Transcriptomics

= subfield of functional genomics that focuses on gene expression

-> typically with a focus on mRNA (transcripts),

-> non-coding RNA can also be analyzed

= amount of mRNA from each gene that is present in a particular cell, tissue, or organism; considered as a phenotype

-> sometimes “intermediate phenotype” since it is very close to the genotype in the pathway which determines the final phenotype of an organism

DNA -> mRNA -> protein -> interactions with internal/external environment -> phenotype

Because mRNAs correspond to particular genes in the genome, it is often possible to establish a link between a genotype and an expression phenotype

• Large-scale, high-throughput methods

  • EST sequencing

  • Microarrays

  • Affymetrix "Affy" GeneChips™

  • SAGE

  • RNA-seq

• Questions (examples):

  • How much transcript is there from each gene (expression level)?

  • How does expression level change over development (expression profile)?

  • How does environment/treatment affect gene expression?

= first genome-wide method used to investigate gene expression

-> has been applied on e.g. Drosophila, human, etc.

Usage

• EST databases used to estimate expression levels when there is no other experimental evidence.

• the more times an EST corresponding to a particular gene was sequenced, the higher is the expression level of that gene.

Steps

• reverse transcription of mRNA (typically not all of it) to cDNA

• sequence large number of cDNA

  -> resulting sequence fragments = EST (Expressed Sequence Tags)

Pro

• gives an estimate of absolute mRNA abundance (if cDNA library is random)

• very useful for gene discovery (annotation of expressed regions of genomes and intron/exon boundaries)

Con

• expensive, time-consuming, requires large-scale sequencing

• redundat sequencing (ESTs from highly-expressed genes are sequenced many times)

• must sequence 100,000's of ESTs to get a good representation of genes expressed at low levels

• the ESTs only reveal gene expression levels in the particular tissue or sample that was used for mRNA preparation

• possibly missing genes expressed in specific tissues, cells, developmental stages, etc.

Construction

• attach specific DNA sequences (probes/spots) to a solid surface (often a glass microscope slide)

  -> many thousands of probes, each matching a different gene

Pro

• can quickly, cost-efficiently do many comparisons and replicates

Con

• do not measure absolute, but relative abundance; statistical interpretation may be difficult

Types of DNA as probes:

• cDNA or EST sequences

  • (-) need to be cloned for each gene that is used

  • (+) longer probe sequences, which may give better hybridization signals (especially for cross-species comparisons)

• PCR-amplified genomic DNA

  • (+) can be made to all predicted genes in the genome

  • (+) specifically designed to reduce crosshybridization (by avoiding sequence regions that are similar in two or more genes)

• Synthesized oligonucleotides (typically 36–80 bases long)

  • advantages same as PCR-amplified genomic DNA

• Purification of RNA from the samples to be campared

• Reverse transcription of mRNA to cDNA while labeling with fluorescent dye (one sample “red”, the other “green”)

• Place labeled cDNA solution on same microarray (under a coverslip) in equal amounts and hybridized overnight

• Removal of excess and unbound DNA by washing

• Let array dry and scan with laser to create graphical image

• Analyze image to determine relative expression difference (red/green signal intensity for each spot)

1. Fold-change

   • chose an arbitrary fold-difference to define genes that are differentially expressed

   • e.g. fold chance = 2: gene must have an expression level that is at least 2 times higher in one sample than in the other to be considered differentially expressed

   • often usage of log-scale: if log ratio of expression is >1 (or < –1) would be considered as differentially expressed

2. Statistical cutoff

   • = statistical method, such as a t-test, binomial test, ANOVA, or a Bayesian method is used to calculate a p-value for each gene

   • null hypothesis: gene is expressed equally in the two samples

     -> reject, if p-val < cutoff and consider gene as differentially expressed

   • typical p-val cutoff = 0.05 cannot be used for multiple testing, instead use False Discovery Rate (FDR) of 5%

-> The approaches can be displayed graphically with a “volcano plot”, which shows the fold-change in expression between two samples (on a log2 scale) on the X-axis, and the pvalue (typically on a –log10 scale) on the Y-axis.

• Affymetrix produces and sells microarrays (GeneChips) for several model species (e.g. human, Drosophila)

Construction

• arrays made by photolithography

  -> = specific oligonucleotide probes (25 bases long) are synthesized directly on the array surface

• For each gene, 20 different probes corresponding to different regions of the transcript are present on the array

• 20 “mismatch controls” that are identical to the above probes except for a single mismatched nucleotide at the center of the sequence (base 13)

Expression level estimation

• Estimate expression level of each gene by the intensity difference between match and mismatch probes, averaged over all 20 probes per genes.

  -> not a competitive hybridization of two different samples. Only one sample is hybridized per array.

Pro

• can buy pre-made chips

• high quality control

• high standardization

• easy to use

Con

• can be expensive

• requires Affymetrix machines

• short probes (25 bases) are not good for divergent species

• useful only for (model) species for which a GeneChip is commercially available

= method that is similar to EST sequencing, but more efficient because only short “tags” of around 10–15 bases are sequenced from each cDNA

Steps

• Before sequencing, the tags are concatenated so that many of them can be sequenced in a single Sanger sequencing reaction

  -> requires annotation of the genome, so that the tags can be accurately mapped back to their corresponding genes

• Purify mRNA (poly-A) from sample

• Use biotinylated oligo dT primer to synthesize double-stranded cDNA

• cut cDNA with a restriction enzyme, such as NlaIII which recognizes the sequence CATG and cuts, on average, every 256 bp

• purify only the 3' poly dT ends of the cut cDNA in a streptavidin column (binds to biotin attached to the oligo dT primer)

• ligate an adapter (short synthesized DNA sequence) to the cut end. The adapter contains a restriction site for the restriction enzyme BsmFI (recognizes GGGAC, but cuts 15 bp away from this sequence into the cDNA fragment)

• ligate two adapter ends to each other tail-to-tail to create “ditags”. PCR amplify the ditags with primers complementary to the to adapter sequence

• cut again with NlaIII to remove adaptors, leaving a 30 bp ditag

• ligate many ditags end-to-end (up to 1 Kb total length), then sequence 1000's of these. (typically sequence 30-40 tags per Sanger sequencing reaction)

-> Each 15-bp tag should give a unique match to a transcript in the genome (random match is very unlikely) and should always be after the 3' most NlaIII site (if not: genes may be missed)

-> To quantify the expression level of a gene: count the number of times that the tag for that gene is sequenced. At least 10,000–50,000 tags should be sequenced to get an accurate estimate of expression (≈300–1000 sequencing reactions).

Pro

• gives an estimate of absolute transcript abundance

• more efficient than large-scale EST sequencing, because many fewer sequencing reactions are required.

Con

• still requires much sequencing, which can be expensive

• not accurate for rare transcripts

• sometimes difficult to map tags to genes

• must be repeated for each sample (tissue, sex, treatment, etc.)

Steps

• follows the same scheme outlined above for EST sequencing.

• difference is that a pool of cDNA is used for next generation sequencing.

• With this approach hundreds of millions of short EST sequences (usually 50–300 bases in length) can be generated quickly and at low cost.

• These sequences can be mapped back to their corresponding gene in the genome and used to quantify gene expression.

• RNAseq can also be applied to species with unknown or un-annotated genomes to assemble the transcriptome de novo.

Pro

• produces direct "counts" of gene expression with very large sample sizes (number of reads per gene), which is good for statistical analysis and comparison of expression between samples

• Can be applied to non-model organisms.

• Can assemble transcriptomes de novo.

Con

• may be more expensive than microarray technologies (but the costs of RNA-seq are dropping rapidly)

• Requires more complex bioinformatic analysis than arrays