Manjima
Dhar
a,
Reem
Khojah
a,
Andy
Tay
a and
Dino
Di Carlo
*abc
aDepartment of Bioengineering, University of California, Los Angeles, CA 90095, USA. E-mail: dicarlo@ucla.edu
bCalifornia NanoSystems Institute, Los Angeles, California, USA
cJonsson Comprehensive Cancer Center, Los Angeles, California, USA
First published on 25th September 2015
Individual cells are the fundamental unit of life with diverse functions from metabolism to motility. In multicellular organisms, a single genome can give rise to tremendous variability across tissues at the single-cell level due to epigenetic differences in the genes that are expressed. Signals from the local environment or a history of signals can drive these variations, and tissues have many cell types that play separate roles. This epigenetic heterogeneity is of biological importance in normal functions such as tissue morphogenesis and can contribute to development or resistance of cancer, or other disease states. Therefore, an improved understanding of variations at the single cell level are fundamental to understanding biology and developing new approaches to combating disease. Traditional approaches to characterize epigenetic modifications of chromatin or the transcriptome of cells have often focused on blended responses of many cells in a tissue; however, such bulk measures lose spatial and temporal differences that occur from cell to cell, and cannot uncover novel or rare populations of cells. Here we highlight a flurry of recent activity to identify the mRNA profiles from thousands of single-cells as well as chromatin accessibility and histone marks on single to few hundreds of cells. Microfluidics and microfabrication have played a central role in the range of new techniques, and will likely continue to impact their further development towards routine single-cell epigenetic analysis.
The two methods include indexing droplet (inDrop) by Klein et al.,2 and Drop-seq by Macosko et al.5 Both systems apply similar techniques to prepare mRNA from thousands of cells while maintaining a record of the cell of origin through barcoding introduced on a solid-phase particle or gel, even though samples are pooled for mRNA sequencing. Differences between these systems lie in the specific reactions that occur inside the droplets prior to RNA sequencing.
The inDrop method as described in Fig. 1 extracts mRNA content from cells and tags each cDNA reaction with a barcode specific for an individual cell. In this way the method is able to preserve the heterogeneity of the population of cells in the pooled sequencing step. Three main components are needed in each droplet: (i) a hydrogel particle with barcoded primers, (ii) lysis buffer RT mix, and (iii) cells. First the hydrogel particles with attached barcodes were fabricated. A pinched flow droplet generator device was used to generate barcoded hydrogel microspheres (BHMs) containing single stranded DNA (ssDNA) primers. These ssDNA primers are constructed with an acrylic phosphoroamidite moiety, photo-cleavable spacer, T7 RNA polymerase promoter sequence, sequencing primer, a unique molecular identifier sequence and 2 barcoded regions. The two separate barcode regions allowed for 147456 unique barcodes to be made through a 2-step split and pool synthesis approach in two 384-well plates. The authors estimate 99% of the 3000 cells will have unique barcodes with this level of diversity. The unique molecular identifier allows for improved sequencing read alignments. A second device is used to generate droplets containing lysis buffer, reverse transcription (RT) reagents, BHMs and cells. BHMs are packed tightly and release regularly into droplets. Once the droplets are formed, a UV source (365 nm at 10 mW cm−2, UVP B-100 lamp) is used to release the primers from the BHMs while on ice. Meanwhile the cell contents are already released by the lysis agents in the droplet. Synthesis of cDNA occurs in the droplet, where the cDNA retains a barcode. After barcoding, all of the droplets are broken and contents are pooled. At this point all the cell contents are linearly amplified, and an amplified cDNA library is created. Lastly the cDNA library is sequenced using conventional CEL-seq.
![]() | ||
Fig. 1 The sequence of steps in forming reaction droplets leading to mRNA sequencing for inDrop: (i) cells, barcoded primers, lysis agents and RT mix are encapsulated in droplets. (ii) Cells are lysed to release mRNA and UV light releases the primers. (iii) Reverse transcription occurs in the droplets. Each cDNA gets tagged with a unique barcode during the RT process. (iv) An amplified cDNA library is prepared and sequenced using CEL-seq. Reprinted with permission from Cell, 161, Klein et al.2 |
The Drop-seq method also uses pinching flow junctions to create droplets. The first junction introduces cell suspension, barcoded primer microparticles and lysis buffer, followed by a second junction which pinches off droplets consisting of the first solutions with oil (Fig. 2). The microparticles are similar to the BHMs in that they both carry a cell-specific barcode, a unique molecular identifier, and an oligo-dT for capturing polyadenylated mRNAs. A larger number of 16777
216 barcodes were made on millions of particles using repeated split-and-pool synthesis techniques, generating greater diversity than inDrop. However, unlike the regular delivery of hydrogel particles for inDrop, the delivery of microspheres followed a Poisson distribution leading to a higher percentage of cells without a partner particle. Improvements in regular delivery of particles to the droplet generator could enhance the yield of such 1 cell/1 particle drops.6 Within seconds of droplet formation cells are lysed by the lysis buffer in the droplet. Unlike inDrop, Drop-seq does not require a UV source to release primers; the barcoded primers are readily available on the microparticle for hybridization with the released mRNA. As for the inDrop method, after barcoding, the droplet emulsion is broken by addition of perfluorooctanol which destabilizes the oil–water interface. The mRNA on the particle is reverse transcribed, amplified by PCR and sequenced. The unique molecular identifiers are used to identify any duplicate signal introduced by PCR.
![]() | ||
Fig. 2 Schematic describing sequence of steps used to sequence mRNA in Drop-seq: (i) the first T-junction is used to introduce cells, microparticles with barcoded primers and lysis buffer. (ii) The second junction introduces oil to form the droplets. (iii) The cells are lysed in the droplet and RNA hybridization occurs on the microparticle with barcoded capture nucleic acids. (iv) The droplets are broken and reverse transcription occurs in bulk extending the barcoded capture strands with cDNA sequences. (v) PCR is performed in bulk, followed by sequencing. Each set of sequence data can be mapped to the original cell through the microparticle barcodes. Reprinted with permission from Cell, 161, Macosko et al.5 |
An important aspect of these systems to consider is the occurrence of multiple cells in one droplet, a cell and no barcoded particle, or a cell with two or more barcoded particles. In the case of more than one cell per droplet, mRNA from both cells would be barcoded identically giving an incorrect merged mRNA profile with abnormal magnitude. Both droplet systems tested for this problem by measuring mixtures of mouse and human cells. In the case of inDrop 4% of the droplets contained mixed species of human and mouse cells. In contrast the Drop-seq had a doublet occurrence of 0.36–11%. Microfluidic approaches to control the spacing between cells or prevent cell–cell adhesion could help to reduce this double occupancy.7 Another problem is a cell without a barcoded particle or more than one particle, which would lead to loss of information from a fraction of cells as discussed above or double counting of a population.
The two systems also differ slightly in their mRNA capture efficiency. The inDrop device captures 7% of the original mRNA, while Drop-seq captures 12.8% of the mRNA. Despite the lower capture inDrop was still successfully used to trace embryonic stem cells during early differentiation. Nevertheless, low mRNA capture would indicate that the system will work best in situations with abundant mRNA content, and low expressed transcripts may be lost. Both systems may be limited in performance by its mRNA capture capability.
Overall, both systems can be used to quickly profile single cell gene expression variations and identify cell types from thousands of cells or even whole tissues. Continued work in this field can lead to identification of specific new cell populations within tissues, rare genetic variations that may lead to diseased states, or help us understand the cross-talk between cells in a tissue. These approaches begin to link the population level gating and other data analyses usually performed with flow cytometry with the massive information content of the single cell transcriptome. We anticipate many new cell types will be discovered using such approaches, as has been the case with flow cytometry in the past.
In 2013, the Greenleaf group invented the Assay for Transposase-Accessible Chromatin using sequencing (ATAC-seq) to map the chromatin accessibility in the genome. This method utilizes the hyperactive prokaryotic Tn5 transposase to tag regulatory regions by inserting sequencing adapters into accessible regions of the genome.9 The complete analysis of 50000 pooled cells required only about 3 hours.
In their paper, Buenrostro et al. describe the integration of ATAC-seq with the single-cell microfluidic analysis platform from Fluidigm.10 In the integrated fluidics circuit, single human embryonic stem cells and K562 chronic myelogenous leukaemia cells were captured followed by cell lysis and ATAC reaction. After which, the Tn5 transposase was released for tagging accessible chromatin regions. After the transposition reaction, magnesium chloride was introduced to quench the activity of the transposase and to facilitate the elongation of the Tn5 end. On-chip PCR was then performed. After the PCR step, the libraries of generated fragments can be purified, barcoded on a 96-well plate, and sequenced to identify regulatory elements (Fig. 3).
The advantages conferred by integrating the microfluidic system with ATAC-seq include: (1) analysis of individual cells; (2) fully integrated machine-controlled steps to minimize user error and the time gap in between sample preparation. For instance, transposition steps need to occur immediately after cell lysis due to fragility of the genomic material and an integrated microfluidic platform can be timed to perform sequenced reaction steps; (3) gentle fluidics control that obviates the need for user mixing which can break up DNA fragments and generate ‘false’ DNA fragments that misleadingly suggest higher fragment yield; (4) efficient Tn5 release protocol from the integrated fluidics circuit permit downstream enzymatic reaction for PCR without DNA purification that can lead to DNA loss and more operating time.
Improvements in the technique will be needed to recover more DNA fragments and increase throughput, potentially using droplet-based approaches. The described single cell ATAC-seq approach could potentially be made compatible with the single-cell RNA-seq techniques described above, leading to unique connections between chromatin accessibility and RNA levels that could help deconstruct system-level information flow within a single-cell.
Microfluidic oscillatory washing-based ChIP-seq (MOWChIP-seq) attempts to address these issues, performing genome wide analysis of histone modification regions (Fig. 4).12 In this study, a microfluidic device is used to orchestrate ChIP sequencing of down to 100 cells. Reaction and mixing takes place in an ellipsoid chamber supported by pillars. ChIP antibody coated 2.8 micrometer magnetic beads are filled in the chamber. Then, chromatin fragments from cells are flowed into the packed beads. A magnet is used to retain magnetic IP beads while washing unbound chromatin. Oscillatory washes with IP buffer were tested in terms of duration (1, 5, and 15 min) to balance between non-specific adsorption, chromatin trapping and DNA loss. In addition, fold enrichment was optimized by amount of beads and antibody concentration used for beads coating. Fold enrichment of H3K4me3 histone modifications were measured in two known positive (UNKL and C9orf3) and two negative loci. This method captured 6.2% of total chromatin, within range of the theoretical limit for H3K4me3 after DNA amplification from 1000 human lymphoblastoid cells. The on-chip process required 1.5 hours and ChIP DNA was directly used for sequencing library construction. Receiver operating characteristic curves (ROC) suggest high quality data from small chromatin quantities extracted from 100–600 cells. After optimization, the authors mapped H3K4me and H3K27ac in hematopoietic stem and progenitor cells (HSPCs) isolated from mouse fetal liver. They uncovered new super enhancers with dynamic activity that play a regulatory role in early hematopoiesis.
![]() | ||
Fig. 4 Overview of microfluidic device setup and protocol. (a) Microfluidic device layout with controlled valves. The device has one inlet, one outlet and an on-chip pneumatic micro-valve outlet. LabVIEW is employed to control the switching solenoid valves. (b) Illustration of the MOWChIP-seq protocol. Five main protocol steps on the chip are: (i) formation of a packed bed of IP beads; (ii) ChIP of chromatin fragments flowing through the packed bed; (iii) oscillatory washing; (iv) flushing to remove unbound chromatin fragments and debris; (v) IP bead collection. Reprinted with permission from Nature Methods, Cao et al.12 |
It is very important to control the sensitivity levels of IP antibodies, due to potential cross-reactivity and their impact on downstream analysis of the modified chromatin sites. Scaling down ChIP-seq to a single-cell level will require further microengineering or assay innovations, perhaps in combination with local amplification schemes prior to IP capture.
This journal is © The Royal Society of Chemistry 2015 |