1. Ito et al. Genome-Scale In Vivo Protein Interaction Testing
A Comprehensive Two-Hybrid Analysis to Explore the Yeast Protein Interactome
Ito et al.
Genome-Scale In Vivo Protein Interaction Testing

2. Background
So you sequenced a genome, and now have thousands of gene candidates of unknown function. We’ve all been there.
When bioinformatic analysis of the open reading frame fails to identify a gene’s function, functional genomic techniques are used.
Functional genomics is a field of techniques used to find the functions or interactions of previously found genes.
In vivo testing is naturally preferable to in vitro for functional genomics.

3. After sequencing and running against databases, 47% of the genes are either hypothetical or unidentified.
Hypothetical: there is a predicted gene there based on statistical analysis, but it may not be expressed.
Unidentified genes are known to exist experimentally, but nothing is known about what they do.

4. Background - Methods
As a test of two-hybrid analysis at the genome scale, Ito wanted to analyze how all ~6000 proteins in Saccharomyces Cervesiae, baker’s yeast, interacted with each other.
Microarrays (in this case protein chips) are one-to-many, but not a good representation of in vivo conditions.
6000 proteins * 1 microarray per protein = $$$
Yeast two-hybrid analysis with two modified proteins was established for large-scale use at this point, but had not been tested on the genome scale.

5. Background – Why Yeast?
How yeast breeds makes two-hybrid analysis easy: Two haploid strains mate to create one diploid strain, and their plasmids mix.
Yeast have the genders A and alpha, which do not fuse with themselves.
Ito used strains (a) PR69-2A for bait plasmids and (alpha) MaV204K for prey plasmids.
PR69-2A used adenine and histidine auxotrophic reporter genes, MaV204K used uracil and histidine, all driven by the Gal4 promoter.

6. Hypotheses
This was a pilot project to show the feasibility of genome-scale analysis of protein interactions using the two-hybrid method.
Goals were to investigate protein interactions and compare the results to a replicate using the same method.
Ito hypothesized the replicate would confirm the same protein interactions he found.
The knowledge of protein interactions gained from this study will lead to future experiments regarding interesting protein interactions.

7. The Yeast Two-Hybrid Protein Interaction Test
First, Ito purchased yeast strains that can only survive with a certain gene expressed: Adenine, uracil, and histidine-producing enzymes.
He gave that ‘reporter’ gene a custom promoter with a known sequence, which he took from the gene Gal4.
Created a custom transcription factor for that promoter in two parts: the DNA-binding domain and the activation domain. The promoter will activate when both parts are interacting with each other, but they do not interact on their own.
Cassette, homologous recombination, CRISPR-Cas9, etc.

8. Yeast Two-Hybrid Test: Protein Interactions
The two halves of the transcription factor are spliced into the plasmid along with the two proteins of interest. The plasmids are called the “bait” and the “prey”, and each has a unique ID tag called an IST, or interaction sequence tag.
If the proteins interact, they will bind each other, creating a whole transcription factor that will go on to bind the promoter for the reporter gene. Four reporter genes were used to eliminate false positives, and the genes were checked for activation using two methods.
If the proteins do not interact, the TF is never formed and the reporter gene is not activated.
Assay is by observing the action of the reporter gene. In Ito’s case, colonies would form in a petri dish of deficient media after the yeast had mated.

9. RNA Polymerase II Complex
The bait and prey sequences interact…
which causes the active and binding domains to interact, forming a whole promoter...
Which recruits the transcription complex to the reporter gene.
IST for ID
Bait Sequence – one ORF
Plasmid: pGAK-RC
Prey Sequence – another ORF
Plasmid: pGBK-RC
Gal4 Active Domain
Gal4 Binding Domain
Gal4-Homologous Promoter
Reporter gene: Produces a necessary amino acid
RNA Polymerase II Complex

10. Two-Hybrid Tests at Genome Scale
Ito made a bait plasmid and a prey plasmid out of each protein to be tested. All baits were gender a, all preys were gender alpha.
Now Ito was ready to observe interactions between any two proteins by crossing any bait and prey samples, and watching for the reporter genes.
How could this method be automated or scaled-up from two proteins to the whole transcriptome? By doing the tests 100 at a time.
3.5x107 matings were performed, 96 at a time. 62 pools of bait and prey clones were mixed from individual strains of one gender arbitrarily, and those pools were mated to each other in a Millipore sampling manifold.

11. Genome Scale Methodology
3,844 matings were performed by mixing a bait strain pool (with a Gal4 DNA-binding domain) and a prey strain pool(with a Gal4 activation domain), then selecting for the reporter genes ADE2, URA3, and HIS3 on deficient media, which all use GAL4 promoters in the custom strain of Cervesiae used for the experiment.
Interactions that activated the reporter genes (and so survived) were transplanted to another plate containing 5-bromo-4-chloro-3-indolyl-a-D- galactopyranoside, which confirmed activation of the endogenous gene Mel1, which is normally activated by Gal4. This verifies there is no loss-of- function associated with the exogenous Gal4.
Once an interaction is found, the IST in the plasmid was sequenced and BLAST and the Yeast Protein Database were checked for the protein.

12. Results
3,844 mating reactions produced 15,523 surviving colonies in which all three reporter genes were activated (Histidine+, Uracil+, Adenine+).
The ISTs of each clone were sequenced, reducing the dataset to 4,549 unique high-certainty interactions involving 3,278 proteins. Some proteins were part of two or three reactions.
The amount of interactions the screen found that are already known (in the database) indicates the screen’s sensitivity.
Uetz
691 Interactions
88 Known
12.7% Known
Ito
841 Interactions
105 Known
12.5% Known

13. Results – the Bad News
Strangely, with equally sensitive screens, only 135 (28.3%) of all known interactions were found independently by both studies.
Neither study found even a fifth of the known interactions they were looking for: There is substantial randomness in this method at the genome scale.

14. Results – the Good News: Protein Networks
Protein network sizes vary wildly: One network of 417 proteins, and 132 networks of 2-12 proteins. The large network probably reflects crosstalk inside the cell, since Ito checked the YPD to look for networks of similar size, and found networks containing up to proteins in projects that did not use two-hybrid methods.
This is excluding RNA polymerase proteins. Including them, 87% of all proteins are in one cluster.
The authors have developed a tool to query databases to indicate if two proteins are found in the same complex, are co-expressed, their interactions, and if those interactions have been independently confirmed.

15. Interesting Protein Networks
Autophagy. Arrows in one direction indicate an interaction from bait to prey, arrows in both directions indicate two-way interactions. Circular interactions indicate a protein complex.
Apg16-Mec3 is involved in the G2 DNA damage checkpoint.
Spindle pole function. Black proteins are hypothetical, of unknown function. The authors would like to do knockout analysis of these proteins: Interactions do not tell you everything.

16. Conclusions
A genome-scale two-hybrid analysis of the transcriptome of saccharomyces Cervesiae found novel protein networks for further study.
However, a replicate of this project failed to duplicate these results.
Errors in PCR while amplifying ORFs.
Different plasmids for inserting ORFs may affect folding between replicates.
Different reporter genes in the replicate study.
Reporter genes activate for no apparent reason sometimes. The authors may not have corrected for this perfectly.
Mating may not occur between every pair of yeast strains in a pool of 200, leading to false negatives.
Not all interactions cause reporter gene activation to the same degree, but that is difficult to correlate to interaction strength.

17. Further Reading
An earlier test of this two-hybrid method: Ito, T., et al. "Toward a Protein-protein Interaction Map of the Budding Yeast: A Comprehensive System to Examine Two-hybrid Interactions in All Possible Combinations between the Yeast Proteins." PROCEEDINGS OF THE NATIONAL ACADEMY OF SCIENCES OF THE UNITED STATES OF AMERICA 97.3 (2000): Web. 28 Feb
The independent replicate of this study: Uetz, P., et al. "A Comprehensive Analysis of Protein-protein Interactions in Saccharomyces Cerevisiae." Nature  (2000): Web. 28 Feb