Limits...
Scalable microfluidics for single-cell RNA printing and sequencing.

Bose S, Wan Z, Carr A, Rizvi AH, Vieira G, Pe'er D, Sims PA - Genome Biol. (2015)

Bottom Line: We then develop a scalable technology for genome-wide, single-cell RNA-Seq.Our device generates pooled libraries from hundreds of individual cells with consumable costs of $0.10-$0.20 per cell and includes five lanes for simultaneous experiments.We anticipate that this system will serve as a general platform for single-cell imaging and sequencing.

View Article: PubMed Central - PubMed

Affiliation: Department of Systems Biology, Columbia University Medical Center, New York, NY, 10032, USA. sb3438@columbia.edu.

ABSTRACT
Many important biological questions demand single-cell transcriptomics on a large scale. Hence, new tools are urgently needed for efficient, inexpensive manipulation of RNA from individual cells. We report a simple platform for trapping single-cell lysates in sealed, picoliter microwells capable of printing RNA on glass or capturing RNA on beads. We then develop a scalable technology for genome-wide, single-cell RNA-Seq. Our device generates pooled libraries from hundreds of individual cells with consumable costs of $0.10-$0.20 per cell and includes five lanes for simultaneous experiments. We anticipate that this system will serve as a general platform for single-cell imaging and sequencing.

No MeSH data available.


Related in: MedlinePlus

Cell type separation by single-cell RNA-Seq. a iPAGE gene ontology/pathway analysis based on rank-ordering of differentially expressed genes using +/−(1-p) where p is the P value for differential expression between the U87- and MCF10a-exclusive lanes given by the Wilcoxin rank-sum test. Values are positive for genes more highly expressed in U87 and negative for genes more highly expressed in MCF10a. b t-SNE clustering of 396 single-cell profiles based on the differentially expressed genes color-coated by the lane-of-origin of each profile. Two clear spatial clusters form and each is predominantly associated with a specific cell type-exclusive lane. c The same t-SNE clustering shown in (b) but color-coated with a score indicating expression of the U87-specific genes vs. the MCF10a-specific genes. The score is based on the relative rank ordering of U87- and MCF10a-specific genes in each cell (see Methods)
© Copyright Policy - open-access
Related In: Results  -  Collection

License 1 - License 2
getmorefigures.php?uid=PMC4487847&req=5

Fig7: Cell type separation by single-cell RNA-Seq. a iPAGE gene ontology/pathway analysis based on rank-ordering of differentially expressed genes using +/−(1-p) where p is the P value for differential expression between the U87- and MCF10a-exclusive lanes given by the Wilcoxin rank-sum test. Values are positive for genes more highly expressed in U87 and negative for genes more highly expressed in MCF10a. b t-SNE clustering of 396 single-cell profiles based on the differentially expressed genes color-coated by the lane-of-origin of each profile. Two clear spatial clusters form and each is predominantly associated with a specific cell type-exclusive lane. c The same t-SNE clustering shown in (b) but color-coated with a score indicating expression of the U87-specific genes vs. the MCF10a-specific genes. The score is based on the relative rank ordering of U87- and MCF10a-specific genes in each cell (see Methods)

Mentions: Figure 7a shows a pathway analysis of gene ontologies enriched across >11,600 genes that were both detected across our 396 single-cell profiles and available in the iPAGE database and ranked based on differential expression [35] in Experiment 1. We generated a matrix of Spearman correlation coefficients across our 396 profiles based on rank-ordering the 189 differentially expressed genes in each cell. We then clustered the data spatially using the t-stochastic neighborhood embedding (t-SNE) algorithm [36], a powerful clustering algorithm that has recently been applied to high-dimensional single-cell analysis data (Fig. 7b, c) [37]. Our t-SNE result contains two closely associated clusters of individual cells. To understand the origin of these two clusters, we displayed our t-SNE clustering with two different color-schemes. In Fig. 7b, we show how single-cell profiles from the various lanes of our device are distributed. As expected, one of the two clusters contains all of the cells from the MCF10a-exclusive lane, while the other contains nearly all of the cells from the U87 lane with a few exceptions. Single-cell profiles from the mixed lanes are distributed throughout the two clusters, although not with perfect uniformity. While the single-cell profiles from mixed lanes are distributed uniformly throughout the MFC10a cluster, there is some separation between a subset of mixed lane profiles and the U87-exclusive lane profiles in the U87 cluster. In Fig. 7c, we display the same clustering result with a different color scheme that indicates the relative rank ordering of the U87 vs. MCF10a gene sets in each profile. This metric clearly associates the cells in each of the two clusters with the expected cell type-specific expression pattern. However, the subset of the cells from the mixed lane in the U87 cluster that are somewhat separated from those in the U87-exclusive lane exhibit more ambiguous gene expression than others. This phenomenon could arise from one or some combination of several sources including: (1) a lane-specific batch effect; (2) the occasional presence of more than one cell in a microwell containing a single bead; (3) the association of a single cell-identifying barcode with more than one cell (which will occur for a few percent of cells given the current size of our barcode library); (4) low levels of cross-talk between wells or imperfections in the sealing of our device at specific locations in the microwell array; and (5) actual phenotypic differences between U87 cells in the presence of other U87 cells vs. MCF10a cells, which may secrete distinct and stimulatory diffusible factors. We also note that a small subpopulation of U87 cells in the U87-exclusive lane actually cluster with the MCF10a cells, highlighting the possibility of phenotypic heterogeneity within the U87 population. We conducted a similar differential expression and clustering analysis for Experiment 2 (Additional file 1: Figure S4) with closely related results. In any case, while there is certainly substantial opportunity to improve this technology from several angles, this analysis provides a compelling demonstration of the initial capabilities of this technology.Fig. 7


Scalable microfluidics for single-cell RNA printing and sequencing.

Bose S, Wan Z, Carr A, Rizvi AH, Vieira G, Pe'er D, Sims PA - Genome Biol. (2015)

Cell type separation by single-cell RNA-Seq. a iPAGE gene ontology/pathway analysis based on rank-ordering of differentially expressed genes using +/−(1-p) where p is the P value for differential expression between the U87- and MCF10a-exclusive lanes given by the Wilcoxin rank-sum test. Values are positive for genes more highly expressed in U87 and negative for genes more highly expressed in MCF10a. b t-SNE clustering of 396 single-cell profiles based on the differentially expressed genes color-coated by the lane-of-origin of each profile. Two clear spatial clusters form and each is predominantly associated with a specific cell type-exclusive lane. c The same t-SNE clustering shown in (b) but color-coated with a score indicating expression of the U87-specific genes vs. the MCF10a-specific genes. The score is based on the relative rank ordering of U87- and MCF10a-specific genes in each cell (see Methods)
© Copyright Policy - open-access
Related In: Results  -  Collection

License 1 - License 2
Show All Figures
getmorefigures.php?uid=PMC4487847&req=5

Fig7: Cell type separation by single-cell RNA-Seq. a iPAGE gene ontology/pathway analysis based on rank-ordering of differentially expressed genes using +/−(1-p) where p is the P value for differential expression between the U87- and MCF10a-exclusive lanes given by the Wilcoxin rank-sum test. Values are positive for genes more highly expressed in U87 and negative for genes more highly expressed in MCF10a. b t-SNE clustering of 396 single-cell profiles based on the differentially expressed genes color-coated by the lane-of-origin of each profile. Two clear spatial clusters form and each is predominantly associated with a specific cell type-exclusive lane. c The same t-SNE clustering shown in (b) but color-coated with a score indicating expression of the U87-specific genes vs. the MCF10a-specific genes. The score is based on the relative rank ordering of U87- and MCF10a-specific genes in each cell (see Methods)
Mentions: Figure 7a shows a pathway analysis of gene ontologies enriched across >11,600 genes that were both detected across our 396 single-cell profiles and available in the iPAGE database and ranked based on differential expression [35] in Experiment 1. We generated a matrix of Spearman correlation coefficients across our 396 profiles based on rank-ordering the 189 differentially expressed genes in each cell. We then clustered the data spatially using the t-stochastic neighborhood embedding (t-SNE) algorithm [36], a powerful clustering algorithm that has recently been applied to high-dimensional single-cell analysis data (Fig. 7b, c) [37]. Our t-SNE result contains two closely associated clusters of individual cells. To understand the origin of these two clusters, we displayed our t-SNE clustering with two different color-schemes. In Fig. 7b, we show how single-cell profiles from the various lanes of our device are distributed. As expected, one of the two clusters contains all of the cells from the MCF10a-exclusive lane, while the other contains nearly all of the cells from the U87 lane with a few exceptions. Single-cell profiles from the mixed lanes are distributed throughout the two clusters, although not with perfect uniformity. While the single-cell profiles from mixed lanes are distributed uniformly throughout the MFC10a cluster, there is some separation between a subset of mixed lane profiles and the U87-exclusive lane profiles in the U87 cluster. In Fig. 7c, we display the same clustering result with a different color scheme that indicates the relative rank ordering of the U87 vs. MCF10a gene sets in each profile. This metric clearly associates the cells in each of the two clusters with the expected cell type-specific expression pattern. However, the subset of the cells from the mixed lane in the U87 cluster that are somewhat separated from those in the U87-exclusive lane exhibit more ambiguous gene expression than others. This phenomenon could arise from one or some combination of several sources including: (1) a lane-specific batch effect; (2) the occasional presence of more than one cell in a microwell containing a single bead; (3) the association of a single cell-identifying barcode with more than one cell (which will occur for a few percent of cells given the current size of our barcode library); (4) low levels of cross-talk between wells or imperfections in the sealing of our device at specific locations in the microwell array; and (5) actual phenotypic differences between U87 cells in the presence of other U87 cells vs. MCF10a cells, which may secrete distinct and stimulatory diffusible factors. We also note that a small subpopulation of U87 cells in the U87-exclusive lane actually cluster with the MCF10a cells, highlighting the possibility of phenotypic heterogeneity within the U87 population. We conducted a similar differential expression and clustering analysis for Experiment 2 (Additional file 1: Figure S4) with closely related results. In any case, while there is certainly substantial opportunity to improve this technology from several angles, this analysis provides a compelling demonstration of the initial capabilities of this technology.Fig. 7

Bottom Line: We then develop a scalable technology for genome-wide, single-cell RNA-Seq.Our device generates pooled libraries from hundreds of individual cells with consumable costs of $0.10-$0.20 per cell and includes five lanes for simultaneous experiments.We anticipate that this system will serve as a general platform for single-cell imaging and sequencing.

View Article: PubMed Central - PubMed

Affiliation: Department of Systems Biology, Columbia University Medical Center, New York, NY, 10032, USA. sb3438@columbia.edu.

ABSTRACT
Many important biological questions demand single-cell transcriptomics on a large scale. Hence, new tools are urgently needed for efficient, inexpensive manipulation of RNA from individual cells. We report a simple platform for trapping single-cell lysates in sealed, picoliter microwells capable of printing RNA on glass or capturing RNA on beads. We then develop a scalable technology for genome-wide, single-cell RNA-Seq. Our device generates pooled libraries from hundreds of individual cells with consumable costs of $0.10-$0.20 per cell and includes five lanes for simultaneous experiments. We anticipate that this system will serve as a general platform for single-cell imaging and sequencing.

No MeSH data available.


Related in: MedlinePlus