. However, the beads differ in their material, which influences the characteristics of the technology. The use of brittle resin for Drop-seq beads results in the beads being encapsulated with a typical Poisson distribution, while the deformable InDrop and 10X genomics beads enable bead occupancies to reach over 80%.

The use of surface-tethered primers in Drop-seq, as opposed to primers that are released via photocleavage (in InDrop) or dissolving of the beads (in 10X genomics), could influence capture efficiency. This also affects where reverse transcription takes place; in Drop-seq, reverse transcription takes place after the beads are released from the droplets, while in InDrop and 10X genomics, the reverse transcription must take place in droplets.

To compare the three different methods, the authors sequenced the same cell line and developed a workflow capable of processing data from all three methods.

10X genomics outperforms both InDrop and Drop-seq in terms of bead quality, with more than half of the cell barcodes in the latter two systems containing obvious mismatches. Additionally, the proportion of effective reads (from valid barcodes) was ~75% for 10X Genomics, but merely ~25% and ~30% for InDrop and Drop-seq respectively.

The raw read levels for the different samples were normalised before gene expression analysis, and it was found that 10X Genomics had the highest sensitivity (capturing 17,000 transcripts from ~3,000 genes on average), followed by Drop-seq (~8,000 transcripts from ~2,500 genes) and InDrop (~2,700 transcripts from ~1,250 genes). Moreover, technical noise is more severe in inDrop data, followed by Drop-seq and then 10X Genomics data.

Comparison of the data generated using the different methods show a large technology-based bias, suggesting that there is system-specific quantification bias present. The authors found that 10X favoured the capture and amplification of shorter genes and genes with higher GC content, while Drop-seq favoured genes with lower GC content.

 

Discussion

The authors compared the three main droplet-based single-cell RNA-seq systems using the same cell line and a unified data processing pipeline to enable a fair comparison. They found that 10X genomics outperforms the other two technologies in various aspects, such as sensitivity, precision and cell barcode quality. However, sequencing cells using 10X genomics is more expensive ($0.87 per cell) compared to InDrop and Drop-seq ($0.44-$0.47 per cell).

Drop-seq performs only slightly worse than 10X Genomics, but is substantially cheaper, making it an attractive choice for labs. The largely open-source nature of Drop-seq (except for the beads), also enables technical modification and development. However, the fact that bead encapsulation here follows a Poisson distribution (unlike in the other two methods) makes the technology less desirable for the study of precious and limited cell samples.

In contrast, InDrop is completely open-source, where even the beads can be manufactured in labs. The authors believe that inDrop does not perform well due to its excessive cDNA amplification, and the fact that the protocol has not been completely optimised. However, inDrop is extremely flexible and amenable to modification. The authors demonstrate this by successfully implementing Smart-seq2, the most widely used scRNA-seq protocol, in the inDrop system, suggesting that inDrop could be ideal for technical development.

 

What I like about this work

The use of droplet microfluidics for single-cell RNA sequencing has revolutionised the field, but there was a lack of proper comparative studies between these three methods. This thorough and systematic comparison could inform potential users of the advantages and disadvantages of the different methods, enabling them to pick a suitable system for their application.

 

Questions

• What aspects of the InDrop protocol do you believe requires optimisation to improve its performance?
• What do you think about the new members of the single-cell RNA sequencing field – sci-RNA-seq and SPLiT-seq? Do you think that droplet-based methods will continue to be relevant?

 

Further reading

InDrop: https://www.cell.com/cell/fulltext/S0092-8674(15)00500-0, https://www.nature.com/articles/nprot.2016.154

Drop-seq: https://www.cell.com/fulltext/S0092-8674(15)00549-8

10x genomics: https://www.nature.com/articles/ncomms14049,  https://www.10xgenomics.com/

 

References

1. Tanay A, Regev A. Scaling single-cell genomics from phenomenology to mechanism. Nature. 2017 18;541(7637):331–8.
2. Krjutškov K, Katayama S, Saare M, Vera-Rodriguez M, Lubenets D, Samuel K, et al. Single-cell transcriptome analysis of endometrial tissue. Hum Reprod Oxf Engl. 2016 Apr;31(4):844–53.
3. Gong H, Do D, Ramakrishnan R. Single-Cell mRNA-Seq Using the Fluidigm C1 System and Integrated Fluidics Circuits. Methods Mol Biol Clifton NJ. 2018;1783:193–207.
4. Sakai S, Kawabata K, Ono T, Ijima H, Kawakami K. Preparation of mammalian cell-enclosing subsieve-sized capsules (<100 microm) in a coflowing stream. Biotechnol Bioeng. 2004 Apr 20;86(2):168–73.
5. Klein AM, Mazutis L, Akartuna I, Tallapragada N, Veres A, Li V, et al. Droplet Barcoding for Single-Cell Transcriptomics Applied to Embryonic Stem Cells. Cell. 2015 May 21;161(5):1187–201.
6. Macosko EZ, Basu A, Satija R, Nemesh J, Shekhar K, Goldman M, et al. Highly Parallel Genome-wide Expression Profiling of Individual Cells Using Nanoliter Droplets. Cell. 2015 May 21;161(5):1202–14.

Tags: 10x genomics, drop-seq, droplet microfluidics, indrop, rna sequencing, single-cell sequencing

doi: https://doi.org/10.1242/prelights.5180

Read preprint (No Ratings Yet)