Measuring single cell RNA expression levels find considerable transcriptional differences among phenotypically identical cells - BMC Genomics

Summary:
Single cell gene expression profiling. Shift towards understanding similarities and differences between individual cells at the transcriptional and translational level. Gene expression levels may be distorted by sampling effects (copying and amplying mRNA pool). Particularly problematic for low copy number transcripts in single cell samples (random dropouts of low abundance transcripts from amplified single cell cDNA populations). Magnitude of distortion will also depend on transcript abundance distribution. High number of genes with transcript abundances lower then 10-20 and relatively few genes with high transcript. Study finds that the majority (44%) of genes are represented by limited number of mRNA copies (less 25), and this may account for the large cell-to-cell variations in mRNA copy number that we have observed. They also conclude that sampling effects do not impede our ability to extract reliable gene expression profiles from single cells and that significant differences in gene expression levels exist between phenotypically identical cells. (Published: 03/06/08)

Notes:

• Single-cell gene expression profiling provides a powerful tool to analyze the composition of complex cell populations
• many contexts in which the focus is shifting towards understanding the cellular networks of individual cells and the similarities and differences between individual cells at the transcriptional and translational level

• Limitations to the sensitivity and resolution of current technologies for studying gene expression mean that when using samples as small as those generated from single cells we are inevitably faced with amplifying cellular mRNA.
• amplification stage may introduce significant distortions in the measured gene expression levels
• especially for genes with small numbers of transcripts in the material under study
• this distortion is introduced by sampling effects that arise from inefficiencies in the processes of copying and amplifying the original mRNA pool.

• In a complex mRNA population with small absolute numbers of individual transcripts, such as that from a single eukaryotic cell, sampling effects can result in only a subset of the population of starting RNA molecules being represented in the final amplified population

• particularly problematic for low copy number transcripts in single cell samples:
• in the first step of the process, reverse transcription may fail for a small proportion of the original mRNA molecules
• eliminated from subsequent amplification and detection
• For genes with only a small number of transcripts in the starting material, this will create a variable (assuming the failures are random) distortion in the relative representation of transcript abundances in the final experimental sample
• potentially leading to the absence of such low abundance transcripts in the final amplified population.
• first round of PCR amplification will have a similar effect, and subsequent rounds will have effects of diminishing importance, in terms of complete dropout of lowabundance transcripts

• overall effect of random dropouts of low abundance transcripts from amplified single cell cDNA populations would be that random sets of transcripts would be called as absent in different cells

• one estimate is that there is a lower limit of 80 copies of a single mRNA per cell for detection of two-fold differences between samples

• magnitude of the overall sampling effect will, in theory, depend on two factors:
1. the transcript abundance distribution which is the variation of transcript number among genes being expressed in a cell (and in particular the relative numbers of genes with low transcript numbers);
2. and the copying and amplification efficiencies for conversion of the original population of mRNA molecules into DNA or RNA detectable by the expression profiling platform in use

• The copying and amplification efficiencies can be estimated from experimental data. However, the estimation of the transcript abundance distribution poses two distinct problems: knowing the form of the distribution; and evaluating the shape and scale parameters for the distribution.

• conflicting reports of the transcript abundance distribution in a typical eukaryotic cell
• ranging from a distribution with a median value for mRNA transcript copies per gene of less then one
  
• to a distribution with a median of approximately 100 copies

• difficulty is that, in general, the transcript abundance distributions of real single cells are not known but are inferred from population measurements

• Based on published data, a simple approximation is that the transcript abundance distribution is log-log-normal, as this distribution captures certain key features of our current understanding of the single cell transcript abundance distribution:
• there is a high number of genes with transcript abundances lower then 10-20 and relatively few genes with high transcript

Discussion

• The main findings of this study are that the contribution of sampling effects to observed single cell expression data is likely to be minor and that substantial transcriptional differences exist between phenotypically identical cells.
• indicates that one can generate reliable gene expression profiles from single cells using microarrays to interrogate globally amplified RNA populations
• However, the considerable variation in gene expression levels between similar cells is likely to dictate that relatively high numbers of cells would need to be analysed to robustly identify significant and consistent differences in gene expression between cell populations.
• Alternatively, these findings argue that single cell expression profiling will be particularly useful for identifying absolute differences in gene expression between cell types.

• A second implication of this study is that one important limit on the use of amplification techniques for single cell expression profiling is that if amplification efficiency drops significantly below 90% then the sampling effect may considerably distort the measured expression profile
• One promising technique for mRNA amplification from individual cells, which combines global exponential and linear amplification, has been shown to produce very low levels of noise and highly reproducible data and may limit the significance of sampling effects when profiling rare transcripts [22].

• Our results demonstrate that the actual transcript abundance distribution for the tested cell type has a peak at approximately 5-20 copies per gene.
• We recognize that our experiments are based on a particular type of mouse neural stem cell, but in the absence of any reason to suppose that the transcript distributions of most other cell types are radically different from this, we believe the result should generally apply to expression experiments performed on a wide range of cell types.
• Although our method did not allow us to discriminate between different models of overall gene and transcript numbers in the cell, we believe it strongly suggests that more then 85% of transcripts are present in relatively low copy numbers (less then 100 copies per cell).

• Insight into the variability of the gene expression profiles of single cells has been obtained using a number of technical approaches, incuding microarray analysis following linear T7-based amplification [16, 25], multiplexed FISH (fluorescence in situ hybridization) [26] and quantitative PCR [27].

• Transcriptional bursting has been observed in Escherichia coli, in which protein levels have very little correlation with mRNA levels, particularly for younger cells [28], as well as Dictyostelium [29] and mammalian cells [30].
  

• Overall, those findings are consistent with a model for cellular phenotypes that are underwritten by transcriptional programs that appear inherently noisy when total cellular transcript levels are measured at the single cell level.
• It has been suggested that because in the individual cell the transcriptional machinery is controlled by a relatively small number of transcription factors, it may result in stochastic behavior in gene activity.

Conclusions

• Our current results revealed that the majority (44%) of genes are represented by limited number of mRNA copies (less 25), and this may account for the large cell-to-cell variations in mRNA copy number that we have observed.
• also concluded that sampling effects do not impede our ability to extract reliable gene expression profiles from single cells and that significant differences in gene expression levels exist between phenotypically identical cells

Expand notes

Summary only...

Strategies For Microarray Analysis of Limiting Amounts of RNA - Brief Funct Genomic Proteomic.

Summary:
This review evaluates current signal and sample amplification technologies, including those that can be used to generate labelled cDNA populations for array analysis from as little as a single cell. Options for expression profiling are to increase cDNA labelling and hybridisation efficiency, or to use an amplification strategy to generate enough RNA/cDNA for use with a standard labelling method. Sample amplification approaches must preserve the representation of the relative abundances of the different RNAs within the starting population and must also be highly reproducible. (Briefings in Functional Genomics and Proteomics, Vol 2, No 1, 31-36, April 2003)


Notes:

Introduction

• microarrays
• have become a standard technology for measuring relative and absolute levels of gene expression
• interest in increasing resolving power of this technology has grown
• particularly in terms of input material required to generate robust data
• drive for this
  
• partly technical
• partly motivated by biological and clinical concerns
• main goal is to use defined populations of cells or small pieces of complex tissue (e.g. clinical biopsy) for expression profiling
  
• associated with a reduction in the amount of cells that can be harvested
• e.g. laser capture microdissection: possible to collect defined cells from fesh and fixed tissue sections
• ultimate aim of this increase in resolution:
  
• to enable reproducible expression profiling at the level of single cells
• several reports that this is currently feasible
• likely to be in general use in the near future
• current methods
• require microgram amounts of total RNA for generating labelled cDNA populations for microarray analysis
• equivalent of over 1 million cells
• efforts to reduce this requirement focus on two complementary approaches:
  
• signal amplification and detection
  
• allowing the use of smaller amounts of input RNA
• RNA sample amplification
• to generate enough material for standard labelled cDNA synthesis, hybridisation and detection
• combination of both

Signal versus sample amplification: theory

• ideally:
  
• extract the RNA from a single cell, directly label that RNA and hybridise it to some form of microarray
• many practical issues
• from: difficulty of harvesting picogram quantities of RNA contained in a typical cell
• to: hybridisation kinetics for very small numbers of molecules at relatively low concentrations
• mRNA abundance: three classes (tissue based estimates from brain cDNA libraries)
1. high abundance transcripts
• ~1/6th of the mass of mRNA
• represents 100 different transcripts
2. medium abundance
• ~45% of mass of mRNA
• 2,000 different transcripts
3. low abundance
• ~40% of mass of mRNA
• 45,000 different transcripts
• inherent technical challenge in labelling all of these low abundance transcripts for microarray hybridisation under any circumstances and developing hybridisation conditions that would ensure that all molecules hybridise in a reasonable timeframe
• under conditions where the input RNA and the corresponding absolute numbers of each low abundance transcript are low, these problems become more accute, with less room for errors in each step of the generation of labelled cDNA population
• a final technical hurdle is the detection of the extremely small numbers of molecules harvested from single cells

Signal amplification

• currently: two main methods for generating labelled cDNA populations for array analysis
1. direct incorporation of fluorescent label-conjugated nucleotides
2. incorporation of modified nucleotides followed by dye coupling to those modified nucleotides
• amino-allyl labeling method
• introduced for
• relative cost reasons
• reduce the biases in incorporation rates of different fluorophore-labelled nucleotides
  
• in widespread use and commercial kits available
• novel strategies
  
• labelling cDNA populations as well as amplifying that label such that smaller numbers of hybridised molecules can be reproducibly detected and quantified
• e.g.
• enzymatic amplification
• e.g. tyramide signal amplification
• use of dendrimers
• increases amount of label per nucleotide and thus per labelled cDNA molecule
• several hundred fluorescent tags per dendrimer
• input amounts of RNA down to 0.5ug
• still considerable amount
• hybridisation takes far longer than with standard methods
  
• due to size of molecules
  
• typically of order of several days
• alternative detection methods, e.g.
• quantum dots
• rolling circle amplification

Sample amplification

• amplification of the input RNA to generate enough material for standard labelled cDNA synthesis
  
• alternative to signal amplification
• currently: two approaches
1. PCR-based or exponential amplification
2. linear amplification
• linear amplification
  
• first described by Eberwine et al. as a method for single cell analysis; now common method
  
• antisense RNA synthesis from a population of double-stranded cDNA molecules, all carrying a standard recognition site for T7 RNA polymerase
• used in Affymetrix system
• curretnly, amplification of nanogram quantities of total RNA (equivalent of 50 - 1000 cells) requires two rounds of T7 linear amplification
• feature: shortening of the amplified transcripts, compared to their parent mRNA population, with the associated 3'-bias in the amplified material
• disadvantages:
• labour intensive
• requires synthesis and purification of double-stranded cDNA from the starting RNA, followed by at least one round of RNA synthesis and amplification
• this RNA is in turn used to synthesise double-stranded cDNA, followed by a second round of RNA synthesis
• typical time taken to generate amplified RNA from picogram quantities of input total RNA is of the order of 3-5 days
  
• PCR-based amplification
• general principle: introduction of PCR-priming sites at either end of each reverse-transcribed cDNA molecule, followed by global amplification of the entire population of molecules
• potential pitfalls (sources of sampling, non-representative amplification):
  
1. during each step
   
• failure to introduce priming sites to the ends of every RNA/cDNA molecule in the starting population will introduce sampling into the amplification process with under-representation and possible amplification of those molecules
2. during the oligo-dT primed reverse transcription steps
3. during the PCR itself
• when the exponential nature of the process amplifies any variations in the amplification efficiency of particular templates
• most significant source of error during PCR based amplification
• Clontech's SMART system
• has been succesfully used for generating labelled cDNA for array analysis from limiting amounts of RNA
• has been shown to preserve the relative abundance of RNA molecules in the amplified population
  
• advantages of PCR
• rapid (exponential) amplification of cDNA population: less than 1 day
  
• short, relatively simple protocols
• particularly useful in medium- and high-throughput situations where many smaples are to be studied
• amplified material can be labelled to generate labelled cDNA populations for array analysis in a number of different ways
• RNA generated by linear amplification can be labelled using standard direct and indirect labelling methods, or with signal amplification methods
• amplified cDNA can be labelled by random primer-mediated incorporation of either directly or indirectly labelled nucleotides

Pushing the system: the challenge of single cell expression profiling

• even with current labelling technologies, generating enough cDNA from a single cell for a single microarray hybridisation requires around 10^6-fold amplification of th emRNA content of that cell
• total degree of amplification depends on the cell type used, given the wide range of total RNA content in different cell types
• from as little as 1pg to as much as 50pg
• only 1-5% of this mass of RNA is composed of mRNA
• containing an estimated total of 100,000 - 300,000 molecules of mRNA
• amplifying 300,000 molecules of different abundances to generate this mass of material represents a considerable challenge
• the particular acute problems for amplifying single-cell material are
  
• the efficiency of priming the intial RT and
  
• the efficiency of the subsequent steps to prepare the cDNA for amplification
• be they the introduction of a second priming site for PCR amplification or production of dsDNA from the single stranded material
• failure of either step for a sub-population of the cDNA will result in the absence of detection of low abundance transcripts
• assuming that all amplification methods introduce some degree of error over the million-fold amplification procedure, it is likely that amplification from single cells is an inherently noisy procedure

Expand notes

Summary only...