A few abundant transcripts dominate RNA populations

: Klick image to enlarge

Typical eukaryotic mRNA populations consist of three abundance classes:

High abundance transcripts

High abundance transcripts include mainly housekeeping genes, such as GAPDH, actins and components of the transcriptional and translational machinery. They are expressed at very high levels - housekeeping genes make up 80-90% of all mRNAs in a cell. They represent the most frequent class of false positive interactions in yeast two-hybrid screens.

Medium abundance transcripts

These transcripts vary depending on the cell type and make up 10-20% of all mRNAs.

Low abundance transcripts

Low abundance transcripts encode many of the key signalling proteins of a cell. They are transcribed at very low levels, making up less than 1% of the total mRNA population. Yet, they encode key components and are often the main targets of protein interaction hunts.

Constructing cDNA libaries which represent low abundance transcripts is the central challenge of yeast two-hybrid screening.

Normalization of mRNA samples

A double stranded cDNA sample is normalized based on differential re-association kinetics after denaturation, since abundant transcripts re-associate faster than rare ones. Transcripts are denatured and then reannealed under carefully controlled conditions. The reannealing process is stopped after a short time, when the majority of abundant transcripts have reformed into double stranded cDNA, whereas rare transcripts are still single stranded.

Subsequently, double stranded cDNA is removed by digestion, the remaining transcripts are denatured again, and annealed to double stranded cDNA. The resulting cDNA population is normalized, i.e. the majority of abundant transcripts has been removed.

Finally, the normalized cDNA sample is size selected and ligated into the appropriate yeast two-hybrid vector.

Yeast two-hybrid screen using a normalized human universal cDNA library

Two sets of yeast two-hybrid screens were carried out using an N-terminally truncated form of human p53 as a bait. For screening, we used two libraries:

A standard (non-normalized) human colon library
A normalized human universal library

The parameters of the screens and the resulting interactors are shown in the table below.

Parameter	Non-normalized screen	Normalized screen
Bait	human p53	human p53
Library	human colon	human universal normalized
Screens carried out	3	1
Number of clones screened	43 million	3.5 million
Number of interactors	436	43
High stringency interactions	0	12
Known interactors	1	3

The results show that significantly fewer clones have to be screened with a normalized library to identify relevant, high stringency interactions. Simultaneously, many low stringency false positive interactions are removed (compare the number of primary interactions), resulting in an overall "cleaner" screen.

Despite a significantly lower coverage (3.5 million clones vs. 43 million clones), three known interactors of p53 were identified from the normalized cDNA library: