Pico Profiling – The Road to Single-cell Transcriptomics

Biowire Fall 2011 — Screening — microRNA Target Identification

Herbert Auer, Functional Genomics Core IRB Barcelona, Barcelona, Spain


It has become increasingly apparent that seemingly homogenous cell populations in fact consist of many different subpopulations. This in-depth understanding of cell populations necessitates the characterization of these smaller subpopulations on a molecular level. We show that TransPlex® Whole Transcriptome Amplification (WTA2) chemistry is perfectly suited to generate large amounts of cDNA that faithfully represent biological differences between interrogated samples of very small cell populations.

back to top Back to Top


The Functional Genomics Core (FGC) within the Institute for Research in Biomedicine (IRB), Barcelona, provides genomic analysis tools like microarrays and Next Generation Sequencing to researchers at the IRB, as well as to research groups across Spain and other countries within Europe. A request that repeatedly came up was whether it is possible to characterize the transcriptome (all transcripts) of very small cell populations, individual cells, or even subcellular structures like the nucleus of an individual cell. There are numerous examples of research areas in need of characterizing individual cells or very small populations including: i) developmental biology, to study the early steps of development where a few cells change their fate and develop a specific organ or part of the body; ii) stem cell research, since multi- or omni-potent stem cells normally make up a minute fraction of cells in a specific tissue; iii) oncology, for characterizing individual cells since tumors are very heterogeneous containing a small fraction of cells with the capacity to metastasize.

back to top Back to Top

Available chemistries to generate large amounts of cDNA from small cell populations

Since FGC tries to fulfill the needs of research groups it works for, we evaluated several commercially available kits and screened the scientific literature for publications describing expression profiling from very small cell populations. Except for single molecule sequencing of transcripts1, which theoretically could directly sequence the RNA content of a single cell, the majority of proposed methods for transcriptomic characterization of very small cell populations contain an amplification step after cDNA synthesis to obtain sufficient material for downstream analysis by qPCR, microarrays or sequencing. The first method for quantitative transcriptional profiling was published almost 20 years ago2 and numerous others followed3-10. The amplification principles underlying these methods can be divided into linear and logarithmic amplifications. In linear amplification, one end of the cDNA is used to start amplification while logarithmic amplification binds primers to both ends of the cDNA and generates an exponential increase of molecules.

The question to be asked was whether the vast amounts of cDNA generated during the amplification process could still provide quantitative information about transcript abundances in the starting material, namely the few picograms of total RNA contained in an average cell11. It is well known that absolute levels of transcripts get altered during an amplification procedure12 but this should effect samples in an equal manner; consequently, relative differences between samples should be maintained, leading to measurements of differential expression representing biological differences between samples, as long as the amplification method works reproducibly across samples.

We developed a checklist of questions (Table 1) that had to be answered to establish whether a proposed amplification method for small amounts of RNA or few cells faithfully represents the underlying biology of the interrogated cell(s). Unfortunately, we were not able to find a single proposed method for expression profiling of small cell populations with well-documented positive evaluation of the criteria of our checklist. Additionally, the scientific literature provided evidence that the rates of false positive and false negative discoveries were sufficiently high to eliminate several of the proposed methods as viable discovery tools3,8,10,13. Our independent reanalysis of the raw data for some of the proposed methods came to the same conclusion14.

Due to our long-lasting positive experience with the Whole Genome Amplification (WGA) chemistry, we decided to evaluate WTA2 against the criteria described in the checklist. To evaluate how well the amplified cDNA represents differential expression between different biological samples, we used Human Gene ST 1.0 arrays (Affymetrix) as our readout for gene expression.

Table 1
Questions to be answered when an expression profiling method for small cell populations is evaluated.
Do picograms of RNAs provide results similar to standard methods from micrograms of the same RNAs?
Do expression profiles from picograms of RNA correlate as well as standard methods to qPCR results generated without pre-amplification?
Do technical replicates of the profiling method from picograms of RNA correlate well to each other?
Do expression profiles from small subpopulations correlate well to the results from bigger populations of the same cell type?


back to top Back to Top

Material and Methods

Materials and methods are described in detail in reference14 and can be found at www.dnaarrays.org.

back to top Back to Top

Adapting WTA2 for Affymetrix expression arrays

Since Affymetrix uses biotin to detect cDNA (or cRNA) hybridized to their microarrays, the first adaptation was to label the WTA2 amplified cDNA with this label. This was accomplished by fragmenting the WTA2 amplification products followed by biotinylation using terminal transferase. This labeling method was formerly developed by us for quantification of genomic DNA on Affymetrix expression arrays to perform Comparative Genome Hybridization (CGH)15. To evaluate how the WTA2 results for differential expression compare to results using the amplification/labeling protocol of the microarray manufacturer, we first evaluated results for the two best characterized RNA samples, namely the samples A and B of the Microarray Quality Control study (MAQC study)16 at the concentrations recommended by the manufacturers of both chemistries. We recommend using WTA2 for 25 ng total RNA and Affymetrix recommends using its chemistry for Gene ST arrays for 100 ng total RNA. Results of differential expression between MAQC samples A and B were highly similar between both chemistries (Figure 1).

Figure 1. Evaluation of WTA2 for expression profiling.A, Principle Component Analysis (PCA) of expression profiles of MAQC samples A (red) and B (blue) processed in triplicate following the manufacturer’s recommendations (cubes, data re-analyzed from17) and processed seven times on different days using WTA2 (globes); the contribution of the specific component is shown next to its axis. B, correlation of differential expression between MAQC samples A and B measured by the microarray manufacturer’s method and using WTA2; average values of triplicates are displayed and all measurements of all probe sets are displayed. Image reprinted from14.


back to top Back to Top

Expression profiling from RNA equivalents of a few cells

Next we evaluated how comparable the results for differential expression are between 25 ng total RNA (as recommended by Sigma for WTA2) and 100 pg, the equivalent of approximately 10 cells11. To do so, we added SYBR®-Green to the amplification reaction to stop amplification once sufficient cDNA was generated. For 25 ng RNA, SYBR-Green signals reached the stationary phase after 17 cycles, for 100 pg, 23 cycles were performed (Figure 2A). The entire workflow from RNA isolation to measuring the entire transcriptome on Affymetrix® arrays is outlined in Figure 3. We named this procedure Pico Profiling. Results of differential expression between MAQC samples A and B were highly similar between measurements from 25 ng and 100 pg total RNA (Figure 2B and 2C). Technical replicates of 100 pg RNA showed only slightly higher variability across replicates than replicates of 25 ng RNA processed using WTA2 or 100 ng RNA using Affymetrix’s recommended chemistry (compare Figures 1A and 2B).

Figure 2. Evaluation of expression profiles from pg of RNA (Pico Profiling).A, SYBR Green amplification signals from 1,000, 100, 10, 1, and 0 cells. B, PCA of expression profiles of MAQC samples A (red) and B (blue) processed in triplicate starting from 25 ng RNA (cubes) and 100 pg (globes); the contribution of the specific component is shown next to its axis. C, correlation of differential expression between MAQC samples A and B measured from 25 ng and 100 pg RNA; average values of triplicates are displayed, all measurements of all probe sets are displayed. D, correlation of differential expression measured by Pico Profiling versus qPCR. Image reprinted from14.

Figure 3. Workflow of Pico Profiling. — Cells are lysed, RNA is purified by magnetic beads, cDNA is synthesized followed by library preparation and amplification; after column purification, cDNA is fragmented and biotinylated, followed by hybridization to a microarray. Image reprinted from14.

back to top Back to Top

Correlation of results from WTA amplified material to qPCR data

Quantitative Real-Time PCR (qPCR) is considered the gold standard for measurements of differential gene expression. We evaluated the correlation of differential expression measured from 100 pg total RNA on microarrays relative to qPCR data and compared the results to the correlation of microarray measurements from standard amounts of RNA relative to qPCR. The MAQC study16 performed measurements of close to 1,000 transcripts and we utilized this data for comparison. Microarray measurements performed from micrograms and nanograms RNA according to Affymetrix’s protocols showed a correlation of r2 = 0.76 (16 and data not shown). Figure 2D shows that the correlation for minute amounts of RNA like 100 pg is minimally reduced.

back to top Back to Top

Expression profiling from a few cells

Based on our results obtained from the above-described experiments, we are confident that WTA2 can be utilized for expression profiling from minute amounts of highly purified RNA. To utilize it in biological research projects, it would be necessary to isolate RNA from very small cell populations. Since column based RNA isolation methods often lose a fixed amount of RNA due to non-specific adsorption on the column surface, we developed an RNA isolation procedure using magnetic beads14. We used SW480 and SW620, two cell lines derived from a primary tumor and a metastasis of the same patient. We sorted 2,000 cells of each of the two cell lines and 10 cells of each. We performed 17 and 23 cycles of amplification respectively. Differential gene expression between both cell lines was virtually identical independent of the number of cells used for RNA isolation (Figure 4).

Figure 4. Pico Profiling from 10 cells.A, PCA of expression profiles of SW480 cells (red) and SW620 cells (blue) processed in triplicate starting from 2,000 cells (cubes) and 10 cells (globes); the contribution of the specific component is shown next to its axis. B, correlation of differential expression between SW480 and SW620 cells measured from 2,000 and 10 cells; average values of triplicates are displayed. C, quantification of false positive and false negative rates for expression profiling from 10 cells versus 2,000 cells. Image reprinted from14.

back to top Back to Top


We evaluated the capability of WTA2 amplification chemistry to faithfully represent transcriptional differences between samples from very small cell populations in the micrograms of cDNA that can be generated. Using well-characterized, highly purified RNAs as well as RNA samples isolated from cell populations as small as 10 cells, we have shown that differential gene expression can be determined after WTA2 amplification with basically the same accuracy as from much higher amounts of RNA or much higher cell numbers without amplification.

cDNA from WTA2 amplification can be used for a variety of downstream analysis methods like qPCR, microarray analysis or Next-Generation-Sequencing (NGS). NGS analysis of WTA2 amplified cDNA is currently constrained by two facts. First, during the amplification procedure adapters are added to both ends of each cDNA fragment and the sequences of these adapters would flood the acquired data with biologically non-relevant sequences. Second, WTA2 amplifies ribosomal RNA (rRNA) with comparable efficiency as mRNA (data not shown). rRNA makes approximately 70% of total RNA, leading to 70% of the generated sequences to be derived from this one class of molecules. rRNA removal prior to amplification could eliminate the second constrain while increasing length of NGS-reads could minimize the percentage of sequenced nucleotides derived from the amplification adapters.

We also performed expression profiling after WTA2 amplification on individual cells and we observed hundreds of differentially expressed genes between individual cells of seemingly homogenous cell populations. From a philosophical point of view, it is almost impossible to evaluate if these observed differences between individual cells represent biological differences or are a result of amplification artifacts. Since the individual cells are destroyed during RNA isolation, no biological material is left of the same individual (cell) to confirm findings by alternative methods.

In summary, WTA2 amplification together with our novel RNA isolation procedure allows to generate almost indefinite amounts of cDNA that faithfully represent biological differences between very small cell populations, if not even individual cells.

back to top Back to Top




  1. Lipson D, Raz T, Kieu A, et al. Quantification of the yeast transcriptome by single-molecule sequencing. Nat Biotechnol. 2009;27:652–658.
  2. Eberwine J, Yeh H, Miyashiro K, et al. Analysis of gene expression in single live neurons. Proc Natl Acad Sci U S A. 1992;89:3010–3014.
  3. Jensen KB, Watt FM. Single-cell expression profiling of human epidermal stem and transit-amplifying cells: Lrig1 is a regulator of stem cell quiescence. Proc Natl Acad Sci U S A. 2006;103:11958–11963.
  4. Tietjen I, Rihel JM, Cao Y, Koentges G, Zakhary L, Dulac C. Single-cell transcriptional analysis of neuronal progenitors. Neuron. 2003;38:161–175.
  5. Kurn N, Chen P, Heath JD, Kopf-Sill A, Stephens KM, Wang S. Novel isothermal, linear nucleic acid amplification systems for highly multiplexed applications. Clin Chem. 2005;51:1973–1981.
  6. Tomlins SA, Mehra R, Rhodes DR, et al. Whole transcriptome amplification for gene expression profiling and development of molecular archives. Neoplasia. 2006;8:153–162.
  7. Iscove NN, Barbara M, Gu M, Gibson M, Modi C, Winegarden N. Representation is faithfully preserved in global cDNA amplified exponentially from sub-picogram quantities of mRNA. Nat Biotechnol. 2002;20:940–943.
  8. Subkhankulova T, Livesey FJ. Comparative evaluation of linear and exponential amplification techniques for expression profiling at the single-cell level. Genome Biol. 2006;7:R18.
  9. Kurimoto K, Yabuta Y, Ohinata Y, et al. An improved single-cell cDNA amplification method for efficient high-density oligonucleotide microarray analysis. Nucleic Acids Res. 2006;34:e42.
  10. Tang F, Barbacioru C, Wang Y, et al. mRNA-Seq whole-transcriptome analysis of a single cell. Nat Methods. 2009;6:377–382.
  11. Sambrook J, Russell DW. Molecular cloning : a laboratory manual (ed 3rd). Cold Spring Harbor, N.Y.: Cold Spring Harbor Laboratory Press; 2001.
  12. Viale A, Li J, Tiesman J, et al. Big results from small samples: evaluation of amplification protocols for gene expression profiling. J Biomol Tech. 2007;18:150–161.
  13. Clement-Ziza M, Gentien D, Lyonnet S, Thiery JP, Besmond C, Decraene C. Evaluation of methods for amplification of picogram amounts of total RNA for whole genome expression profiling. BMC Genomics. 2009;10:246.
  14. Gonzalez-Roca E, Garcia-Albeniz X, Rodriguez-Mulero S, Gomis RR, Kornacker K, Auer H. Accurate expression profiling of very small cell populations. PLoS One;5:e14418.
  15. Auer H, Newsom DL, Nowak NJ, et al. Gene-resolution analysis of DNA copy number variation using oligonucleotide expression microarrays. BMC Genomics. 2007;8:111.
  16. Shi L, Reid LH, Jones WD, et al. The MicroArray Quality Control (MAQC) project shows inter- and intraplatform reproducibility of gene expression measurements. Nat Biotechnol. 2006;24:1151–1161.
  17. Pradervand S, Paillusson A, Thomas J, et al. Affymetrix Whole-Transcript Human Gene 1.0 ST array is highly concordant with standard 3' expression arrays. Biotechniques. 2008;44:759–762.

back to top Back to Top

Related Links